xref: /linux/net/l2tp/l2tp_core.c (revision 6fdcba32711044c35c0e1b094cbd8f3f0b4472c9)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * L2TP core.
4  *
5  * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
6  *
7  * This file contains some code of the original L2TPv2 pppol2tp
8  * driver, which has the following copyright:
9  *
10  * Authors:	Martijn van Oosterhout <kleptog@svana.org>
11  *		James Chapman (jchapman@katalix.com)
12  * Contributors:
13  *		Michal Ostrowski <mostrows@speakeasy.net>
14  *		Arnaldo Carvalho de Melo <acme@xconectiva.com.br>
15  *		David S. Miller (davem@redhat.com)
16  */
17 
18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
19 
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/list.h>
23 #include <linux/rculist.h>
24 #include <linux/uaccess.h>
25 
26 #include <linux/kernel.h>
27 #include <linux/spinlock.h>
28 #include <linux/kthread.h>
29 #include <linux/sched.h>
30 #include <linux/slab.h>
31 #include <linux/errno.h>
32 #include <linux/jiffies.h>
33 
34 #include <linux/netdevice.h>
35 #include <linux/net.h>
36 #include <linux/inetdevice.h>
37 #include <linux/skbuff.h>
38 #include <linux/init.h>
39 #include <linux/in.h>
40 #include <linux/ip.h>
41 #include <linux/udp.h>
42 #include <linux/l2tp.h>
43 #include <linux/hash.h>
44 #include <linux/sort.h>
45 #include <linux/file.h>
46 #include <linux/nsproxy.h>
47 #include <net/net_namespace.h>
48 #include <net/netns/generic.h>
49 #include <net/dst.h>
50 #include <net/ip.h>
51 #include <net/udp.h>
52 #include <net/udp_tunnel.h>
53 #include <net/inet_common.h>
54 #include <net/xfrm.h>
55 #include <net/protocol.h>
56 #include <net/inet6_connection_sock.h>
57 #include <net/inet_ecn.h>
58 #include <net/ip6_route.h>
59 #include <net/ip6_checksum.h>
60 
61 #include <asm/byteorder.h>
62 #include <linux/atomic.h>
63 
64 #include "l2tp_core.h"
65 
66 #define L2TP_DRV_VERSION	"V2.0"
67 
68 /* L2TP header constants */
69 #define L2TP_HDRFLAG_T	   0x8000
70 #define L2TP_HDRFLAG_L	   0x4000
71 #define L2TP_HDRFLAG_S	   0x0800
72 #define L2TP_HDRFLAG_O	   0x0200
73 #define L2TP_HDRFLAG_P	   0x0100
74 
75 #define L2TP_HDR_VER_MASK  0x000F
76 #define L2TP_HDR_VER_2	   0x0002
77 #define L2TP_HDR_VER_3	   0x0003
78 
79 /* L2TPv3 default L2-specific sublayer */
80 #define L2TP_SLFLAG_S	   0x40000000
81 #define L2TP_SL_SEQ_MASK   0x00ffffff
82 
83 #define L2TP_HDR_SIZE_MAX		14
84 
85 /* Default trace flags */
86 #define L2TP_DEFAULT_DEBUG_FLAGS	0
87 
88 /* Private data stored for received packets in the skb.
89  */
90 struct l2tp_skb_cb {
91 	u32			ns;
92 	u16			has_seq;
93 	u16			length;
94 	unsigned long		expires;
95 };
96 
97 #define L2TP_SKB_CB(skb)	((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
98 
99 static struct workqueue_struct *l2tp_wq;
100 
101 /* per-net private data for this module */
102 static unsigned int l2tp_net_id;
103 struct l2tp_net {
104 	struct list_head l2tp_tunnel_list;
105 	spinlock_t l2tp_tunnel_list_lock;
106 	struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
107 	spinlock_t l2tp_session_hlist_lock;
108 };
109 
110 #if IS_ENABLED(CONFIG_IPV6)
111 static bool l2tp_sk_is_v6(struct sock *sk)
112 {
113 	return sk->sk_family == PF_INET6 &&
114 	       !ipv6_addr_v4mapped(&sk->sk_v6_daddr);
115 }
116 #endif
117 
118 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
119 {
120 	return sk->sk_user_data;
121 }
122 
123 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
124 {
125 	BUG_ON(!net);
126 
127 	return net_generic(net, l2tp_net_id);
128 }
129 
130 /* Session hash global list for L2TPv3.
131  * The session_id SHOULD be random according to RFC3931, but several
132  * L2TP implementations use incrementing session_ids.  So we do a real
133  * hash on the session_id, rather than a simple bitmask.
134  */
135 static inline struct hlist_head *
136 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
137 {
138 	return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
139 
140 }
141 
142 /* Session hash list.
143  * The session_id SHOULD be random according to RFC2661, but several
144  * L2TP implementations (Cisco and Microsoft) use incrementing
145  * session_ids.  So we do a real hash on the session_id, rather than a
146  * simple bitmask.
147  */
148 static inline struct hlist_head *
149 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
150 {
151 	return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
152 }
153 
154 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
155 {
156 	sock_put(tunnel->sock);
157 	/* the tunnel is freed in the socket destructor */
158 }
159 EXPORT_SYMBOL(l2tp_tunnel_free);
160 
161 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
162 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
163 {
164 	const struct l2tp_net *pn = l2tp_pernet(net);
165 	struct l2tp_tunnel *tunnel;
166 
167 	rcu_read_lock_bh();
168 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
169 		if (tunnel->tunnel_id == tunnel_id &&
170 		    refcount_inc_not_zero(&tunnel->ref_count)) {
171 			rcu_read_unlock_bh();
172 
173 			return tunnel;
174 		}
175 	}
176 	rcu_read_unlock_bh();
177 
178 	return NULL;
179 }
180 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
181 
182 struct l2tp_tunnel *l2tp_tunnel_get_nth(const struct net *net, int nth)
183 {
184 	const struct l2tp_net *pn = l2tp_pernet(net);
185 	struct l2tp_tunnel *tunnel;
186 	int count = 0;
187 
188 	rcu_read_lock_bh();
189 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
190 		if (++count > nth &&
191 		    refcount_inc_not_zero(&tunnel->ref_count)) {
192 			rcu_read_unlock_bh();
193 			return tunnel;
194 		}
195 	}
196 	rcu_read_unlock_bh();
197 
198 	return NULL;
199 }
200 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_nth);
201 
202 struct l2tp_session *l2tp_tunnel_get_session(struct l2tp_tunnel *tunnel,
203 					     u32 session_id)
204 {
205 	struct hlist_head *session_list;
206 	struct l2tp_session *session;
207 
208 	session_list = l2tp_session_id_hash(tunnel, session_id);
209 
210 	read_lock_bh(&tunnel->hlist_lock);
211 	hlist_for_each_entry(session, session_list, hlist)
212 		if (session->session_id == session_id) {
213 			l2tp_session_inc_refcount(session);
214 			read_unlock_bh(&tunnel->hlist_lock);
215 
216 			return session;
217 		}
218 	read_unlock_bh(&tunnel->hlist_lock);
219 
220 	return NULL;
221 }
222 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_session);
223 
224 struct l2tp_session *l2tp_session_get(const struct net *net, u32 session_id)
225 {
226 	struct hlist_head *session_list;
227 	struct l2tp_session *session;
228 
229 	session_list = l2tp_session_id_hash_2(l2tp_pernet(net), session_id);
230 
231 	rcu_read_lock_bh();
232 	hlist_for_each_entry_rcu(session, session_list, global_hlist)
233 		if (session->session_id == session_id) {
234 			l2tp_session_inc_refcount(session);
235 			rcu_read_unlock_bh();
236 
237 			return session;
238 		}
239 	rcu_read_unlock_bh();
240 
241 	return NULL;
242 }
243 EXPORT_SYMBOL_GPL(l2tp_session_get);
244 
245 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth)
246 {
247 	int hash;
248 	struct l2tp_session *session;
249 	int count = 0;
250 
251 	read_lock_bh(&tunnel->hlist_lock);
252 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
253 		hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
254 			if (++count > nth) {
255 				l2tp_session_inc_refcount(session);
256 				read_unlock_bh(&tunnel->hlist_lock);
257 				return session;
258 			}
259 		}
260 	}
261 
262 	read_unlock_bh(&tunnel->hlist_lock);
263 
264 	return NULL;
265 }
266 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
267 
268 /* Lookup a session by interface name.
269  * This is very inefficient but is only used by management interfaces.
270  */
271 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
272 						const char *ifname)
273 {
274 	struct l2tp_net *pn = l2tp_pernet(net);
275 	int hash;
276 	struct l2tp_session *session;
277 
278 	rcu_read_lock_bh();
279 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
280 		hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
281 			if (!strcmp(session->ifname, ifname)) {
282 				l2tp_session_inc_refcount(session);
283 				rcu_read_unlock_bh();
284 
285 				return session;
286 			}
287 		}
288 	}
289 
290 	rcu_read_unlock_bh();
291 
292 	return NULL;
293 }
294 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
295 
296 int l2tp_session_register(struct l2tp_session *session,
297 			  struct l2tp_tunnel *tunnel)
298 {
299 	struct l2tp_session *session_walk;
300 	struct hlist_head *g_head;
301 	struct hlist_head *head;
302 	struct l2tp_net *pn;
303 	int err;
304 
305 	head = l2tp_session_id_hash(tunnel, session->session_id);
306 
307 	write_lock_bh(&tunnel->hlist_lock);
308 	if (!tunnel->acpt_newsess) {
309 		err = -ENODEV;
310 		goto err_tlock;
311 	}
312 
313 	hlist_for_each_entry(session_walk, head, hlist)
314 		if (session_walk->session_id == session->session_id) {
315 			err = -EEXIST;
316 			goto err_tlock;
317 		}
318 
319 	if (tunnel->version == L2TP_HDR_VER_3) {
320 		pn = l2tp_pernet(tunnel->l2tp_net);
321 		g_head = l2tp_session_id_hash_2(pn, session->session_id);
322 
323 		spin_lock_bh(&pn->l2tp_session_hlist_lock);
324 
325 		hlist_for_each_entry(session_walk, g_head, global_hlist)
326 			if (session_walk->session_id == session->session_id) {
327 				err = -EEXIST;
328 				goto err_tlock_pnlock;
329 			}
330 
331 		l2tp_tunnel_inc_refcount(tunnel);
332 		hlist_add_head_rcu(&session->global_hlist, g_head);
333 
334 		spin_unlock_bh(&pn->l2tp_session_hlist_lock);
335 	} else {
336 		l2tp_tunnel_inc_refcount(tunnel);
337 	}
338 
339 	hlist_add_head(&session->hlist, head);
340 	write_unlock_bh(&tunnel->hlist_lock);
341 
342 	return 0;
343 
344 err_tlock_pnlock:
345 	spin_unlock_bh(&pn->l2tp_session_hlist_lock);
346 err_tlock:
347 	write_unlock_bh(&tunnel->hlist_lock);
348 
349 	return err;
350 }
351 EXPORT_SYMBOL_GPL(l2tp_session_register);
352 
353 /*****************************************************************************
354  * Receive data handling
355  *****************************************************************************/
356 
357 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
358  * number.
359  */
360 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
361 {
362 	struct sk_buff *skbp;
363 	struct sk_buff *tmp;
364 	u32 ns = L2TP_SKB_CB(skb)->ns;
365 
366 	spin_lock_bh(&session->reorder_q.lock);
367 	skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
368 		if (L2TP_SKB_CB(skbp)->ns > ns) {
369 			__skb_queue_before(&session->reorder_q, skbp, skb);
370 			l2tp_dbg(session, L2TP_MSG_SEQ,
371 				 "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
372 				 session->name, ns, L2TP_SKB_CB(skbp)->ns,
373 				 skb_queue_len(&session->reorder_q));
374 			atomic_long_inc(&session->stats.rx_oos_packets);
375 			goto out;
376 		}
377 	}
378 
379 	__skb_queue_tail(&session->reorder_q, skb);
380 
381 out:
382 	spin_unlock_bh(&session->reorder_q.lock);
383 }
384 
385 /* Dequeue a single skb.
386  */
387 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
388 {
389 	struct l2tp_tunnel *tunnel = session->tunnel;
390 	int length = L2TP_SKB_CB(skb)->length;
391 
392 	/* We're about to requeue the skb, so return resources
393 	 * to its current owner (a socket receive buffer).
394 	 */
395 	skb_orphan(skb);
396 
397 	atomic_long_inc(&tunnel->stats.rx_packets);
398 	atomic_long_add(length, &tunnel->stats.rx_bytes);
399 	atomic_long_inc(&session->stats.rx_packets);
400 	atomic_long_add(length, &session->stats.rx_bytes);
401 
402 	if (L2TP_SKB_CB(skb)->has_seq) {
403 		/* Bump our Nr */
404 		session->nr++;
405 		session->nr &= session->nr_max;
406 
407 		l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
408 			 session->name, session->nr);
409 	}
410 
411 	/* call private receive handler */
412 	if (session->recv_skb != NULL)
413 		(*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
414 	else
415 		kfree_skb(skb);
416 }
417 
418 /* Dequeue skbs from the session's reorder_q, subject to packet order.
419  * Skbs that have been in the queue for too long are simply discarded.
420  */
421 static void l2tp_recv_dequeue(struct l2tp_session *session)
422 {
423 	struct sk_buff *skb;
424 	struct sk_buff *tmp;
425 
426 	/* If the pkt at the head of the queue has the nr that we
427 	 * expect to send up next, dequeue it and any other
428 	 * in-sequence packets behind it.
429 	 */
430 start:
431 	spin_lock_bh(&session->reorder_q.lock);
432 	skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
433 		if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
434 			atomic_long_inc(&session->stats.rx_seq_discards);
435 			atomic_long_inc(&session->stats.rx_errors);
436 			l2tp_dbg(session, L2TP_MSG_SEQ,
437 				 "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
438 				 session->name, L2TP_SKB_CB(skb)->ns,
439 				 L2TP_SKB_CB(skb)->length, session->nr,
440 				 skb_queue_len(&session->reorder_q));
441 			session->reorder_skip = 1;
442 			__skb_unlink(skb, &session->reorder_q);
443 			kfree_skb(skb);
444 			continue;
445 		}
446 
447 		if (L2TP_SKB_CB(skb)->has_seq) {
448 			if (session->reorder_skip) {
449 				l2tp_dbg(session, L2TP_MSG_SEQ,
450 					 "%s: advancing nr to next pkt: %u -> %u",
451 					 session->name, session->nr,
452 					 L2TP_SKB_CB(skb)->ns);
453 				session->reorder_skip = 0;
454 				session->nr = L2TP_SKB_CB(skb)->ns;
455 			}
456 			if (L2TP_SKB_CB(skb)->ns != session->nr) {
457 				l2tp_dbg(session, L2TP_MSG_SEQ,
458 					 "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
459 					 session->name, L2TP_SKB_CB(skb)->ns,
460 					 L2TP_SKB_CB(skb)->length, session->nr,
461 					 skb_queue_len(&session->reorder_q));
462 				goto out;
463 			}
464 		}
465 		__skb_unlink(skb, &session->reorder_q);
466 
467 		/* Process the skb. We release the queue lock while we
468 		 * do so to let other contexts process the queue.
469 		 */
470 		spin_unlock_bh(&session->reorder_q.lock);
471 		l2tp_recv_dequeue_skb(session, skb);
472 		goto start;
473 	}
474 
475 out:
476 	spin_unlock_bh(&session->reorder_q.lock);
477 }
478 
479 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
480 {
481 	u32 nws;
482 
483 	if (nr >= session->nr)
484 		nws = nr - session->nr;
485 	else
486 		nws = (session->nr_max + 1) - (session->nr - nr);
487 
488 	return nws < session->nr_window_size;
489 }
490 
491 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
492  * acceptable, else non-zero.
493  */
494 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
495 {
496 	if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
497 		/* Packet sequence number is outside allowed window.
498 		 * Discard it.
499 		 */
500 		l2tp_dbg(session, L2TP_MSG_SEQ,
501 			 "%s: pkt %u len %d discarded, outside window, nr=%u\n",
502 			 session->name, L2TP_SKB_CB(skb)->ns,
503 			 L2TP_SKB_CB(skb)->length, session->nr);
504 		goto discard;
505 	}
506 
507 	if (session->reorder_timeout != 0) {
508 		/* Packet reordering enabled. Add skb to session's
509 		 * reorder queue, in order of ns.
510 		 */
511 		l2tp_recv_queue_skb(session, skb);
512 		goto out;
513 	}
514 
515 	/* Packet reordering disabled. Discard out-of-sequence packets, while
516 	 * tracking the number if in-sequence packets after the first OOS packet
517 	 * is seen. After nr_oos_count_max in-sequence packets, reset the
518 	 * sequence number to re-enable packet reception.
519 	 */
520 	if (L2TP_SKB_CB(skb)->ns == session->nr) {
521 		skb_queue_tail(&session->reorder_q, skb);
522 	} else {
523 		u32 nr_oos = L2TP_SKB_CB(skb)->ns;
524 		u32 nr_next = (session->nr_oos + 1) & session->nr_max;
525 
526 		if (nr_oos == nr_next)
527 			session->nr_oos_count++;
528 		else
529 			session->nr_oos_count = 0;
530 
531 		session->nr_oos = nr_oos;
532 		if (session->nr_oos_count > session->nr_oos_count_max) {
533 			session->reorder_skip = 1;
534 			l2tp_dbg(session, L2TP_MSG_SEQ,
535 				 "%s: %d oos packets received. Resetting sequence numbers\n",
536 				 session->name, session->nr_oos_count);
537 		}
538 		if (!session->reorder_skip) {
539 			atomic_long_inc(&session->stats.rx_seq_discards);
540 			l2tp_dbg(session, L2TP_MSG_SEQ,
541 				 "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
542 				 session->name, L2TP_SKB_CB(skb)->ns,
543 				 L2TP_SKB_CB(skb)->length, session->nr,
544 				 skb_queue_len(&session->reorder_q));
545 			goto discard;
546 		}
547 		skb_queue_tail(&session->reorder_q, skb);
548 	}
549 
550 out:
551 	return 0;
552 
553 discard:
554 	return 1;
555 }
556 
557 /* Do receive processing of L2TP data frames. We handle both L2TPv2
558  * and L2TPv3 data frames here.
559  *
560  * L2TPv2 Data Message Header
561  *
562  *  0                   1                   2                   3
563  *  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
564  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
565  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
566  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
567  * |           Tunnel ID           |           Session ID          |
568  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
569  * |             Ns (opt)          |             Nr (opt)          |
570  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
571  * |      Offset Size (opt)        |    Offset pad... (opt)
572  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
573  *
574  * Data frames are marked by T=0. All other fields are the same as
575  * those in L2TP control frames.
576  *
577  * L2TPv3 Data Message Header
578  *
579  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
580  * |                      L2TP Session Header                      |
581  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
582  * |                      L2-Specific Sublayer                     |
583  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
584  * |                        Tunnel Payload                      ...
585  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
586  *
587  * L2TPv3 Session Header Over IP
588  *
589  *  0                   1                   2                   3
590  *  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
591  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
592  * |                           Session ID                          |
593  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
594  * |               Cookie (optional, maximum 64 bits)...
595  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
596  *                                                                 |
597  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
598  *
599  * L2TPv3 L2-Specific Sublayer Format
600  *
601  *  0                   1                   2                   3
602  *  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
603  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
604  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
605  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
606  *
607  * Cookie value and sublayer format are negotiated with the peer when
608  * the session is set up. Unlike L2TPv2, we do not need to parse the
609  * packet header to determine if optional fields are present.
610  *
611  * Caller must already have parsed the frame and determined that it is
612  * a data (not control) frame before coming here. Fields up to the
613  * session-id have already been parsed and ptr points to the data
614  * after the session-id.
615  */
616 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
617 		      unsigned char *ptr, unsigned char *optr, u16 hdrflags,
618 		      int length)
619 {
620 	struct l2tp_tunnel *tunnel = session->tunnel;
621 	int offset;
622 	u32 ns, nr;
623 
624 	/* Parse and check optional cookie */
625 	if (session->peer_cookie_len > 0) {
626 		if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
627 			l2tp_info(tunnel, L2TP_MSG_DATA,
628 				  "%s: cookie mismatch (%u/%u). Discarding.\n",
629 				  tunnel->name, tunnel->tunnel_id,
630 				  session->session_id);
631 			atomic_long_inc(&session->stats.rx_cookie_discards);
632 			goto discard;
633 		}
634 		ptr += session->peer_cookie_len;
635 	}
636 
637 	/* Handle the optional sequence numbers. Sequence numbers are
638 	 * in different places for L2TPv2 and L2TPv3.
639 	 *
640 	 * If we are the LAC, enable/disable sequence numbers under
641 	 * the control of the LNS.  If no sequence numbers present but
642 	 * we were expecting them, discard frame.
643 	 */
644 	ns = nr = 0;
645 	L2TP_SKB_CB(skb)->has_seq = 0;
646 	if (tunnel->version == L2TP_HDR_VER_2) {
647 		if (hdrflags & L2TP_HDRFLAG_S) {
648 			ns = ntohs(*(__be16 *) ptr);
649 			ptr += 2;
650 			nr = ntohs(*(__be16 *) ptr);
651 			ptr += 2;
652 
653 			/* Store L2TP info in the skb */
654 			L2TP_SKB_CB(skb)->ns = ns;
655 			L2TP_SKB_CB(skb)->has_seq = 1;
656 
657 			l2tp_dbg(session, L2TP_MSG_SEQ,
658 				 "%s: recv data ns=%u, nr=%u, session nr=%u\n",
659 				 session->name, ns, nr, session->nr);
660 		}
661 	} else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
662 		u32 l2h = ntohl(*(__be32 *) ptr);
663 
664 		if (l2h & 0x40000000) {
665 			ns = l2h & 0x00ffffff;
666 
667 			/* Store L2TP info in the skb */
668 			L2TP_SKB_CB(skb)->ns = ns;
669 			L2TP_SKB_CB(skb)->has_seq = 1;
670 
671 			l2tp_dbg(session, L2TP_MSG_SEQ,
672 				 "%s: recv data ns=%u, session nr=%u\n",
673 				 session->name, ns, session->nr);
674 		}
675 		ptr += 4;
676 	}
677 
678 	if (L2TP_SKB_CB(skb)->has_seq) {
679 		/* Received a packet with sequence numbers. If we're the LNS,
680 		 * check if we sre sending sequence numbers and if not,
681 		 * configure it so.
682 		 */
683 		if ((!session->lns_mode) && (!session->send_seq)) {
684 			l2tp_info(session, L2TP_MSG_SEQ,
685 				  "%s: requested to enable seq numbers by LNS\n",
686 				  session->name);
687 			session->send_seq = 1;
688 			l2tp_session_set_header_len(session, tunnel->version);
689 		}
690 	} else {
691 		/* No sequence numbers.
692 		 * If user has configured mandatory sequence numbers, discard.
693 		 */
694 		if (session->recv_seq) {
695 			l2tp_warn(session, L2TP_MSG_SEQ,
696 				  "%s: recv data has no seq numbers when required. Discarding.\n",
697 				  session->name);
698 			atomic_long_inc(&session->stats.rx_seq_discards);
699 			goto discard;
700 		}
701 
702 		/* If we're the LAC and we're sending sequence numbers, the
703 		 * LNS has requested that we no longer send sequence numbers.
704 		 * If we're the LNS and we're sending sequence numbers, the
705 		 * LAC is broken. Discard the frame.
706 		 */
707 		if ((!session->lns_mode) && (session->send_seq)) {
708 			l2tp_info(session, L2TP_MSG_SEQ,
709 				  "%s: requested to disable seq numbers by LNS\n",
710 				  session->name);
711 			session->send_seq = 0;
712 			l2tp_session_set_header_len(session, tunnel->version);
713 		} else if (session->send_seq) {
714 			l2tp_warn(session, L2TP_MSG_SEQ,
715 				  "%s: recv data has no seq numbers when required. Discarding.\n",
716 				  session->name);
717 			atomic_long_inc(&session->stats.rx_seq_discards);
718 			goto discard;
719 		}
720 	}
721 
722 	/* Session data offset is defined only for L2TPv2 and is
723 	 * indicated by an optional 16-bit value in the header.
724 	 */
725 	if (tunnel->version == L2TP_HDR_VER_2) {
726 		/* If offset bit set, skip it. */
727 		if (hdrflags & L2TP_HDRFLAG_O) {
728 			offset = ntohs(*(__be16 *)ptr);
729 			ptr += 2 + offset;
730 		}
731 	}
732 
733 	offset = ptr - optr;
734 	if (!pskb_may_pull(skb, offset))
735 		goto discard;
736 
737 	__skb_pull(skb, offset);
738 
739 	/* Prepare skb for adding to the session's reorder_q.  Hold
740 	 * packets for max reorder_timeout or 1 second if not
741 	 * reordering.
742 	 */
743 	L2TP_SKB_CB(skb)->length = length;
744 	L2TP_SKB_CB(skb)->expires = jiffies +
745 		(session->reorder_timeout ? session->reorder_timeout : HZ);
746 
747 	/* Add packet to the session's receive queue. Reordering is done here, if
748 	 * enabled. Saved L2TP protocol info is stored in skb->sb[].
749 	 */
750 	if (L2TP_SKB_CB(skb)->has_seq) {
751 		if (l2tp_recv_data_seq(session, skb))
752 			goto discard;
753 	} else {
754 		/* No sequence numbers. Add the skb to the tail of the
755 		 * reorder queue. This ensures that it will be
756 		 * delivered after all previous sequenced skbs.
757 		 */
758 		skb_queue_tail(&session->reorder_q, skb);
759 	}
760 
761 	/* Try to dequeue as many skbs from reorder_q as we can. */
762 	l2tp_recv_dequeue(session);
763 
764 	return;
765 
766 discard:
767 	atomic_long_inc(&session->stats.rx_errors);
768 	kfree_skb(skb);
769 }
770 EXPORT_SYMBOL(l2tp_recv_common);
771 
772 /* Drop skbs from the session's reorder_q
773  */
774 static int l2tp_session_queue_purge(struct l2tp_session *session)
775 {
776 	struct sk_buff *skb = NULL;
777 	BUG_ON(!session);
778 	BUG_ON(session->magic != L2TP_SESSION_MAGIC);
779 	while ((skb = skb_dequeue(&session->reorder_q))) {
780 		atomic_long_inc(&session->stats.rx_errors);
781 		kfree_skb(skb);
782 	}
783 	return 0;
784 }
785 
786 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
787  * here. The skb is not on a list when we get here.
788  * Returns 0 if the packet was a data packet and was successfully passed on.
789  * Returns 1 if the packet was not a good data packet and could not be
790  * forwarded.  All such packets are passed up to userspace to deal with.
791  */
792 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb)
793 {
794 	struct l2tp_session *session = NULL;
795 	unsigned char *ptr, *optr;
796 	u16 hdrflags;
797 	u32 tunnel_id, session_id;
798 	u16 version;
799 	int length;
800 
801 	/* UDP has verifed checksum */
802 
803 	/* UDP always verifies the packet length. */
804 	__skb_pull(skb, sizeof(struct udphdr));
805 
806 	/* Short packet? */
807 	if (!pskb_may_pull(skb, L2TP_HDR_SIZE_MAX)) {
808 		l2tp_info(tunnel, L2TP_MSG_DATA,
809 			  "%s: recv short packet (len=%d)\n",
810 			  tunnel->name, skb->len);
811 		goto error;
812 	}
813 
814 	/* Trace packet contents, if enabled */
815 	if (tunnel->debug & L2TP_MSG_DATA) {
816 		length = min(32u, skb->len);
817 		if (!pskb_may_pull(skb, length))
818 			goto error;
819 
820 		pr_debug("%s: recv\n", tunnel->name);
821 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
822 	}
823 
824 	/* Point to L2TP header */
825 	optr = ptr = skb->data;
826 
827 	/* Get L2TP header flags */
828 	hdrflags = ntohs(*(__be16 *) ptr);
829 
830 	/* Check protocol version */
831 	version = hdrflags & L2TP_HDR_VER_MASK;
832 	if (version != tunnel->version) {
833 		l2tp_info(tunnel, L2TP_MSG_DATA,
834 			  "%s: recv protocol version mismatch: got %d expected %d\n",
835 			  tunnel->name, version, tunnel->version);
836 		goto error;
837 	}
838 
839 	/* Get length of L2TP packet */
840 	length = skb->len;
841 
842 	/* If type is control packet, it is handled by userspace. */
843 	if (hdrflags & L2TP_HDRFLAG_T) {
844 		l2tp_dbg(tunnel, L2TP_MSG_DATA,
845 			 "%s: recv control packet, len=%d\n",
846 			 tunnel->name, length);
847 		goto error;
848 	}
849 
850 	/* Skip flags */
851 	ptr += 2;
852 
853 	if (tunnel->version == L2TP_HDR_VER_2) {
854 		/* If length is present, skip it */
855 		if (hdrflags & L2TP_HDRFLAG_L)
856 			ptr += 2;
857 
858 		/* Extract tunnel and session ID */
859 		tunnel_id = ntohs(*(__be16 *) ptr);
860 		ptr += 2;
861 		session_id = ntohs(*(__be16 *) ptr);
862 		ptr += 2;
863 	} else {
864 		ptr += 2;	/* skip reserved bits */
865 		tunnel_id = tunnel->tunnel_id;
866 		session_id = ntohl(*(__be32 *) ptr);
867 		ptr += 4;
868 	}
869 
870 	/* Find the session context */
871 	session = l2tp_tunnel_get_session(tunnel, session_id);
872 	if (!session || !session->recv_skb) {
873 		if (session)
874 			l2tp_session_dec_refcount(session);
875 
876 		/* Not found? Pass to userspace to deal with */
877 		l2tp_info(tunnel, L2TP_MSG_DATA,
878 			  "%s: no session found (%u/%u). Passing up.\n",
879 			  tunnel->name, tunnel_id, session_id);
880 		goto error;
881 	}
882 
883 	if (tunnel->version == L2TP_HDR_VER_3 &&
884 	    l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
885 		goto error;
886 
887 	l2tp_recv_common(session, skb, ptr, optr, hdrflags, length);
888 	l2tp_session_dec_refcount(session);
889 
890 	return 0;
891 
892 error:
893 	/* Put UDP header back */
894 	__skb_push(skb, sizeof(struct udphdr));
895 
896 	return 1;
897 }
898 
899 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
900  * Return codes:
901  * 0 : success.
902  * <0: error
903  * >0: skb should be passed up to userspace as UDP.
904  */
905 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
906 {
907 	struct l2tp_tunnel *tunnel;
908 
909 	tunnel = rcu_dereference_sk_user_data(sk);
910 	if (tunnel == NULL)
911 		goto pass_up;
912 
913 	l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
914 		 tunnel->name, skb->len);
915 
916 	if (l2tp_udp_recv_core(tunnel, skb))
917 		goto pass_up;
918 
919 	return 0;
920 
921 pass_up:
922 	return 1;
923 }
924 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
925 
926 /************************************************************************
927  * Transmit handling
928  ***********************************************************************/
929 
930 /* Build an L2TP header for the session into the buffer provided.
931  */
932 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
933 {
934 	struct l2tp_tunnel *tunnel = session->tunnel;
935 	__be16 *bufp = buf;
936 	__be16 *optr = buf;
937 	u16 flags = L2TP_HDR_VER_2;
938 	u32 tunnel_id = tunnel->peer_tunnel_id;
939 	u32 session_id = session->peer_session_id;
940 
941 	if (session->send_seq)
942 		flags |= L2TP_HDRFLAG_S;
943 
944 	/* Setup L2TP header. */
945 	*bufp++ = htons(flags);
946 	*bufp++ = htons(tunnel_id);
947 	*bufp++ = htons(session_id);
948 	if (session->send_seq) {
949 		*bufp++ = htons(session->ns);
950 		*bufp++ = 0;
951 		session->ns++;
952 		session->ns &= 0xffff;
953 		l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
954 			 session->name, session->ns);
955 	}
956 
957 	return bufp - optr;
958 }
959 
960 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
961 {
962 	struct l2tp_tunnel *tunnel = session->tunnel;
963 	char *bufp = buf;
964 	char *optr = bufp;
965 
966 	/* Setup L2TP header. The header differs slightly for UDP and
967 	 * IP encapsulations. For UDP, there is 4 bytes of flags.
968 	 */
969 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
970 		u16 flags = L2TP_HDR_VER_3;
971 		*((__be16 *) bufp) = htons(flags);
972 		bufp += 2;
973 		*((__be16 *) bufp) = 0;
974 		bufp += 2;
975 	}
976 
977 	*((__be32 *) bufp) = htonl(session->peer_session_id);
978 	bufp += 4;
979 	if (session->cookie_len) {
980 		memcpy(bufp, &session->cookie[0], session->cookie_len);
981 		bufp += session->cookie_len;
982 	}
983 	if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
984 		u32 l2h = 0;
985 
986 		if (session->send_seq) {
987 			l2h = 0x40000000 | session->ns;
988 			session->ns++;
989 			session->ns &= 0xffffff;
990 			l2tp_dbg(session, L2TP_MSG_SEQ,
991 				 "%s: updated ns to %u\n",
992 				 session->name, session->ns);
993 		}
994 
995 		*((__be32 *)bufp) = htonl(l2h);
996 		bufp += 4;
997 	}
998 
999 	return bufp - optr;
1000 }
1001 
1002 static void l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1003 			   struct flowi *fl, size_t data_len)
1004 {
1005 	struct l2tp_tunnel *tunnel = session->tunnel;
1006 	unsigned int len = skb->len;
1007 	int error;
1008 
1009 	/* Debug */
1010 	if (session->send_seq)
1011 		l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes, ns=%u\n",
1012 			 session->name, data_len, session->ns - 1);
1013 	else
1014 		l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes\n",
1015 			 session->name, data_len);
1016 
1017 	if (session->debug & L2TP_MSG_DATA) {
1018 		int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1019 		unsigned char *datap = skb->data + uhlen;
1020 
1021 		pr_debug("%s: xmit\n", session->name);
1022 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1023 				     datap, min_t(size_t, 32, len - uhlen));
1024 	}
1025 
1026 	/* Queue the packet to IP for output */
1027 	skb->ignore_df = 1;
1028 #if IS_ENABLED(CONFIG_IPV6)
1029 	if (l2tp_sk_is_v6(tunnel->sock))
1030 		error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1031 	else
1032 #endif
1033 		error = ip_queue_xmit(tunnel->sock, skb, fl);
1034 
1035 	/* Update stats */
1036 	if (error >= 0) {
1037 		atomic_long_inc(&tunnel->stats.tx_packets);
1038 		atomic_long_add(len, &tunnel->stats.tx_bytes);
1039 		atomic_long_inc(&session->stats.tx_packets);
1040 		atomic_long_add(len, &session->stats.tx_bytes);
1041 	} else {
1042 		atomic_long_inc(&tunnel->stats.tx_errors);
1043 		atomic_long_inc(&session->stats.tx_errors);
1044 	}
1045 }
1046 
1047 /* If caller requires the skb to have a ppp header, the header must be
1048  * inserted in the skb data before calling this function.
1049  */
1050 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1051 {
1052 	int data_len = skb->len;
1053 	struct l2tp_tunnel *tunnel = session->tunnel;
1054 	struct sock *sk = tunnel->sock;
1055 	struct flowi *fl;
1056 	struct udphdr *uh;
1057 	struct inet_sock *inet;
1058 	int headroom;
1059 	int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1060 	int udp_len;
1061 	int ret = NET_XMIT_SUCCESS;
1062 
1063 	/* Check that there's enough headroom in the skb to insert IP,
1064 	 * UDP and L2TP headers. If not enough, expand it to
1065 	 * make room. Adjust truesize.
1066 	 */
1067 	headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1068 		uhlen + hdr_len;
1069 	if (skb_cow_head(skb, headroom)) {
1070 		kfree_skb(skb);
1071 		return NET_XMIT_DROP;
1072 	}
1073 
1074 	/* Setup L2TP header */
1075 	session->build_header(session, __skb_push(skb, hdr_len));
1076 
1077 	/* Reset skb netfilter state */
1078 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1079 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1080 			      IPSKB_REROUTED);
1081 	nf_reset_ct(skb);
1082 
1083 	bh_lock_sock(sk);
1084 	if (sock_owned_by_user(sk)) {
1085 		kfree_skb(skb);
1086 		ret = NET_XMIT_DROP;
1087 		goto out_unlock;
1088 	}
1089 
1090 	/* The user-space may change the connection status for the user-space
1091 	 * provided socket at run time: we must check it under the socket lock
1092 	 */
1093 	if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1094 		kfree_skb(skb);
1095 		ret = NET_XMIT_DROP;
1096 		goto out_unlock;
1097 	}
1098 
1099 	/* Get routing info from the tunnel socket */
1100 	skb_dst_drop(skb);
1101 	skb_dst_set(skb, sk_dst_check(sk, 0));
1102 
1103 	inet = inet_sk(sk);
1104 	fl = &inet->cork.fl;
1105 	switch (tunnel->encap) {
1106 	case L2TP_ENCAPTYPE_UDP:
1107 		/* Setup UDP header */
1108 		__skb_push(skb, sizeof(*uh));
1109 		skb_reset_transport_header(skb);
1110 		uh = udp_hdr(skb);
1111 		uh->source = inet->inet_sport;
1112 		uh->dest = inet->inet_dport;
1113 		udp_len = uhlen + hdr_len + data_len;
1114 		uh->len = htons(udp_len);
1115 
1116 		/* Calculate UDP checksum if configured to do so */
1117 #if IS_ENABLED(CONFIG_IPV6)
1118 		if (l2tp_sk_is_v6(sk))
1119 			udp6_set_csum(udp_get_no_check6_tx(sk),
1120 				      skb, &inet6_sk(sk)->saddr,
1121 				      &sk->sk_v6_daddr, udp_len);
1122 		else
1123 #endif
1124 		udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1125 			     inet->inet_daddr, udp_len);
1126 		break;
1127 
1128 	case L2TP_ENCAPTYPE_IP:
1129 		break;
1130 	}
1131 
1132 	l2tp_xmit_core(session, skb, fl, data_len);
1133 out_unlock:
1134 	bh_unlock_sock(sk);
1135 
1136 	return ret;
1137 }
1138 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1139 
1140 /*****************************************************************************
1141  * Tinnel and session create/destroy.
1142  *****************************************************************************/
1143 
1144 /* Tunnel socket destruct hook.
1145  * The tunnel context is deleted only when all session sockets have been
1146  * closed.
1147  */
1148 static void l2tp_tunnel_destruct(struct sock *sk)
1149 {
1150 	struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1151 
1152 	if (tunnel == NULL)
1153 		goto end;
1154 
1155 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1156 
1157 	/* Disable udp encapsulation */
1158 	switch (tunnel->encap) {
1159 	case L2TP_ENCAPTYPE_UDP:
1160 		/* No longer an encapsulation socket. See net/ipv4/udp.c */
1161 		(udp_sk(sk))->encap_type = 0;
1162 		(udp_sk(sk))->encap_rcv = NULL;
1163 		(udp_sk(sk))->encap_destroy = NULL;
1164 		break;
1165 	case L2TP_ENCAPTYPE_IP:
1166 		break;
1167 	}
1168 
1169 	/* Remove hooks into tunnel socket */
1170 	sk->sk_destruct = tunnel->old_sk_destruct;
1171 	sk->sk_user_data = NULL;
1172 
1173 	/* Call the original destructor */
1174 	if (sk->sk_destruct)
1175 		(*sk->sk_destruct)(sk);
1176 
1177 	kfree_rcu(tunnel, rcu);
1178 end:
1179 	return;
1180 }
1181 
1182 /* When the tunnel is closed, all the attached sessions need to go too.
1183  */
1184 static void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1185 {
1186 	int hash;
1187 	struct hlist_node *walk;
1188 	struct hlist_node *tmp;
1189 	struct l2tp_session *session;
1190 
1191 	BUG_ON(tunnel == NULL);
1192 
1193 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1194 		  tunnel->name);
1195 
1196 	write_lock_bh(&tunnel->hlist_lock);
1197 	tunnel->acpt_newsess = false;
1198 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1199 again:
1200 		hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1201 			session = hlist_entry(walk, struct l2tp_session, hlist);
1202 
1203 			l2tp_info(session, L2TP_MSG_CONTROL,
1204 				  "%s: closing session\n", session->name);
1205 
1206 			hlist_del_init(&session->hlist);
1207 
1208 			if (test_and_set_bit(0, &session->dead))
1209 				goto again;
1210 
1211 			write_unlock_bh(&tunnel->hlist_lock);
1212 
1213 			__l2tp_session_unhash(session);
1214 			l2tp_session_queue_purge(session);
1215 
1216 			if (session->session_close != NULL)
1217 				(*session->session_close)(session);
1218 
1219 			l2tp_session_dec_refcount(session);
1220 
1221 			write_lock_bh(&tunnel->hlist_lock);
1222 
1223 			/* Now restart from the beginning of this hash
1224 			 * chain.  We always remove a session from the
1225 			 * list so we are guaranteed to make forward
1226 			 * progress.
1227 			 */
1228 			goto again;
1229 		}
1230 	}
1231 	write_unlock_bh(&tunnel->hlist_lock);
1232 }
1233 
1234 /* Tunnel socket destroy hook for UDP encapsulation */
1235 static void l2tp_udp_encap_destroy(struct sock *sk)
1236 {
1237 	struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1238 
1239 	if (tunnel)
1240 		l2tp_tunnel_delete(tunnel);
1241 }
1242 
1243 /* Workqueue tunnel deletion function */
1244 static void l2tp_tunnel_del_work(struct work_struct *work)
1245 {
1246 	struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1247 						  del_work);
1248 	struct sock *sk = tunnel->sock;
1249 	struct socket *sock = sk->sk_socket;
1250 	struct l2tp_net *pn;
1251 
1252 	l2tp_tunnel_closeall(tunnel);
1253 
1254 	/* If the tunnel socket was created within the kernel, use
1255 	 * the sk API to release it here.
1256 	 */
1257 	if (tunnel->fd < 0) {
1258 		if (sock) {
1259 			kernel_sock_shutdown(sock, SHUT_RDWR);
1260 			sock_release(sock);
1261 		}
1262 	}
1263 
1264 	/* Remove the tunnel struct from the tunnel list */
1265 	pn = l2tp_pernet(tunnel->l2tp_net);
1266 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1267 	list_del_rcu(&tunnel->list);
1268 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1269 
1270 	/* drop initial ref */
1271 	l2tp_tunnel_dec_refcount(tunnel);
1272 
1273 	/* drop workqueue ref */
1274 	l2tp_tunnel_dec_refcount(tunnel);
1275 }
1276 
1277 /* Create a socket for the tunnel, if one isn't set up by
1278  * userspace. This is used for static tunnels where there is no
1279  * managing L2TP daemon.
1280  *
1281  * Since we don't want these sockets to keep a namespace alive by
1282  * themselves, we drop the socket's namespace refcount after creation.
1283  * These sockets are freed when the namespace exits using the pernet
1284  * exit hook.
1285  */
1286 static int l2tp_tunnel_sock_create(struct net *net,
1287 				u32 tunnel_id,
1288 				u32 peer_tunnel_id,
1289 				struct l2tp_tunnel_cfg *cfg,
1290 				struct socket **sockp)
1291 {
1292 	int err = -EINVAL;
1293 	struct socket *sock = NULL;
1294 	struct udp_port_cfg udp_conf;
1295 
1296 	switch (cfg->encap) {
1297 	case L2TP_ENCAPTYPE_UDP:
1298 		memset(&udp_conf, 0, sizeof(udp_conf));
1299 
1300 #if IS_ENABLED(CONFIG_IPV6)
1301 		if (cfg->local_ip6 && cfg->peer_ip6) {
1302 			udp_conf.family = AF_INET6;
1303 			memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1304 			       sizeof(udp_conf.local_ip6));
1305 			memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1306 			       sizeof(udp_conf.peer_ip6));
1307 			udp_conf.use_udp6_tx_checksums =
1308 			  ! cfg->udp6_zero_tx_checksums;
1309 			udp_conf.use_udp6_rx_checksums =
1310 			  ! cfg->udp6_zero_rx_checksums;
1311 		} else
1312 #endif
1313 		{
1314 			udp_conf.family = AF_INET;
1315 			udp_conf.local_ip = cfg->local_ip;
1316 			udp_conf.peer_ip = cfg->peer_ip;
1317 			udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1318 		}
1319 
1320 		udp_conf.local_udp_port = htons(cfg->local_udp_port);
1321 		udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1322 
1323 		err = udp_sock_create(net, &udp_conf, &sock);
1324 		if (err < 0)
1325 			goto out;
1326 
1327 		break;
1328 
1329 	case L2TP_ENCAPTYPE_IP:
1330 #if IS_ENABLED(CONFIG_IPV6)
1331 		if (cfg->local_ip6 && cfg->peer_ip6) {
1332 			struct sockaddr_l2tpip6 ip6_addr = {0};
1333 
1334 			err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1335 					  IPPROTO_L2TP, &sock);
1336 			if (err < 0)
1337 				goto out;
1338 
1339 			ip6_addr.l2tp_family = AF_INET6;
1340 			memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1341 			       sizeof(ip6_addr.l2tp_addr));
1342 			ip6_addr.l2tp_conn_id = tunnel_id;
1343 			err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1344 					  sizeof(ip6_addr));
1345 			if (err < 0)
1346 				goto out;
1347 
1348 			ip6_addr.l2tp_family = AF_INET6;
1349 			memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1350 			       sizeof(ip6_addr.l2tp_addr));
1351 			ip6_addr.l2tp_conn_id = peer_tunnel_id;
1352 			err = kernel_connect(sock,
1353 					     (struct sockaddr *) &ip6_addr,
1354 					     sizeof(ip6_addr), 0);
1355 			if (err < 0)
1356 				goto out;
1357 		} else
1358 #endif
1359 		{
1360 			struct sockaddr_l2tpip ip_addr = {0};
1361 
1362 			err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1363 					  IPPROTO_L2TP, &sock);
1364 			if (err < 0)
1365 				goto out;
1366 
1367 			ip_addr.l2tp_family = AF_INET;
1368 			ip_addr.l2tp_addr = cfg->local_ip;
1369 			ip_addr.l2tp_conn_id = tunnel_id;
1370 			err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1371 					  sizeof(ip_addr));
1372 			if (err < 0)
1373 				goto out;
1374 
1375 			ip_addr.l2tp_family = AF_INET;
1376 			ip_addr.l2tp_addr = cfg->peer_ip;
1377 			ip_addr.l2tp_conn_id = peer_tunnel_id;
1378 			err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1379 					     sizeof(ip_addr), 0);
1380 			if (err < 0)
1381 				goto out;
1382 		}
1383 		break;
1384 
1385 	default:
1386 		goto out;
1387 	}
1388 
1389 out:
1390 	*sockp = sock;
1391 	if ((err < 0) && sock) {
1392 		kernel_sock_shutdown(sock, SHUT_RDWR);
1393 		sock_release(sock);
1394 		*sockp = NULL;
1395 	}
1396 
1397 	return err;
1398 }
1399 
1400 static struct lock_class_key l2tp_socket_class;
1401 
1402 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)
1403 {
1404 	struct l2tp_tunnel *tunnel = NULL;
1405 	int err;
1406 	enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1407 
1408 	if (cfg != NULL)
1409 		encap = cfg->encap;
1410 
1411 	tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1412 	if (tunnel == NULL) {
1413 		err = -ENOMEM;
1414 		goto err;
1415 	}
1416 
1417 	tunnel->version = version;
1418 	tunnel->tunnel_id = tunnel_id;
1419 	tunnel->peer_tunnel_id = peer_tunnel_id;
1420 	tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1421 
1422 	tunnel->magic = L2TP_TUNNEL_MAGIC;
1423 	sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1424 	rwlock_init(&tunnel->hlist_lock);
1425 	tunnel->acpt_newsess = true;
1426 
1427 	if (cfg != NULL)
1428 		tunnel->debug = cfg->debug;
1429 
1430 	tunnel->encap = encap;
1431 
1432 	refcount_set(&tunnel->ref_count, 1);
1433 	tunnel->fd = fd;
1434 
1435 	/* Init delete workqueue struct */
1436 	INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1437 
1438 	INIT_LIST_HEAD(&tunnel->list);
1439 
1440 	err = 0;
1441 err:
1442 	if (tunnelp)
1443 		*tunnelp = tunnel;
1444 
1445 	return err;
1446 }
1447 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1448 
1449 static int l2tp_validate_socket(const struct sock *sk, const struct net *net,
1450 				enum l2tp_encap_type encap)
1451 {
1452 	if (!net_eq(sock_net(sk), net))
1453 		return -EINVAL;
1454 
1455 	if (sk->sk_type != SOCK_DGRAM)
1456 		return -EPROTONOSUPPORT;
1457 
1458 	if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) ||
1459 	    (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP))
1460 		return -EPROTONOSUPPORT;
1461 
1462 	if (sk->sk_user_data)
1463 		return -EBUSY;
1464 
1465 	return 0;
1466 }
1467 
1468 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net,
1469 			 struct l2tp_tunnel_cfg *cfg)
1470 {
1471 	struct l2tp_tunnel *tunnel_walk;
1472 	struct l2tp_net *pn;
1473 	struct socket *sock;
1474 	struct sock *sk;
1475 	int ret;
1476 
1477 	if (tunnel->fd < 0) {
1478 		ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id,
1479 					      tunnel->peer_tunnel_id, cfg,
1480 					      &sock);
1481 		if (ret < 0)
1482 			goto err;
1483 	} else {
1484 		sock = sockfd_lookup(tunnel->fd, &ret);
1485 		if (!sock)
1486 			goto err;
1487 
1488 		ret = l2tp_validate_socket(sock->sk, net, tunnel->encap);
1489 		if (ret < 0)
1490 			goto err_sock;
1491 	}
1492 
1493 	tunnel->l2tp_net = net;
1494 	pn = l2tp_pernet(net);
1495 
1496 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1497 	list_for_each_entry(tunnel_walk, &pn->l2tp_tunnel_list, list) {
1498 		if (tunnel_walk->tunnel_id == tunnel->tunnel_id) {
1499 			spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1500 
1501 			ret = -EEXIST;
1502 			goto err_sock;
1503 		}
1504 	}
1505 	list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1506 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1507 
1508 	sk = sock->sk;
1509 	sock_hold(sk);
1510 	tunnel->sock = sk;
1511 
1512 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1513 		struct udp_tunnel_sock_cfg udp_cfg = {
1514 			.sk_user_data = tunnel,
1515 			.encap_type = UDP_ENCAP_L2TPINUDP,
1516 			.encap_rcv = l2tp_udp_encap_recv,
1517 			.encap_destroy = l2tp_udp_encap_destroy,
1518 		};
1519 
1520 		setup_udp_tunnel_sock(net, sock, &udp_cfg);
1521 	} else {
1522 		sk->sk_user_data = tunnel;
1523 	}
1524 
1525 	tunnel->old_sk_destruct = sk->sk_destruct;
1526 	sk->sk_destruct = &l2tp_tunnel_destruct;
1527 	lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class,
1528 				   "l2tp_sock");
1529 	sk->sk_allocation = GFP_ATOMIC;
1530 
1531 	if (tunnel->fd >= 0)
1532 		sockfd_put(sock);
1533 
1534 	return 0;
1535 
1536 err_sock:
1537 	if (tunnel->fd < 0)
1538 		sock_release(sock);
1539 	else
1540 		sockfd_put(sock);
1541 err:
1542 	return ret;
1543 }
1544 EXPORT_SYMBOL_GPL(l2tp_tunnel_register);
1545 
1546 /* This function is used by the netlink TUNNEL_DELETE command.
1547  */
1548 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1549 {
1550 	if (!test_and_set_bit(0, &tunnel->dead)) {
1551 		l2tp_tunnel_inc_refcount(tunnel);
1552 		queue_work(l2tp_wq, &tunnel->del_work);
1553 	}
1554 }
1555 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1556 
1557 /* Really kill the session.
1558  */
1559 void l2tp_session_free(struct l2tp_session *session)
1560 {
1561 	struct l2tp_tunnel *tunnel = session->tunnel;
1562 
1563 	BUG_ON(refcount_read(&session->ref_count) != 0);
1564 
1565 	if (tunnel) {
1566 		BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1567 		l2tp_tunnel_dec_refcount(tunnel);
1568 	}
1569 
1570 	kfree(session);
1571 }
1572 EXPORT_SYMBOL_GPL(l2tp_session_free);
1573 
1574 /* Remove an l2tp session from l2tp_core's hash lists.
1575  * Provides a tidyup interface for pseudowire code which can't just route all
1576  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1577  * callback.
1578  */
1579 void __l2tp_session_unhash(struct l2tp_session *session)
1580 {
1581 	struct l2tp_tunnel *tunnel = session->tunnel;
1582 
1583 	/* Remove the session from core hashes */
1584 	if (tunnel) {
1585 		/* Remove from the per-tunnel hash */
1586 		write_lock_bh(&tunnel->hlist_lock);
1587 		hlist_del_init(&session->hlist);
1588 		write_unlock_bh(&tunnel->hlist_lock);
1589 
1590 		/* For L2TPv3 we have a per-net hash: remove from there, too */
1591 		if (tunnel->version != L2TP_HDR_VER_2) {
1592 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1593 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1594 			hlist_del_init_rcu(&session->global_hlist);
1595 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1596 			synchronize_rcu();
1597 		}
1598 	}
1599 }
1600 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1601 
1602 /* This function is used by the netlink SESSION_DELETE command and by
1603    pseudowire modules.
1604  */
1605 int l2tp_session_delete(struct l2tp_session *session)
1606 {
1607 	if (test_and_set_bit(0, &session->dead))
1608 		return 0;
1609 
1610 	__l2tp_session_unhash(session);
1611 	l2tp_session_queue_purge(session);
1612 	if (session->session_close != NULL)
1613 		(*session->session_close)(session);
1614 
1615 	l2tp_session_dec_refcount(session);
1616 
1617 	return 0;
1618 }
1619 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1620 
1621 /* We come here whenever a session's send_seq, cookie_len or
1622  * l2specific_type parameters are set.
1623  */
1624 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1625 {
1626 	if (version == L2TP_HDR_VER_2) {
1627 		session->hdr_len = 6;
1628 		if (session->send_seq)
1629 			session->hdr_len += 4;
1630 	} else {
1631 		session->hdr_len = 4 + session->cookie_len;
1632 		session->hdr_len += l2tp_get_l2specific_len(session);
1633 		if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1634 			session->hdr_len += 4;
1635 	}
1636 
1637 }
1638 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1639 
1640 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)
1641 {
1642 	struct l2tp_session *session;
1643 
1644 	session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1645 	if (session != NULL) {
1646 		session->magic = L2TP_SESSION_MAGIC;
1647 		session->tunnel = tunnel;
1648 
1649 		session->session_id = session_id;
1650 		session->peer_session_id = peer_session_id;
1651 		session->nr = 0;
1652 		if (tunnel->version == L2TP_HDR_VER_2)
1653 			session->nr_max = 0xffff;
1654 		else
1655 			session->nr_max = 0xffffff;
1656 		session->nr_window_size = session->nr_max / 2;
1657 		session->nr_oos_count_max = 4;
1658 
1659 		/* Use NR of first received packet */
1660 		session->reorder_skip = 1;
1661 
1662 		sprintf(&session->name[0], "sess %u/%u",
1663 			tunnel->tunnel_id, session->session_id);
1664 
1665 		skb_queue_head_init(&session->reorder_q);
1666 
1667 		INIT_HLIST_NODE(&session->hlist);
1668 		INIT_HLIST_NODE(&session->global_hlist);
1669 
1670 		/* Inherit debug options from tunnel */
1671 		session->debug = tunnel->debug;
1672 
1673 		if (cfg) {
1674 			session->pwtype = cfg->pw_type;
1675 			session->debug = cfg->debug;
1676 			session->send_seq = cfg->send_seq;
1677 			session->recv_seq = cfg->recv_seq;
1678 			session->lns_mode = cfg->lns_mode;
1679 			session->reorder_timeout = cfg->reorder_timeout;
1680 			session->l2specific_type = cfg->l2specific_type;
1681 			session->cookie_len = cfg->cookie_len;
1682 			memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1683 			session->peer_cookie_len = cfg->peer_cookie_len;
1684 			memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1685 		}
1686 
1687 		if (tunnel->version == L2TP_HDR_VER_2)
1688 			session->build_header = l2tp_build_l2tpv2_header;
1689 		else
1690 			session->build_header = l2tp_build_l2tpv3_header;
1691 
1692 		l2tp_session_set_header_len(session, tunnel->version);
1693 
1694 		refcount_set(&session->ref_count, 1);
1695 
1696 		return session;
1697 	}
1698 
1699 	return ERR_PTR(-ENOMEM);
1700 }
1701 EXPORT_SYMBOL_GPL(l2tp_session_create);
1702 
1703 /*****************************************************************************
1704  * Init and cleanup
1705  *****************************************************************************/
1706 
1707 static __net_init int l2tp_init_net(struct net *net)
1708 {
1709 	struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1710 	int hash;
1711 
1712 	INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1713 	spin_lock_init(&pn->l2tp_tunnel_list_lock);
1714 
1715 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1716 		INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1717 
1718 	spin_lock_init(&pn->l2tp_session_hlist_lock);
1719 
1720 	return 0;
1721 }
1722 
1723 static __net_exit void l2tp_exit_net(struct net *net)
1724 {
1725 	struct l2tp_net *pn = l2tp_pernet(net);
1726 	struct l2tp_tunnel *tunnel = NULL;
1727 	int hash;
1728 
1729 	rcu_read_lock_bh();
1730 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1731 		l2tp_tunnel_delete(tunnel);
1732 	}
1733 	rcu_read_unlock_bh();
1734 
1735 	if (l2tp_wq)
1736 		flush_workqueue(l2tp_wq);
1737 	rcu_barrier();
1738 
1739 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1740 		WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1741 }
1742 
1743 static struct pernet_operations l2tp_net_ops = {
1744 	.init = l2tp_init_net,
1745 	.exit = l2tp_exit_net,
1746 	.id   = &l2tp_net_id,
1747 	.size = sizeof(struct l2tp_net),
1748 };
1749 
1750 static int __init l2tp_init(void)
1751 {
1752 	int rc = 0;
1753 
1754 	rc = register_pernet_device(&l2tp_net_ops);
1755 	if (rc)
1756 		goto out;
1757 
1758 	l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1759 	if (!l2tp_wq) {
1760 		pr_err("alloc_workqueue failed\n");
1761 		unregister_pernet_device(&l2tp_net_ops);
1762 		rc = -ENOMEM;
1763 		goto out;
1764 	}
1765 
1766 	pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1767 
1768 out:
1769 	return rc;
1770 }
1771 
1772 static void __exit l2tp_exit(void)
1773 {
1774 	unregister_pernet_device(&l2tp_net_ops);
1775 	if (l2tp_wq) {
1776 		destroy_workqueue(l2tp_wq);
1777 		l2tp_wq = NULL;
1778 	}
1779 }
1780 
1781 module_init(l2tp_init);
1782 module_exit(l2tp_exit);
1783 
1784 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1785 MODULE_DESCRIPTION("L2TP core");
1786 MODULE_LICENSE("GPL");
1787 MODULE_VERSION(L2TP_DRV_VERSION);
1788