scsi: ufs: fix race between clock gating and devfreq scaling work
[linux/fpc-iii.git] / net / l2tp / l2tp_ip.c
blob9d77a54e8854ea15184a811c20bc0aff5b29c51d
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 struct inet_sock *inet = inet_sk(sk);
58 struct l2tp_ip_sock *l2tp = l2tp_ip_sk(sk);
60 if (l2tp == NULL)
61 continue;
63 if ((l2tp->conn_id == tunnel_id) &&
64 net_eq(sock_net(sk), net) &&
65 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
66 (!inet->inet_daddr || !raddr || inet->inet_daddr == raddr) &&
67 (!sk->sk_bound_dev_if || !dif ||
68 sk->sk_bound_dev_if == dif))
69 goto found;
72 sk = NULL;
73 found:
74 return sk;
77 /* When processing receive frames, there are two cases to
78 * consider. Data frames consist of a non-zero session-id and an
79 * optional cookie. Control frames consist of a regular L2TP header
80 * preceded by 32-bits of zeros.
82 * L2TPv3 Session Header Over IP
84 * 0 1 2 3
85 * 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
86 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
87 * | Session ID |
88 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
89 * | Cookie (optional, maximum 64 bits)...
90 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
91 * |
92 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
94 * L2TPv3 Control Message Header Over IP
96 * 0 1 2 3
97 * 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
98 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
99 * | (32 bits of zeros) |
100 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
101 * |T|L|x|x|S|x|x|x|x|x|x|x| Ver | Length |
102 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
103 * | Control Connection ID |
104 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
105 * | Ns | Nr |
106 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
108 * All control frames are passed to userspace.
110 static int l2tp_ip_recv(struct sk_buff *skb)
112 struct net *net = dev_net(skb->dev);
113 struct sock *sk;
114 u32 session_id;
115 u32 tunnel_id;
116 unsigned char *ptr, *optr;
117 struct l2tp_session *session;
118 struct l2tp_tunnel *tunnel = NULL;
119 int length;
121 if (!pskb_may_pull(skb, 4))
122 goto discard;
124 /* Point to L2TP header */
125 optr = ptr = skb->data;
126 session_id = ntohl(*((__be32 *) ptr));
127 ptr += 4;
129 /* RFC3931: L2TP/IP packets have the first 4 bytes containing
130 * the session_id. If it is 0, the packet is a L2TP control
131 * frame and the session_id value can be discarded.
133 if (session_id == 0) {
134 __skb_pull(skb, 4);
135 goto pass_up;
138 /* Ok, this is a data packet. Lookup the session. */
139 session = l2tp_session_get(net, NULL, session_id, true);
140 if (!session)
141 goto discard;
143 tunnel = session->tunnel;
144 if (!tunnel)
145 goto discard_sess;
147 /* Trace packet contents, if enabled */
148 if (tunnel->debug & L2TP_MSG_DATA) {
149 length = min(32u, skb->len);
150 if (!pskb_may_pull(skb, length))
151 goto discard_sess;
153 /* Point to L2TP header */
154 optr = ptr = skb->data;
155 ptr += 4;
156 pr_debug("%s: ip recv\n", tunnel->name);
157 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, ptr, length);
160 l2tp_recv_common(session, skb, ptr, optr, 0, skb->len, tunnel->recv_payload_hook);
161 l2tp_session_dec_refcount(session);
163 return 0;
165 pass_up:
166 /* Get the tunnel_id from the L2TP header */
167 if (!pskb_may_pull(skb, 12))
168 goto discard;
170 if ((skb->data[0] & 0xc0) != 0xc0)
171 goto discard;
173 tunnel_id = ntohl(*(__be32 *) &skb->data[4]);
174 tunnel = l2tp_tunnel_find(net, tunnel_id);
175 if (tunnel) {
176 sk = tunnel->sock;
177 sock_hold(sk);
178 } else {
179 struct iphdr *iph = (struct iphdr *) skb_network_header(skb);
181 read_lock_bh(&l2tp_ip_lock);
182 sk = __l2tp_ip_bind_lookup(net, iph->daddr, iph->saddr,
183 inet_iif(skb), tunnel_id);
184 if (!sk) {
185 read_unlock_bh(&l2tp_ip_lock);
186 goto discard;
189 sock_hold(sk);
190 read_unlock_bh(&l2tp_ip_lock);
193 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
194 goto discard_put;
196 nf_reset(skb);
198 return sk_receive_skb(sk, skb, 1);
200 discard_sess:
201 if (session->deref)
202 session->deref(session);
203 l2tp_session_dec_refcount(session);
204 goto discard;
206 discard_put:
207 sock_put(sk);
209 discard:
210 kfree_skb(skb);
211 return 0;
214 static int l2tp_ip_open(struct sock *sk)
216 /* Prevent autobind. We don't have ports. */
217 inet_sk(sk)->inet_num = IPPROTO_L2TP;
219 write_lock_bh(&l2tp_ip_lock);
220 sk_add_node(sk, &l2tp_ip_table);
221 write_unlock_bh(&l2tp_ip_lock);
223 return 0;
226 static void l2tp_ip_close(struct sock *sk, long timeout)
228 write_lock_bh(&l2tp_ip_lock);
229 hlist_del_init(&sk->sk_bind_node);
230 sk_del_node_init(sk);
231 write_unlock_bh(&l2tp_ip_lock);
232 sk_common_release(sk);
235 static void l2tp_ip_destroy_sock(struct sock *sk)
237 struct sk_buff *skb;
238 struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
240 while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL)
241 kfree_skb(skb);
243 if (tunnel) {
244 l2tp_tunnel_closeall(tunnel);
245 sock_put(sk);
248 sk_refcnt_debug_dec(sk);
251 static int l2tp_ip_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
253 struct inet_sock *inet = inet_sk(sk);
254 struct sockaddr_l2tpip *addr = (struct sockaddr_l2tpip *) uaddr;
255 struct net *net = sock_net(sk);
256 int ret;
257 int chk_addr_ret;
259 if (addr_len < sizeof(struct sockaddr_l2tpip))
260 return -EINVAL;
261 if (addr->l2tp_family != AF_INET)
262 return -EINVAL;
264 lock_sock(sk);
266 ret = -EINVAL;
267 if (!sock_flag(sk, SOCK_ZAPPED))
268 goto out;
270 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_l2tpip))
271 goto out;
273 chk_addr_ret = inet_addr_type(net, addr->l2tp_addr.s_addr);
274 ret = -EADDRNOTAVAIL;
275 if (addr->l2tp_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
276 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
277 goto out;
279 if (addr->l2tp_addr.s_addr)
280 inet->inet_rcv_saddr = inet->inet_saddr = addr->l2tp_addr.s_addr;
281 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
282 inet->inet_saddr = 0; /* Use device */
284 write_lock_bh(&l2tp_ip_lock);
285 if (__l2tp_ip_bind_lookup(net, addr->l2tp_addr.s_addr, 0,
286 sk->sk_bound_dev_if, addr->l2tp_conn_id)) {
287 write_unlock_bh(&l2tp_ip_lock);
288 ret = -EADDRINUSE;
289 goto out;
292 sk_dst_reset(sk);
293 l2tp_ip_sk(sk)->conn_id = addr->l2tp_conn_id;
295 sk_add_bind_node(sk, &l2tp_ip_bind_table);
296 sk_del_node_init(sk);
297 write_unlock_bh(&l2tp_ip_lock);
299 ret = 0;
300 sock_reset_flag(sk, SOCK_ZAPPED);
302 out:
303 release_sock(sk);
305 return ret;
308 static int l2tp_ip_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
310 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *) uaddr;
311 int rc;
313 if (addr_len < sizeof(*lsa))
314 return -EINVAL;
316 if (ipv4_is_multicast(lsa->l2tp_addr.s_addr))
317 return -EINVAL;
319 lock_sock(sk);
321 /* Must bind first - autobinding does not work */
322 if (sock_flag(sk, SOCK_ZAPPED)) {
323 rc = -EINVAL;
324 goto out_sk;
327 rc = __ip4_datagram_connect(sk, uaddr, addr_len);
328 if (rc < 0)
329 goto out_sk;
331 l2tp_ip_sk(sk)->peer_conn_id = lsa->l2tp_conn_id;
333 write_lock_bh(&l2tp_ip_lock);
334 hlist_del_init(&sk->sk_bind_node);
335 sk_add_bind_node(sk, &l2tp_ip_bind_table);
336 write_unlock_bh(&l2tp_ip_lock);
338 out_sk:
339 release_sock(sk);
341 return rc;
344 static int l2tp_ip_disconnect(struct sock *sk, int flags)
346 if (sock_flag(sk, SOCK_ZAPPED))
347 return 0;
349 return __udp_disconnect(sk, flags);
352 static int l2tp_ip_getname(struct socket *sock, struct sockaddr *uaddr,
353 int *uaddr_len, int peer)
355 struct sock *sk = sock->sk;
356 struct inet_sock *inet = inet_sk(sk);
357 struct l2tp_ip_sock *lsk = l2tp_ip_sk(sk);
358 struct sockaddr_l2tpip *lsa = (struct sockaddr_l2tpip *)uaddr;
360 memset(lsa, 0, sizeof(*lsa));
361 lsa->l2tp_family = AF_INET;
362 if (peer) {
363 if (!inet->inet_dport)
364 return -ENOTCONN;
365 lsa->l2tp_conn_id = lsk->peer_conn_id;
366 lsa->l2tp_addr.s_addr = inet->inet_daddr;
367 } else {
368 __be32 addr = inet->inet_rcv_saddr;
369 if (!addr)
370 addr = inet->inet_saddr;
371 lsa->l2tp_conn_id = lsk->conn_id;
372 lsa->l2tp_addr.s_addr = addr;
374 *uaddr_len = sizeof(*lsa);
375 return 0;
378 static int l2tp_ip_backlog_recv(struct sock *sk, struct sk_buff *skb)
380 int rc;
382 /* Charge it to the socket, dropping if the queue is full. */
383 rc = sock_queue_rcv_skb(sk, skb);
384 if (rc < 0)
385 goto drop;
387 return 0;
389 drop:
390 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_INDISCARDS);
391 kfree_skb(skb);
392 return 0;
395 /* Userspace will call sendmsg() on the tunnel socket to send L2TP
396 * control frames.
398 static int l2tp_ip_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
400 struct sk_buff *skb;
401 int rc;
402 struct inet_sock *inet = inet_sk(sk);
403 struct rtable *rt = NULL;
404 struct flowi4 *fl4;
405 int connected = 0;
406 __be32 daddr;
408 lock_sock(sk);
410 rc = -ENOTCONN;
411 if (sock_flag(sk, SOCK_DEAD))
412 goto out;
414 /* Get and verify the address. */
415 if (msg->msg_name) {
416 DECLARE_SOCKADDR(struct sockaddr_l2tpip *, lip, msg->msg_name);
417 rc = -EINVAL;
418 if (msg->msg_namelen < sizeof(*lip))
419 goto out;
421 if (lip->l2tp_family != AF_INET) {
422 rc = -EAFNOSUPPORT;
423 if (lip->l2tp_family != AF_UNSPEC)
424 goto out;
427 daddr = lip->l2tp_addr.s_addr;
428 } else {
429 rc = -EDESTADDRREQ;
430 if (sk->sk_state != TCP_ESTABLISHED)
431 goto out;
433 daddr = inet->inet_daddr;
434 connected = 1;
437 /* Allocate a socket buffer */
438 rc = -ENOMEM;
439 skb = sock_wmalloc(sk, 2 + NET_SKB_PAD + sizeof(struct iphdr) +
440 4 + len, 0, GFP_KERNEL);
441 if (!skb)
442 goto error;
444 /* Reserve space for headers, putting IP header on 4-byte boundary. */
445 skb_reserve(skb, 2 + NET_SKB_PAD);
446 skb_reset_network_header(skb);
447 skb_reserve(skb, sizeof(struct iphdr));
448 skb_reset_transport_header(skb);
450 /* Insert 0 session_id */
451 *((__be32 *) skb_put(skb, 4)) = 0;
453 /* Copy user data into skb */
454 rc = memcpy_from_msg(skb_put(skb, len), msg, len);
455 if (rc < 0) {
456 kfree_skb(skb);
457 goto error;
460 fl4 = &inet->cork.fl.u.ip4;
461 if (connected)
462 rt = (struct rtable *) __sk_dst_check(sk, 0);
464 rcu_read_lock();
465 if (rt == NULL) {
466 const struct ip_options_rcu *inet_opt;
468 inet_opt = rcu_dereference(inet->inet_opt);
470 /* Use correct destination address if we have options. */
471 if (inet_opt && inet_opt->opt.srr)
472 daddr = inet_opt->opt.faddr;
474 /* If this fails, retransmit mechanism of transport layer will
475 * keep trying until route appears or the connection times
476 * itself out.
478 rt = ip_route_output_ports(sock_net(sk), fl4, sk,
479 daddr, inet->inet_saddr,
480 inet->inet_dport, inet->inet_sport,
481 sk->sk_protocol, RT_CONN_FLAGS(sk),
482 sk->sk_bound_dev_if);
483 if (IS_ERR(rt))
484 goto no_route;
485 if (connected) {
486 sk_setup_caps(sk, &rt->dst);
487 } else {
488 skb_dst_set(skb, &rt->dst);
489 goto xmit;
493 /* We dont need to clone dst here, it is guaranteed to not disappear.
494 * __dev_xmit_skb() might force a refcount if needed.
496 skb_dst_set_noref(skb, &rt->dst);
498 xmit:
499 /* Queue the packet to IP for output */
500 rc = ip_queue_xmit(sk, skb, &inet->cork.fl);
501 rcu_read_unlock();
503 error:
504 if (rc >= 0)
505 rc = len;
507 out:
508 release_sock(sk);
509 return rc;
511 no_route:
512 rcu_read_unlock();
513 IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
514 kfree_skb(skb);
515 rc = -EHOSTUNREACH;
516 goto out;
519 static int l2tp_ip_recvmsg(struct sock *sk, struct msghdr *msg,
520 size_t len, int noblock, int flags, int *addr_len)
522 struct inet_sock *inet = inet_sk(sk);
523 size_t copied = 0;
524 int err = -EOPNOTSUPP;
525 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
526 struct sk_buff *skb;
528 if (flags & MSG_OOB)
529 goto out;
531 skb = skb_recv_datagram(sk, flags, noblock, &err);
532 if (!skb)
533 goto out;
535 copied = skb->len;
536 if (len < copied) {
537 msg->msg_flags |= MSG_TRUNC;
538 copied = len;
541 err = skb_copy_datagram_msg(skb, 0, msg, copied);
542 if (err)
543 goto done;
545 sock_recv_timestamp(msg, sk, skb);
547 /* Copy the address. */
548 if (sin) {
549 sin->sin_family = AF_INET;
550 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
551 sin->sin_port = 0;
552 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
553 *addr_len = sizeof(*sin);
555 if (inet->cmsg_flags)
556 ip_cmsg_recv(msg, skb);
557 if (flags & MSG_TRUNC)
558 copied = skb->len;
559 done:
560 skb_free_datagram(sk, skb);
561 out:
562 return err ? err : copied;
565 int l2tp_ioctl(struct sock *sk, int cmd, unsigned long arg)
567 struct sk_buff *skb;
568 int amount;
570 switch (cmd) {
571 case SIOCOUTQ:
572 amount = sk_wmem_alloc_get(sk);
573 break;
574 case SIOCINQ:
575 spin_lock_bh(&sk->sk_receive_queue.lock);
576 skb = skb_peek(&sk->sk_receive_queue);
577 amount = skb ? skb->len : 0;
578 spin_unlock_bh(&sk->sk_receive_queue.lock);
579 break;
581 default:
582 return -ENOIOCTLCMD;
585 return put_user(amount, (int __user *)arg);
587 EXPORT_SYMBOL(l2tp_ioctl);
589 static struct proto l2tp_ip_prot = {
590 .name = "L2TP/IP",
591 .owner = THIS_MODULE,
592 .init = l2tp_ip_open,
593 .close = l2tp_ip_close,
594 .bind = l2tp_ip_bind,
595 .connect = l2tp_ip_connect,
596 .disconnect = l2tp_ip_disconnect,
597 .ioctl = l2tp_ioctl,
598 .destroy = l2tp_ip_destroy_sock,
599 .setsockopt = ip_setsockopt,
600 .getsockopt = ip_getsockopt,
601 .sendmsg = l2tp_ip_sendmsg,
602 .recvmsg = l2tp_ip_recvmsg,
603 .backlog_rcv = l2tp_ip_backlog_recv,
604 .hash = inet_hash,
605 .unhash = inet_unhash,
606 .obj_size = sizeof(struct l2tp_ip_sock),
607 #ifdef CONFIG_COMPAT
608 .compat_setsockopt = compat_ip_setsockopt,
609 .compat_getsockopt = compat_ip_getsockopt,
610 #endif
613 static const struct proto_ops l2tp_ip_ops = {
614 .family = PF_INET,
615 .owner = THIS_MODULE,
616 .release = inet_release,
617 .bind = inet_bind,
618 .connect = inet_dgram_connect,
619 .socketpair = sock_no_socketpair,
620 .accept = sock_no_accept,
621 .getname = l2tp_ip_getname,
622 .poll = datagram_poll,
623 .ioctl = inet_ioctl,
624 .listen = sock_no_listen,
625 .shutdown = inet_shutdown,
626 .setsockopt = sock_common_setsockopt,
627 .getsockopt = sock_common_getsockopt,
628 .sendmsg = inet_sendmsg,
629 .recvmsg = sock_common_recvmsg,
630 .mmap = sock_no_mmap,
631 .sendpage = sock_no_sendpage,
632 #ifdef CONFIG_COMPAT
633 .compat_setsockopt = compat_sock_common_setsockopt,
634 .compat_getsockopt = compat_sock_common_getsockopt,
635 #endif
638 static struct inet_protosw l2tp_ip_protosw = {
639 .type = SOCK_DGRAM,
640 .protocol = IPPROTO_L2TP,
641 .prot = &l2tp_ip_prot,
642 .ops = &l2tp_ip_ops,
645 static struct net_protocol l2tp_ip_protocol __read_mostly = {
646 .handler = l2tp_ip_recv,
647 .netns_ok = 1,
650 static int __init l2tp_ip_init(void)
652 int err;
654 pr_info("L2TP IP encapsulation support (L2TPv3)\n");
656 err = proto_register(&l2tp_ip_prot, 1);
657 if (err != 0)
658 goto out;
660 err = inet_add_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
661 if (err)
662 goto out1;
664 inet_register_protosw(&l2tp_ip_protosw);
665 return 0;
667 out1:
668 proto_unregister(&l2tp_ip_prot);
669 out:
670 return err;
673 static void __exit l2tp_ip_exit(void)
675 inet_unregister_protosw(&l2tp_ip_protosw);
676 inet_del_protocol(&l2tp_ip_protocol, IPPROTO_L2TP);
677 proto_unregister(&l2tp_ip_prot);
680 module_init(l2tp_ip_init);
681 module_exit(l2tp_ip_exit);
683 MODULE_LICENSE("GPL");
684 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
685 MODULE_DESCRIPTION("L2TP over IP");
686 MODULE_VERSION("1.0");
688 /* Use the value of SOCK_DGRAM (2) directory, because __stringify doesn't like
689 * enums
691 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 2, IPPROTO_L2TP);
692 MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_L2TP);