xref: /linux/net/l2tp/l2tp_core.c (revision d39d0ed196aa1685bb24771e92f78633c66ac9cb)
1 /*
2  * L2TP core.
3  *
4  * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
5  *
6  * This file contains some code of the original L2TPv2 pppol2tp
7  * driver, which has the following copyright:
8  *
9  * Authors:	Martijn van Oosterhout <kleptog@svana.org>
10  *		James Chapman (jchapman@katalix.com)
11  * Contributors:
12  *		Michal Ostrowski <mostrows@speakeasy.net>
13  *		Arnaldo Carvalho de Melo <acme@xconectiva.com.br>
14  *		David S. Miller (davem@redhat.com)
15  *
16  * This program is free software; you can redistribute it and/or modify
17  * it under the terms of the GNU General Public License version 2 as
18  * published by the Free Software Foundation.
19  */
20 
21 #include <linux/module.h>
22 #include <linux/string.h>
23 #include <linux/list.h>
24 #include <linux/rculist.h>
25 #include <linux/uaccess.h>
26 
27 #include <linux/kernel.h>
28 #include <linux/spinlock.h>
29 #include <linux/kthread.h>
30 #include <linux/sched.h>
31 #include <linux/slab.h>
32 #include <linux/errno.h>
33 #include <linux/jiffies.h>
34 
35 #include <linux/netdevice.h>
36 #include <linux/net.h>
37 #include <linux/inetdevice.h>
38 #include <linux/skbuff.h>
39 #include <linux/init.h>
40 #include <linux/in.h>
41 #include <linux/ip.h>
42 #include <linux/udp.h>
43 #include <linux/l2tp.h>
44 #include <linux/hash.h>
45 #include <linux/sort.h>
46 #include <linux/file.h>
47 #include <linux/nsproxy.h>
48 #include <net/net_namespace.h>
49 #include <net/netns/generic.h>
50 #include <net/dst.h>
51 #include <net/ip.h>
52 #include <net/udp.h>
53 #include <net/inet_common.h>
54 #include <net/xfrm.h>
55 #include <net/protocol.h>
56 
57 #include <asm/byteorder.h>
58 #include <asm/atomic.h>
59 
60 #include "l2tp_core.h"
61 
62 #define L2TP_DRV_VERSION	"V2.0"
63 
64 /* L2TP header constants */
65 #define L2TP_HDRFLAG_T	   0x8000
66 #define L2TP_HDRFLAG_L	   0x4000
67 #define L2TP_HDRFLAG_S	   0x0800
68 #define L2TP_HDRFLAG_O	   0x0200
69 #define L2TP_HDRFLAG_P	   0x0100
70 
71 #define L2TP_HDR_VER_MASK  0x000F
72 #define L2TP_HDR_VER_2	   0x0002
73 #define L2TP_HDR_VER_3	   0x0003
74 
75 /* L2TPv3 default L2-specific sublayer */
76 #define L2TP_SLFLAG_S	   0x40000000
77 #define L2TP_SL_SEQ_MASK   0x00ffffff
78 
79 #define L2TP_HDR_SIZE_SEQ		10
80 #define L2TP_HDR_SIZE_NOSEQ		6
81 
82 /* Default trace flags */
83 #define L2TP_DEFAULT_DEBUG_FLAGS	0
84 
85 #define PRINTK(_mask, _type, _lvl, _fmt, args...)			\
86 	do {								\
87 		if ((_mask) & (_type))					\
88 			printk(_lvl "L2TP: " _fmt, ##args);		\
89 	} while (0)
90 
91 /* Private data stored for received packets in the skb.
92  */
93 struct l2tp_skb_cb {
94 	u32			ns;
95 	u16			has_seq;
96 	u16			length;
97 	unsigned long		expires;
98 };
99 
100 #define L2TP_SKB_CB(skb)	((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
101 
102 static atomic_t l2tp_tunnel_count;
103 static atomic_t l2tp_session_count;
104 
105 /* per-net private data for this module */
106 static unsigned int l2tp_net_id;
107 struct l2tp_net {
108 	struct list_head l2tp_tunnel_list;
109 	spinlock_t l2tp_tunnel_list_lock;
110 	struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
111 	spinlock_t l2tp_session_hlist_lock;
112 };
113 
114 static inline struct l2tp_net *l2tp_pernet(struct net *net)
115 {
116 	BUG_ON(!net);
117 
118 	return net_generic(net, l2tp_net_id);
119 }
120 
121 /* Session hash global list for L2TPv3.
122  * The session_id SHOULD be random according to RFC3931, but several
123  * L2TP implementations use incrementing session_ids.  So we do a real
124  * hash on the session_id, rather than a simple bitmask.
125  */
126 static inline struct hlist_head *
127 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
128 {
129 	return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
130 
131 }
132 
133 /* Lookup a session by id in the global session list
134  */
135 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
136 {
137 	struct l2tp_net *pn = l2tp_pernet(net);
138 	struct hlist_head *session_list =
139 		l2tp_session_id_hash_2(pn, session_id);
140 	struct l2tp_session *session;
141 	struct hlist_node *walk;
142 
143 	rcu_read_lock_bh();
144 	hlist_for_each_entry_rcu(session, walk, session_list, global_hlist) {
145 		if (session->session_id == session_id) {
146 			rcu_read_unlock_bh();
147 			return session;
148 		}
149 	}
150 	rcu_read_unlock_bh();
151 
152 	return NULL;
153 }
154 
155 /* Session hash list.
156  * The session_id SHOULD be random according to RFC2661, but several
157  * L2TP implementations (Cisco and Microsoft) use incrementing
158  * session_ids.  So we do a real hash on the session_id, rather than a
159  * simple bitmask.
160  */
161 static inline struct hlist_head *
162 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
163 {
164 	return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
165 }
166 
167 /* Lookup a session by id
168  */
169 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
170 {
171 	struct hlist_head *session_list;
172 	struct l2tp_session *session;
173 	struct hlist_node *walk;
174 
175 	/* In L2TPv3, session_ids are unique over all tunnels and we
176 	 * sometimes need to look them up before we know the
177 	 * tunnel.
178 	 */
179 	if (tunnel == NULL)
180 		return l2tp_session_find_2(net, session_id);
181 
182 	session_list = l2tp_session_id_hash(tunnel, session_id);
183 	read_lock_bh(&tunnel->hlist_lock);
184 	hlist_for_each_entry(session, walk, session_list, hlist) {
185 		if (session->session_id == session_id) {
186 			read_unlock_bh(&tunnel->hlist_lock);
187 			return session;
188 		}
189 	}
190 	read_unlock_bh(&tunnel->hlist_lock);
191 
192 	return NULL;
193 }
194 EXPORT_SYMBOL_GPL(l2tp_session_find);
195 
196 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
197 {
198 	int hash;
199 	struct hlist_node *walk;
200 	struct l2tp_session *session;
201 	int count = 0;
202 
203 	read_lock_bh(&tunnel->hlist_lock);
204 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
205 		hlist_for_each_entry(session, walk, &tunnel->session_hlist[hash], hlist) {
206 			if (++count > nth) {
207 				read_unlock_bh(&tunnel->hlist_lock);
208 				return session;
209 			}
210 		}
211 	}
212 
213 	read_unlock_bh(&tunnel->hlist_lock);
214 
215 	return NULL;
216 }
217 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
218 
219 /* Lookup a session by interface name.
220  * This is very inefficient but is only used by management interfaces.
221  */
222 struct l2tp_session *l2tp_session_find_by_ifname(struct net *net, char *ifname)
223 {
224 	struct l2tp_net *pn = l2tp_pernet(net);
225 	int hash;
226 	struct hlist_node *walk;
227 	struct l2tp_session *session;
228 
229 	rcu_read_lock_bh();
230 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
231 		hlist_for_each_entry_rcu(session, walk, &pn->l2tp_session_hlist[hash], global_hlist) {
232 			if (!strcmp(session->ifname, ifname)) {
233 				rcu_read_unlock_bh();
234 				return session;
235 			}
236 		}
237 	}
238 
239 	rcu_read_unlock_bh();
240 
241 	return NULL;
242 }
243 EXPORT_SYMBOL_GPL(l2tp_session_find_by_ifname);
244 
245 /* Lookup a tunnel by id
246  */
247 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
248 {
249 	struct l2tp_tunnel *tunnel;
250 	struct l2tp_net *pn = l2tp_pernet(net);
251 
252 	rcu_read_lock_bh();
253 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
254 		if (tunnel->tunnel_id == tunnel_id) {
255 			rcu_read_unlock_bh();
256 			return tunnel;
257 		}
258 	}
259 	rcu_read_unlock_bh();
260 
261 	return NULL;
262 }
263 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
264 
265 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
266 {
267 	struct l2tp_net *pn = l2tp_pernet(net);
268 	struct l2tp_tunnel *tunnel;
269 	int count = 0;
270 
271 	rcu_read_lock_bh();
272 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
273 		if (++count > nth) {
274 			rcu_read_unlock_bh();
275 			return tunnel;
276 		}
277 	}
278 
279 	rcu_read_unlock_bh();
280 
281 	return NULL;
282 }
283 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
284 
285 /*****************************************************************************
286  * Receive data handling
287  *****************************************************************************/
288 
289 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
290  * number.
291  */
292 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
293 {
294 	struct sk_buff *skbp;
295 	struct sk_buff *tmp;
296 	u32 ns = L2TP_SKB_CB(skb)->ns;
297 
298 	spin_lock_bh(&session->reorder_q.lock);
299 	skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
300 		if (L2TP_SKB_CB(skbp)->ns > ns) {
301 			__skb_queue_before(&session->reorder_q, skbp, skb);
302 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
303 			       "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
304 			       session->name, ns, L2TP_SKB_CB(skbp)->ns,
305 			       skb_queue_len(&session->reorder_q));
306 			session->stats.rx_oos_packets++;
307 			goto out;
308 		}
309 	}
310 
311 	__skb_queue_tail(&session->reorder_q, skb);
312 
313 out:
314 	spin_unlock_bh(&session->reorder_q.lock);
315 }
316 
317 /* Dequeue a single skb.
318  */
319 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
320 {
321 	struct l2tp_tunnel *tunnel = session->tunnel;
322 	int length = L2TP_SKB_CB(skb)->length;
323 
324 	/* We're about to requeue the skb, so return resources
325 	 * to its current owner (a socket receive buffer).
326 	 */
327 	skb_orphan(skb);
328 
329 	tunnel->stats.rx_packets++;
330 	tunnel->stats.rx_bytes += length;
331 	session->stats.rx_packets++;
332 	session->stats.rx_bytes += length;
333 
334 	if (L2TP_SKB_CB(skb)->has_seq) {
335 		/* Bump our Nr */
336 		session->nr++;
337 		if (tunnel->version == L2TP_HDR_VER_2)
338 			session->nr &= 0xffff;
339 		else
340 			session->nr &= 0xffffff;
341 
342 		PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
343 		       "%s: updated nr to %hu\n", session->name, session->nr);
344 	}
345 
346 	/* call private receive handler */
347 	if (session->recv_skb != NULL)
348 		(*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
349 	else
350 		kfree_skb(skb);
351 
352 	if (session->deref)
353 		(*session->deref)(session);
354 }
355 
356 /* Dequeue skbs from the session's reorder_q, subject to packet order.
357  * Skbs that have been in the queue for too long are simply discarded.
358  */
359 static void l2tp_recv_dequeue(struct l2tp_session *session)
360 {
361 	struct sk_buff *skb;
362 	struct sk_buff *tmp;
363 
364 	/* If the pkt at the head of the queue has the nr that we
365 	 * expect to send up next, dequeue it and any other
366 	 * in-sequence packets behind it.
367 	 */
368 	spin_lock_bh(&session->reorder_q.lock);
369 	skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
370 		if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
371 			session->stats.rx_seq_discards++;
372 			session->stats.rx_errors++;
373 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
374 			       "%s: oos pkt %u len %d discarded (too old), "
375 			       "waiting for %u, reorder_q_len=%d\n",
376 			       session->name, L2TP_SKB_CB(skb)->ns,
377 			       L2TP_SKB_CB(skb)->length, session->nr,
378 			       skb_queue_len(&session->reorder_q));
379 			__skb_unlink(skb, &session->reorder_q);
380 			kfree_skb(skb);
381 			if (session->deref)
382 				(*session->deref)(session);
383 			continue;
384 		}
385 
386 		if (L2TP_SKB_CB(skb)->has_seq) {
387 			if (L2TP_SKB_CB(skb)->ns != session->nr) {
388 				PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
389 				       "%s: holding oos pkt %u len %d, "
390 				       "waiting for %u, reorder_q_len=%d\n",
391 				       session->name, L2TP_SKB_CB(skb)->ns,
392 				       L2TP_SKB_CB(skb)->length, session->nr,
393 				       skb_queue_len(&session->reorder_q));
394 				goto out;
395 			}
396 		}
397 		__skb_unlink(skb, &session->reorder_q);
398 
399 		/* Process the skb. We release the queue lock while we
400 		 * do so to let other contexts process the queue.
401 		 */
402 		spin_unlock_bh(&session->reorder_q.lock);
403 		l2tp_recv_dequeue_skb(session, skb);
404 		spin_lock_bh(&session->reorder_q.lock);
405 	}
406 
407 out:
408 	spin_unlock_bh(&session->reorder_q.lock);
409 }
410 
411 static inline int l2tp_verify_udp_checksum(struct sock *sk,
412 					   struct sk_buff *skb)
413 {
414 	struct udphdr *uh = udp_hdr(skb);
415 	u16 ulen = ntohs(uh->len);
416 	struct inet_sock *inet;
417 	__wsum psum;
418 
419 	if (sk->sk_no_check || skb_csum_unnecessary(skb) || !uh->check)
420 		return 0;
421 
422 	inet = inet_sk(sk);
423 	psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr, ulen,
424 				  IPPROTO_UDP, 0);
425 
426 	if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
427 	    !csum_fold(csum_add(psum, skb->csum)))
428 		return 0;
429 
430 	skb->csum = psum;
431 
432 	return __skb_checksum_complete(skb);
433 }
434 
435 /* Do receive processing of L2TP data frames. We handle both L2TPv2
436  * and L2TPv3 data frames here.
437  *
438  * L2TPv2 Data Message Header
439  *
440  *  0                   1                   2                   3
441  *  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
442  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
443  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
444  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
445  * |           Tunnel ID           |           Session ID          |
446  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
447  * |             Ns (opt)          |             Nr (opt)          |
448  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
449  * |      Offset Size (opt)        |    Offset pad... (opt)
450  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
451  *
452  * Data frames are marked by T=0. All other fields are the same as
453  * those in L2TP control frames.
454  *
455  * L2TPv3 Data Message Header
456  *
457  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
458  * |                      L2TP Session Header                      |
459  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
460  * |                      L2-Specific Sublayer                     |
461  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
462  * |                        Tunnel Payload                      ...
463  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
464  *
465  * L2TPv3 Session Header Over IP
466  *
467  *  0                   1                   2                   3
468  *  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
469  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
470  * |                           Session ID                          |
471  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
472  * |               Cookie (optional, maximum 64 bits)...
473  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
474  *                                                                 |
475  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
476  *
477  * L2TPv3 L2-Specific Sublayer Format
478  *
479  *  0                   1                   2                   3
480  *  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
481  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
482  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
483  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
484  *
485  * Cookie value, sublayer format and offset (pad) are negotiated with
486  * the peer when the session is set up. Unlike L2TPv2, we do not need
487  * to parse the packet header to determine if optional fields are
488  * present.
489  *
490  * Caller must already have parsed the frame and determined that it is
491  * a data (not control) frame before coming here. Fields up to the
492  * session-id have already been parsed and ptr points to the data
493  * after the session-id.
494  */
495 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
496 		      unsigned char *ptr, unsigned char *optr, u16 hdrflags,
497 		      int length, int (*payload_hook)(struct sk_buff *skb))
498 {
499 	struct l2tp_tunnel *tunnel = session->tunnel;
500 	int offset;
501 	u32 ns, nr;
502 
503 	/* The ref count is increased since we now hold a pointer to
504 	 * the session. Take care to decrement the refcnt when exiting
505 	 * this function from now on...
506 	 */
507 	l2tp_session_inc_refcount(session);
508 	if (session->ref)
509 		(*session->ref)(session);
510 
511 	/* Parse and check optional cookie */
512 	if (session->peer_cookie_len > 0) {
513 		if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
514 			PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
515 			       "%s: cookie mismatch (%u/%u). Discarding.\n",
516 			       tunnel->name, tunnel->tunnel_id, session->session_id);
517 			session->stats.rx_cookie_discards++;
518 			goto discard;
519 		}
520 		ptr += session->peer_cookie_len;
521 	}
522 
523 	/* Handle the optional sequence numbers. Sequence numbers are
524 	 * in different places for L2TPv2 and L2TPv3.
525 	 *
526 	 * If we are the LAC, enable/disable sequence numbers under
527 	 * the control of the LNS.  If no sequence numbers present but
528 	 * we were expecting them, discard frame.
529 	 */
530 	ns = nr = 0;
531 	L2TP_SKB_CB(skb)->has_seq = 0;
532 	if (tunnel->version == L2TP_HDR_VER_2) {
533 		if (hdrflags & L2TP_HDRFLAG_S) {
534 			ns = ntohs(*(__be16 *) ptr);
535 			ptr += 2;
536 			nr = ntohs(*(__be16 *) ptr);
537 			ptr += 2;
538 
539 			/* Store L2TP info in the skb */
540 			L2TP_SKB_CB(skb)->ns = ns;
541 			L2TP_SKB_CB(skb)->has_seq = 1;
542 
543 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
544 			       "%s: recv data ns=%u, nr=%u, session nr=%u\n",
545 			       session->name, ns, nr, session->nr);
546 		}
547 	} else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
548 		u32 l2h = ntohl(*(__be32 *) ptr);
549 
550 		if (l2h & 0x40000000) {
551 			ns = l2h & 0x00ffffff;
552 
553 			/* Store L2TP info in the skb */
554 			L2TP_SKB_CB(skb)->ns = ns;
555 			L2TP_SKB_CB(skb)->has_seq = 1;
556 
557 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
558 			       "%s: recv data ns=%u, session nr=%u\n",
559 			       session->name, ns, session->nr);
560 		}
561 	}
562 
563 	/* Advance past L2-specific header, if present */
564 	ptr += session->l2specific_len;
565 
566 	if (L2TP_SKB_CB(skb)->has_seq) {
567 		/* Received a packet with sequence numbers. If we're the LNS,
568 		 * check if we sre sending sequence numbers and if not,
569 		 * configure it so.
570 		 */
571 		if ((!session->lns_mode) && (!session->send_seq)) {
572 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_INFO,
573 			       "%s: requested to enable seq numbers by LNS\n",
574 			       session->name);
575 			session->send_seq = -1;
576 			l2tp_session_set_header_len(session, tunnel->version);
577 		}
578 	} else {
579 		/* No sequence numbers.
580 		 * If user has configured mandatory sequence numbers, discard.
581 		 */
582 		if (session->recv_seq) {
583 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_WARNING,
584 			       "%s: recv data has no seq numbers when required. "
585 			       "Discarding\n", session->name);
586 			session->stats.rx_seq_discards++;
587 			goto discard;
588 		}
589 
590 		/* If we're the LAC and we're sending sequence numbers, the
591 		 * LNS has requested that we no longer send sequence numbers.
592 		 * If we're the LNS and we're sending sequence numbers, the
593 		 * LAC is broken. Discard the frame.
594 		 */
595 		if ((!session->lns_mode) && (session->send_seq)) {
596 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_INFO,
597 			       "%s: requested to disable seq numbers by LNS\n",
598 			       session->name);
599 			session->send_seq = 0;
600 			l2tp_session_set_header_len(session, tunnel->version);
601 		} else if (session->send_seq) {
602 			PRINTK(session->debug, L2TP_MSG_SEQ, KERN_WARNING,
603 			       "%s: recv data has no seq numbers when required. "
604 			       "Discarding\n", session->name);
605 			session->stats.rx_seq_discards++;
606 			goto discard;
607 		}
608 	}
609 
610 	/* Session data offset is handled differently for L2TPv2 and
611 	 * L2TPv3. For L2TPv2, there is an optional 16-bit value in
612 	 * the header. For L2TPv3, the offset is negotiated using AVPs
613 	 * in the session setup control protocol.
614 	 */
615 	if (tunnel->version == L2TP_HDR_VER_2) {
616 		/* If offset bit set, skip it. */
617 		if (hdrflags & L2TP_HDRFLAG_O) {
618 			offset = ntohs(*(__be16 *)ptr);
619 			ptr += 2 + offset;
620 		}
621 	} else
622 		ptr += session->offset;
623 
624 	offset = ptr - optr;
625 	if (!pskb_may_pull(skb, offset))
626 		goto discard;
627 
628 	__skb_pull(skb, offset);
629 
630 	/* If caller wants to process the payload before we queue the
631 	 * packet, do so now.
632 	 */
633 	if (payload_hook)
634 		if ((*payload_hook)(skb))
635 			goto discard;
636 
637 	/* Prepare skb for adding to the session's reorder_q.  Hold
638 	 * packets for max reorder_timeout or 1 second if not
639 	 * reordering.
640 	 */
641 	L2TP_SKB_CB(skb)->length = length;
642 	L2TP_SKB_CB(skb)->expires = jiffies +
643 		(session->reorder_timeout ? session->reorder_timeout : HZ);
644 
645 	/* Add packet to the session's receive queue. Reordering is done here, if
646 	 * enabled. Saved L2TP protocol info is stored in skb->sb[].
647 	 */
648 	if (L2TP_SKB_CB(skb)->has_seq) {
649 		if (session->reorder_timeout != 0) {
650 			/* Packet reordering enabled. Add skb to session's
651 			 * reorder queue, in order of ns.
652 			 */
653 			l2tp_recv_queue_skb(session, skb);
654 		} else {
655 			/* Packet reordering disabled. Discard out-of-sequence
656 			 * packets
657 			 */
658 			if (L2TP_SKB_CB(skb)->ns != session->nr) {
659 				session->stats.rx_seq_discards++;
660 				PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
661 				       "%s: oos pkt %u len %d discarded, "
662 				       "waiting for %u, reorder_q_len=%d\n",
663 				       session->name, L2TP_SKB_CB(skb)->ns,
664 				       L2TP_SKB_CB(skb)->length, session->nr,
665 				       skb_queue_len(&session->reorder_q));
666 				goto discard;
667 			}
668 			skb_queue_tail(&session->reorder_q, skb);
669 		}
670 	} else {
671 		/* No sequence numbers. Add the skb to the tail of the
672 		 * reorder queue. This ensures that it will be
673 		 * delivered after all previous sequenced skbs.
674 		 */
675 		skb_queue_tail(&session->reorder_q, skb);
676 	}
677 
678 	/* Try to dequeue as many skbs from reorder_q as we can. */
679 	l2tp_recv_dequeue(session);
680 
681 	l2tp_session_dec_refcount(session);
682 
683 	return;
684 
685 discard:
686 	session->stats.rx_errors++;
687 	kfree_skb(skb);
688 
689 	if (session->deref)
690 		(*session->deref)(session);
691 
692 	l2tp_session_dec_refcount(session);
693 }
694 EXPORT_SYMBOL(l2tp_recv_common);
695 
696 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
697  * here. The skb is not on a list when we get here.
698  * Returns 0 if the packet was a data packet and was successfully passed on.
699  * Returns 1 if the packet was not a good data packet and could not be
700  * forwarded.  All such packets are passed up to userspace to deal with.
701  */
702 int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
703 		       int (*payload_hook)(struct sk_buff *skb))
704 {
705 	struct l2tp_session *session = NULL;
706 	unsigned char *ptr, *optr;
707 	u16 hdrflags;
708 	u32 tunnel_id, session_id;
709 	int offset;
710 	u16 version;
711 	int length;
712 
713 	if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
714 		goto discard_bad_csum;
715 
716 	/* UDP always verifies the packet length. */
717 	__skb_pull(skb, sizeof(struct udphdr));
718 
719 	/* Short packet? */
720 	if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
721 		PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
722 		       "%s: recv short packet (len=%d)\n", tunnel->name, skb->len);
723 		goto error;
724 	}
725 
726 	/* Point to L2TP header */
727 	optr = ptr = skb->data;
728 
729 	/* Trace packet contents, if enabled */
730 	if (tunnel->debug & L2TP_MSG_DATA) {
731 		length = min(32u, skb->len);
732 		if (!pskb_may_pull(skb, length))
733 			goto error;
734 
735 		printk(KERN_DEBUG "%s: recv: ", tunnel->name);
736 
737 		offset = 0;
738 		do {
739 			printk(" %02X", ptr[offset]);
740 		} while (++offset < length);
741 
742 		printk("\n");
743 	}
744 
745 	/* Get L2TP header flags */
746 	hdrflags = ntohs(*(__be16 *) ptr);
747 
748 	/* Check protocol version */
749 	version = hdrflags & L2TP_HDR_VER_MASK;
750 	if (version != tunnel->version) {
751 		PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
752 		       "%s: recv protocol version mismatch: got %d expected %d\n",
753 		       tunnel->name, version, tunnel->version);
754 		goto error;
755 	}
756 
757 	/* Get length of L2TP packet */
758 	length = skb->len;
759 
760 	/* If type is control packet, it is handled by userspace. */
761 	if (hdrflags & L2TP_HDRFLAG_T) {
762 		PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_DEBUG,
763 		       "%s: recv control packet, len=%d\n", tunnel->name, length);
764 		goto error;
765 	}
766 
767 	/* Skip flags */
768 	ptr += 2;
769 
770 	if (tunnel->version == L2TP_HDR_VER_2) {
771 		/* If length is present, skip it */
772 		if (hdrflags & L2TP_HDRFLAG_L)
773 			ptr += 2;
774 
775 		/* Extract tunnel and session ID */
776 		tunnel_id = ntohs(*(__be16 *) ptr);
777 		ptr += 2;
778 		session_id = ntohs(*(__be16 *) ptr);
779 		ptr += 2;
780 	} else {
781 		ptr += 2;	/* skip reserved bits */
782 		tunnel_id = tunnel->tunnel_id;
783 		session_id = ntohl(*(__be32 *) ptr);
784 		ptr += 4;
785 	}
786 
787 	/* Find the session context */
788 	session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
789 	if (!session || !session->recv_skb) {
790 		/* Not found? Pass to userspace to deal with */
791 		PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
792 		       "%s: no session found (%u/%u). Passing up.\n",
793 		       tunnel->name, tunnel_id, session_id);
794 		goto error;
795 	}
796 
797 	l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
798 
799 	return 0;
800 
801 discard_bad_csum:
802 	LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
803 	UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
804 	tunnel->stats.rx_errors++;
805 	kfree_skb(skb);
806 
807 	return 0;
808 
809 error:
810 	/* Put UDP header back */
811 	__skb_push(skb, sizeof(struct udphdr));
812 
813 	return 1;
814 }
815 EXPORT_SYMBOL_GPL(l2tp_udp_recv_core);
816 
817 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
818  * Return codes:
819  * 0 : success.
820  * <0: error
821  * >0: skb should be passed up to userspace as UDP.
822  */
823 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
824 {
825 	struct l2tp_tunnel *tunnel;
826 
827 	tunnel = l2tp_sock_to_tunnel(sk);
828 	if (tunnel == NULL)
829 		goto pass_up;
830 
831 	PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_DEBUG,
832 	       "%s: received %d bytes\n", tunnel->name, skb->len);
833 
834 	if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
835 		goto pass_up_put;
836 
837 	sock_put(sk);
838 	return 0;
839 
840 pass_up_put:
841 	sock_put(sk);
842 pass_up:
843 	return 1;
844 }
845 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
846 
847 /************************************************************************
848  * Transmit handling
849  ***********************************************************************/
850 
851 /* Build an L2TP header for the session into the buffer provided.
852  */
853 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
854 {
855 	struct l2tp_tunnel *tunnel = session->tunnel;
856 	__be16 *bufp = buf;
857 	__be16 *optr = buf;
858 	u16 flags = L2TP_HDR_VER_2;
859 	u32 tunnel_id = tunnel->peer_tunnel_id;
860 	u32 session_id = session->peer_session_id;
861 
862 	if (session->send_seq)
863 		flags |= L2TP_HDRFLAG_S;
864 
865 	/* Setup L2TP header. */
866 	*bufp++ = htons(flags);
867 	*bufp++ = htons(tunnel_id);
868 	*bufp++ = htons(session_id);
869 	if (session->send_seq) {
870 		*bufp++ = htons(session->ns);
871 		*bufp++ = 0;
872 		session->ns++;
873 		session->ns &= 0xffff;
874 		PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
875 		       "%s: updated ns to %u\n", session->name, session->ns);
876 	}
877 
878 	return bufp - optr;
879 }
880 
881 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
882 {
883 	struct l2tp_tunnel *tunnel = session->tunnel;
884 	char *bufp = buf;
885 	char *optr = bufp;
886 
887 	/* Setup L2TP header. The header differs slightly for UDP and
888 	 * IP encapsulations. For UDP, there is 4 bytes of flags.
889 	 */
890 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
891 		u16 flags = L2TP_HDR_VER_3;
892 		*((__be16 *) bufp) = htons(flags);
893 		bufp += 2;
894 		*((__be16 *) bufp) = 0;
895 		bufp += 2;
896 	}
897 
898 	*((__be32 *) bufp) = htonl(session->peer_session_id);
899 	bufp += 4;
900 	if (session->cookie_len) {
901 		memcpy(bufp, &session->cookie[0], session->cookie_len);
902 		bufp += session->cookie_len;
903 	}
904 	if (session->l2specific_len) {
905 		if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
906 			u32 l2h = 0;
907 			if (session->send_seq) {
908 				l2h = 0x40000000 | session->ns;
909 				session->ns++;
910 				session->ns &= 0xffffff;
911 				PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
912 				       "%s: updated ns to %u\n", session->name, session->ns);
913 			}
914 
915 			*((__be32 *) bufp) = htonl(l2h);
916 		}
917 		bufp += session->l2specific_len;
918 	}
919 	if (session->offset)
920 		bufp += session->offset;
921 
922 	return bufp - optr;
923 }
924 
925 int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, size_t data_len)
926 {
927 	struct l2tp_tunnel *tunnel = session->tunnel;
928 	unsigned int len = skb->len;
929 	int error;
930 
931 	/* Debug */
932 	if (session->send_seq)
933 		PRINTK(session->debug, L2TP_MSG_DATA, KERN_DEBUG,
934 		       "%s: send %Zd bytes, ns=%u\n", session->name,
935 		       data_len, session->ns - 1);
936 	else
937 		PRINTK(session->debug, L2TP_MSG_DATA, KERN_DEBUG,
938 		       "%s: send %Zd bytes\n", session->name, data_len);
939 
940 	if (session->debug & L2TP_MSG_DATA) {
941 		int i;
942 		int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
943 		unsigned char *datap = skb->data + uhlen;
944 
945 		printk(KERN_DEBUG "%s: xmit:", session->name);
946 		for (i = 0; i < (len - uhlen); i++) {
947 			printk(" %02X", *datap++);
948 			if (i == 31) {
949 				printk(" ...");
950 				break;
951 			}
952 		}
953 		printk("\n");
954 	}
955 
956 	/* Queue the packet to IP for output */
957 	skb->local_df = 1;
958 	error = ip_queue_xmit(skb);
959 
960 	/* Update stats */
961 	if (error >= 0) {
962 		tunnel->stats.tx_packets++;
963 		tunnel->stats.tx_bytes += len;
964 		session->stats.tx_packets++;
965 		session->stats.tx_bytes += len;
966 	} else {
967 		tunnel->stats.tx_errors++;
968 		session->stats.tx_errors++;
969 	}
970 
971 	return 0;
972 }
973 EXPORT_SYMBOL_GPL(l2tp_xmit_core);
974 
975 /* Automatically called when the skb is freed.
976  */
977 static void l2tp_sock_wfree(struct sk_buff *skb)
978 {
979 	sock_put(skb->sk);
980 }
981 
982 /* For data skbs that we transmit, we associate with the tunnel socket
983  * but don't do accounting.
984  */
985 static inline void l2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
986 {
987 	sock_hold(sk);
988 	skb->sk = sk;
989 	skb->destructor = l2tp_sock_wfree;
990 }
991 
992 /* If caller requires the skb to have a ppp header, the header must be
993  * inserted in the skb data before calling this function.
994  */
995 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
996 {
997 	int data_len = skb->len;
998 	struct l2tp_tunnel *tunnel = session->tunnel;
999 	struct sock *sk = tunnel->sock;
1000 	struct udphdr *uh;
1001 	struct inet_sock *inet;
1002 	__wsum csum;
1003 	int old_headroom;
1004 	int new_headroom;
1005 	int headroom;
1006 	int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1007 	int udp_len;
1008 
1009 	/* Check that there's enough headroom in the skb to insert IP,
1010 	 * UDP and L2TP headers. If not enough, expand it to
1011 	 * make room. Adjust truesize.
1012 	 */
1013 	headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1014 		uhlen + hdr_len;
1015 	old_headroom = skb_headroom(skb);
1016 	if (skb_cow_head(skb, headroom))
1017 		goto abort;
1018 
1019 	new_headroom = skb_headroom(skb);
1020 	skb_orphan(skb);
1021 	skb->truesize += new_headroom - old_headroom;
1022 
1023 	/* Setup L2TP header */
1024 	session->build_header(session, __skb_push(skb, hdr_len));
1025 
1026 	/* Reset skb netfilter state */
1027 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1028 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1029 			      IPSKB_REROUTED);
1030 	nf_reset(skb);
1031 
1032 	/* Get routing info from the tunnel socket */
1033 	skb_dst_drop(skb);
1034 	skb_dst_set(skb, dst_clone(__sk_dst_get(sk)));
1035 
1036 	switch (tunnel->encap) {
1037 	case L2TP_ENCAPTYPE_UDP:
1038 		/* Setup UDP header */
1039 		inet = inet_sk(sk);
1040 		__skb_push(skb, sizeof(*uh));
1041 		skb_reset_transport_header(skb);
1042 		uh = udp_hdr(skb);
1043 		uh->source = inet->inet_sport;
1044 		uh->dest = inet->inet_dport;
1045 		udp_len = uhlen + hdr_len + data_len;
1046 		uh->len = htons(udp_len);
1047 		uh->check = 0;
1048 
1049 		/* Calculate UDP checksum if configured to do so */
1050 		if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1051 			skb->ip_summed = CHECKSUM_NONE;
1052 		else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1053 			 (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1054 			skb->ip_summed = CHECKSUM_COMPLETE;
1055 			csum = skb_checksum(skb, 0, udp_len, 0);
1056 			uh->check = csum_tcpudp_magic(inet->inet_saddr,
1057 						      inet->inet_daddr,
1058 						      udp_len, IPPROTO_UDP, csum);
1059 			if (uh->check == 0)
1060 				uh->check = CSUM_MANGLED_0;
1061 		} else {
1062 			skb->ip_summed = CHECKSUM_PARTIAL;
1063 			skb->csum_start = skb_transport_header(skb) - skb->head;
1064 			skb->csum_offset = offsetof(struct udphdr, check);
1065 			uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1066 						       inet->inet_daddr,
1067 						       udp_len, IPPROTO_UDP, 0);
1068 		}
1069 		break;
1070 
1071 	case L2TP_ENCAPTYPE_IP:
1072 		break;
1073 	}
1074 
1075 	l2tp_skb_set_owner_w(skb, sk);
1076 
1077 	l2tp_xmit_core(session, skb, data_len);
1078 
1079 abort:
1080 	return 0;
1081 }
1082 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1083 
1084 /*****************************************************************************
1085  * Tinnel and session create/destroy.
1086  *****************************************************************************/
1087 
1088 /* Tunnel socket destruct hook.
1089  * The tunnel context is deleted only when all session sockets have been
1090  * closed.
1091  */
1092 void l2tp_tunnel_destruct(struct sock *sk)
1093 {
1094 	struct l2tp_tunnel *tunnel;
1095 
1096 	tunnel = sk->sk_user_data;
1097 	if (tunnel == NULL)
1098 		goto end;
1099 
1100 	PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1101 	       "%s: closing...\n", tunnel->name);
1102 
1103 	/* Close all sessions */
1104 	l2tp_tunnel_closeall(tunnel);
1105 
1106 	switch (tunnel->encap) {
1107 	case L2TP_ENCAPTYPE_UDP:
1108 		/* No longer an encapsulation socket. See net/ipv4/udp.c */
1109 		(udp_sk(sk))->encap_type = 0;
1110 		(udp_sk(sk))->encap_rcv = NULL;
1111 		break;
1112 	case L2TP_ENCAPTYPE_IP:
1113 		break;
1114 	}
1115 
1116 	/* Remove hooks into tunnel socket */
1117 	tunnel->sock = NULL;
1118 	sk->sk_destruct = tunnel->old_sk_destruct;
1119 	sk->sk_user_data = NULL;
1120 
1121 	/* Call the original destructor */
1122 	if (sk->sk_destruct)
1123 		(*sk->sk_destruct)(sk);
1124 
1125 	/* We're finished with the socket */
1126 	l2tp_tunnel_dec_refcount(tunnel);
1127 
1128 end:
1129 	return;
1130 }
1131 EXPORT_SYMBOL(l2tp_tunnel_destruct);
1132 
1133 /* When the tunnel is closed, all the attached sessions need to go too.
1134  */
1135 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1136 {
1137 	int hash;
1138 	struct hlist_node *walk;
1139 	struct hlist_node *tmp;
1140 	struct l2tp_session *session;
1141 
1142 	BUG_ON(tunnel == NULL);
1143 
1144 	PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1145 	       "%s: closing all sessions...\n", tunnel->name);
1146 
1147 	write_lock_bh(&tunnel->hlist_lock);
1148 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1149 again:
1150 		hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1151 			session = hlist_entry(walk, struct l2tp_session, hlist);
1152 
1153 			PRINTK(session->debug, L2TP_MSG_CONTROL, KERN_INFO,
1154 			       "%s: closing session\n", session->name);
1155 
1156 			hlist_del_init(&session->hlist);
1157 
1158 			/* Since we should hold the sock lock while
1159 			 * doing any unbinding, we need to release the
1160 			 * lock we're holding before taking that lock.
1161 			 * Hold a reference to the sock so it doesn't
1162 			 * disappear as we're jumping between locks.
1163 			 */
1164 			if (session->ref != NULL)
1165 				(*session->ref)(session);
1166 
1167 			write_unlock_bh(&tunnel->hlist_lock);
1168 
1169 			if (tunnel->version != L2TP_HDR_VER_2) {
1170 				struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1171 
1172 				spin_lock_bh(&pn->l2tp_session_hlist_lock);
1173 				hlist_del_init_rcu(&session->global_hlist);
1174 				spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1175 				synchronize_rcu();
1176 			}
1177 
1178 			if (session->session_close != NULL)
1179 				(*session->session_close)(session);
1180 
1181 			if (session->deref != NULL)
1182 				(*session->deref)(session);
1183 
1184 			write_lock_bh(&tunnel->hlist_lock);
1185 
1186 			/* Now restart from the beginning of this hash
1187 			 * chain.  We always remove a session from the
1188 			 * list so we are guaranteed to make forward
1189 			 * progress.
1190 			 */
1191 			goto again;
1192 		}
1193 	}
1194 	write_unlock_bh(&tunnel->hlist_lock);
1195 }
1196 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1197 
1198 /* Really kill the tunnel.
1199  * Come here only when all sessions have been cleared from the tunnel.
1200  */
1201 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1202 {
1203 	struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1204 
1205 	BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1206 	BUG_ON(tunnel->sock != NULL);
1207 
1208 	PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1209 	       "%s: free...\n", tunnel->name);
1210 
1211 	/* Remove from tunnel list */
1212 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1213 	list_del_rcu(&tunnel->list);
1214 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1215 	synchronize_rcu();
1216 
1217 	atomic_dec(&l2tp_tunnel_count);
1218 	kfree(tunnel);
1219 }
1220 EXPORT_SYMBOL_GPL(l2tp_tunnel_free);
1221 
1222 /* Create a socket for the tunnel, if one isn't set up by
1223  * userspace. This is used for static tunnels where there is no
1224  * managing L2TP daemon.
1225  */
1226 static int l2tp_tunnel_sock_create(u32 tunnel_id, u32 peer_tunnel_id, struct l2tp_tunnel_cfg *cfg, struct socket **sockp)
1227 {
1228 	int err = -EINVAL;
1229 	struct sockaddr_in udp_addr;
1230 	struct sockaddr_l2tpip ip_addr;
1231 	struct socket *sock = NULL;
1232 
1233 	switch (cfg->encap) {
1234 	case L2TP_ENCAPTYPE_UDP:
1235 		err = sock_create(AF_INET, SOCK_DGRAM, 0, sockp);
1236 		if (err < 0)
1237 			goto out;
1238 
1239 		sock = *sockp;
1240 
1241 		memset(&udp_addr, 0, sizeof(udp_addr));
1242 		udp_addr.sin_family = AF_INET;
1243 		udp_addr.sin_addr = cfg->local_ip;
1244 		udp_addr.sin_port = htons(cfg->local_udp_port);
1245 		err = kernel_bind(sock, (struct sockaddr *) &udp_addr, sizeof(udp_addr));
1246 		if (err < 0)
1247 			goto out;
1248 
1249 		udp_addr.sin_family = AF_INET;
1250 		udp_addr.sin_addr = cfg->peer_ip;
1251 		udp_addr.sin_port = htons(cfg->peer_udp_port);
1252 		err = kernel_connect(sock, (struct sockaddr *) &udp_addr, sizeof(udp_addr), 0);
1253 		if (err < 0)
1254 			goto out;
1255 
1256 		if (!cfg->use_udp_checksums)
1257 			sock->sk->sk_no_check = UDP_CSUM_NOXMIT;
1258 
1259 		break;
1260 
1261 	case L2TP_ENCAPTYPE_IP:
1262 		err = sock_create(AF_INET, SOCK_DGRAM, IPPROTO_L2TP, sockp);
1263 		if (err < 0)
1264 			goto out;
1265 
1266 		sock = *sockp;
1267 
1268 		memset(&ip_addr, 0, sizeof(ip_addr));
1269 		ip_addr.l2tp_family = AF_INET;
1270 		ip_addr.l2tp_addr = cfg->local_ip;
1271 		ip_addr.l2tp_conn_id = tunnel_id;
1272 		err = kernel_bind(sock, (struct sockaddr *) &ip_addr, sizeof(ip_addr));
1273 		if (err < 0)
1274 			goto out;
1275 
1276 		ip_addr.l2tp_family = AF_INET;
1277 		ip_addr.l2tp_addr = cfg->peer_ip;
1278 		ip_addr.l2tp_conn_id = peer_tunnel_id;
1279 		err = kernel_connect(sock, (struct sockaddr *) &ip_addr, sizeof(ip_addr), 0);
1280 		if (err < 0)
1281 			goto out;
1282 
1283 		break;
1284 
1285 	default:
1286 		goto out;
1287 	}
1288 
1289 out:
1290 	if ((err < 0) && sock) {
1291 		sock_release(sock);
1292 		*sockp = NULL;
1293 	}
1294 
1295 	return err;
1296 }
1297 
1298 int l2tp_tunnel_create(struct net *net, int fd, int version, u32 tunnel_id, u32 peer_tunnel_id, struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1299 {
1300 	struct l2tp_tunnel *tunnel = NULL;
1301 	int err;
1302 	struct socket *sock = NULL;
1303 	struct sock *sk = NULL;
1304 	struct l2tp_net *pn;
1305 	enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1306 
1307 	/* Get the tunnel socket from the fd, which was opened by
1308 	 * the userspace L2TP daemon. If not specified, create a
1309 	 * kernel socket.
1310 	 */
1311 	if (fd < 0) {
1312 		err = l2tp_tunnel_sock_create(tunnel_id, peer_tunnel_id, cfg, &sock);
1313 		if (err < 0)
1314 			goto err;
1315 	} else {
1316 		err = -EBADF;
1317 		sock = sockfd_lookup(fd, &err);
1318 		if (!sock) {
1319 			printk(KERN_ERR "tunl %hu: sockfd_lookup(fd=%d) returned %d\n",
1320 			       tunnel_id, fd, err);
1321 			goto err;
1322 		}
1323 	}
1324 
1325 	sk = sock->sk;
1326 
1327 	if (cfg != NULL)
1328 		encap = cfg->encap;
1329 
1330 	/* Quick sanity checks */
1331 	switch (encap) {
1332 	case L2TP_ENCAPTYPE_UDP:
1333 		err = -EPROTONOSUPPORT;
1334 		if (sk->sk_protocol != IPPROTO_UDP) {
1335 			printk(KERN_ERR "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1336 			       tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1337 			goto err;
1338 		}
1339 		break;
1340 	case L2TP_ENCAPTYPE_IP:
1341 		err = -EPROTONOSUPPORT;
1342 		if (sk->sk_protocol != IPPROTO_L2TP) {
1343 			printk(KERN_ERR "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1344 			       tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1345 			goto err;
1346 		}
1347 		break;
1348 	}
1349 
1350 	/* Check if this socket has already been prepped */
1351 	tunnel = (struct l2tp_tunnel *)sk->sk_user_data;
1352 	if (tunnel != NULL) {
1353 		/* This socket has already been prepped */
1354 		err = -EBUSY;
1355 		goto err;
1356 	}
1357 
1358 	tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1359 	if (tunnel == NULL) {
1360 		err = -ENOMEM;
1361 		goto err;
1362 	}
1363 
1364 	tunnel->version = version;
1365 	tunnel->tunnel_id = tunnel_id;
1366 	tunnel->peer_tunnel_id = peer_tunnel_id;
1367 	tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1368 
1369 	tunnel->magic = L2TP_TUNNEL_MAGIC;
1370 	sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1371 	rwlock_init(&tunnel->hlist_lock);
1372 
1373 	/* The net we belong to */
1374 	tunnel->l2tp_net = net;
1375 	pn = l2tp_pernet(net);
1376 
1377 	if (cfg != NULL)
1378 		tunnel->debug = cfg->debug;
1379 
1380 	/* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1381 	tunnel->encap = encap;
1382 	if (encap == L2TP_ENCAPTYPE_UDP) {
1383 		/* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1384 		udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1385 		udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1386 	}
1387 
1388 	sk->sk_user_data = tunnel;
1389 
1390 	/* Hook on the tunnel socket destructor so that we can cleanup
1391 	 * if the tunnel socket goes away.
1392 	 */
1393 	tunnel->old_sk_destruct = sk->sk_destruct;
1394 	sk->sk_destruct = &l2tp_tunnel_destruct;
1395 	tunnel->sock = sk;
1396 	sk->sk_allocation = GFP_ATOMIC;
1397 
1398 	/* Add tunnel to our list */
1399 	INIT_LIST_HEAD(&tunnel->list);
1400 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1401 	list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1402 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1403 	synchronize_rcu();
1404 	atomic_inc(&l2tp_tunnel_count);
1405 
1406 	/* Bump the reference count. The tunnel context is deleted
1407 	 * only when this drops to zero.
1408 	 */
1409 	l2tp_tunnel_inc_refcount(tunnel);
1410 
1411 	err = 0;
1412 err:
1413 	if (tunnelp)
1414 		*tunnelp = tunnel;
1415 
1416 	/* If tunnel's socket was created by the kernel, it doesn't
1417 	 *  have a file.
1418 	 */
1419 	if (sock && sock->file)
1420 		sockfd_put(sock);
1421 
1422 	return err;
1423 }
1424 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1425 
1426 /* This function is used by the netlink TUNNEL_DELETE command.
1427  */
1428 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1429 {
1430 	int err = 0;
1431 	struct socket *sock = tunnel->sock ? tunnel->sock->sk_socket : NULL;
1432 
1433 	/* Force the tunnel socket to close. This will eventually
1434 	 * cause the tunnel to be deleted via the normal socket close
1435 	 * mechanisms when userspace closes the tunnel socket.
1436 	 */
1437 	if (sock != NULL) {
1438 		err = inet_shutdown(sock, 2);
1439 
1440 		/* If the tunnel's socket was created by the kernel,
1441 		 * close the socket here since the socket was not
1442 		 * created by userspace.
1443 		 */
1444 		if (sock->file == NULL)
1445 			err = inet_release(sock);
1446 	}
1447 
1448 	return err;
1449 }
1450 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1451 
1452 /* Really kill the session.
1453  */
1454 void l2tp_session_free(struct l2tp_session *session)
1455 {
1456 	struct l2tp_tunnel *tunnel;
1457 
1458 	BUG_ON(atomic_read(&session->ref_count) != 0);
1459 
1460 	tunnel = session->tunnel;
1461 	if (tunnel != NULL) {
1462 		BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1463 
1464 		/* Delete the session from the hash */
1465 		write_lock_bh(&tunnel->hlist_lock);
1466 		hlist_del_init(&session->hlist);
1467 		write_unlock_bh(&tunnel->hlist_lock);
1468 
1469 		/* Unlink from the global hash if not L2TPv2 */
1470 		if (tunnel->version != L2TP_HDR_VER_2) {
1471 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1472 
1473 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1474 			hlist_del_init_rcu(&session->global_hlist);
1475 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1476 			synchronize_rcu();
1477 		}
1478 
1479 		if (session->session_id != 0)
1480 			atomic_dec(&l2tp_session_count);
1481 
1482 		sock_put(tunnel->sock);
1483 
1484 		/* This will delete the tunnel context if this
1485 		 * is the last session on the tunnel.
1486 		 */
1487 		session->tunnel = NULL;
1488 		l2tp_tunnel_dec_refcount(tunnel);
1489 	}
1490 
1491 	kfree(session);
1492 
1493 	return;
1494 }
1495 EXPORT_SYMBOL_GPL(l2tp_session_free);
1496 
1497 /* This function is used by the netlink SESSION_DELETE command and by
1498    pseudowire modules.
1499  */
1500 int l2tp_session_delete(struct l2tp_session *session)
1501 {
1502 	if (session->session_close != NULL)
1503 		(*session->session_close)(session);
1504 
1505 	l2tp_session_dec_refcount(session);
1506 
1507 	return 0;
1508 }
1509 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1510 
1511 
1512 /* We come here whenever a session's send_seq, cookie_len or
1513  * l2specific_len parameters are set.
1514  */
1515 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1516 {
1517 	if (version == L2TP_HDR_VER_2) {
1518 		session->hdr_len = 6;
1519 		if (session->send_seq)
1520 			session->hdr_len += 4;
1521 	} else {
1522 		session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1523 		if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1524 			session->hdr_len += 4;
1525 	}
1526 
1527 }
1528 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1529 
1530 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id, u32 peer_session_id, struct l2tp_session_cfg *cfg)
1531 {
1532 	struct l2tp_session *session;
1533 
1534 	session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1535 	if (session != NULL) {
1536 		session->magic = L2TP_SESSION_MAGIC;
1537 		session->tunnel = tunnel;
1538 
1539 		session->session_id = session_id;
1540 		session->peer_session_id = peer_session_id;
1541 		session->nr = 1;
1542 
1543 		sprintf(&session->name[0], "sess %u/%u",
1544 			tunnel->tunnel_id, session->session_id);
1545 
1546 		skb_queue_head_init(&session->reorder_q);
1547 
1548 		INIT_HLIST_NODE(&session->hlist);
1549 		INIT_HLIST_NODE(&session->global_hlist);
1550 
1551 		/* Inherit debug options from tunnel */
1552 		session->debug = tunnel->debug;
1553 
1554 		if (cfg) {
1555 			session->pwtype = cfg->pw_type;
1556 			session->debug = cfg->debug;
1557 			session->mtu = cfg->mtu;
1558 			session->mru = cfg->mru;
1559 			session->send_seq = cfg->send_seq;
1560 			session->recv_seq = cfg->recv_seq;
1561 			session->lns_mode = cfg->lns_mode;
1562 			session->reorder_timeout = cfg->reorder_timeout;
1563 			session->offset = cfg->offset;
1564 			session->l2specific_type = cfg->l2specific_type;
1565 			session->l2specific_len = cfg->l2specific_len;
1566 			session->cookie_len = cfg->cookie_len;
1567 			memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1568 			session->peer_cookie_len = cfg->peer_cookie_len;
1569 			memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1570 		}
1571 
1572 		if (tunnel->version == L2TP_HDR_VER_2)
1573 			session->build_header = l2tp_build_l2tpv2_header;
1574 		else
1575 			session->build_header = l2tp_build_l2tpv3_header;
1576 
1577 		l2tp_session_set_header_len(session, tunnel->version);
1578 
1579 		/* Bump the reference count. The session context is deleted
1580 		 * only when this drops to zero.
1581 		 */
1582 		l2tp_session_inc_refcount(session);
1583 		l2tp_tunnel_inc_refcount(tunnel);
1584 
1585 		/* Ensure tunnel socket isn't deleted */
1586 		sock_hold(tunnel->sock);
1587 
1588 		/* Add session to the tunnel's hash list */
1589 		write_lock_bh(&tunnel->hlist_lock);
1590 		hlist_add_head(&session->hlist,
1591 			       l2tp_session_id_hash(tunnel, session_id));
1592 		write_unlock_bh(&tunnel->hlist_lock);
1593 
1594 		/* And to the global session list if L2TPv3 */
1595 		if (tunnel->version != L2TP_HDR_VER_2) {
1596 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1597 
1598 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1599 			hlist_add_head_rcu(&session->global_hlist,
1600 					   l2tp_session_id_hash_2(pn, session_id));
1601 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1602 			synchronize_rcu();
1603 		}
1604 
1605 		/* Ignore management session in session count value */
1606 		if (session->session_id != 0)
1607 			atomic_inc(&l2tp_session_count);
1608 	}
1609 
1610 	return session;
1611 }
1612 EXPORT_SYMBOL_GPL(l2tp_session_create);
1613 
1614 /*****************************************************************************
1615  * Init and cleanup
1616  *****************************************************************************/
1617 
1618 static __net_init int l2tp_init_net(struct net *net)
1619 {
1620 	struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1621 	int hash;
1622 
1623 	INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1624 	spin_lock_init(&pn->l2tp_tunnel_list_lock);
1625 
1626 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1627 		INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1628 
1629 	spin_lock_init(&pn->l2tp_session_hlist_lock);
1630 
1631 	return 0;
1632 }
1633 
1634 static struct pernet_operations l2tp_net_ops = {
1635 	.init = l2tp_init_net,
1636 	.id   = &l2tp_net_id,
1637 	.size = sizeof(struct l2tp_net),
1638 };
1639 
1640 static int __init l2tp_init(void)
1641 {
1642 	int rc = 0;
1643 
1644 	rc = register_pernet_device(&l2tp_net_ops);
1645 	if (rc)
1646 		goto out;
1647 
1648 	printk(KERN_INFO "L2TP core driver, %s\n", L2TP_DRV_VERSION);
1649 
1650 out:
1651 	return rc;
1652 }
1653 
1654 static void __exit l2tp_exit(void)
1655 {
1656 	unregister_pernet_device(&l2tp_net_ops);
1657 }
1658 
1659 module_init(l2tp_init);
1660 module_exit(l2tp_exit);
1661 
1662 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1663 MODULE_DESCRIPTION("L2TP core");
1664 MODULE_LICENSE("GPL");
1665 MODULE_VERSION(L2TP_DRV_VERSION);
1666 
1667