tcp: mitigate scheduling jitter in EDT pacing model
[linux/fpc-iii.git] / net / l2tp / l2tp_ip.c
blob35f6f86d4dcce898f0deacbeba695cfe438cdd91
1 /*
2 * L2TPv3 IP encapsulation support
4 * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
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.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 #include <asm/ioctls.h>
15 #include <linux/icmp.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/random.h>
19 #include <linux/socket.h>
20 #include <linux/l2tp.h>
21 #include <linux/in.h>
22 #include <net/sock.h>
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/udp.h>
26 #include <net/inet_common.h>
27 #include <net/inet_hashtables.h>
28 #include <net/tcp_states.h>
29 #include <net/protocol.h>
30 #include <net/xfrm.h>
32 #include "l2tp_core.h"
34 struct l2tp_ip_sock {
35 /* inet_sock has to be the first member of l2tp_ip_sock */
36 struct inet_sock inet;
38 u32 conn_id;
39 u32 peer_conn_id;
42 static DEFINE_RWLOCK(l2tp_ip_lock);
43 static struct hlist_head l2tp_ip_table;
44 static struct hlist_head l2tp_ip_bind_table;
46 static inline struct l2tp_ip_sock *l2tp_ip_sk(const struct sock *sk)
48 return (struct l2tp_ip_sock *)sk;
51 static struct sock *__l2tp_ip_bind_lookup(const struct net *net, __be32 laddr,
52 __be32 raddr, int dif, u32 tunnel_id)
54 struct sock *sk;
56 sk_for_each_bound(sk, &l2tp_ip_bind_table) {
57 const struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
58 const struct inet_sock *inet = inet_sk(sk);
60 if (!net_eq(sock_net(sk), net))
61 continue;
63 if (sk->sk_bound_dev_if && dif && sk->sk_bound_dev_if != dif)
64 continue;
66 if (inet->inet_rcv_saddr && laddr &&
67 inet->inet_rcv_saddr != laddr)
68 continue;
70 if (inet->inet_daddr && raddr && inet->inet_daddr != raddr)
71 continue;
73 if (l2tp->conn_id != tunnel_id)
74 continue;
76 goto found;
79 sk = NULL;
80 found:
81 return sk;
84 /* When processing receive frames, there are two cases to
85 * consider. Data frames consist of a non-zero session-id and an
86 * optional cookie. Control frames consist of a regular L2TP header
87 * preceded by 32-bits of zeros.
89 * L2TPv3 Session Header Over IP
91 * 0 1 2 3
92 * 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
93 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94 * | Session ID |
95 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
96 * | Cookie (optional, maximum 64 bits)...
97 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
98 * |
99 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * L2TPv3 Control Message Header Over IP
103 * 0 1 2 3
104 * 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
105 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
106 * | (32 bits of zeros) |
107 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
109 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
110 * | Control Connection ID |
111 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
112 * | Ns | Nr |
113 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
115 * All control frames are passed to userspace.
117 static int l2tp_ip_recv(struct sk_buff *skb)
119 struct net *net = dev_net(skb->dev);
120 struct sock *sk;
121 u32 session_id;
122 u32 tunnel_id;
123 unsigned char *ptr, *optr;
124 struct l2tp_session *session;
125 struct l2tp_tunnel *tunnel = NULL;
126 struct iphdr *iph;
127 int length;
129 if (!pskb_may_pull(skb, 4))
130 goto discard;
132 /* Point to L2TP header */
133 optr = ptr = skb->data;
134 session_id = ntohl(*((__be32 *) ptr));
135 ptr += 4;
137 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
138 * the session_id. If it is 0, the packet is a L2TP control
139 * frame and the session_id value can be discarded.
141 if (session_id == 0) {
142 __skb_pull(skb, 4);
143 goto pass_up;
146 /* Ok, this is a data packet. Lookup the session. */
147 session = l2tp_session_get(net, session_id);
148 if (!session)
149 goto discard;
151 tunnel = session->tunnel;
152 if (!tunnel)
153 goto discard_sess;
155 /* Trace packet contents, if enabled */
156 if (tunnel->debug & L2TP_MSG_DATA) {
157 length = min(32u, skb->len);
158 if (!pskb_may_pull(skb, length))
159 goto discard_sess;
161 /* Point to L2TP header */
162 optr = ptr = skb->data;
163 ptr += 4;
164 pr_debug("%s: ip recv\n", tunnel->name);
165 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
168 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len);
169 l2tp_session_dec_refcount(session);
171 return 0;
173 pass_up:
174 /* Get the tunnel_id from the L2TP header */
175 if (!pskb_may_pull(skb, 12))
176 goto discard;
178 if ((skb->data[0] & 0xc0) != 0xc0)
179 goto discard;
181 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
182 iph = (struct iphdr *)skb_network_header(skb);
184 read_lock_bh(&l2tp_ip_lock);
185 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr, inet_iif(skb),
186 tunnel_id);
187 if (!sk) {
188 read_unlock_bh(&l2tp_ip_lock);
189 goto discard;
191 sock_hold(sk);
192 read_unlock_bh(&l2tp_ip_lock);
194 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
195 goto discard_put;
197 nf_reset(skb);
199 return sk_receive_skb(sk, skb, 1);
201 discard_sess:
202 l2tp_session_dec_refcount(session);
203 goto discard;
205 discard_put:
206 sock_put(sk);
208 discard:
209 kfree_skb(skb);
210 return 0;
213 static int l2tp_ip_open(struct sock *sk)
215 /* Prevent autobind. We don't have ports. */
216 inet_sk(sk)->inet_num = IPPROTO_L2TP;
218 write_lock_bh(&l2tp_ip_lock);
219 sk_add_node(sk, &l2tp_ip_table);
220 write_unlock_bh(&l2tp_ip_lock);
222 return 0;
225 static void l2tp_ip_close(struct sock *sk, long timeout)
227 write_lock_bh(&l2tp_ip_lock);
228 hlist_del_init(&sk->sk_bind_node);
229 sk_del_node_init(sk);
230 write_unlock_bh(&l2tp_ip_lock);
231 sk_common_release(sk);
234 static void l2tp_ip_destroy_sock(struct sock *sk)
236 struct sk_buff *skb;
237 struct l2tp_tunnel *tunnel = sk->sk_user_data;
239 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
240 kfree_skb(skb);
242 if (tunnel)
243 l2tp_tunnel_delete(tunnel);
246 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
248 struct inet_sock *inet = inet_sk(sk);
249 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
250 struct net *net = sock_net(sk);
251 int ret;
252 int chk_addr_ret;
254 if (addr_len < sizeof(struct sockaddr_l2tpip))
255 return -EINVAL;
256 if (addr->l2tp_family != AF_INET)
257 return -EINVAL;
259 lock_sock(sk);
261 ret = -EINVAL;
262 if (!sock_flag(sk, SOCK_ZAPPED))
263 goto out;
265 if (sk->sk_state != TCP_CLOSE)
266 goto out;
268 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
269 ret = -EADDRNOTAVAIL;
270 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
271 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
272 goto out;
274 if (addr->l2tp_addr.s_addr)
275 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
276 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
277 inet->inet_saddr = 0; /* Use device */
279 write_lock_bh(&l2tp_ip_lock);
280 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
281 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
282 write_unlock_bh(&l2tp_ip_lock);
283 ret = -EADDRINUSE;
284 goto out;
287 sk_dst_reset(sk);
288 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
290 sk_add_bind_node(sk, &l2tp_ip_bind_table);
291 sk_del_node_init(sk);
292 write_unlock_bh(&l2tp_ip_lock);
294 ret = 0;
295 sock_reset_flag(sk, SOCK_ZAPPED);
297 out:
298 release_sock(sk);
300 return ret;
303 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
305 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
306 int rc;
308 if (addr_len < sizeof(*lsa))
309 return -EINVAL;
311 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
312 return -EINVAL;
314 lock_sock(sk);
316 /* Must bind first - autobinding does not work */
317 if (sock_flag(sk, SOCK_ZAPPED)) {
318 rc = -EINVAL;
319 goto out_sk;
322 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
323 if (rc < 0)
324 goto out_sk;
326 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
328 write_lock_bh(&l2tp_ip_lock);
329 hlist_del_init(&sk->sk_bind_node);
330 sk_add_bind_node(sk, &l2tp_ip_bind_table);
331 write_unlock_bh(&l2tp_ip_lock);
333 out_sk:
334 release_sock(sk);
336 return rc;
339 static int l2tp_ip_disconnect(struct sock *sk, int flags)
341 if (sock_flag(sk, SOCK_ZAPPED))
342 return 0;
344 return __udp_disconnect(sk, flags);
347 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
348 int peer)
350 struct sock *sk = sock->sk;
351 struct inet_sock *inet = inet_sk(sk);
352 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
353 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
355 memset(lsa, 0, sizeof(*lsa));
356 lsa->l2tp_family = AF_INET;
357 if (peer) {
358 if (!inet->inet_dport)
359 return -ENOTCONN;
360 lsa->l2tp_conn_id = lsk->peer_conn_id;
361 lsa->l2tp_addr.s_addr = inet->inet_daddr;
362 } else {
363 __be32 addr = inet->inet_rcv_saddr;
364 if (!addr)
365 addr = inet->inet_saddr;
366 lsa->l2tp_conn_id = lsk->conn_id;
367 lsa->l2tp_addr.s_addr = addr;
369 return sizeof(*lsa);
372 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
374 int rc;
376 /* Charge it to the socket, dropping if the queue is full. */
377 rc = sock_queue_rcv_skb(sk, skb);
378 if (rc < 0)
379 goto drop;
381 return 0;
383 drop:
384 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
385 kfree_skb(skb);
386 return 0;
389 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
390 * control frames.
392 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
394 struct sk_buff *skb;
395 int rc;
396 struct inet_sock *inet = inet_sk(sk);
397 struct rtable *rt = NULL;
398 struct flowi4 *fl4;
399 int connected = 0;
400 __be32 daddr;
402 lock_sock(sk);
404 rc = -ENOTCONN;
405 if (sock_flag(sk, SOCK_DEAD))
406 goto out;
408 /* Get and verify the address. */
409 if (msg->msg_name) {
410 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
411 rc = -EINVAL;
412 if (msg->msg_namelen < sizeof(*lip))
413 goto out;
415 if (lip->l2tp_family != AF_INET) {
416 rc = -EAFNOSUPPORT;
417 if (lip->l2tp_family != AF_UNSPEC)
418 goto out;
421 daddr = lip->l2tp_addr.s_addr;
422 } else {
423 rc = -EDESTADDRREQ;
424 if (sk->sk_state != TCP_ESTABLISHED)
425 goto out;
427 daddr = inet->inet_daddr;
428 connected = 1;
431 /* Allocate a socket buffer */
432 rc = -ENOMEM;
433 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
434 4 + len, 0, GFP_KERNEL);
435 if (!skb)
436 goto error;
438 /* Reserve space for headers, putting IP header on 4-byte boundary. */
439 skb_reserve(skb, 2 + NET_SKB_PAD);
440 skb_reset_network_header(skb);
441 skb_reserve(skb, sizeof(struct iphdr));
442 skb_reset_transport_header(skb);
444 /* Insert 0 session_id */
445 *((__be32 *) skb_put(skb, 4)) = 0;
447 /* Copy user data into skb */
448 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
449 if (rc < 0) {
450 kfree_skb(skb);
451 goto error;
454 fl4 = &inet->cork.fl.u.ip4;
455 if (connected)
456 rt = (struct rtable *) __sk_dst_check(sk, 0);
458 rcu_read_lock();
459 if (rt == NULL) {
460 const struct ip_options_rcu *inet_opt;
462 inet_opt = rcu_dereference(inet->inet_opt);
464 /* Use correct destination address if we have options. */
465 if (inet_opt && inet_opt->opt.srr)
466 daddr = inet_opt->opt.faddr;
468 /* If this fails, retransmit mechanism of transport layer will
469 * keep trying until route appears or the connection times
470 * itself out.
472 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
473 daddr, inet->inet_saddr,
474 inet->inet_dport, inet->inet_sport,
475 sk->sk_protocol, RT_CONN_FLAGS(sk),
476 sk->sk_bound_dev_if);
477 if (IS_ERR(rt))
478 goto no_route;
479 if (connected) {
480 sk_setup_caps(sk, &rt->dst);
481 } else {
482 skb_dst_set(skb, &rt->dst);
483 goto xmit;
487 /* We dont need to clone dst here, it is guaranteed to not disappear.
488 * __dev_xmit_skb() might force a refcount if needed.
490 skb_dst_set_noref(skb, &rt->dst);
492 xmit:
493 /* Queue the packet to IP for output */
494 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
495 rcu_read_unlock();
497 error:
498 if (rc >= 0)
499 rc = len;
501 out:
502 release_sock(sk);
503 return rc;
505 no_route:
506 rcu_read_unlock();
507 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
508 kfree_skb(skb);
509 rc = -EHOSTUNREACH;
510 goto out;
513 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
514 size_t len, int noblock, int flags, int *addr_len)
516 struct inet_sock *inet = inet_sk(sk);
517 size_t copied = 0;
518 int err = -EOPNOTSUPP;
519 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
520 struct sk_buff *skb;
522 if (flags & MSG_OOB)
523 goto out;
525 skb = skb_recv_datagram(sk, flags, noblock, &err);
526 if (!skb)
527 goto out;
529 copied = skb->len;
530 if (len < copied) {
531 msg->msg_flags |= MSG_TRUNC;
532 copied = len;
535 err = skb_copy_datagram_msg(skb, 0, msg, copied);
536 if (err)
537 goto done;
539 sock_recv_timestamp(msg, sk, skb);
541 /* Copy the address. */
542 if (sin) {
543 sin->sin_family = AF_INET;
544 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
545 sin->sin_port = 0;
546 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
547 *addr_len = sizeof(*sin);
549 if (inet->cmsg_flags)
550 ip_cmsg_recv(msg, skb);
551 if (flags & MSG_TRUNC)
552 copied = skb->len;
553 done:
554 skb_free_datagram(sk, skb);
555 out:
556 return err ? err : copied;
559 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
561 struct sk_buff *skb;
562 int amount;
564 switch (cmd) {
565 case SIOCOUTQ:
566 amount = sk_wmem_alloc_get(sk);
567 break;
568 case SIOCINQ:
569 spin_lock_bh(&sk->sk_receive_queue.lock);
570 skb = skb_peek(&sk->sk_receive_queue);
571 amount = skb ? skb->len : 0;
572 spin_unlock_bh(&sk->sk_receive_queue.lock);
573 break;
575 default:
576 return -ENOIOCTLCMD;
579 return put_user(amount, (int __user *)arg);
581 EXPORT_SYMBOL(l2tp_ioctl);
583 static struct proto l2tp_ip_prot = {
584 .name = "L2TP/IP",
585 .owner = THIS_MODULE,
586 .init = l2tp_ip_open,
587 .close = l2tp_ip_close,
588 .bind = l2tp_ip_bind,
589 .connect = l2tp_ip_connect,
590 .disconnect = l2tp_ip_disconnect,
591 .ioctl = l2tp_ioctl,
592 .destroy = l2tp_ip_destroy_sock,
593 .setsockopt = ip_setsockopt,
594 .getsockopt = ip_getsockopt,
595 .sendmsg = l2tp_ip_sendmsg,
596 .recvmsg = l2tp_ip_recvmsg,
597 .backlog_rcv = l2tp_ip_backlog_recv,
598 .hash = inet_hash,
599 .unhash = inet_unhash,
600 .obj_size = sizeof(struct l2tp_ip_sock),
601 #ifdef CONFIG_COMPAT
602 .compat_setsockopt = compat_ip_setsockopt,
603 .compat_getsockopt = compat_ip_getsockopt,
604 #endif
607 static const struct proto_ops l2tp_ip_ops = {
608 .family = PF_INET,
609 .owner = THIS_MODULE,
610 .release = inet_release,
611 .bind = inet_bind,
612 .connect = inet_dgram_connect,
613 .socketpair = sock_no_socketpair,
614 .accept = sock_no_accept,
615 .getname = l2tp_ip_getname,
616 .poll = datagram_poll,
617 .ioctl = inet_ioctl,
618 .listen = sock_no_listen,
619 .shutdown = inet_shutdown,
620 .setsockopt = sock_common_setsockopt,
621 .getsockopt = sock_common_getsockopt,
622 .sendmsg = inet_sendmsg,
623 .recvmsg = sock_common_recvmsg,
624 .mmap = sock_no_mmap,
625 .sendpage = sock_no_sendpage,
626 #ifdef CONFIG_COMPAT
627 .compat_setsockopt = compat_sock_common_setsockopt,
628 .compat_getsockopt = compat_sock_common_getsockopt,
629 #endif
632 static struct inet_protosw l2tp_ip_protosw = {
633 .type = SOCK_DGRAM,
634 .protocol = IPPROTO_L2TP,
635 .prot = &l2tp_ip_prot,
636 .ops = &l2tp_ip_ops,
639 static struct net_protocol l2tp_ip_protocol __read_mostly = {
640 .handler = l2tp_ip_recv,
641 .netns_ok = 1,
644 static int __init l2tp_ip_init(void)
646 int err;
648 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
650 err = proto_register(&l2tp_ip_prot, 1);
651 if (err != 0)
652 goto out;
654 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
655 if (err)
656 goto out1;
658 inet_register_protosw(&l2tp_ip_protosw);
659 return 0;
661 out1:
662 proto_unregister(&l2tp_ip_prot);
663 out:
664 return err;
667 static void __exit l2tp_ip_exit(void)
669 inet_unregister_protosw(&l2tp_ip_protosw);
670 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
671 proto_unregister(&l2tp_ip_prot);
674 module_init(l2tp_ip_init);
675 module_exit(l2tp_ip_exit);
677 MODULE_LICENSE("GPL");
678 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
679 MODULE_DESCRIPTION("L2TP over IP");
680 MODULE_VERSION("1.0");
682 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
683 * enums
685 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
686 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);