1 /* 2 * L2TPv3 IP encapsulation support 3 * 4 * Copyright (c) 2008,2009,2010 Katalix Systems Ltd 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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 13 14 #include <linux/icmp.h> 15 #include <linux/module.h> 16 #include <linux/skbuff.h> 17 #include <linux/random.h> 18 #include <linux/socket.h> 19 #include <linux/l2tp.h> 20 #include <linux/in.h> 21 #include <net/sock.h> 22 #include <net/ip.h> 23 #include <net/icmp.h> 24 #include <net/udp.h> 25 #include <net/inet_common.h> 26 #include <net/inet_hashtables.h> 27 #include <net/tcp_states.h> 28 #include <net/protocol.h> 29 #include <net/xfrm.h> 30 31 #include "l2tp_core.h" 32 33 struct l2tp_ip_sock { 34 /* inet_sock has to be the first member of l2tp_ip_sock */ 35 struct inet_sock inet; 36 37 u32 conn_id; 38 u32 peer_conn_id; 39 }; 40 41 static DEFINE_RWLOCK(l2tp_ip_lock); 42 static struct hlist_head l2tp_ip_table; 43 static struct hlist_head l2tp_ip_bind_table; 44 45 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk) 46 { 47 return (struct l2tp_ip_sock *)sk; 48 } 49 50 static struct sock *__l2tp_ip_bind_lookup(const struct net *net, __be32 laddr, 51 __be32 raddr, int dif, u32 tunnel_id) 52 { 53 struct sock *sk; 54 55 sk_for_each_bound(sk, &l2tp_ip_bind_table) { 56 struct inet_sock *inet = inet_sk(sk); 57 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk); 58 59 if (l2tp == NULL) 60 continue; 61 62 if ((l2tp->conn_id == tunnel_id) && 63 net_eq(sock_net(sk), net) && 64 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) && 65 (!inet->inet_daddr || !raddr || inet->inet_daddr == raddr) && 66 (!sk->sk_bound_dev_if || !dif || 67 sk->sk_bound_dev_if == dif)) 68 goto found; 69 } 70 71 sk = NULL; 72 found: 73 return sk; 74 } 75 76 /* When processing receive frames, there are two cases to 77 * consider. Data frames consist of a non-zero session-id and an 78 * optional cookie. Control frames consist of a regular L2TP header 79 * preceded by 32-bits of zeros. 80 * 81 * L2TPv3 Session Header Over IP 82 * 83 * 0 1 2 3 84 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 85 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 86 * | Session ID | 87 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 88 * | Cookie (optional, maximum 64 bits)... 89 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 90 * | 91 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 92 * 93 * L2TPv3 Control Message Header Over IP 94 * 95 * 0 1 2 3 96 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 98 * | (32 bits of zeros) | 99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 100 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length | 101 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 102 * | Control Connection ID | 103 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 104 * | Ns | Nr | 105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 106 * 107 * All control frames are passed to userspace. 108 */ 109 static int l2tp_ip_recv(struct sk_buff *skb) 110 { 111 struct net *net = dev_net(skb->dev); 112 struct sock *sk; 113 u32 session_id; 114 u32 tunnel_id; 115 unsigned char *ptr, *optr; 116 struct l2tp_session *session; 117 struct l2tp_tunnel *tunnel = NULL; 118 int length; 119 120 if (!pskb_may_pull(skb, 4)) 121 goto discard; 122 123 /* Point to L2TP header */ 124 optr = ptr = skb->data; 125 session_id = ntohl(*((__be32 *) ptr)); 126 ptr += 4; 127 128 /* RFC3931: L2TP/IP packets have the first 4 bytes containing 129 * the session_id. If it is 0, the packet is a L2TP control 130 * frame and the session_id value can be discarded. 131 */ 132 if (session_id == 0) { 133 __skb_pull(skb, 4); 134 goto pass_up; 135 } 136 137 /* Ok, this is a data packet. Lookup the session. */ 138 session = l2tp_session_find(net, NULL, session_id); 139 if (session == NULL) 140 goto discard; 141 142 tunnel = session->tunnel; 143 if (tunnel == NULL) 144 goto discard; 145 146 /* Trace packet contents, if enabled */ 147 if (tunnel->debug & L2TP_MSG_DATA) { 148 length = min(32u, skb->len); 149 if (!pskb_may_pull(skb, length)) 150 goto discard; 151 152 /* Point to L2TP header */ 153 optr = ptr = skb->data; 154 ptr += 4; 155 pr_debug("%s: ip recv\n", tunnel->name); 156 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length); 157 } 158 159 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook); 160 161 return 0; 162 163 pass_up: 164 /* Get the tunnel_id from the L2TP header */ 165 if (!pskb_may_pull(skb, 12)) 166 goto discard; 167 168 if ((skb->data[0] & 0xc0) != 0xc0) 169 goto discard; 170 171 tunnel_id = ntohl(*(__be32 *) &skb->data[4]); 172 tunnel = l2tp_tunnel_find(net, tunnel_id); 173 if (tunnel != NULL) 174 sk = tunnel->sock; 175 else { 176 struct iphdr *iph = (struct iphdr *) skb_network_header(skb); 177 178 read_lock_bh(&l2tp_ip_lock); 179 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr, 180 inet_iif(skb), tunnel_id); 181 if (!sk) { 182 read_unlock_bh(&l2tp_ip_lock); 183 goto discard; 184 } 185 186 sock_hold(sk); 187 read_unlock_bh(&l2tp_ip_lock); 188 } 189 190 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) 191 goto discard_put; 192 193 nf_reset(skb); 194 195 return sk_receive_skb(sk, skb, 1); 196 197 discard_put: 198 sock_put(sk); 199 200 discard: 201 kfree_skb(skb); 202 return 0; 203 } 204 205 static int l2tp_ip_open(struct sock *sk) 206 { 207 /* Prevent autobind. We don't have ports. */ 208 inet_sk(sk)->inet_num = IPPROTO_L2TP; 209 210 write_lock_bh(&l2tp_ip_lock); 211 sk_add_node(sk, &l2tp_ip_table); 212 write_unlock_bh(&l2tp_ip_lock); 213 214 return 0; 215 } 216 217 static void l2tp_ip_close(struct sock *sk, long timeout) 218 { 219 write_lock_bh(&l2tp_ip_lock); 220 hlist_del_init(&sk->sk_bind_node); 221 sk_del_node_init(sk); 222 write_unlock_bh(&l2tp_ip_lock); 223 sk_common_release(sk); 224 } 225 226 static void l2tp_ip_destroy_sock(struct sock *sk) 227 { 228 struct sk_buff *skb; 229 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk); 230 231 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) 232 kfree_skb(skb); 233 234 if (tunnel) { 235 l2tp_tunnel_closeall(tunnel); 236 sock_put(sk); 237 } 238 239 sk_refcnt_debug_dec(sk); 240 } 241 242 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 243 { 244 struct inet_sock *inet = inet_sk(sk); 245 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr; 246 struct net *net = sock_net(sk); 247 int ret; 248 int chk_addr_ret; 249 250 if (addr_len < sizeof(struct sockaddr_l2tpip)) 251 return -EINVAL; 252 if (addr->l2tp_family != AF_INET) 253 return -EINVAL; 254 255 lock_sock(sk); 256 257 ret = -EINVAL; 258 if (!sock_flag(sk, SOCK_ZAPPED)) 259 goto out; 260 261 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip)) 262 goto out; 263 264 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr); 265 ret = -EADDRNOTAVAIL; 266 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL && 267 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 268 goto out; 269 270 if (addr->l2tp_addr.s_addr) 271 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr; 272 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 273 inet->inet_saddr = 0; /* Use device */ 274 275 write_lock_bh(&l2tp_ip_lock); 276 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0, 277 sk->sk_bound_dev_if, addr->l2tp_conn_id)) { 278 write_unlock_bh(&l2tp_ip_lock); 279 ret = -EADDRINUSE; 280 goto out; 281 } 282 283 sk_dst_reset(sk); 284 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id; 285 286 sk_add_bind_node(sk, &l2tp_ip_bind_table); 287 sk_del_node_init(sk); 288 write_unlock_bh(&l2tp_ip_lock); 289 290 ret = 0; 291 sock_reset_flag(sk, SOCK_ZAPPED); 292 293 out: 294 release_sock(sk); 295 296 return ret; 297 } 298 299 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) 300 { 301 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr; 302 int rc; 303 304 if (addr_len < sizeof(*lsa)) 305 return -EINVAL; 306 307 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr)) 308 return -EINVAL; 309 310 lock_sock(sk); 311 312 /* Must bind first - autobinding does not work */ 313 if (sock_flag(sk, SOCK_ZAPPED)) { 314 rc = -EINVAL; 315 goto out_sk; 316 } 317 318 rc = __ip4_datagram_connect(sk, uaddr, addr_len); 319 if (rc < 0) 320 goto out_sk; 321 322 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id; 323 324 write_lock_bh(&l2tp_ip_lock); 325 hlist_del_init(&sk->sk_bind_node); 326 sk_add_bind_node(sk, &l2tp_ip_bind_table); 327 write_unlock_bh(&l2tp_ip_lock); 328 329 out_sk: 330 release_sock(sk); 331 332 return rc; 333 } 334 335 static int l2tp_ip_disconnect(struct sock *sk, int flags) 336 { 337 if (sock_flag(sk, SOCK_ZAPPED)) 338 return 0; 339 340 return __udp_disconnect(sk, flags); 341 } 342 343 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr, 344 int *uaddr_len, int peer) 345 { 346 struct sock *sk = sock->sk; 347 struct inet_sock *inet = inet_sk(sk); 348 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk); 349 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr; 350 351 memset(lsa, 0, sizeof(*lsa)); 352 lsa->l2tp_family = AF_INET; 353 if (peer) { 354 if (!inet->inet_dport) 355 return -ENOTCONN; 356 lsa->l2tp_conn_id = lsk->peer_conn_id; 357 lsa->l2tp_addr.s_addr = inet->inet_daddr; 358 } else { 359 __be32 addr = inet->inet_rcv_saddr; 360 if (!addr) 361 addr = inet->inet_saddr; 362 lsa->l2tp_conn_id = lsk->conn_id; 363 lsa->l2tp_addr.s_addr = addr; 364 } 365 *uaddr_len = sizeof(*lsa); 366 return 0; 367 } 368 369 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb) 370 { 371 int rc; 372 373 /* Charge it to the socket, dropping if the queue is full. */ 374 rc = sock_queue_rcv_skb(sk, skb); 375 if (rc < 0) 376 goto drop; 377 378 return 0; 379 380 drop: 381 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS); 382 kfree_skb(skb); 383 return -1; 384 } 385 386 /* Userspace will call sendmsg() on the tunnel socket to send L2TP 387 * control frames. 388 */ 389 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 390 { 391 struct sk_buff *skb; 392 int rc; 393 struct inet_sock *inet = inet_sk(sk); 394 struct rtable *rt = NULL; 395 struct flowi4 *fl4; 396 int connected = 0; 397 __be32 daddr; 398 399 lock_sock(sk); 400 401 rc = -ENOTCONN; 402 if (sock_flag(sk, SOCK_DEAD)) 403 goto out; 404 405 /* Get and verify the address. */ 406 if (msg->msg_name) { 407 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name); 408 rc = -EINVAL; 409 if (msg->msg_namelen < sizeof(*lip)) 410 goto out; 411 412 if (lip->l2tp_family != AF_INET) { 413 rc = -EAFNOSUPPORT; 414 if (lip->l2tp_family != AF_UNSPEC) 415 goto out; 416 } 417 418 daddr = lip->l2tp_addr.s_addr; 419 } else { 420 rc = -EDESTADDRREQ; 421 if (sk->sk_state != TCP_ESTABLISHED) 422 goto out; 423 424 daddr = inet->inet_daddr; 425 connected = 1; 426 } 427 428 /* Allocate a socket buffer */ 429 rc = -ENOMEM; 430 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) + 431 4 + len, 0, GFP_KERNEL); 432 if (!skb) 433 goto error; 434 435 /* Reserve space for headers, putting IP header on 4-byte boundary. */ 436 skb_reserve(skb, 2 + NET_SKB_PAD); 437 skb_reset_network_header(skb); 438 skb_reserve(skb, sizeof(struct iphdr)); 439 skb_reset_transport_header(skb); 440 441 /* Insert 0 session_id */ 442 *((__be32 *) skb_put(skb, 4)) = 0; 443 444 /* Copy user data into skb */ 445 rc = memcpy_from_msg(skb_put(skb, len), msg, len); 446 if (rc < 0) { 447 kfree_skb(skb); 448 goto error; 449 } 450 451 fl4 = &inet->cork.fl.u.ip4; 452 if (connected) 453 rt = (struct rtable *) __sk_dst_check(sk, 0); 454 455 rcu_read_lock(); 456 if (rt == NULL) { 457 const struct ip_options_rcu *inet_opt; 458 459 inet_opt = rcu_dereference(inet->inet_opt); 460 461 /* Use correct destination address if we have options. */ 462 if (inet_opt && inet_opt->opt.srr) 463 daddr = inet_opt->opt.faddr; 464 465 /* If this fails, retransmit mechanism of transport layer will 466 * keep trying until route appears or the connection times 467 * itself out. 468 */ 469 rt = ip_route_output_ports(sock_net(sk), fl4, sk, 470 daddr, inet->inet_saddr, 471 inet->inet_dport, inet->inet_sport, 472 sk->sk_protocol, RT_CONN_FLAGS(sk), 473 sk->sk_bound_dev_if); 474 if (IS_ERR(rt)) 475 goto no_route; 476 if (connected) { 477 sk_setup_caps(sk, &rt->dst); 478 } else { 479 skb_dst_set(skb, &rt->dst); 480 goto xmit; 481 } 482 } 483 484 /* We dont need to clone dst here, it is guaranteed to not disappear. 485 * __dev_xmit_skb() might force a refcount if needed. 486 */ 487 skb_dst_set_noref(skb, &rt->dst); 488 489 xmit: 490 /* Queue the packet to IP for output */ 491 rc = ip_queue_xmit(sk, skb, &inet->cork.fl); 492 rcu_read_unlock(); 493 494 error: 495 if (rc >= 0) 496 rc = len; 497 498 out: 499 release_sock(sk); 500 return rc; 501 502 no_route: 503 rcu_read_unlock(); 504 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); 505 kfree_skb(skb); 506 rc = -EHOSTUNREACH; 507 goto out; 508 } 509 510 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg, 511 size_t len, int noblock, int flags, int *addr_len) 512 { 513 struct inet_sock *inet = inet_sk(sk); 514 size_t copied = 0; 515 int err = -EOPNOTSUPP; 516 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 517 struct sk_buff *skb; 518 519 if (flags & MSG_OOB) 520 goto out; 521 522 skb = skb_recv_datagram(sk, flags, noblock, &err); 523 if (!skb) 524 goto out; 525 526 copied = skb->len; 527 if (len < copied) { 528 msg->msg_flags |= MSG_TRUNC; 529 copied = len; 530 } 531 532 err = skb_copy_datagram_msg(skb, 0, msg, copied); 533 if (err) 534 goto done; 535 536 sock_recv_timestamp(msg, sk, skb); 537 538 /* Copy the address. */ 539 if (sin) { 540 sin->sin_family = AF_INET; 541 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 542 sin->sin_port = 0; 543 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 544 *addr_len = sizeof(*sin); 545 } 546 if (inet->cmsg_flags) 547 ip_cmsg_recv(msg, skb); 548 if (flags & MSG_TRUNC) 549 copied = skb->len; 550 done: 551 skb_free_datagram(sk, skb); 552 out: 553 return err ? err : copied; 554 } 555 556 static struct proto l2tp_ip_prot = { 557 .name = "L2TP/IP", 558 .owner = THIS_MODULE, 559 .init = l2tp_ip_open, 560 .close = l2tp_ip_close, 561 .bind = l2tp_ip_bind, 562 .connect = l2tp_ip_connect, 563 .disconnect = l2tp_ip_disconnect, 564 .ioctl = udp_ioctl, 565 .destroy = l2tp_ip_destroy_sock, 566 .setsockopt = ip_setsockopt, 567 .getsockopt = ip_getsockopt, 568 .sendmsg = l2tp_ip_sendmsg, 569 .recvmsg = l2tp_ip_recvmsg, 570 .backlog_rcv = l2tp_ip_backlog_recv, 571 .hash = inet_hash, 572 .unhash = inet_unhash, 573 .obj_size = sizeof(struct l2tp_ip_sock), 574 #ifdef CONFIG_COMPAT 575 .compat_setsockopt = compat_ip_setsockopt, 576 .compat_getsockopt = compat_ip_getsockopt, 577 #endif 578 }; 579 580 static const struct proto_ops l2tp_ip_ops = { 581 .family = PF_INET, 582 .owner = THIS_MODULE, 583 .release = inet_release, 584 .bind = inet_bind, 585 .connect = inet_dgram_connect, 586 .socketpair = sock_no_socketpair, 587 .accept = sock_no_accept, 588 .getname = l2tp_ip_getname, 589 .poll = datagram_poll, 590 .ioctl = inet_ioctl, 591 .listen = sock_no_listen, 592 .shutdown = inet_shutdown, 593 .setsockopt = sock_common_setsockopt, 594 .getsockopt = sock_common_getsockopt, 595 .sendmsg = inet_sendmsg, 596 .recvmsg = sock_common_recvmsg, 597 .mmap = sock_no_mmap, 598 .sendpage = sock_no_sendpage, 599 #ifdef CONFIG_COMPAT 600 .compat_setsockopt = compat_sock_common_setsockopt, 601 .compat_getsockopt = compat_sock_common_getsockopt, 602 #endif 603 }; 604 605 static struct inet_protosw l2tp_ip_protosw = { 606 .type = SOCK_DGRAM, 607 .protocol = IPPROTO_L2TP, 608 .prot = &l2tp_ip_prot, 609 .ops = &l2tp_ip_ops, 610 }; 611 612 static struct net_protocol l2tp_ip_protocol __read_mostly = { 613 .handler = l2tp_ip_recv, 614 .netns_ok = 1, 615 }; 616 617 static int __init l2tp_ip_init(void) 618 { 619 int err; 620 621 pr_info("L2TP IP encapsulation support (L2TPv3)\n"); 622 623 err = proto_register(&l2tp_ip_prot, 1); 624 if (err != 0) 625 goto out; 626 627 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP); 628 if (err) 629 goto out1; 630 631 inet_register_protosw(&l2tp_ip_protosw); 632 return 0; 633 634 out1: 635 proto_unregister(&l2tp_ip_prot); 636 out: 637 return err; 638 } 639 640 static void __exit l2tp_ip_exit(void) 641 { 642 inet_unregister_protosw(&l2tp_ip_protosw); 643 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP); 644 proto_unregister(&l2tp_ip_prot); 645 } 646 647 module_init(l2tp_ip_init); 648 module_exit(l2tp_ip_exit); 649 650 MODULE_LICENSE("GPL"); 651 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>"); 652 MODULE_DESCRIPTION("L2TP over IP"); 653 MODULE_VERSION("1.0"); 654 655 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like 656 * enums 657 */ 658 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP); 659 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP); 660