xref: /openbmc/linux/net/l2tp/l2tp_core.c (revision 078a55fc)
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 
23 #include <linux/module.h>
24 #include <linux/string.h>
25 #include <linux/list.h>
26 #include <linux/rculist.h>
27 #include <linux/uaccess.h>
28 
29 #include <linux/kernel.h>
30 #include <linux/spinlock.h>
31 #include <linux/kthread.h>
32 #include <linux/sched.h>
33 #include <linux/slab.h>
34 #include <linux/errno.h>
35 #include <linux/jiffies.h>
36 
37 #include <linux/netdevice.h>
38 #include <linux/net.h>
39 #include <linux/inetdevice.h>
40 #include <linux/skbuff.h>
41 #include <linux/init.h>
42 #include <linux/in.h>
43 #include <linux/ip.h>
44 #include <linux/udp.h>
45 #include <linux/l2tp.h>
46 #include <linux/hash.h>
47 #include <linux/sort.h>
48 #include <linux/file.h>
49 #include <linux/nsproxy.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/dst.h>
53 #include <net/ip.h>
54 #include <net/udp.h>
55 #include <net/inet_common.h>
56 #include <net/xfrm.h>
57 #include <net/protocol.h>
58 #include <net/inet6_connection_sock.h>
59 #include <net/inet_ecn.h>
60 #include <net/ip6_route.h>
61 #include <net/ip6_checksum.h>
62 
63 #include <asm/byteorder.h>
64 #include <linux/atomic.h>
65 
66 #include "l2tp_core.h"
67 
68 #define L2TP_DRV_VERSION	"V2.0"
69 
70 /* L2TP header constants */
71 #define L2TP_HDRFLAG_T	   0x8000
72 #define L2TP_HDRFLAG_L	   0x4000
73 #define L2TP_HDRFLAG_S	   0x0800
74 #define L2TP_HDRFLAG_O	   0x0200
75 #define L2TP_HDRFLAG_P	   0x0100
76 
77 #define L2TP_HDR_VER_MASK  0x000F
78 #define L2TP_HDR_VER_2	   0x0002
79 #define L2TP_HDR_VER_3	   0x0003
80 
81 /* L2TPv3 default L2-specific sublayer */
82 #define L2TP_SLFLAG_S	   0x40000000
83 #define L2TP_SL_SEQ_MASK   0x00ffffff
84 
85 #define L2TP_HDR_SIZE_SEQ		10
86 #define L2TP_HDR_SIZE_NOSEQ		6
87 
88 /* Default trace flags */
89 #define L2TP_DEFAULT_DEBUG_FLAGS	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 static struct workqueue_struct *l2tp_wq;
105 
106 /* per-net private data for this module */
107 static unsigned int l2tp_net_id;
108 struct l2tp_net {
109 	struct list_head l2tp_tunnel_list;
110 	spinlock_t l2tp_tunnel_list_lock;
111 	struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
112 	spinlock_t l2tp_session_hlist_lock;
113 };
114 
115 static void l2tp_session_set_header_len(struct l2tp_session *session, int version);
116 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel);
117 
118 static inline struct l2tp_net *l2tp_pernet(struct net *net)
119 {
120 	BUG_ON(!net);
121 
122 	return net_generic(net, l2tp_net_id);
123 }
124 
125 /* Tunnel reference counts. Incremented per session that is added to
126  * the tunnel.
127  */
128 static inline void l2tp_tunnel_inc_refcount_1(struct l2tp_tunnel *tunnel)
129 {
130 	atomic_inc(&tunnel->ref_count);
131 }
132 
133 static inline void l2tp_tunnel_dec_refcount_1(struct l2tp_tunnel *tunnel)
134 {
135 	if (atomic_dec_and_test(&tunnel->ref_count))
136 		l2tp_tunnel_free(tunnel);
137 }
138 #ifdef L2TP_REFCNT_DEBUG
139 #define l2tp_tunnel_inc_refcount(_t)					\
140 do {									\
141 	pr_debug("l2tp_tunnel_inc_refcount: %s:%d %s: cnt=%d\n",	\
142 		 __func__, __LINE__, (_t)->name,			\
143 		 atomic_read(&_t->ref_count));				\
144 	l2tp_tunnel_inc_refcount_1(_t);					\
145 } while (0)
146 #define l2tp_tunnel_dec_refcount(_t)
147 do {									\
148 	pr_debug("l2tp_tunnel_dec_refcount: %s:%d %s: cnt=%d\n",	\
149 		 __func__, __LINE__, (_t)->name,			\
150 		 atomic_read(&_t->ref_count));				\
151 	l2tp_tunnel_dec_refcount_1(_t);					\
152 } while (0)
153 #else
154 #define l2tp_tunnel_inc_refcount(t) l2tp_tunnel_inc_refcount_1(t)
155 #define l2tp_tunnel_dec_refcount(t) l2tp_tunnel_dec_refcount_1(t)
156 #endif
157 
158 /* Session hash global list for L2TPv3.
159  * The session_id SHOULD be random according to RFC3931, but several
160  * L2TP implementations use incrementing session_ids.  So we do a real
161  * hash on the session_id, rather than a simple bitmask.
162  */
163 static inline struct hlist_head *
164 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
165 {
166 	return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
167 
168 }
169 
170 /* Lookup the tunnel socket, possibly involving the fs code if the socket is
171  * owned by userspace.  A struct sock returned from this function must be
172  * released using l2tp_tunnel_sock_put once you're done with it.
173  */
174 struct sock *l2tp_tunnel_sock_lookup(struct l2tp_tunnel *tunnel)
175 {
176 	int err = 0;
177 	struct socket *sock = NULL;
178 	struct sock *sk = NULL;
179 
180 	if (!tunnel)
181 		goto out;
182 
183 	if (tunnel->fd >= 0) {
184 		/* Socket is owned by userspace, who might be in the process
185 		 * of closing it.  Look the socket up using the fd to ensure
186 		 * consistency.
187 		 */
188 		sock = sockfd_lookup(tunnel->fd, &err);
189 		if (sock)
190 			sk = sock->sk;
191 	} else {
192 		/* Socket is owned by kernelspace */
193 		sk = tunnel->sock;
194 		sock_hold(sk);
195 	}
196 
197 out:
198 	return sk;
199 }
200 EXPORT_SYMBOL_GPL(l2tp_tunnel_sock_lookup);
201 
202 /* Drop a reference to a tunnel socket obtained via. l2tp_tunnel_sock_put */
203 void l2tp_tunnel_sock_put(struct sock *sk)
204 {
205 	struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
206 	if (tunnel) {
207 		if (tunnel->fd >= 0) {
208 			/* Socket is owned by userspace */
209 			sockfd_put(sk->sk_socket);
210 		}
211 		sock_put(sk);
212 	}
213 	sock_put(sk);
214 }
215 EXPORT_SYMBOL_GPL(l2tp_tunnel_sock_put);
216 
217 /* Lookup a session by id in the global session list
218  */
219 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
220 {
221 	struct l2tp_net *pn = l2tp_pernet(net);
222 	struct hlist_head *session_list =
223 		l2tp_session_id_hash_2(pn, session_id);
224 	struct l2tp_session *session;
225 
226 	rcu_read_lock_bh();
227 	hlist_for_each_entry_rcu(session, session_list, global_hlist) {
228 		if (session->session_id == session_id) {
229 			rcu_read_unlock_bh();
230 			return session;
231 		}
232 	}
233 	rcu_read_unlock_bh();
234 
235 	return NULL;
236 }
237 
238 /* Session hash list.
239  * The session_id SHOULD be random according to RFC2661, but several
240  * L2TP implementations (Cisco and Microsoft) use incrementing
241  * session_ids.  So we do a real hash on the session_id, rather than a
242  * simple bitmask.
243  */
244 static inline struct hlist_head *
245 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
246 {
247 	return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
248 }
249 
250 /* Lookup a session by id
251  */
252 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
253 {
254 	struct hlist_head *session_list;
255 	struct l2tp_session *session;
256 
257 	/* In L2TPv3, session_ids are unique over all tunnels and we
258 	 * sometimes need to look them up before we know the
259 	 * tunnel.
260 	 */
261 	if (tunnel == NULL)
262 		return l2tp_session_find_2(net, session_id);
263 
264 	session_list = l2tp_session_id_hash(tunnel, session_id);
265 	read_lock_bh(&tunnel->hlist_lock);
266 	hlist_for_each_entry(session, session_list, hlist) {
267 		if (session->session_id == session_id) {
268 			read_unlock_bh(&tunnel->hlist_lock);
269 			return session;
270 		}
271 	}
272 	read_unlock_bh(&tunnel->hlist_lock);
273 
274 	return NULL;
275 }
276 EXPORT_SYMBOL_GPL(l2tp_session_find);
277 
278 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
279 {
280 	int hash;
281 	struct l2tp_session *session;
282 	int count = 0;
283 
284 	read_lock_bh(&tunnel->hlist_lock);
285 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
286 		hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
287 			if (++count > nth) {
288 				read_unlock_bh(&tunnel->hlist_lock);
289 				return session;
290 			}
291 		}
292 	}
293 
294 	read_unlock_bh(&tunnel->hlist_lock);
295 
296 	return NULL;
297 }
298 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
299 
300 /* Lookup a session by interface name.
301  * This is very inefficient but is only used by management interfaces.
302  */
303 struct l2tp_session *l2tp_session_find_by_ifname(struct net *net, char *ifname)
304 {
305 	struct l2tp_net *pn = l2tp_pernet(net);
306 	int hash;
307 	struct l2tp_session *session;
308 
309 	rcu_read_lock_bh();
310 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
311 		hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
312 			if (!strcmp(session->ifname, ifname)) {
313 				rcu_read_unlock_bh();
314 				return session;
315 			}
316 		}
317 	}
318 
319 	rcu_read_unlock_bh();
320 
321 	return NULL;
322 }
323 EXPORT_SYMBOL_GPL(l2tp_session_find_by_ifname);
324 
325 /* Lookup a tunnel by id
326  */
327 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
328 {
329 	struct l2tp_tunnel *tunnel;
330 	struct l2tp_net *pn = l2tp_pernet(net);
331 
332 	rcu_read_lock_bh();
333 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
334 		if (tunnel->tunnel_id == tunnel_id) {
335 			rcu_read_unlock_bh();
336 			return tunnel;
337 		}
338 	}
339 	rcu_read_unlock_bh();
340 
341 	return NULL;
342 }
343 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
344 
345 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
346 {
347 	struct l2tp_net *pn = l2tp_pernet(net);
348 	struct l2tp_tunnel *tunnel;
349 	int count = 0;
350 
351 	rcu_read_lock_bh();
352 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
353 		if (++count > nth) {
354 			rcu_read_unlock_bh();
355 			return tunnel;
356 		}
357 	}
358 
359 	rcu_read_unlock_bh();
360 
361 	return NULL;
362 }
363 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
364 
365 /*****************************************************************************
366  * Receive data handling
367  *****************************************************************************/
368 
369 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
370  * number.
371  */
372 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
373 {
374 	struct sk_buff *skbp;
375 	struct sk_buff *tmp;
376 	u32 ns = L2TP_SKB_CB(skb)->ns;
377 
378 	spin_lock_bh(&session->reorder_q.lock);
379 	skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
380 		if (L2TP_SKB_CB(skbp)->ns > ns) {
381 			__skb_queue_before(&session->reorder_q, skbp, skb);
382 			l2tp_dbg(session, L2TP_MSG_SEQ,
383 				 "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
384 				 session->name, ns, L2TP_SKB_CB(skbp)->ns,
385 				 skb_queue_len(&session->reorder_q));
386 			atomic_long_inc(&session->stats.rx_oos_packets);
387 			goto out;
388 		}
389 	}
390 
391 	__skb_queue_tail(&session->reorder_q, skb);
392 
393 out:
394 	spin_unlock_bh(&session->reorder_q.lock);
395 }
396 
397 /* Dequeue a single skb.
398  */
399 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
400 {
401 	struct l2tp_tunnel *tunnel = session->tunnel;
402 	int length = L2TP_SKB_CB(skb)->length;
403 
404 	/* We're about to requeue the skb, so return resources
405 	 * to its current owner (a socket receive buffer).
406 	 */
407 	skb_orphan(skb);
408 
409 	atomic_long_inc(&tunnel->stats.rx_packets);
410 	atomic_long_add(length, &tunnel->stats.rx_bytes);
411 	atomic_long_inc(&session->stats.rx_packets);
412 	atomic_long_add(length, &session->stats.rx_bytes);
413 
414 	if (L2TP_SKB_CB(skb)->has_seq) {
415 		/* Bump our Nr */
416 		session->nr++;
417 		session->nr &= session->nr_max;
418 
419 		l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
420 			 session->name, session->nr);
421 	}
422 
423 	/* call private receive handler */
424 	if (session->recv_skb != NULL)
425 		(*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
426 	else
427 		kfree_skb(skb);
428 
429 	if (session->deref)
430 		(*session->deref)(session);
431 }
432 
433 /* Dequeue skbs from the session's reorder_q, subject to packet order.
434  * Skbs that have been in the queue for too long are simply discarded.
435  */
436 static void l2tp_recv_dequeue(struct l2tp_session *session)
437 {
438 	struct sk_buff *skb;
439 	struct sk_buff *tmp;
440 
441 	/* If the pkt at the head of the queue has the nr that we
442 	 * expect to send up next, dequeue it and any other
443 	 * in-sequence packets behind it.
444 	 */
445 start:
446 	spin_lock_bh(&session->reorder_q.lock);
447 	skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
448 		if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
449 			atomic_long_inc(&session->stats.rx_seq_discards);
450 			atomic_long_inc(&session->stats.rx_errors);
451 			l2tp_dbg(session, L2TP_MSG_SEQ,
452 				 "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
453 				 session->name, L2TP_SKB_CB(skb)->ns,
454 				 L2TP_SKB_CB(skb)->length, session->nr,
455 				 skb_queue_len(&session->reorder_q));
456 			session->reorder_skip = 1;
457 			__skb_unlink(skb, &session->reorder_q);
458 			kfree_skb(skb);
459 			if (session->deref)
460 				(*session->deref)(session);
461 			continue;
462 		}
463 
464 		if (L2TP_SKB_CB(skb)->has_seq) {
465 			if (session->reorder_skip) {
466 				l2tp_dbg(session, L2TP_MSG_SEQ,
467 					 "%s: advancing nr to next pkt: %u -> %u",
468 					 session->name, session->nr,
469 					 L2TP_SKB_CB(skb)->ns);
470 				session->reorder_skip = 0;
471 				session->nr = L2TP_SKB_CB(skb)->ns;
472 			}
473 			if (L2TP_SKB_CB(skb)->ns != session->nr) {
474 				l2tp_dbg(session, L2TP_MSG_SEQ,
475 					 "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
476 					 session->name, L2TP_SKB_CB(skb)->ns,
477 					 L2TP_SKB_CB(skb)->length, session->nr,
478 					 skb_queue_len(&session->reorder_q));
479 				goto out;
480 			}
481 		}
482 		__skb_unlink(skb, &session->reorder_q);
483 
484 		/* Process the skb. We release the queue lock while we
485 		 * do so to let other contexts process the queue.
486 		 */
487 		spin_unlock_bh(&session->reorder_q.lock);
488 		l2tp_recv_dequeue_skb(session, skb);
489 		goto start;
490 	}
491 
492 out:
493 	spin_unlock_bh(&session->reorder_q.lock);
494 }
495 
496 static inline int l2tp_verify_udp_checksum(struct sock *sk,
497 					   struct sk_buff *skb)
498 {
499 	struct udphdr *uh = udp_hdr(skb);
500 	u16 ulen = ntohs(uh->len);
501 	__wsum psum;
502 
503 	if (sk->sk_no_check || skb_csum_unnecessary(skb))
504 		return 0;
505 
506 #if IS_ENABLED(CONFIG_IPV6)
507 	if (sk->sk_family == PF_INET6) {
508 		if (!uh->check) {
509 			LIMIT_NETDEBUG(KERN_INFO "L2TP: IPv6: checksum is 0\n");
510 			return 1;
511 		}
512 		if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
513 		    !csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
514 				     &ipv6_hdr(skb)->daddr, ulen,
515 				     IPPROTO_UDP, skb->csum)) {
516 			skb->ip_summed = CHECKSUM_UNNECESSARY;
517 			return 0;
518 		}
519 		skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
520 							 &ipv6_hdr(skb)->daddr,
521 							 skb->len, IPPROTO_UDP,
522 							 0));
523 	} else
524 #endif
525 	{
526 		struct inet_sock *inet;
527 		if (!uh->check)
528 			return 0;
529 		inet = inet_sk(sk);
530 		psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr,
531 					  ulen, IPPROTO_UDP, 0);
532 
533 		if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
534 		    !csum_fold(csum_add(psum, skb->csum)))
535 			return 0;
536 		skb->csum = psum;
537 	}
538 
539 	return __skb_checksum_complete(skb);
540 }
541 
542 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
543 {
544 	u32 nws;
545 
546 	if (nr >= session->nr)
547 		nws = nr - session->nr;
548 	else
549 		nws = (session->nr_max + 1) - (session->nr - nr);
550 
551 	return nws < session->nr_window_size;
552 }
553 
554 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
555  * acceptable, else non-zero.
556  */
557 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
558 {
559 	if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
560 		/* Packet sequence number is outside allowed window.
561 		 * Discard it.
562 		 */
563 		l2tp_dbg(session, L2TP_MSG_SEQ,
564 			 "%s: pkt %u len %d discarded, outside window, nr=%u\n",
565 			 session->name, L2TP_SKB_CB(skb)->ns,
566 			 L2TP_SKB_CB(skb)->length, session->nr);
567 		goto discard;
568 	}
569 
570 	if (session->reorder_timeout != 0) {
571 		/* Packet reordering enabled. Add skb to session's
572 		 * reorder queue, in order of ns.
573 		 */
574 		l2tp_recv_queue_skb(session, skb);
575 		goto out;
576 	}
577 
578 	/* Packet reordering disabled. Discard out-of-sequence packets, while
579 	 * tracking the number if in-sequence packets after the first OOS packet
580 	 * is seen. After nr_oos_count_max in-sequence packets, reset the
581 	 * sequence number to re-enable packet reception.
582 	 */
583 	if (L2TP_SKB_CB(skb)->ns == session->nr) {
584 		skb_queue_tail(&session->reorder_q, skb);
585 	} else {
586 		u32 nr_oos = L2TP_SKB_CB(skb)->ns;
587 		u32 nr_next = (session->nr_oos + 1) & session->nr_max;
588 
589 		if (nr_oos == nr_next)
590 			session->nr_oos_count++;
591 		else
592 			session->nr_oos_count = 0;
593 
594 		session->nr_oos = nr_oos;
595 		if (session->nr_oos_count > session->nr_oos_count_max) {
596 			session->reorder_skip = 1;
597 			l2tp_dbg(session, L2TP_MSG_SEQ,
598 				 "%s: %d oos packets received. Resetting sequence numbers\n",
599 				 session->name, session->nr_oos_count);
600 		}
601 		if (!session->reorder_skip) {
602 			atomic_long_inc(&session->stats.rx_seq_discards);
603 			l2tp_dbg(session, L2TP_MSG_SEQ,
604 				 "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
605 				 session->name, L2TP_SKB_CB(skb)->ns,
606 				 L2TP_SKB_CB(skb)->length, session->nr,
607 				 skb_queue_len(&session->reorder_q));
608 			goto discard;
609 		}
610 		skb_queue_tail(&session->reorder_q, skb);
611 	}
612 
613 out:
614 	return 0;
615 
616 discard:
617 	return 1;
618 }
619 
620 /* Do receive processing of L2TP data frames. We handle both L2TPv2
621  * and L2TPv3 data frames here.
622  *
623  * L2TPv2 Data Message Header
624  *
625  *  0                   1                   2                   3
626  *  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
627  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
628  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
629  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
630  * |           Tunnel ID           |           Session ID          |
631  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
632  * |             Ns (opt)          |             Nr (opt)          |
633  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
634  * |      Offset Size (opt)        |    Offset pad... (opt)
635  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
636  *
637  * Data frames are marked by T=0. All other fields are the same as
638  * those in L2TP control frames.
639  *
640  * L2TPv3 Data Message Header
641  *
642  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
643  * |                      L2TP Session Header                      |
644  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
645  * |                      L2-Specific Sublayer                     |
646  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
647  * |                        Tunnel Payload                      ...
648  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
649  *
650  * L2TPv3 Session Header Over IP
651  *
652  *  0                   1                   2                   3
653  *  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
654  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
655  * |                           Session ID                          |
656  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
657  * |               Cookie (optional, maximum 64 bits)...
658  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
659  *                                                                 |
660  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
661  *
662  * L2TPv3 L2-Specific Sublayer Format
663  *
664  *  0                   1                   2                   3
665  *  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
666  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
667  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
668  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
669  *
670  * Cookie value, sublayer format and offset (pad) are negotiated with
671  * the peer when the session is set up. Unlike L2TPv2, we do not need
672  * to parse the packet header to determine if optional fields are
673  * present.
674  *
675  * Caller must already have parsed the frame and determined that it is
676  * a data (not control) frame before coming here. Fields up to the
677  * session-id have already been parsed and ptr points to the data
678  * after the session-id.
679  */
680 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
681 		      unsigned char *ptr, unsigned char *optr, u16 hdrflags,
682 		      int length, int (*payload_hook)(struct sk_buff *skb))
683 {
684 	struct l2tp_tunnel *tunnel = session->tunnel;
685 	int offset;
686 	u32 ns, nr;
687 
688 	/* The ref count is increased since we now hold a pointer to
689 	 * the session. Take care to decrement the refcnt when exiting
690 	 * this function from now on...
691 	 */
692 	l2tp_session_inc_refcount(session);
693 	if (session->ref)
694 		(*session->ref)(session);
695 
696 	/* Parse and check optional cookie */
697 	if (session->peer_cookie_len > 0) {
698 		if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
699 			l2tp_info(tunnel, L2TP_MSG_DATA,
700 				  "%s: cookie mismatch (%u/%u). Discarding.\n",
701 				  tunnel->name, tunnel->tunnel_id,
702 				  session->session_id);
703 			atomic_long_inc(&session->stats.rx_cookie_discards);
704 			goto discard;
705 		}
706 		ptr += session->peer_cookie_len;
707 	}
708 
709 	/* Handle the optional sequence numbers. Sequence numbers are
710 	 * in different places for L2TPv2 and L2TPv3.
711 	 *
712 	 * If we are the LAC, enable/disable sequence numbers under
713 	 * the control of the LNS.  If no sequence numbers present but
714 	 * we were expecting them, discard frame.
715 	 */
716 	ns = nr = 0;
717 	L2TP_SKB_CB(skb)->has_seq = 0;
718 	if (tunnel->version == L2TP_HDR_VER_2) {
719 		if (hdrflags & L2TP_HDRFLAG_S) {
720 			ns = ntohs(*(__be16 *) ptr);
721 			ptr += 2;
722 			nr = ntohs(*(__be16 *) ptr);
723 			ptr += 2;
724 
725 			/* Store L2TP info in the skb */
726 			L2TP_SKB_CB(skb)->ns = ns;
727 			L2TP_SKB_CB(skb)->has_seq = 1;
728 
729 			l2tp_dbg(session, L2TP_MSG_SEQ,
730 				 "%s: recv data ns=%u, nr=%u, session nr=%u\n",
731 				 session->name, ns, nr, session->nr);
732 		}
733 	} else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
734 		u32 l2h = ntohl(*(__be32 *) ptr);
735 
736 		if (l2h & 0x40000000) {
737 			ns = l2h & 0x00ffffff;
738 
739 			/* Store L2TP info in the skb */
740 			L2TP_SKB_CB(skb)->ns = ns;
741 			L2TP_SKB_CB(skb)->has_seq = 1;
742 
743 			l2tp_dbg(session, L2TP_MSG_SEQ,
744 				 "%s: recv data ns=%u, session nr=%u\n",
745 				 session->name, ns, session->nr);
746 		}
747 	}
748 
749 	/* Advance past L2-specific header, if present */
750 	ptr += session->l2specific_len;
751 
752 	if (L2TP_SKB_CB(skb)->has_seq) {
753 		/* Received a packet with sequence numbers. If we're the LNS,
754 		 * check if we sre sending sequence numbers and if not,
755 		 * configure it so.
756 		 */
757 		if ((!session->lns_mode) && (!session->send_seq)) {
758 			l2tp_info(session, L2TP_MSG_SEQ,
759 				  "%s: requested to enable seq numbers by LNS\n",
760 				  session->name);
761 			session->send_seq = -1;
762 			l2tp_session_set_header_len(session, tunnel->version);
763 		}
764 	} else {
765 		/* No sequence numbers.
766 		 * If user has configured mandatory sequence numbers, discard.
767 		 */
768 		if (session->recv_seq) {
769 			l2tp_warn(session, L2TP_MSG_SEQ,
770 				  "%s: recv data has no seq numbers when required. Discarding.\n",
771 				  session->name);
772 			atomic_long_inc(&session->stats.rx_seq_discards);
773 			goto discard;
774 		}
775 
776 		/* If we're the LAC and we're sending sequence numbers, the
777 		 * LNS has requested that we no longer send sequence numbers.
778 		 * If we're the LNS and we're sending sequence numbers, the
779 		 * LAC is broken. Discard the frame.
780 		 */
781 		if ((!session->lns_mode) && (session->send_seq)) {
782 			l2tp_info(session, L2TP_MSG_SEQ,
783 				  "%s: requested to disable seq numbers by LNS\n",
784 				  session->name);
785 			session->send_seq = 0;
786 			l2tp_session_set_header_len(session, tunnel->version);
787 		} else if (session->send_seq) {
788 			l2tp_warn(session, L2TP_MSG_SEQ,
789 				  "%s: recv data has no seq numbers when required. Discarding.\n",
790 				  session->name);
791 			atomic_long_inc(&session->stats.rx_seq_discards);
792 			goto discard;
793 		}
794 	}
795 
796 	/* Session data offset is handled differently for L2TPv2 and
797 	 * L2TPv3. For L2TPv2, there is an optional 16-bit value in
798 	 * the header. For L2TPv3, the offset is negotiated using AVPs
799 	 * in the session setup control protocol.
800 	 */
801 	if (tunnel->version == L2TP_HDR_VER_2) {
802 		/* If offset bit set, skip it. */
803 		if (hdrflags & L2TP_HDRFLAG_O) {
804 			offset = ntohs(*(__be16 *)ptr);
805 			ptr += 2 + offset;
806 		}
807 	} else
808 		ptr += session->offset;
809 
810 	offset = ptr - optr;
811 	if (!pskb_may_pull(skb, offset))
812 		goto discard;
813 
814 	__skb_pull(skb, offset);
815 
816 	/* If caller wants to process the payload before we queue the
817 	 * packet, do so now.
818 	 */
819 	if (payload_hook)
820 		if ((*payload_hook)(skb))
821 			goto discard;
822 
823 	/* Prepare skb for adding to the session's reorder_q.  Hold
824 	 * packets for max reorder_timeout or 1 second if not
825 	 * reordering.
826 	 */
827 	L2TP_SKB_CB(skb)->length = length;
828 	L2TP_SKB_CB(skb)->expires = jiffies +
829 		(session->reorder_timeout ? session->reorder_timeout : HZ);
830 
831 	/* Add packet to the session's receive queue. Reordering is done here, if
832 	 * enabled. Saved L2TP protocol info is stored in skb->sb[].
833 	 */
834 	if (L2TP_SKB_CB(skb)->has_seq) {
835 		if (l2tp_recv_data_seq(session, skb))
836 			goto discard;
837 	} else {
838 		/* No sequence numbers. Add the skb to the tail of the
839 		 * reorder queue. This ensures that it will be
840 		 * delivered after all previous sequenced skbs.
841 		 */
842 		skb_queue_tail(&session->reorder_q, skb);
843 	}
844 
845 	/* Try to dequeue as many skbs from reorder_q as we can. */
846 	l2tp_recv_dequeue(session);
847 
848 	l2tp_session_dec_refcount(session);
849 
850 	return;
851 
852 discard:
853 	atomic_long_inc(&session->stats.rx_errors);
854 	kfree_skb(skb);
855 
856 	if (session->deref)
857 		(*session->deref)(session);
858 
859 	l2tp_session_dec_refcount(session);
860 }
861 EXPORT_SYMBOL(l2tp_recv_common);
862 
863 /* Drop skbs from the session's reorder_q
864  */
865 int l2tp_session_queue_purge(struct l2tp_session *session)
866 {
867 	struct sk_buff *skb = NULL;
868 	BUG_ON(!session);
869 	BUG_ON(session->magic != L2TP_SESSION_MAGIC);
870 	while ((skb = skb_dequeue(&session->reorder_q))) {
871 		atomic_long_inc(&session->stats.rx_errors);
872 		kfree_skb(skb);
873 		if (session->deref)
874 			(*session->deref)(session);
875 	}
876 	return 0;
877 }
878 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
879 
880 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
881  * here. The skb is not on a list when we get here.
882  * Returns 0 if the packet was a data packet and was successfully passed on.
883  * Returns 1 if the packet was not a good data packet and could not be
884  * forwarded.  All such packets are passed up to userspace to deal with.
885  */
886 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
887 			      int (*payload_hook)(struct sk_buff *skb))
888 {
889 	struct l2tp_session *session = NULL;
890 	unsigned char *ptr, *optr;
891 	u16 hdrflags;
892 	u32 tunnel_id, session_id;
893 	u16 version;
894 	int length;
895 
896 	if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
897 		goto discard_bad_csum;
898 
899 	/* UDP always verifies the packet length. */
900 	__skb_pull(skb, sizeof(struct udphdr));
901 
902 	/* Short packet? */
903 	if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
904 		l2tp_info(tunnel, L2TP_MSG_DATA,
905 			  "%s: recv short packet (len=%d)\n",
906 			  tunnel->name, skb->len);
907 		goto error;
908 	}
909 
910 	/* Trace packet contents, if enabled */
911 	if (tunnel->debug & L2TP_MSG_DATA) {
912 		length = min(32u, skb->len);
913 		if (!pskb_may_pull(skb, length))
914 			goto error;
915 
916 		pr_debug("%s: recv\n", tunnel->name);
917 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
918 	}
919 
920 	/* Point to L2TP header */
921 	optr = ptr = skb->data;
922 
923 	/* Get L2TP header flags */
924 	hdrflags = ntohs(*(__be16 *) ptr);
925 
926 	/* Check protocol version */
927 	version = hdrflags & L2TP_HDR_VER_MASK;
928 	if (version != tunnel->version) {
929 		l2tp_info(tunnel, L2TP_MSG_DATA,
930 			  "%s: recv protocol version mismatch: got %d expected %d\n",
931 			  tunnel->name, version, tunnel->version);
932 		goto error;
933 	}
934 
935 	/* Get length of L2TP packet */
936 	length = skb->len;
937 
938 	/* If type is control packet, it is handled by userspace. */
939 	if (hdrflags & L2TP_HDRFLAG_T) {
940 		l2tp_dbg(tunnel, L2TP_MSG_DATA,
941 			 "%s: recv control packet, len=%d\n",
942 			 tunnel->name, length);
943 		goto error;
944 	}
945 
946 	/* Skip flags */
947 	ptr += 2;
948 
949 	if (tunnel->version == L2TP_HDR_VER_2) {
950 		/* If length is present, skip it */
951 		if (hdrflags & L2TP_HDRFLAG_L)
952 			ptr += 2;
953 
954 		/* Extract tunnel and session ID */
955 		tunnel_id = ntohs(*(__be16 *) ptr);
956 		ptr += 2;
957 		session_id = ntohs(*(__be16 *) ptr);
958 		ptr += 2;
959 	} else {
960 		ptr += 2;	/* skip reserved bits */
961 		tunnel_id = tunnel->tunnel_id;
962 		session_id = ntohl(*(__be32 *) ptr);
963 		ptr += 4;
964 	}
965 
966 	/* Find the session context */
967 	session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
968 	if (!session || !session->recv_skb) {
969 		/* Not found? Pass to userspace to deal with */
970 		l2tp_info(tunnel, L2TP_MSG_DATA,
971 			  "%s: no session found (%u/%u). Passing up.\n",
972 			  tunnel->name, tunnel_id, session_id);
973 		goto error;
974 	}
975 
976 	l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
977 
978 	return 0;
979 
980 discard_bad_csum:
981 	LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
982 	UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
983 	atomic_long_inc(&tunnel->stats.rx_errors);
984 	kfree_skb(skb);
985 
986 	return 0;
987 
988 error:
989 	/* Put UDP header back */
990 	__skb_push(skb, sizeof(struct udphdr));
991 
992 	return 1;
993 }
994 
995 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
996  * Return codes:
997  * 0 : success.
998  * <0: error
999  * >0: skb should be passed up to userspace as UDP.
1000  */
1001 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1002 {
1003 	struct l2tp_tunnel *tunnel;
1004 
1005 	tunnel = l2tp_sock_to_tunnel(sk);
1006 	if (tunnel == NULL)
1007 		goto pass_up;
1008 
1009 	l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
1010 		 tunnel->name, skb->len);
1011 
1012 	if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
1013 		goto pass_up_put;
1014 
1015 	sock_put(sk);
1016 	return 0;
1017 
1018 pass_up_put:
1019 	sock_put(sk);
1020 pass_up:
1021 	return 1;
1022 }
1023 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
1024 
1025 /************************************************************************
1026  * Transmit handling
1027  ***********************************************************************/
1028 
1029 /* Build an L2TP header for the session into the buffer provided.
1030  */
1031 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
1032 {
1033 	struct l2tp_tunnel *tunnel = session->tunnel;
1034 	__be16 *bufp = buf;
1035 	__be16 *optr = buf;
1036 	u16 flags = L2TP_HDR_VER_2;
1037 	u32 tunnel_id = tunnel->peer_tunnel_id;
1038 	u32 session_id = session->peer_session_id;
1039 
1040 	if (session->send_seq)
1041 		flags |= L2TP_HDRFLAG_S;
1042 
1043 	/* Setup L2TP header. */
1044 	*bufp++ = htons(flags);
1045 	*bufp++ = htons(tunnel_id);
1046 	*bufp++ = htons(session_id);
1047 	if (session->send_seq) {
1048 		*bufp++ = htons(session->ns);
1049 		*bufp++ = 0;
1050 		session->ns++;
1051 		session->ns &= 0xffff;
1052 		l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
1053 			 session->name, session->ns);
1054 	}
1055 
1056 	return bufp - optr;
1057 }
1058 
1059 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1060 {
1061 	struct l2tp_tunnel *tunnel = session->tunnel;
1062 	char *bufp = buf;
1063 	char *optr = bufp;
1064 
1065 	/* Setup L2TP header. The header differs slightly for UDP and
1066 	 * IP encapsulations. For UDP, there is 4 bytes of flags.
1067 	 */
1068 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1069 		u16 flags = L2TP_HDR_VER_3;
1070 		*((__be16 *) bufp) = htons(flags);
1071 		bufp += 2;
1072 		*((__be16 *) bufp) = 0;
1073 		bufp += 2;
1074 	}
1075 
1076 	*((__be32 *) bufp) = htonl(session->peer_session_id);
1077 	bufp += 4;
1078 	if (session->cookie_len) {
1079 		memcpy(bufp, &session->cookie[0], session->cookie_len);
1080 		bufp += session->cookie_len;
1081 	}
1082 	if (session->l2specific_len) {
1083 		if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1084 			u32 l2h = 0;
1085 			if (session->send_seq) {
1086 				l2h = 0x40000000 | session->ns;
1087 				session->ns++;
1088 				session->ns &= 0xffffff;
1089 				l2tp_dbg(session, L2TP_MSG_SEQ,
1090 					 "%s: updated ns to %u\n",
1091 					 session->name, session->ns);
1092 			}
1093 
1094 			*((__be32 *) bufp) = htonl(l2h);
1095 		}
1096 		bufp += session->l2specific_len;
1097 	}
1098 	if (session->offset)
1099 		bufp += session->offset;
1100 
1101 	return bufp - optr;
1102 }
1103 
1104 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1105 			  struct flowi *fl, size_t data_len)
1106 {
1107 	struct l2tp_tunnel *tunnel = session->tunnel;
1108 	unsigned int len = skb->len;
1109 	int error;
1110 
1111 	/* Debug */
1112 	if (session->send_seq)
1113 		l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes, ns=%u\n",
1114 			 session->name, data_len, session->ns - 1);
1115 	else
1116 		l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes\n",
1117 			 session->name, data_len);
1118 
1119 	if (session->debug & L2TP_MSG_DATA) {
1120 		int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1121 		unsigned char *datap = skb->data + uhlen;
1122 
1123 		pr_debug("%s: xmit\n", session->name);
1124 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1125 				     datap, min_t(size_t, 32, len - uhlen));
1126 	}
1127 
1128 	/* Queue the packet to IP for output */
1129 	skb->local_df = 1;
1130 #if IS_ENABLED(CONFIG_IPV6)
1131 	if (skb->sk->sk_family == PF_INET6)
1132 		error = inet6_csk_xmit(skb, NULL);
1133 	else
1134 #endif
1135 		error = ip_queue_xmit(skb, fl);
1136 
1137 	/* Update stats */
1138 	if (error >= 0) {
1139 		atomic_long_inc(&tunnel->stats.tx_packets);
1140 		atomic_long_add(len, &tunnel->stats.tx_bytes);
1141 		atomic_long_inc(&session->stats.tx_packets);
1142 		atomic_long_add(len, &session->stats.tx_bytes);
1143 	} else {
1144 		atomic_long_inc(&tunnel->stats.tx_errors);
1145 		atomic_long_inc(&session->stats.tx_errors);
1146 	}
1147 
1148 	return 0;
1149 }
1150 
1151 /* Automatically called when the skb is freed.
1152  */
1153 static void l2tp_sock_wfree(struct sk_buff *skb)
1154 {
1155 	sock_put(skb->sk);
1156 }
1157 
1158 /* For data skbs that we transmit, we associate with the tunnel socket
1159  * but don't do accounting.
1160  */
1161 static inline void l2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1162 {
1163 	sock_hold(sk);
1164 	skb->sk = sk;
1165 	skb->destructor = l2tp_sock_wfree;
1166 }
1167 
1168 #if IS_ENABLED(CONFIG_IPV6)
1169 static void l2tp_xmit_ipv6_csum(struct sock *sk, struct sk_buff *skb,
1170 				int udp_len)
1171 {
1172 	struct ipv6_pinfo *np = inet6_sk(sk);
1173 	struct udphdr *uh = udp_hdr(skb);
1174 
1175 	if (!skb_dst(skb) || !skb_dst(skb)->dev ||
1176 	    !(skb_dst(skb)->dev->features & NETIF_F_IPV6_CSUM)) {
1177 		__wsum csum = skb_checksum(skb, 0, udp_len, 0);
1178 		skb->ip_summed = CHECKSUM_UNNECESSARY;
1179 		uh->check = csum_ipv6_magic(&np->saddr, &np->daddr, udp_len,
1180 					    IPPROTO_UDP, csum);
1181 		if (uh->check == 0)
1182 			uh->check = CSUM_MANGLED_0;
1183 	} else {
1184 		skb->ip_summed = CHECKSUM_PARTIAL;
1185 		skb->csum_start = skb_transport_header(skb) - skb->head;
1186 		skb->csum_offset = offsetof(struct udphdr, check);
1187 		uh->check = ~csum_ipv6_magic(&np->saddr, &np->daddr,
1188 					     udp_len, IPPROTO_UDP, 0);
1189 	}
1190 }
1191 #endif
1192 
1193 /* If caller requires the skb to have a ppp header, the header must be
1194  * inserted in the skb data before calling this function.
1195  */
1196 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1197 {
1198 	int data_len = skb->len;
1199 	struct l2tp_tunnel *tunnel = session->tunnel;
1200 	struct sock *sk = tunnel->sock;
1201 	struct flowi *fl;
1202 	struct udphdr *uh;
1203 	struct inet_sock *inet;
1204 	__wsum csum;
1205 	int headroom;
1206 	int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1207 	int udp_len;
1208 	int ret = NET_XMIT_SUCCESS;
1209 
1210 	/* Check that there's enough headroom in the skb to insert IP,
1211 	 * UDP and L2TP headers. If not enough, expand it to
1212 	 * make room. Adjust truesize.
1213 	 */
1214 	headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1215 		uhlen + hdr_len;
1216 	if (skb_cow_head(skb, headroom)) {
1217 		kfree_skb(skb);
1218 		return NET_XMIT_DROP;
1219 	}
1220 
1221 	skb_orphan(skb);
1222 	/* Setup L2TP header */
1223 	session->build_header(session, __skb_push(skb, hdr_len));
1224 
1225 	/* Reset skb netfilter state */
1226 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1227 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1228 			      IPSKB_REROUTED);
1229 	nf_reset(skb);
1230 
1231 	bh_lock_sock(sk);
1232 	if (sock_owned_by_user(sk)) {
1233 		kfree_skb(skb);
1234 		ret = NET_XMIT_DROP;
1235 		goto out_unlock;
1236 	}
1237 
1238 	/* Get routing info from the tunnel socket */
1239 	skb_dst_drop(skb);
1240 	skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1241 
1242 	inet = inet_sk(sk);
1243 	fl = &inet->cork.fl;
1244 	switch (tunnel->encap) {
1245 	case L2TP_ENCAPTYPE_UDP:
1246 		/* Setup UDP header */
1247 		__skb_push(skb, sizeof(*uh));
1248 		skb_reset_transport_header(skb);
1249 		uh = udp_hdr(skb);
1250 		uh->source = inet->inet_sport;
1251 		uh->dest = inet->inet_dport;
1252 		udp_len = uhlen + hdr_len + data_len;
1253 		uh->len = htons(udp_len);
1254 		uh->check = 0;
1255 
1256 		/* Calculate UDP checksum if configured to do so */
1257 #if IS_ENABLED(CONFIG_IPV6)
1258 		if (sk->sk_family == PF_INET6)
1259 			l2tp_xmit_ipv6_csum(sk, skb, udp_len);
1260 		else
1261 #endif
1262 		if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1263 			skb->ip_summed = CHECKSUM_NONE;
1264 		else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1265 			 (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1266 			skb->ip_summed = CHECKSUM_COMPLETE;
1267 			csum = skb_checksum(skb, 0, udp_len, 0);
1268 			uh->check = csum_tcpudp_magic(inet->inet_saddr,
1269 						      inet->inet_daddr,
1270 						      udp_len, IPPROTO_UDP, csum);
1271 			if (uh->check == 0)
1272 				uh->check = CSUM_MANGLED_0;
1273 		} else {
1274 			skb->ip_summed = CHECKSUM_PARTIAL;
1275 			skb->csum_start = skb_transport_header(skb) - skb->head;
1276 			skb->csum_offset = offsetof(struct udphdr, check);
1277 			uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1278 						       inet->inet_daddr,
1279 						       udp_len, IPPROTO_UDP, 0);
1280 		}
1281 		break;
1282 
1283 	case L2TP_ENCAPTYPE_IP:
1284 		break;
1285 	}
1286 
1287 	l2tp_skb_set_owner_w(skb, sk);
1288 
1289 	l2tp_xmit_core(session, skb, fl, data_len);
1290 out_unlock:
1291 	bh_unlock_sock(sk);
1292 
1293 	return ret;
1294 }
1295 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1296 
1297 /*****************************************************************************
1298  * Tinnel and session create/destroy.
1299  *****************************************************************************/
1300 
1301 /* Tunnel socket destruct hook.
1302  * The tunnel context is deleted only when all session sockets have been
1303  * closed.
1304  */
1305 static void l2tp_tunnel_destruct(struct sock *sk)
1306 {
1307 	struct l2tp_tunnel *tunnel;
1308 	struct l2tp_net *pn;
1309 
1310 	tunnel = sk->sk_user_data;
1311 	if (tunnel == NULL)
1312 		goto end;
1313 
1314 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1315 
1316 
1317 	/* Disable udp encapsulation */
1318 	switch (tunnel->encap) {
1319 	case L2TP_ENCAPTYPE_UDP:
1320 		/* No longer an encapsulation socket. See net/ipv4/udp.c */
1321 		(udp_sk(sk))->encap_type = 0;
1322 		(udp_sk(sk))->encap_rcv = NULL;
1323 		(udp_sk(sk))->encap_destroy = NULL;
1324 		break;
1325 	case L2TP_ENCAPTYPE_IP:
1326 		break;
1327 	}
1328 
1329 	/* Remove hooks into tunnel socket */
1330 	sk->sk_destruct = tunnel->old_sk_destruct;
1331 	sk->sk_user_data = NULL;
1332 	tunnel->sock = NULL;
1333 
1334 	/* Remove the tunnel struct from the tunnel list */
1335 	pn = l2tp_pernet(tunnel->l2tp_net);
1336 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1337 	list_del_rcu(&tunnel->list);
1338 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1339 	atomic_dec(&l2tp_tunnel_count);
1340 
1341 	l2tp_tunnel_closeall(tunnel);
1342 	l2tp_tunnel_dec_refcount(tunnel);
1343 
1344 	/* Call the original destructor */
1345 	if (sk->sk_destruct)
1346 		(*sk->sk_destruct)(sk);
1347 end:
1348 	return;
1349 }
1350 
1351 /* When the tunnel is closed, all the attached sessions need to go too.
1352  */
1353 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1354 {
1355 	int hash;
1356 	struct hlist_node *walk;
1357 	struct hlist_node *tmp;
1358 	struct l2tp_session *session;
1359 
1360 	BUG_ON(tunnel == NULL);
1361 
1362 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1363 		  tunnel->name);
1364 
1365 	write_lock_bh(&tunnel->hlist_lock);
1366 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1367 again:
1368 		hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1369 			session = hlist_entry(walk, struct l2tp_session, hlist);
1370 
1371 			l2tp_info(session, L2TP_MSG_CONTROL,
1372 				  "%s: closing session\n", session->name);
1373 
1374 			hlist_del_init(&session->hlist);
1375 
1376 			if (session->ref != NULL)
1377 				(*session->ref)(session);
1378 
1379 			write_unlock_bh(&tunnel->hlist_lock);
1380 
1381 			__l2tp_session_unhash(session);
1382 			l2tp_session_queue_purge(session);
1383 
1384 			if (session->session_close != NULL)
1385 				(*session->session_close)(session);
1386 
1387 			if (session->deref != NULL)
1388 				(*session->deref)(session);
1389 
1390 			l2tp_session_dec_refcount(session);
1391 
1392 			write_lock_bh(&tunnel->hlist_lock);
1393 
1394 			/* Now restart from the beginning of this hash
1395 			 * chain.  We always remove a session from the
1396 			 * list so we are guaranteed to make forward
1397 			 * progress.
1398 			 */
1399 			goto again;
1400 		}
1401 	}
1402 	write_unlock_bh(&tunnel->hlist_lock);
1403 }
1404 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1405 
1406 /* Tunnel socket destroy hook for UDP encapsulation */
1407 static void l2tp_udp_encap_destroy(struct sock *sk)
1408 {
1409 	struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
1410 	if (tunnel) {
1411 		l2tp_tunnel_closeall(tunnel);
1412 		sock_put(sk);
1413 	}
1414 }
1415 
1416 /* Really kill the tunnel.
1417  * Come here only when all sessions have been cleared from the tunnel.
1418  */
1419 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1420 {
1421 	BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1422 	BUG_ON(tunnel->sock != NULL);
1423 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: free...\n", tunnel->name);
1424 	kfree_rcu(tunnel, rcu);
1425 }
1426 
1427 /* Workqueue tunnel deletion function */
1428 static void l2tp_tunnel_del_work(struct work_struct *work)
1429 {
1430 	struct l2tp_tunnel *tunnel = NULL;
1431 	struct socket *sock = NULL;
1432 	struct sock *sk = NULL;
1433 
1434 	tunnel = container_of(work, struct l2tp_tunnel, del_work);
1435 	sk = l2tp_tunnel_sock_lookup(tunnel);
1436 	if (!sk)
1437 		return;
1438 
1439 	sock = sk->sk_socket;
1440 
1441 	/* If the tunnel socket was created by userspace, then go through the
1442 	 * inet layer to shut the socket down, and let userspace close it.
1443 	 * Otherwise, if we created the socket directly within the kernel, use
1444 	 * the sk API to release it here.
1445 	 * In either case the tunnel resources are freed in the socket
1446 	 * destructor when the tunnel socket goes away.
1447 	 */
1448 	if (tunnel->fd >= 0) {
1449 		if (sock)
1450 			inet_shutdown(sock, 2);
1451 	} else {
1452 		if (sock)
1453 			kernel_sock_shutdown(sock, SHUT_RDWR);
1454 		sk_release_kernel(sk);
1455 	}
1456 
1457 	l2tp_tunnel_sock_put(sk);
1458 }
1459 
1460 /* Create a socket for the tunnel, if one isn't set up by
1461  * userspace. This is used for static tunnels where there is no
1462  * managing L2TP daemon.
1463  *
1464  * Since we don't want these sockets to keep a namespace alive by
1465  * themselves, we drop the socket's namespace refcount after creation.
1466  * These sockets are freed when the namespace exits using the pernet
1467  * exit hook.
1468  */
1469 static int l2tp_tunnel_sock_create(struct net *net,
1470 				u32 tunnel_id,
1471 				u32 peer_tunnel_id,
1472 				struct l2tp_tunnel_cfg *cfg,
1473 				struct socket **sockp)
1474 {
1475 	int err = -EINVAL;
1476 	struct socket *sock = NULL;
1477 	struct sockaddr_in udp_addr = {0};
1478 	struct sockaddr_l2tpip ip_addr = {0};
1479 #if IS_ENABLED(CONFIG_IPV6)
1480 	struct sockaddr_in6 udp6_addr = {0};
1481 	struct sockaddr_l2tpip6 ip6_addr = {0};
1482 #endif
1483 
1484 	switch (cfg->encap) {
1485 	case L2TP_ENCAPTYPE_UDP:
1486 #if IS_ENABLED(CONFIG_IPV6)
1487 		if (cfg->local_ip6 && cfg->peer_ip6) {
1488 			err = sock_create_kern(AF_INET6, SOCK_DGRAM, 0, &sock);
1489 			if (err < 0)
1490 				goto out;
1491 
1492 			sk_change_net(sock->sk, net);
1493 
1494 			udp6_addr.sin6_family = AF_INET6;
1495 			memcpy(&udp6_addr.sin6_addr, cfg->local_ip6,
1496 			       sizeof(udp6_addr.sin6_addr));
1497 			udp6_addr.sin6_port = htons(cfg->local_udp_port);
1498 			err = kernel_bind(sock, (struct sockaddr *) &udp6_addr,
1499 					  sizeof(udp6_addr));
1500 			if (err < 0)
1501 				goto out;
1502 
1503 			udp6_addr.sin6_family = AF_INET6;
1504 			memcpy(&udp6_addr.sin6_addr, cfg->peer_ip6,
1505 			       sizeof(udp6_addr.sin6_addr));
1506 			udp6_addr.sin6_port = htons(cfg->peer_udp_port);
1507 			err = kernel_connect(sock,
1508 					     (struct sockaddr *) &udp6_addr,
1509 					     sizeof(udp6_addr), 0);
1510 			if (err < 0)
1511 				goto out;
1512 		} else
1513 #endif
1514 		{
1515 			err = sock_create_kern(AF_INET, SOCK_DGRAM, 0, &sock);
1516 			if (err < 0)
1517 				goto out;
1518 
1519 			sk_change_net(sock->sk, net);
1520 
1521 			udp_addr.sin_family = AF_INET;
1522 			udp_addr.sin_addr = cfg->local_ip;
1523 			udp_addr.sin_port = htons(cfg->local_udp_port);
1524 			err = kernel_bind(sock, (struct sockaddr *) &udp_addr,
1525 					  sizeof(udp_addr));
1526 			if (err < 0)
1527 				goto out;
1528 
1529 			udp_addr.sin_family = AF_INET;
1530 			udp_addr.sin_addr = cfg->peer_ip;
1531 			udp_addr.sin_port = htons(cfg->peer_udp_port);
1532 			err = kernel_connect(sock,
1533 					     (struct sockaddr *) &udp_addr,
1534 					     sizeof(udp_addr), 0);
1535 			if (err < 0)
1536 				goto out;
1537 		}
1538 
1539 		if (!cfg->use_udp_checksums)
1540 			sock->sk->sk_no_check = UDP_CSUM_NOXMIT;
1541 
1542 		break;
1543 
1544 	case L2TP_ENCAPTYPE_IP:
1545 #if IS_ENABLED(CONFIG_IPV6)
1546 		if (cfg->local_ip6 && cfg->peer_ip6) {
1547 			err = sock_create_kern(AF_INET6, SOCK_DGRAM,
1548 					  IPPROTO_L2TP, &sock);
1549 			if (err < 0)
1550 				goto out;
1551 
1552 			sk_change_net(sock->sk, net);
1553 
1554 			ip6_addr.l2tp_family = AF_INET6;
1555 			memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1556 			       sizeof(ip6_addr.l2tp_addr));
1557 			ip6_addr.l2tp_conn_id = tunnel_id;
1558 			err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1559 					  sizeof(ip6_addr));
1560 			if (err < 0)
1561 				goto out;
1562 
1563 			ip6_addr.l2tp_family = AF_INET6;
1564 			memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1565 			       sizeof(ip6_addr.l2tp_addr));
1566 			ip6_addr.l2tp_conn_id = peer_tunnel_id;
1567 			err = kernel_connect(sock,
1568 					     (struct sockaddr *) &ip6_addr,
1569 					     sizeof(ip6_addr), 0);
1570 			if (err < 0)
1571 				goto out;
1572 		} else
1573 #endif
1574 		{
1575 			err = sock_create_kern(AF_INET, SOCK_DGRAM,
1576 					  IPPROTO_L2TP, &sock);
1577 			if (err < 0)
1578 				goto out;
1579 
1580 			sk_change_net(sock->sk, net);
1581 
1582 			ip_addr.l2tp_family = AF_INET;
1583 			ip_addr.l2tp_addr = cfg->local_ip;
1584 			ip_addr.l2tp_conn_id = tunnel_id;
1585 			err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1586 					  sizeof(ip_addr));
1587 			if (err < 0)
1588 				goto out;
1589 
1590 			ip_addr.l2tp_family = AF_INET;
1591 			ip_addr.l2tp_addr = cfg->peer_ip;
1592 			ip_addr.l2tp_conn_id = peer_tunnel_id;
1593 			err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1594 					     sizeof(ip_addr), 0);
1595 			if (err < 0)
1596 				goto out;
1597 		}
1598 		break;
1599 
1600 	default:
1601 		goto out;
1602 	}
1603 
1604 out:
1605 	*sockp = sock;
1606 	if ((err < 0) && sock) {
1607 		kernel_sock_shutdown(sock, SHUT_RDWR);
1608 		sk_release_kernel(sock->sk);
1609 		*sockp = NULL;
1610 	}
1611 
1612 	return err;
1613 }
1614 
1615 static struct lock_class_key l2tp_socket_class;
1616 
1617 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)
1618 {
1619 	struct l2tp_tunnel *tunnel = NULL;
1620 	int err;
1621 	struct socket *sock = NULL;
1622 	struct sock *sk = NULL;
1623 	struct l2tp_net *pn;
1624 	enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1625 
1626 	/* Get the tunnel socket from the fd, which was opened by
1627 	 * the userspace L2TP daemon. If not specified, create a
1628 	 * kernel socket.
1629 	 */
1630 	if (fd < 0) {
1631 		err = l2tp_tunnel_sock_create(net, tunnel_id, peer_tunnel_id,
1632 				cfg, &sock);
1633 		if (err < 0)
1634 			goto err;
1635 	} else {
1636 		sock = sockfd_lookup(fd, &err);
1637 		if (!sock) {
1638 			pr_err("tunl %u: sockfd_lookup(fd=%d) returned %d\n",
1639 			       tunnel_id, fd, err);
1640 			err = -EBADF;
1641 			goto err;
1642 		}
1643 
1644 		/* Reject namespace mismatches */
1645 		if (!net_eq(sock_net(sock->sk), net)) {
1646 			pr_err("tunl %u: netns mismatch\n", tunnel_id);
1647 			err = -EINVAL;
1648 			goto err;
1649 		}
1650 	}
1651 
1652 	sk = sock->sk;
1653 
1654 	if (cfg != NULL)
1655 		encap = cfg->encap;
1656 
1657 	/* Quick sanity checks */
1658 	switch (encap) {
1659 	case L2TP_ENCAPTYPE_UDP:
1660 		err = -EPROTONOSUPPORT;
1661 		if (sk->sk_protocol != IPPROTO_UDP) {
1662 			pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1663 			       tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1664 			goto err;
1665 		}
1666 		break;
1667 	case L2TP_ENCAPTYPE_IP:
1668 		err = -EPROTONOSUPPORT;
1669 		if (sk->sk_protocol != IPPROTO_L2TP) {
1670 			pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1671 			       tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1672 			goto err;
1673 		}
1674 		break;
1675 	}
1676 
1677 	/* Check if this socket has already been prepped */
1678 	tunnel = (struct l2tp_tunnel *)sk->sk_user_data;
1679 	if (tunnel != NULL) {
1680 		/* This socket has already been prepped */
1681 		err = -EBUSY;
1682 		goto err;
1683 	}
1684 
1685 	tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1686 	if (tunnel == NULL) {
1687 		err = -ENOMEM;
1688 		goto err;
1689 	}
1690 
1691 	tunnel->version = version;
1692 	tunnel->tunnel_id = tunnel_id;
1693 	tunnel->peer_tunnel_id = peer_tunnel_id;
1694 	tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1695 
1696 	tunnel->magic = L2TP_TUNNEL_MAGIC;
1697 	sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1698 	rwlock_init(&tunnel->hlist_lock);
1699 
1700 	/* The net we belong to */
1701 	tunnel->l2tp_net = net;
1702 	pn = l2tp_pernet(net);
1703 
1704 	if (cfg != NULL)
1705 		tunnel->debug = cfg->debug;
1706 
1707 	/* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1708 	tunnel->encap = encap;
1709 	if (encap == L2TP_ENCAPTYPE_UDP) {
1710 		/* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1711 		udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1712 		udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1713 		udp_sk(sk)->encap_destroy = l2tp_udp_encap_destroy;
1714 #if IS_ENABLED(CONFIG_IPV6)
1715 		if (sk->sk_family == PF_INET6)
1716 			udpv6_encap_enable();
1717 		else
1718 #endif
1719 		udp_encap_enable();
1720 	}
1721 
1722 	sk->sk_user_data = tunnel;
1723 
1724 	/* Hook on the tunnel socket destructor so that we can cleanup
1725 	 * if the tunnel socket goes away.
1726 	 */
1727 	tunnel->old_sk_destruct = sk->sk_destruct;
1728 	sk->sk_destruct = &l2tp_tunnel_destruct;
1729 	tunnel->sock = sk;
1730 	tunnel->fd = fd;
1731 	lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class, "l2tp_sock");
1732 
1733 	sk->sk_allocation = GFP_ATOMIC;
1734 
1735 	/* Init delete workqueue struct */
1736 	INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1737 
1738 	/* Add tunnel to our list */
1739 	INIT_LIST_HEAD(&tunnel->list);
1740 	atomic_inc(&l2tp_tunnel_count);
1741 
1742 	/* Bump the reference count. The tunnel context is deleted
1743 	 * only when this drops to zero. Must be done before list insertion
1744 	 */
1745 	l2tp_tunnel_inc_refcount(tunnel);
1746 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1747 	list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1748 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1749 
1750 	err = 0;
1751 err:
1752 	if (tunnelp)
1753 		*tunnelp = tunnel;
1754 
1755 	/* If tunnel's socket was created by the kernel, it doesn't
1756 	 *  have a file.
1757 	 */
1758 	if (sock && sock->file)
1759 		sockfd_put(sock);
1760 
1761 	return err;
1762 }
1763 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1764 
1765 /* This function is used by the netlink TUNNEL_DELETE command.
1766  */
1767 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1768 {
1769 	l2tp_tunnel_closeall(tunnel);
1770 	return (false == queue_work(l2tp_wq, &tunnel->del_work));
1771 }
1772 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1773 
1774 /* Really kill the session.
1775  */
1776 void l2tp_session_free(struct l2tp_session *session)
1777 {
1778 	struct l2tp_tunnel *tunnel = session->tunnel;
1779 
1780 	BUG_ON(atomic_read(&session->ref_count) != 0);
1781 
1782 	if (tunnel) {
1783 		BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1784 		if (session->session_id != 0)
1785 			atomic_dec(&l2tp_session_count);
1786 		sock_put(tunnel->sock);
1787 		session->tunnel = NULL;
1788 		l2tp_tunnel_dec_refcount(tunnel);
1789 	}
1790 
1791 	kfree(session);
1792 
1793 	return;
1794 }
1795 EXPORT_SYMBOL_GPL(l2tp_session_free);
1796 
1797 /* Remove an l2tp session from l2tp_core's hash lists.
1798  * Provides a tidyup interface for pseudowire code which can't just route all
1799  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1800  * callback.
1801  */
1802 void __l2tp_session_unhash(struct l2tp_session *session)
1803 {
1804 	struct l2tp_tunnel *tunnel = session->tunnel;
1805 
1806 	/* Remove the session from core hashes */
1807 	if (tunnel) {
1808 		/* Remove from the per-tunnel hash */
1809 		write_lock_bh(&tunnel->hlist_lock);
1810 		hlist_del_init(&session->hlist);
1811 		write_unlock_bh(&tunnel->hlist_lock);
1812 
1813 		/* For L2TPv3 we have a per-net hash: remove from there, too */
1814 		if (tunnel->version != L2TP_HDR_VER_2) {
1815 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1816 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1817 			hlist_del_init_rcu(&session->global_hlist);
1818 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1819 			synchronize_rcu();
1820 		}
1821 	}
1822 }
1823 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1824 
1825 /* This function is used by the netlink SESSION_DELETE command and by
1826    pseudowire modules.
1827  */
1828 int l2tp_session_delete(struct l2tp_session *session)
1829 {
1830 	if (session->ref)
1831 		(*session->ref)(session);
1832 	__l2tp_session_unhash(session);
1833 	l2tp_session_queue_purge(session);
1834 	if (session->session_close != NULL)
1835 		(*session->session_close)(session);
1836 	if (session->deref)
1837 		(*session->deref)(session);
1838 	l2tp_session_dec_refcount(session);
1839 	return 0;
1840 }
1841 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1842 
1843 /* We come here whenever a session's send_seq, cookie_len or
1844  * l2specific_len parameters are set.
1845  */
1846 static void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1847 {
1848 	if (version == L2TP_HDR_VER_2) {
1849 		session->hdr_len = 6;
1850 		if (session->send_seq)
1851 			session->hdr_len += 4;
1852 	} else {
1853 		session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1854 		if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1855 			session->hdr_len += 4;
1856 	}
1857 
1858 }
1859 
1860 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)
1861 {
1862 	struct l2tp_session *session;
1863 
1864 	session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1865 	if (session != NULL) {
1866 		session->magic = L2TP_SESSION_MAGIC;
1867 		session->tunnel = tunnel;
1868 
1869 		session->session_id = session_id;
1870 		session->peer_session_id = peer_session_id;
1871 		session->nr = 0;
1872 		if (tunnel->version == L2TP_HDR_VER_2)
1873 			session->nr_max = 0xffff;
1874 		else
1875 			session->nr_max = 0xffffff;
1876 		session->nr_window_size = session->nr_max / 2;
1877 		session->nr_oos_count_max = 4;
1878 
1879 		/* Use NR of first received packet */
1880 		session->reorder_skip = 1;
1881 
1882 		sprintf(&session->name[0], "sess %u/%u",
1883 			tunnel->tunnel_id, session->session_id);
1884 
1885 		skb_queue_head_init(&session->reorder_q);
1886 
1887 		INIT_HLIST_NODE(&session->hlist);
1888 		INIT_HLIST_NODE(&session->global_hlist);
1889 
1890 		/* Inherit debug options from tunnel */
1891 		session->debug = tunnel->debug;
1892 
1893 		if (cfg) {
1894 			session->pwtype = cfg->pw_type;
1895 			session->debug = cfg->debug;
1896 			session->mtu = cfg->mtu;
1897 			session->mru = cfg->mru;
1898 			session->send_seq = cfg->send_seq;
1899 			session->recv_seq = cfg->recv_seq;
1900 			session->lns_mode = cfg->lns_mode;
1901 			session->reorder_timeout = cfg->reorder_timeout;
1902 			session->offset = cfg->offset;
1903 			session->l2specific_type = cfg->l2specific_type;
1904 			session->l2specific_len = cfg->l2specific_len;
1905 			session->cookie_len = cfg->cookie_len;
1906 			memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1907 			session->peer_cookie_len = cfg->peer_cookie_len;
1908 			memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1909 		}
1910 
1911 		if (tunnel->version == L2TP_HDR_VER_2)
1912 			session->build_header = l2tp_build_l2tpv2_header;
1913 		else
1914 			session->build_header = l2tp_build_l2tpv3_header;
1915 
1916 		l2tp_session_set_header_len(session, tunnel->version);
1917 
1918 		/* Bump the reference count. The session context is deleted
1919 		 * only when this drops to zero.
1920 		 */
1921 		l2tp_session_inc_refcount(session);
1922 		l2tp_tunnel_inc_refcount(tunnel);
1923 
1924 		/* Ensure tunnel socket isn't deleted */
1925 		sock_hold(tunnel->sock);
1926 
1927 		/* Add session to the tunnel's hash list */
1928 		write_lock_bh(&tunnel->hlist_lock);
1929 		hlist_add_head(&session->hlist,
1930 			       l2tp_session_id_hash(tunnel, session_id));
1931 		write_unlock_bh(&tunnel->hlist_lock);
1932 
1933 		/* And to the global session list if L2TPv3 */
1934 		if (tunnel->version != L2TP_HDR_VER_2) {
1935 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1936 
1937 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1938 			hlist_add_head_rcu(&session->global_hlist,
1939 					   l2tp_session_id_hash_2(pn, session_id));
1940 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1941 		}
1942 
1943 		/* Ignore management session in session count value */
1944 		if (session->session_id != 0)
1945 			atomic_inc(&l2tp_session_count);
1946 	}
1947 
1948 	return session;
1949 }
1950 EXPORT_SYMBOL_GPL(l2tp_session_create);
1951 
1952 /*****************************************************************************
1953  * Init and cleanup
1954  *****************************************************************************/
1955 
1956 static __net_init int l2tp_init_net(struct net *net)
1957 {
1958 	struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1959 	int hash;
1960 
1961 	INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1962 	spin_lock_init(&pn->l2tp_tunnel_list_lock);
1963 
1964 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1965 		INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1966 
1967 	spin_lock_init(&pn->l2tp_session_hlist_lock);
1968 
1969 	return 0;
1970 }
1971 
1972 static __net_exit void l2tp_exit_net(struct net *net)
1973 {
1974 	struct l2tp_net *pn = l2tp_pernet(net);
1975 	struct l2tp_tunnel *tunnel = NULL;
1976 
1977 	rcu_read_lock_bh();
1978 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1979 		(void)l2tp_tunnel_delete(tunnel);
1980 	}
1981 	rcu_read_unlock_bh();
1982 }
1983 
1984 static struct pernet_operations l2tp_net_ops = {
1985 	.init = l2tp_init_net,
1986 	.exit = l2tp_exit_net,
1987 	.id   = &l2tp_net_id,
1988 	.size = sizeof(struct l2tp_net),
1989 };
1990 
1991 static int __init l2tp_init(void)
1992 {
1993 	int rc = 0;
1994 
1995 	rc = register_pernet_device(&l2tp_net_ops);
1996 	if (rc)
1997 		goto out;
1998 
1999 	l2tp_wq = alloc_workqueue("l2tp", WQ_NON_REENTRANT | WQ_UNBOUND, 0);
2000 	if (!l2tp_wq) {
2001 		pr_err("alloc_workqueue failed\n");
2002 		rc = -ENOMEM;
2003 		goto out;
2004 	}
2005 
2006 	pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
2007 
2008 out:
2009 	return rc;
2010 }
2011 
2012 static void __exit l2tp_exit(void)
2013 {
2014 	unregister_pernet_device(&l2tp_net_ops);
2015 	if (l2tp_wq) {
2016 		destroy_workqueue(l2tp_wq);
2017 		l2tp_wq = NULL;
2018 	}
2019 }
2020 
2021 module_init(l2tp_init);
2022 module_exit(l2tp_exit);
2023 
2024 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
2025 MODULE_DESCRIPTION("L2TP core");
2026 MODULE_LICENSE("GPL");
2027 MODULE_VERSION(L2TP_DRV_VERSION);
2028 
2029