xref: /linux/net/l2tp/l2tp_core.c (revision b1a54551dd9ed5ef1763b97b35a0999ca002b95c)
1 // SPDX-License-Identifier: GPL-2.0-only
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 
17 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
18 
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/list.h>
22 #include <linux/rculist.h>
23 #include <linux/uaccess.h>
24 
25 #include <linux/kernel.h>
26 #include <linux/spinlock.h>
27 #include <linux/kthread.h>
28 #include <linux/sched.h>
29 #include <linux/slab.h>
30 #include <linux/errno.h>
31 #include <linux/jiffies.h>
32 
33 #include <linux/netdevice.h>
34 #include <linux/net.h>
35 #include <linux/inetdevice.h>
36 #include <linux/skbuff.h>
37 #include <linux/init.h>
38 #include <linux/in.h>
39 #include <linux/ip.h>
40 #include <linux/udp.h>
41 #include <linux/l2tp.h>
42 #include <linux/hash.h>
43 #include <linux/sort.h>
44 #include <linux/file.h>
45 #include <linux/nsproxy.h>
46 #include <net/net_namespace.h>
47 #include <net/netns/generic.h>
48 #include <net/dst.h>
49 #include <net/ip.h>
50 #include <net/udp.h>
51 #include <net/udp_tunnel.h>
52 #include <net/inet_common.h>
53 #include <net/xfrm.h>
54 #include <net/protocol.h>
55 #include <net/inet6_connection_sock.h>
56 #include <net/inet_ecn.h>
57 #include <net/ip6_route.h>
58 #include <net/ip6_checksum.h>
59 
60 #include <asm/byteorder.h>
61 #include <linux/atomic.h>
62 
63 #include "l2tp_core.h"
64 #include "trace.h"
65 
66 #define CREATE_TRACE_POINTS
67 #include "trace.h"
68 
69 #define L2TP_DRV_VERSION	"V2.0"
70 
71 /* L2TP header constants */
72 #define L2TP_HDRFLAG_T	   0x8000
73 #define L2TP_HDRFLAG_L	   0x4000
74 #define L2TP_HDRFLAG_S	   0x0800
75 #define L2TP_HDRFLAG_O	   0x0200
76 #define L2TP_HDRFLAG_P	   0x0100
77 
78 #define L2TP_HDR_VER_MASK  0x000F
79 #define L2TP_HDR_VER_2	   0x0002
80 #define L2TP_HDR_VER_3	   0x0003
81 
82 /* L2TPv3 default L2-specific sublayer */
83 #define L2TP_SLFLAG_S	   0x40000000
84 #define L2TP_SL_SEQ_MASK   0x00ffffff
85 
86 #define L2TP_HDR_SIZE_MAX		14
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 struct workqueue_struct *l2tp_wq;
103 
104 /* per-net private data for this module */
105 static unsigned int l2tp_net_id;
106 struct l2tp_net {
107 	/* Lock for write access to l2tp_tunnel_idr */
108 	spinlock_t l2tp_tunnel_idr_lock;
109 	struct idr l2tp_tunnel_idr;
110 	struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
111 	/* Lock for write access to l2tp_session_hlist */
112 	spinlock_t l2tp_session_hlist_lock;
113 };
114 
115 #if IS_ENABLED(CONFIG_IPV6)
116 static bool l2tp_sk_is_v6(struct sock *sk)
117 {
118 	return sk->sk_family == PF_INET6 &&
119 	       !ipv6_addr_v4mapped(&sk->sk_v6_daddr);
120 }
121 #endif
122 
123 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
124 {
125 	return net_generic(net, l2tp_net_id);
126 }
127 
128 /* Session hash global list for L2TPv3.
129  * The session_id SHOULD be random according to RFC3931, but several
130  * L2TP implementations use incrementing session_ids.  So we do a real
131  * hash on the session_id, rather than a simple bitmask.
132  */
133 static inline struct hlist_head *
134 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
135 {
136 	return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
137 }
138 
139 /* Session hash list.
140  * The session_id SHOULD be random according to RFC2661, but several
141  * L2TP implementations (Cisco and Microsoft) use incrementing
142  * session_ids.  So we do a real hash on the session_id, rather than a
143  * simple bitmask.
144  */
145 static inline struct hlist_head *
146 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
147 {
148 	return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
149 }
150 
151 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
152 {
153 	trace_free_tunnel(tunnel);
154 	sock_put(tunnel->sock);
155 	/* the tunnel is freed in the socket destructor */
156 }
157 
158 static void l2tp_session_free(struct l2tp_session *session)
159 {
160 	trace_free_session(session);
161 	if (session->tunnel)
162 		l2tp_tunnel_dec_refcount(session->tunnel);
163 	kfree(session);
164 }
165 
166 struct l2tp_tunnel *l2tp_sk_to_tunnel(struct sock *sk)
167 {
168 	struct l2tp_tunnel *tunnel = sk->sk_user_data;
169 
170 	if (tunnel)
171 		if (WARN_ON(tunnel->magic != L2TP_TUNNEL_MAGIC))
172 			return NULL;
173 
174 	return tunnel;
175 }
176 EXPORT_SYMBOL_GPL(l2tp_sk_to_tunnel);
177 
178 void l2tp_tunnel_inc_refcount(struct l2tp_tunnel *tunnel)
179 {
180 	refcount_inc(&tunnel->ref_count);
181 }
182 EXPORT_SYMBOL_GPL(l2tp_tunnel_inc_refcount);
183 
184 void l2tp_tunnel_dec_refcount(struct l2tp_tunnel *tunnel)
185 {
186 	if (refcount_dec_and_test(&tunnel->ref_count))
187 		l2tp_tunnel_free(tunnel);
188 }
189 EXPORT_SYMBOL_GPL(l2tp_tunnel_dec_refcount);
190 
191 void l2tp_session_inc_refcount(struct l2tp_session *session)
192 {
193 	refcount_inc(&session->ref_count);
194 }
195 EXPORT_SYMBOL_GPL(l2tp_session_inc_refcount);
196 
197 void l2tp_session_dec_refcount(struct l2tp_session *session)
198 {
199 	if (refcount_dec_and_test(&session->ref_count))
200 		l2tp_session_free(session);
201 }
202 EXPORT_SYMBOL_GPL(l2tp_session_dec_refcount);
203 
204 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
205 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
206 {
207 	const struct l2tp_net *pn = l2tp_pernet(net);
208 	struct l2tp_tunnel *tunnel;
209 
210 	rcu_read_lock_bh();
211 	tunnel = idr_find(&pn->l2tp_tunnel_idr, tunnel_id);
212 	if (tunnel && refcount_inc_not_zero(&tunnel->ref_count)) {
213 		rcu_read_unlock_bh();
214 		return tunnel;
215 	}
216 	rcu_read_unlock_bh();
217 
218 	return NULL;
219 }
220 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
221 
222 struct l2tp_tunnel *l2tp_tunnel_get_nth(const struct net *net, int nth)
223 {
224 	struct l2tp_net *pn = l2tp_pernet(net);
225 	unsigned long tunnel_id, tmp;
226 	struct l2tp_tunnel *tunnel;
227 	int count = 0;
228 
229 	rcu_read_lock_bh();
230 	idr_for_each_entry_ul(&pn->l2tp_tunnel_idr, tunnel, tmp, tunnel_id) {
231 		if (tunnel && ++count > nth &&
232 		    refcount_inc_not_zero(&tunnel->ref_count)) {
233 			rcu_read_unlock_bh();
234 			return tunnel;
235 		}
236 	}
237 	rcu_read_unlock_bh();
238 
239 	return NULL;
240 }
241 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_nth);
242 
243 struct l2tp_session *l2tp_tunnel_get_session(struct l2tp_tunnel *tunnel,
244 					     u32 session_id)
245 {
246 	struct hlist_head *session_list;
247 	struct l2tp_session *session;
248 
249 	session_list = l2tp_session_id_hash(tunnel, session_id);
250 
251 	rcu_read_lock_bh();
252 	hlist_for_each_entry_rcu(session, session_list, hlist)
253 		if (session->session_id == session_id) {
254 			l2tp_session_inc_refcount(session);
255 			rcu_read_unlock_bh();
256 
257 			return session;
258 		}
259 	rcu_read_unlock_bh();
260 
261 	return NULL;
262 }
263 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_session);
264 
265 struct l2tp_session *l2tp_session_get(const struct net *net, u32 session_id)
266 {
267 	struct hlist_head *session_list;
268 	struct l2tp_session *session;
269 
270 	session_list = l2tp_session_id_hash_2(l2tp_pernet(net), session_id);
271 
272 	rcu_read_lock_bh();
273 	hlist_for_each_entry_rcu(session, session_list, global_hlist)
274 		if (session->session_id == session_id) {
275 			l2tp_session_inc_refcount(session);
276 			rcu_read_unlock_bh();
277 
278 			return session;
279 		}
280 	rcu_read_unlock_bh();
281 
282 	return NULL;
283 }
284 EXPORT_SYMBOL_GPL(l2tp_session_get);
285 
286 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth)
287 {
288 	int hash;
289 	struct l2tp_session *session;
290 	int count = 0;
291 
292 	rcu_read_lock_bh();
293 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
294 		hlist_for_each_entry_rcu(session, &tunnel->session_hlist[hash], hlist) {
295 			if (++count > nth) {
296 				l2tp_session_inc_refcount(session);
297 				rcu_read_unlock_bh();
298 				return session;
299 			}
300 		}
301 	}
302 
303 	rcu_read_unlock_bh();
304 
305 	return NULL;
306 }
307 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
308 
309 /* Lookup a session by interface name.
310  * This is very inefficient but is only used by management interfaces.
311  */
312 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
313 						const char *ifname)
314 {
315 	struct l2tp_net *pn = l2tp_pernet(net);
316 	int hash;
317 	struct l2tp_session *session;
318 
319 	rcu_read_lock_bh();
320 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
321 		hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
322 			if (!strcmp(session->ifname, ifname)) {
323 				l2tp_session_inc_refcount(session);
324 				rcu_read_unlock_bh();
325 
326 				return session;
327 			}
328 		}
329 	}
330 
331 	rcu_read_unlock_bh();
332 
333 	return NULL;
334 }
335 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
336 
337 int l2tp_session_register(struct l2tp_session *session,
338 			  struct l2tp_tunnel *tunnel)
339 {
340 	struct l2tp_session *session_walk;
341 	struct hlist_head *g_head;
342 	struct hlist_head *head;
343 	struct l2tp_net *pn;
344 	int err;
345 
346 	head = l2tp_session_id_hash(tunnel, session->session_id);
347 
348 	spin_lock_bh(&tunnel->hlist_lock);
349 	if (!tunnel->acpt_newsess) {
350 		err = -ENODEV;
351 		goto err_tlock;
352 	}
353 
354 	hlist_for_each_entry(session_walk, head, hlist)
355 		if (session_walk->session_id == session->session_id) {
356 			err = -EEXIST;
357 			goto err_tlock;
358 		}
359 
360 	if (tunnel->version == L2TP_HDR_VER_3) {
361 		pn = l2tp_pernet(tunnel->l2tp_net);
362 		g_head = l2tp_session_id_hash_2(pn, session->session_id);
363 
364 		spin_lock_bh(&pn->l2tp_session_hlist_lock);
365 
366 		/* IP encap expects session IDs to be globally unique, while
367 		 * UDP encap doesn't.
368 		 */
369 		hlist_for_each_entry(session_walk, g_head, global_hlist)
370 			if (session_walk->session_id == session->session_id &&
371 			    (session_walk->tunnel->encap == L2TP_ENCAPTYPE_IP ||
372 			     tunnel->encap == L2TP_ENCAPTYPE_IP)) {
373 				err = -EEXIST;
374 				goto err_tlock_pnlock;
375 			}
376 
377 		l2tp_tunnel_inc_refcount(tunnel);
378 		hlist_add_head_rcu(&session->global_hlist, g_head);
379 
380 		spin_unlock_bh(&pn->l2tp_session_hlist_lock);
381 	} else {
382 		l2tp_tunnel_inc_refcount(tunnel);
383 	}
384 
385 	hlist_add_head_rcu(&session->hlist, head);
386 	spin_unlock_bh(&tunnel->hlist_lock);
387 
388 	trace_register_session(session);
389 
390 	return 0;
391 
392 err_tlock_pnlock:
393 	spin_unlock_bh(&pn->l2tp_session_hlist_lock);
394 err_tlock:
395 	spin_unlock_bh(&tunnel->hlist_lock);
396 
397 	return err;
398 }
399 EXPORT_SYMBOL_GPL(l2tp_session_register);
400 
401 /*****************************************************************************
402  * Receive data handling
403  *****************************************************************************/
404 
405 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
406  * number.
407  */
408 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
409 {
410 	struct sk_buff *skbp;
411 	struct sk_buff *tmp;
412 	u32 ns = L2TP_SKB_CB(skb)->ns;
413 
414 	spin_lock_bh(&session->reorder_q.lock);
415 	skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
416 		if (L2TP_SKB_CB(skbp)->ns > ns) {
417 			__skb_queue_before(&session->reorder_q, skbp, skb);
418 			atomic_long_inc(&session->stats.rx_oos_packets);
419 			goto out;
420 		}
421 	}
422 
423 	__skb_queue_tail(&session->reorder_q, skb);
424 
425 out:
426 	spin_unlock_bh(&session->reorder_q.lock);
427 }
428 
429 /* Dequeue a single skb.
430  */
431 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
432 {
433 	struct l2tp_tunnel *tunnel = session->tunnel;
434 	int length = L2TP_SKB_CB(skb)->length;
435 
436 	/* We're about to requeue the skb, so return resources
437 	 * to its current owner (a socket receive buffer).
438 	 */
439 	skb_orphan(skb);
440 
441 	atomic_long_inc(&tunnel->stats.rx_packets);
442 	atomic_long_add(length, &tunnel->stats.rx_bytes);
443 	atomic_long_inc(&session->stats.rx_packets);
444 	atomic_long_add(length, &session->stats.rx_bytes);
445 
446 	if (L2TP_SKB_CB(skb)->has_seq) {
447 		/* Bump our Nr */
448 		session->nr++;
449 		session->nr &= session->nr_max;
450 		trace_session_seqnum_update(session);
451 	}
452 
453 	/* call private receive handler */
454 	if (session->recv_skb)
455 		(*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
456 	else
457 		kfree_skb(skb);
458 }
459 
460 /* Dequeue skbs from the session's reorder_q, subject to packet order.
461  * Skbs that have been in the queue for too long are simply discarded.
462  */
463 static void l2tp_recv_dequeue(struct l2tp_session *session)
464 {
465 	struct sk_buff *skb;
466 	struct sk_buff *tmp;
467 
468 	/* If the pkt at the head of the queue has the nr that we
469 	 * expect to send up next, dequeue it and any other
470 	 * in-sequence packets behind it.
471 	 */
472 start:
473 	spin_lock_bh(&session->reorder_q.lock);
474 	skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
475 		struct l2tp_skb_cb *cb = L2TP_SKB_CB(skb);
476 
477 		/* If the packet has been pending on the queue for too long, discard it */
478 		if (time_after(jiffies, cb->expires)) {
479 			atomic_long_inc(&session->stats.rx_seq_discards);
480 			atomic_long_inc(&session->stats.rx_errors);
481 			trace_session_pkt_expired(session, cb->ns);
482 			session->reorder_skip = 1;
483 			__skb_unlink(skb, &session->reorder_q);
484 			kfree_skb(skb);
485 			continue;
486 		}
487 
488 		if (cb->has_seq) {
489 			if (session->reorder_skip) {
490 				session->reorder_skip = 0;
491 				session->nr = cb->ns;
492 				trace_session_seqnum_reset(session);
493 			}
494 			if (cb->ns != session->nr)
495 				goto out;
496 		}
497 		__skb_unlink(skb, &session->reorder_q);
498 
499 		/* Process the skb. We release the queue lock while we
500 		 * do so to let other contexts process the queue.
501 		 */
502 		spin_unlock_bh(&session->reorder_q.lock);
503 		l2tp_recv_dequeue_skb(session, skb);
504 		goto start;
505 	}
506 
507 out:
508 	spin_unlock_bh(&session->reorder_q.lock);
509 }
510 
511 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
512 {
513 	u32 nws;
514 
515 	if (nr >= session->nr)
516 		nws = nr - session->nr;
517 	else
518 		nws = (session->nr_max + 1) - (session->nr - nr);
519 
520 	return nws < session->nr_window_size;
521 }
522 
523 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
524  * acceptable, else non-zero.
525  */
526 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
527 {
528 	struct l2tp_skb_cb *cb = L2TP_SKB_CB(skb);
529 
530 	if (!l2tp_seq_check_rx_window(session, cb->ns)) {
531 		/* Packet sequence number is outside allowed window.
532 		 * Discard it.
533 		 */
534 		trace_session_pkt_outside_rx_window(session, cb->ns);
535 		goto discard;
536 	}
537 
538 	if (session->reorder_timeout != 0) {
539 		/* Packet reordering enabled. Add skb to session's
540 		 * reorder queue, in order of ns.
541 		 */
542 		l2tp_recv_queue_skb(session, skb);
543 		goto out;
544 	}
545 
546 	/* Packet reordering disabled. Discard out-of-sequence packets, while
547 	 * tracking the number if in-sequence packets after the first OOS packet
548 	 * is seen. After nr_oos_count_max in-sequence packets, reset the
549 	 * sequence number to re-enable packet reception.
550 	 */
551 	if (cb->ns == session->nr) {
552 		skb_queue_tail(&session->reorder_q, skb);
553 	} else {
554 		u32 nr_oos = cb->ns;
555 		u32 nr_next = (session->nr_oos + 1) & session->nr_max;
556 
557 		if (nr_oos == nr_next)
558 			session->nr_oos_count++;
559 		else
560 			session->nr_oos_count = 0;
561 
562 		session->nr_oos = nr_oos;
563 		if (session->nr_oos_count > session->nr_oos_count_max) {
564 			session->reorder_skip = 1;
565 		}
566 		if (!session->reorder_skip) {
567 			atomic_long_inc(&session->stats.rx_seq_discards);
568 			trace_session_pkt_oos(session, cb->ns);
569 			goto discard;
570 		}
571 		skb_queue_tail(&session->reorder_q, skb);
572 	}
573 
574 out:
575 	return 0;
576 
577 discard:
578 	return 1;
579 }
580 
581 /* Do receive processing of L2TP data frames. We handle both L2TPv2
582  * and L2TPv3 data frames here.
583  *
584  * L2TPv2 Data Message Header
585  *
586  *  0                   1                   2                   3
587  *  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
588  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
589  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
590  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
591  * |           Tunnel ID           |           Session ID          |
592  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
593  * |             Ns (opt)          |             Nr (opt)          |
594  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
595  * |      Offset Size (opt)        |    Offset pad... (opt)
596  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
597  *
598  * Data frames are marked by T=0. All other fields are the same as
599  * those in L2TP control frames.
600  *
601  * L2TPv3 Data Message Header
602  *
603  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
604  * |                      L2TP Session Header                      |
605  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
606  * |                      L2-Specific Sublayer                     |
607  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
608  * |                        Tunnel Payload                      ...
609  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
610  *
611  * L2TPv3 Session Header Over IP
612  *
613  *  0                   1                   2                   3
614  *  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
615  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
616  * |                           Session ID                          |
617  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
618  * |               Cookie (optional, maximum 64 bits)...
619  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
620  *                                                                 |
621  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
622  *
623  * L2TPv3 L2-Specific Sublayer Format
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  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
629  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
630  *
631  * Cookie value and sublayer format are negotiated with the peer when
632  * the session is set up. Unlike L2TPv2, we do not need to parse the
633  * packet header to determine if optional fields are present.
634  *
635  * Caller must already have parsed the frame and determined that it is
636  * a data (not control) frame before coming here. Fields up to the
637  * session-id have already been parsed and ptr points to the data
638  * after the session-id.
639  */
640 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
641 		      unsigned char *ptr, unsigned char *optr, u16 hdrflags,
642 		      int length)
643 {
644 	struct l2tp_tunnel *tunnel = session->tunnel;
645 	int offset;
646 
647 	/* Parse and check optional cookie */
648 	if (session->peer_cookie_len > 0) {
649 		if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
650 			pr_debug_ratelimited("%s: cookie mismatch (%u/%u). Discarding.\n",
651 					     tunnel->name, tunnel->tunnel_id,
652 					     session->session_id);
653 			atomic_long_inc(&session->stats.rx_cookie_discards);
654 			goto discard;
655 		}
656 		ptr += session->peer_cookie_len;
657 	}
658 
659 	/* Handle the optional sequence numbers. Sequence numbers are
660 	 * in different places for L2TPv2 and L2TPv3.
661 	 *
662 	 * If we are the LAC, enable/disable sequence numbers under
663 	 * the control of the LNS.  If no sequence numbers present but
664 	 * we were expecting them, discard frame.
665 	 */
666 	L2TP_SKB_CB(skb)->has_seq = 0;
667 	if (tunnel->version == L2TP_HDR_VER_2) {
668 		if (hdrflags & L2TP_HDRFLAG_S) {
669 			/* Store L2TP info in the skb */
670 			L2TP_SKB_CB(skb)->ns = ntohs(*(__be16 *)ptr);
671 			L2TP_SKB_CB(skb)->has_seq = 1;
672 			ptr += 2;
673 			/* Skip past nr in the header */
674 			ptr += 2;
675 
676 		}
677 	} else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
678 		u32 l2h = ntohl(*(__be32 *)ptr);
679 
680 		if (l2h & 0x40000000) {
681 			/* Store L2TP info in the skb */
682 			L2TP_SKB_CB(skb)->ns = l2h & 0x00ffffff;
683 			L2TP_SKB_CB(skb)->has_seq = 1;
684 		}
685 		ptr += 4;
686 	}
687 
688 	if (L2TP_SKB_CB(skb)->has_seq) {
689 		/* Received a packet with sequence numbers. If we're the LAC,
690 		 * check if we sre sending sequence numbers and if not,
691 		 * configure it so.
692 		 */
693 		if (!session->lns_mode && !session->send_seq) {
694 			trace_session_seqnum_lns_enable(session);
695 			session->send_seq = 1;
696 			l2tp_session_set_header_len(session, tunnel->version);
697 		}
698 	} else {
699 		/* No sequence numbers.
700 		 * If user has configured mandatory sequence numbers, discard.
701 		 */
702 		if (session->recv_seq) {
703 			pr_debug_ratelimited("%s: recv data has no seq numbers when required. Discarding.\n",
704 					     session->name);
705 			atomic_long_inc(&session->stats.rx_seq_discards);
706 			goto discard;
707 		}
708 
709 		/* If we're the LAC and we're sending sequence numbers, the
710 		 * LNS has requested that we no longer send sequence numbers.
711 		 * If we're the LNS and we're sending sequence numbers, the
712 		 * LAC is broken. Discard the frame.
713 		 */
714 		if (!session->lns_mode && session->send_seq) {
715 			trace_session_seqnum_lns_disable(session);
716 			session->send_seq = 0;
717 			l2tp_session_set_header_len(session, tunnel->version);
718 		} else if (session->send_seq) {
719 			pr_debug_ratelimited("%s: recv data has no seq numbers when required. Discarding.\n",
720 					     session->name);
721 			atomic_long_inc(&session->stats.rx_seq_discards);
722 			goto discard;
723 		}
724 	}
725 
726 	/* Session data offset is defined only for L2TPv2 and is
727 	 * indicated by an optional 16-bit value in the header.
728 	 */
729 	if (tunnel->version == L2TP_HDR_VER_2) {
730 		/* If offset bit set, skip it. */
731 		if (hdrflags & L2TP_HDRFLAG_O) {
732 			offset = ntohs(*(__be16 *)ptr);
733 			ptr += 2 + offset;
734 		}
735 	}
736 
737 	offset = ptr - optr;
738 	if (!pskb_may_pull(skb, offset))
739 		goto discard;
740 
741 	__skb_pull(skb, offset);
742 
743 	/* Prepare skb for adding to the session's reorder_q.  Hold
744 	 * packets for max reorder_timeout or 1 second if not
745 	 * reordering.
746 	 */
747 	L2TP_SKB_CB(skb)->length = length;
748 	L2TP_SKB_CB(skb)->expires = jiffies +
749 		(session->reorder_timeout ? session->reorder_timeout : HZ);
750 
751 	/* Add packet to the session's receive queue. Reordering is done here, if
752 	 * enabled. Saved L2TP protocol info is stored in skb->sb[].
753 	 */
754 	if (L2TP_SKB_CB(skb)->has_seq) {
755 		if (l2tp_recv_data_seq(session, skb))
756 			goto discard;
757 	} else {
758 		/* No sequence numbers. Add the skb to the tail of the
759 		 * reorder queue. This ensures that it will be
760 		 * delivered after all previous sequenced skbs.
761 		 */
762 		skb_queue_tail(&session->reorder_q, skb);
763 	}
764 
765 	/* Try to dequeue as many skbs from reorder_q as we can. */
766 	l2tp_recv_dequeue(session);
767 
768 	return;
769 
770 discard:
771 	atomic_long_inc(&session->stats.rx_errors);
772 	kfree_skb(skb);
773 }
774 EXPORT_SYMBOL_GPL(l2tp_recv_common);
775 
776 /* Drop skbs from the session's reorder_q
777  */
778 static void l2tp_session_queue_purge(struct l2tp_session *session)
779 {
780 	struct sk_buff *skb = NULL;
781 
782 	while ((skb = skb_dequeue(&session->reorder_q))) {
783 		atomic_long_inc(&session->stats.rx_errors);
784 		kfree_skb(skb);
785 	}
786 }
787 
788 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
789  * here. The skb is not on a list when we get here.
790  * Returns 0 if the packet was a data packet and was successfully passed on.
791  * Returns 1 if the packet was not a good data packet and could not be
792  * forwarded.  All such packets are passed up to userspace to deal with.
793  */
794 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb)
795 {
796 	struct l2tp_session *session = NULL;
797 	unsigned char *ptr, *optr;
798 	u16 hdrflags;
799 	u32 tunnel_id, session_id;
800 	u16 version;
801 	int length;
802 
803 	/* UDP has verified checksum */
804 
805 	/* UDP always verifies the packet length. */
806 	__skb_pull(skb, sizeof(struct udphdr));
807 
808 	/* Short packet? */
809 	if (!pskb_may_pull(skb, L2TP_HDR_SIZE_MAX)) {
810 		pr_debug_ratelimited("%s: recv short packet (len=%d)\n",
811 				     tunnel->name, skb->len);
812 		goto invalid;
813 	}
814 
815 	/* Point to L2TP header */
816 	optr = skb->data;
817 	ptr = skb->data;
818 
819 	/* Get L2TP header flags */
820 	hdrflags = ntohs(*(__be16 *)ptr);
821 
822 	/* Check protocol version */
823 	version = hdrflags & L2TP_HDR_VER_MASK;
824 	if (version != tunnel->version) {
825 		pr_debug_ratelimited("%s: recv protocol version mismatch: got %d expected %d\n",
826 				     tunnel->name, version, tunnel->version);
827 		goto invalid;
828 	}
829 
830 	/* Get length of L2TP packet */
831 	length = skb->len;
832 
833 	/* If type is control packet, it is handled by userspace. */
834 	if (hdrflags & L2TP_HDRFLAG_T)
835 		goto pass;
836 
837 	/* Skip flags */
838 	ptr += 2;
839 
840 	if (tunnel->version == L2TP_HDR_VER_2) {
841 		/* If length is present, skip it */
842 		if (hdrflags & L2TP_HDRFLAG_L)
843 			ptr += 2;
844 
845 		/* Extract tunnel and session ID */
846 		tunnel_id = ntohs(*(__be16 *)ptr);
847 		ptr += 2;
848 		session_id = ntohs(*(__be16 *)ptr);
849 		ptr += 2;
850 	} else {
851 		ptr += 2;	/* skip reserved bits */
852 		tunnel_id = tunnel->tunnel_id;
853 		session_id = ntohl(*(__be32 *)ptr);
854 		ptr += 4;
855 	}
856 
857 	/* Find the session context */
858 	session = l2tp_tunnel_get_session(tunnel, session_id);
859 	if (!session || !session->recv_skb) {
860 		if (session)
861 			l2tp_session_dec_refcount(session);
862 
863 		/* Not found? Pass to userspace to deal with */
864 		pr_debug_ratelimited("%s: no session found (%u/%u). Passing up.\n",
865 				     tunnel->name, tunnel_id, session_id);
866 		goto pass;
867 	}
868 
869 	if (tunnel->version == L2TP_HDR_VER_3 &&
870 	    l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr)) {
871 		l2tp_session_dec_refcount(session);
872 		goto invalid;
873 	}
874 
875 	l2tp_recv_common(session, skb, ptr, optr, hdrflags, length);
876 	l2tp_session_dec_refcount(session);
877 
878 	return 0;
879 
880 invalid:
881 	atomic_long_inc(&tunnel->stats.rx_invalid);
882 
883 pass:
884 	/* Put UDP header back */
885 	__skb_push(skb, sizeof(struct udphdr));
886 
887 	return 1;
888 }
889 
890 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
891  * Return codes:
892  * 0 : success.
893  * <0: error
894  * >0: skb should be passed up to userspace as UDP.
895  */
896 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
897 {
898 	struct l2tp_tunnel *tunnel;
899 
900 	/* Note that this is called from the encap_rcv hook inside an
901 	 * RCU-protected region, but without the socket being locked.
902 	 * Hence we use rcu_dereference_sk_user_data to access the
903 	 * tunnel data structure rather the usual l2tp_sk_to_tunnel
904 	 * accessor function.
905 	 */
906 	tunnel = rcu_dereference_sk_user_data(sk);
907 	if (!tunnel)
908 		goto pass_up;
909 	if (WARN_ON(tunnel->magic != L2TP_TUNNEL_MAGIC))
910 		goto pass_up;
911 
912 	if (l2tp_udp_recv_core(tunnel, skb))
913 		goto pass_up;
914 
915 	return 0;
916 
917 pass_up:
918 	return 1;
919 }
920 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
921 
922 /************************************************************************
923  * Transmit handling
924  ***********************************************************************/
925 
926 /* Build an L2TP header for the session into the buffer provided.
927  */
928 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
929 {
930 	struct l2tp_tunnel *tunnel = session->tunnel;
931 	__be16 *bufp = buf;
932 	__be16 *optr = buf;
933 	u16 flags = L2TP_HDR_VER_2;
934 	u32 tunnel_id = tunnel->peer_tunnel_id;
935 	u32 session_id = session->peer_session_id;
936 
937 	if (session->send_seq)
938 		flags |= L2TP_HDRFLAG_S;
939 
940 	/* Setup L2TP header. */
941 	*bufp++ = htons(flags);
942 	*bufp++ = htons(tunnel_id);
943 	*bufp++ = htons(session_id);
944 	if (session->send_seq) {
945 		*bufp++ = htons(session->ns);
946 		*bufp++ = 0;
947 		session->ns++;
948 		session->ns &= 0xffff;
949 		trace_session_seqnum_update(session);
950 	}
951 
952 	return bufp - optr;
953 }
954 
955 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
956 {
957 	struct l2tp_tunnel *tunnel = session->tunnel;
958 	char *bufp = buf;
959 	char *optr = bufp;
960 
961 	/* Setup L2TP header. The header differs slightly for UDP and
962 	 * IP encapsulations. For UDP, there is 4 bytes of flags.
963 	 */
964 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
965 		u16 flags = L2TP_HDR_VER_3;
966 		*((__be16 *)bufp) = htons(flags);
967 		bufp += 2;
968 		*((__be16 *)bufp) = 0;
969 		bufp += 2;
970 	}
971 
972 	*((__be32 *)bufp) = htonl(session->peer_session_id);
973 	bufp += 4;
974 	if (session->cookie_len) {
975 		memcpy(bufp, &session->cookie[0], session->cookie_len);
976 		bufp += session->cookie_len;
977 	}
978 	if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
979 		u32 l2h = 0;
980 
981 		if (session->send_seq) {
982 			l2h = 0x40000000 | session->ns;
983 			session->ns++;
984 			session->ns &= 0xffffff;
985 			trace_session_seqnum_update(session);
986 		}
987 
988 		*((__be32 *)bufp) = htonl(l2h);
989 		bufp += 4;
990 	}
991 
992 	return bufp - optr;
993 }
994 
995 /* Queue the packet to IP for output: tunnel socket lock must be held */
996 static int l2tp_xmit_queue(struct l2tp_tunnel *tunnel, struct sk_buff *skb, struct flowi *fl)
997 {
998 	int err;
999 
1000 	skb->ignore_df = 1;
1001 	skb_dst_drop(skb);
1002 #if IS_ENABLED(CONFIG_IPV6)
1003 	if (l2tp_sk_is_v6(tunnel->sock))
1004 		err = inet6_csk_xmit(tunnel->sock, skb, NULL);
1005 	else
1006 #endif
1007 		err = ip_queue_xmit(tunnel->sock, skb, fl);
1008 
1009 	return err >= 0 ? NET_XMIT_SUCCESS : NET_XMIT_DROP;
1010 }
1011 
1012 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, unsigned int *len)
1013 {
1014 	struct l2tp_tunnel *tunnel = session->tunnel;
1015 	unsigned int data_len = skb->len;
1016 	struct sock *sk = tunnel->sock;
1017 	int headroom, uhlen, udp_len;
1018 	int ret = NET_XMIT_SUCCESS;
1019 	struct inet_sock *inet;
1020 	struct udphdr *uh;
1021 
1022 	/* Check that there's enough headroom in the skb to insert IP,
1023 	 * UDP and L2TP headers. If not enough, expand it to
1024 	 * make room. Adjust truesize.
1025 	 */
1026 	uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(*uh) : 0;
1027 	headroom = NET_SKB_PAD + sizeof(struct iphdr) + uhlen + session->hdr_len;
1028 	if (skb_cow_head(skb, headroom)) {
1029 		kfree_skb(skb);
1030 		return NET_XMIT_DROP;
1031 	}
1032 
1033 	/* Setup L2TP header */
1034 	if (tunnel->version == L2TP_HDR_VER_2)
1035 		l2tp_build_l2tpv2_header(session, __skb_push(skb, session->hdr_len));
1036 	else
1037 		l2tp_build_l2tpv3_header(session, __skb_push(skb, session->hdr_len));
1038 
1039 	/* Reset skb netfilter state */
1040 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1041 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | IPSKB_REROUTED);
1042 	nf_reset_ct(skb);
1043 
1044 	bh_lock_sock_nested(sk);
1045 	if (sock_owned_by_user(sk)) {
1046 		kfree_skb(skb);
1047 		ret = NET_XMIT_DROP;
1048 		goto out_unlock;
1049 	}
1050 
1051 	/* The user-space may change the connection status for the user-space
1052 	 * provided socket at run time: we must check it under the socket lock
1053 	 */
1054 	if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1055 		kfree_skb(skb);
1056 		ret = NET_XMIT_DROP;
1057 		goto out_unlock;
1058 	}
1059 
1060 	/* Report transmitted length before we add encap header, which keeps
1061 	 * statistics consistent for both UDP and IP encap tx/rx paths.
1062 	 */
1063 	*len = skb->len;
1064 
1065 	inet = inet_sk(sk);
1066 	switch (tunnel->encap) {
1067 	case L2TP_ENCAPTYPE_UDP:
1068 		/* Setup UDP header */
1069 		__skb_push(skb, sizeof(*uh));
1070 		skb_reset_transport_header(skb);
1071 		uh = udp_hdr(skb);
1072 		uh->source = inet->inet_sport;
1073 		uh->dest = inet->inet_dport;
1074 		udp_len = uhlen + session->hdr_len + data_len;
1075 		uh->len = htons(udp_len);
1076 
1077 		/* Calculate UDP checksum if configured to do so */
1078 #if IS_ENABLED(CONFIG_IPV6)
1079 		if (l2tp_sk_is_v6(sk))
1080 			udp6_set_csum(udp_get_no_check6_tx(sk),
1081 				      skb, &inet6_sk(sk)->saddr,
1082 				      &sk->sk_v6_daddr, udp_len);
1083 		else
1084 #endif
1085 			udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1086 				     inet->inet_daddr, udp_len);
1087 		break;
1088 
1089 	case L2TP_ENCAPTYPE_IP:
1090 		break;
1091 	}
1092 
1093 	ret = l2tp_xmit_queue(tunnel, skb, &inet->cork.fl);
1094 
1095 out_unlock:
1096 	bh_unlock_sock(sk);
1097 
1098 	return ret;
1099 }
1100 
1101 /* If caller requires the skb to have a ppp header, the header must be
1102  * inserted in the skb data before calling this function.
1103  */
1104 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb)
1105 {
1106 	unsigned int len = 0;
1107 	int ret;
1108 
1109 	ret = l2tp_xmit_core(session, skb, &len);
1110 	if (ret == NET_XMIT_SUCCESS) {
1111 		atomic_long_inc(&session->tunnel->stats.tx_packets);
1112 		atomic_long_add(len, &session->tunnel->stats.tx_bytes);
1113 		atomic_long_inc(&session->stats.tx_packets);
1114 		atomic_long_add(len, &session->stats.tx_bytes);
1115 	} else {
1116 		atomic_long_inc(&session->tunnel->stats.tx_errors);
1117 		atomic_long_inc(&session->stats.tx_errors);
1118 	}
1119 	return ret;
1120 }
1121 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1122 
1123 /*****************************************************************************
1124  * Tinnel and session create/destroy.
1125  *****************************************************************************/
1126 
1127 /* Tunnel socket destruct hook.
1128  * The tunnel context is deleted only when all session sockets have been
1129  * closed.
1130  */
1131 static void l2tp_tunnel_destruct(struct sock *sk)
1132 {
1133 	struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
1134 
1135 	if (!tunnel)
1136 		goto end;
1137 
1138 	/* Disable udp encapsulation */
1139 	switch (tunnel->encap) {
1140 	case L2TP_ENCAPTYPE_UDP:
1141 		/* No longer an encapsulation socket. See net/ipv4/udp.c */
1142 		WRITE_ONCE(udp_sk(sk)->encap_type, 0);
1143 		udp_sk(sk)->encap_rcv = NULL;
1144 		udp_sk(sk)->encap_destroy = NULL;
1145 		break;
1146 	case L2TP_ENCAPTYPE_IP:
1147 		break;
1148 	}
1149 
1150 	/* Remove hooks into tunnel socket */
1151 	write_lock_bh(&sk->sk_callback_lock);
1152 	sk->sk_destruct = tunnel->old_sk_destruct;
1153 	sk->sk_user_data = NULL;
1154 	write_unlock_bh(&sk->sk_callback_lock);
1155 
1156 	/* Call the original destructor */
1157 	if (sk->sk_destruct)
1158 		(*sk->sk_destruct)(sk);
1159 
1160 	kfree_rcu(tunnel, rcu);
1161 end:
1162 	return;
1163 }
1164 
1165 /* Remove an l2tp session from l2tp_core's hash lists. */
1166 static void l2tp_session_unhash(struct l2tp_session *session)
1167 {
1168 	struct l2tp_tunnel *tunnel = session->tunnel;
1169 
1170 	/* Remove the session from core hashes */
1171 	if (tunnel) {
1172 		/* Remove from the per-tunnel hash */
1173 		spin_lock_bh(&tunnel->hlist_lock);
1174 		hlist_del_init_rcu(&session->hlist);
1175 		spin_unlock_bh(&tunnel->hlist_lock);
1176 
1177 		/* For L2TPv3 we have a per-net hash: remove from there, too */
1178 		if (tunnel->version != L2TP_HDR_VER_2) {
1179 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1180 
1181 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1182 			hlist_del_init_rcu(&session->global_hlist);
1183 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1184 		}
1185 
1186 		synchronize_rcu();
1187 	}
1188 }
1189 
1190 /* When the tunnel is closed, all the attached sessions need to go too.
1191  */
1192 static void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1193 {
1194 	struct l2tp_session *session;
1195 	int hash;
1196 
1197 	spin_lock_bh(&tunnel->hlist_lock);
1198 	tunnel->acpt_newsess = false;
1199 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1200 again:
1201 		hlist_for_each_entry_rcu(session, &tunnel->session_hlist[hash], hlist) {
1202 			hlist_del_init_rcu(&session->hlist);
1203 
1204 			spin_unlock_bh(&tunnel->hlist_lock);
1205 			l2tp_session_delete(session);
1206 			spin_lock_bh(&tunnel->hlist_lock);
1207 
1208 			/* Now restart from the beginning of this hash
1209 			 * chain.  We always remove a session from the
1210 			 * list so we are guaranteed to make forward
1211 			 * progress.
1212 			 */
1213 			goto again;
1214 		}
1215 	}
1216 	spin_unlock_bh(&tunnel->hlist_lock);
1217 }
1218 
1219 /* Tunnel socket destroy hook for UDP encapsulation */
1220 static void l2tp_udp_encap_destroy(struct sock *sk)
1221 {
1222 	struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
1223 
1224 	if (tunnel)
1225 		l2tp_tunnel_delete(tunnel);
1226 }
1227 
1228 static void l2tp_tunnel_remove(struct net *net, struct l2tp_tunnel *tunnel)
1229 {
1230 	struct l2tp_net *pn = l2tp_pernet(net);
1231 
1232 	spin_lock_bh(&pn->l2tp_tunnel_idr_lock);
1233 	idr_remove(&pn->l2tp_tunnel_idr, tunnel->tunnel_id);
1234 	spin_unlock_bh(&pn->l2tp_tunnel_idr_lock);
1235 }
1236 
1237 /* Workqueue tunnel deletion function */
1238 static void l2tp_tunnel_del_work(struct work_struct *work)
1239 {
1240 	struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1241 						  del_work);
1242 	struct sock *sk = tunnel->sock;
1243 	struct socket *sock = sk->sk_socket;
1244 
1245 	l2tp_tunnel_closeall(tunnel);
1246 
1247 	/* If the tunnel socket was created within the kernel, use
1248 	 * the sk API to release it here.
1249 	 */
1250 	if (tunnel->fd < 0) {
1251 		if (sock) {
1252 			kernel_sock_shutdown(sock, SHUT_RDWR);
1253 			sock_release(sock);
1254 		}
1255 	}
1256 
1257 	l2tp_tunnel_remove(tunnel->l2tp_net, tunnel);
1258 	/* drop initial ref */
1259 	l2tp_tunnel_dec_refcount(tunnel);
1260 
1261 	/* drop workqueue ref */
1262 	l2tp_tunnel_dec_refcount(tunnel);
1263 }
1264 
1265 /* Create a socket for the tunnel, if one isn't set up by
1266  * userspace. This is used for static tunnels where there is no
1267  * managing L2TP daemon.
1268  *
1269  * Since we don't want these sockets to keep a namespace alive by
1270  * themselves, we drop the socket's namespace refcount after creation.
1271  * These sockets are freed when the namespace exits using the pernet
1272  * exit hook.
1273  */
1274 static int l2tp_tunnel_sock_create(struct net *net,
1275 				   u32 tunnel_id,
1276 				   u32 peer_tunnel_id,
1277 				   struct l2tp_tunnel_cfg *cfg,
1278 				   struct socket **sockp)
1279 {
1280 	int err = -EINVAL;
1281 	struct socket *sock = NULL;
1282 	struct udp_port_cfg udp_conf;
1283 
1284 	switch (cfg->encap) {
1285 	case L2TP_ENCAPTYPE_UDP:
1286 		memset(&udp_conf, 0, sizeof(udp_conf));
1287 
1288 #if IS_ENABLED(CONFIG_IPV6)
1289 		if (cfg->local_ip6 && cfg->peer_ip6) {
1290 			udp_conf.family = AF_INET6;
1291 			memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1292 			       sizeof(udp_conf.local_ip6));
1293 			memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1294 			       sizeof(udp_conf.peer_ip6));
1295 			udp_conf.use_udp6_tx_checksums =
1296 			  !cfg->udp6_zero_tx_checksums;
1297 			udp_conf.use_udp6_rx_checksums =
1298 			  !cfg->udp6_zero_rx_checksums;
1299 		} else
1300 #endif
1301 		{
1302 			udp_conf.family = AF_INET;
1303 			udp_conf.local_ip = cfg->local_ip;
1304 			udp_conf.peer_ip = cfg->peer_ip;
1305 			udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1306 		}
1307 
1308 		udp_conf.local_udp_port = htons(cfg->local_udp_port);
1309 		udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1310 
1311 		err = udp_sock_create(net, &udp_conf, &sock);
1312 		if (err < 0)
1313 			goto out;
1314 
1315 		break;
1316 
1317 	case L2TP_ENCAPTYPE_IP:
1318 #if IS_ENABLED(CONFIG_IPV6)
1319 		if (cfg->local_ip6 && cfg->peer_ip6) {
1320 			struct sockaddr_l2tpip6 ip6_addr = {0};
1321 
1322 			err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1323 					       IPPROTO_L2TP, &sock);
1324 			if (err < 0)
1325 				goto out;
1326 
1327 			ip6_addr.l2tp_family = AF_INET6;
1328 			memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1329 			       sizeof(ip6_addr.l2tp_addr));
1330 			ip6_addr.l2tp_conn_id = tunnel_id;
1331 			err = kernel_bind(sock, (struct sockaddr *)&ip6_addr,
1332 					  sizeof(ip6_addr));
1333 			if (err < 0)
1334 				goto out;
1335 
1336 			ip6_addr.l2tp_family = AF_INET6;
1337 			memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1338 			       sizeof(ip6_addr.l2tp_addr));
1339 			ip6_addr.l2tp_conn_id = peer_tunnel_id;
1340 			err = kernel_connect(sock,
1341 					     (struct sockaddr *)&ip6_addr,
1342 					     sizeof(ip6_addr), 0);
1343 			if (err < 0)
1344 				goto out;
1345 		} else
1346 #endif
1347 		{
1348 			struct sockaddr_l2tpip ip_addr = {0};
1349 
1350 			err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1351 					       IPPROTO_L2TP, &sock);
1352 			if (err < 0)
1353 				goto out;
1354 
1355 			ip_addr.l2tp_family = AF_INET;
1356 			ip_addr.l2tp_addr = cfg->local_ip;
1357 			ip_addr.l2tp_conn_id = tunnel_id;
1358 			err = kernel_bind(sock, (struct sockaddr *)&ip_addr,
1359 					  sizeof(ip_addr));
1360 			if (err < 0)
1361 				goto out;
1362 
1363 			ip_addr.l2tp_family = AF_INET;
1364 			ip_addr.l2tp_addr = cfg->peer_ip;
1365 			ip_addr.l2tp_conn_id = peer_tunnel_id;
1366 			err = kernel_connect(sock, (struct sockaddr *)&ip_addr,
1367 					     sizeof(ip_addr), 0);
1368 			if (err < 0)
1369 				goto out;
1370 		}
1371 		break;
1372 
1373 	default:
1374 		goto out;
1375 	}
1376 
1377 out:
1378 	*sockp = sock;
1379 	if (err < 0 && sock) {
1380 		kernel_sock_shutdown(sock, SHUT_RDWR);
1381 		sock_release(sock);
1382 		*sockp = NULL;
1383 	}
1384 
1385 	return err;
1386 }
1387 
1388 int l2tp_tunnel_create(int fd, int version, u32 tunnel_id, u32 peer_tunnel_id,
1389 		       struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1390 {
1391 	struct l2tp_tunnel *tunnel = NULL;
1392 	int err;
1393 	enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1394 
1395 	if (cfg)
1396 		encap = cfg->encap;
1397 
1398 	tunnel = kzalloc(sizeof(*tunnel), GFP_KERNEL);
1399 	if (!tunnel) {
1400 		err = -ENOMEM;
1401 		goto err;
1402 	}
1403 
1404 	tunnel->version = version;
1405 	tunnel->tunnel_id = tunnel_id;
1406 	tunnel->peer_tunnel_id = peer_tunnel_id;
1407 
1408 	tunnel->magic = L2TP_TUNNEL_MAGIC;
1409 	sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1410 	spin_lock_init(&tunnel->hlist_lock);
1411 	tunnel->acpt_newsess = true;
1412 
1413 	tunnel->encap = encap;
1414 
1415 	refcount_set(&tunnel->ref_count, 1);
1416 	tunnel->fd = fd;
1417 
1418 	/* Init delete workqueue struct */
1419 	INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1420 
1421 	INIT_LIST_HEAD(&tunnel->list);
1422 
1423 	err = 0;
1424 err:
1425 	if (tunnelp)
1426 		*tunnelp = tunnel;
1427 
1428 	return err;
1429 }
1430 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1431 
1432 static int l2tp_validate_socket(const struct sock *sk, const struct net *net,
1433 				enum l2tp_encap_type encap)
1434 {
1435 	if (!net_eq(sock_net(sk), net))
1436 		return -EINVAL;
1437 
1438 	if (sk->sk_type != SOCK_DGRAM)
1439 		return -EPROTONOSUPPORT;
1440 
1441 	if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
1442 		return -EPROTONOSUPPORT;
1443 
1444 	if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) ||
1445 	    (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP))
1446 		return -EPROTONOSUPPORT;
1447 
1448 	if (sk->sk_user_data)
1449 		return -EBUSY;
1450 
1451 	return 0;
1452 }
1453 
1454 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net,
1455 			 struct l2tp_tunnel_cfg *cfg)
1456 {
1457 	struct l2tp_net *pn = l2tp_pernet(net);
1458 	u32 tunnel_id = tunnel->tunnel_id;
1459 	struct socket *sock;
1460 	struct sock *sk;
1461 	int ret;
1462 
1463 	spin_lock_bh(&pn->l2tp_tunnel_idr_lock);
1464 	ret = idr_alloc_u32(&pn->l2tp_tunnel_idr, NULL, &tunnel_id, tunnel_id,
1465 			    GFP_ATOMIC);
1466 	spin_unlock_bh(&pn->l2tp_tunnel_idr_lock);
1467 	if (ret)
1468 		return ret == -ENOSPC ? -EEXIST : ret;
1469 
1470 	if (tunnel->fd < 0) {
1471 		ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id,
1472 					      tunnel->peer_tunnel_id, cfg,
1473 					      &sock);
1474 		if (ret < 0)
1475 			goto err;
1476 	} else {
1477 		sock = sockfd_lookup(tunnel->fd, &ret);
1478 		if (!sock)
1479 			goto err;
1480 	}
1481 
1482 	sk = sock->sk;
1483 	lock_sock(sk);
1484 	write_lock_bh(&sk->sk_callback_lock);
1485 	ret = l2tp_validate_socket(sk, net, tunnel->encap);
1486 	if (ret < 0)
1487 		goto err_inval_sock;
1488 	rcu_assign_sk_user_data(sk, tunnel);
1489 	write_unlock_bh(&sk->sk_callback_lock);
1490 
1491 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1492 		struct udp_tunnel_sock_cfg udp_cfg = {
1493 			.sk_user_data = tunnel,
1494 			.encap_type = UDP_ENCAP_L2TPINUDP,
1495 			.encap_rcv = l2tp_udp_encap_recv,
1496 			.encap_destroy = l2tp_udp_encap_destroy,
1497 		};
1498 
1499 		setup_udp_tunnel_sock(net, sock, &udp_cfg);
1500 	}
1501 
1502 	tunnel->old_sk_destruct = sk->sk_destruct;
1503 	sk->sk_destruct = &l2tp_tunnel_destruct;
1504 	sk->sk_allocation = GFP_ATOMIC;
1505 	release_sock(sk);
1506 
1507 	sock_hold(sk);
1508 	tunnel->sock = sk;
1509 	tunnel->l2tp_net = net;
1510 
1511 	spin_lock_bh(&pn->l2tp_tunnel_idr_lock);
1512 	idr_replace(&pn->l2tp_tunnel_idr, tunnel, tunnel->tunnel_id);
1513 	spin_unlock_bh(&pn->l2tp_tunnel_idr_lock);
1514 
1515 	trace_register_tunnel(tunnel);
1516 
1517 	if (tunnel->fd >= 0)
1518 		sockfd_put(sock);
1519 
1520 	return 0;
1521 
1522 err_inval_sock:
1523 	write_unlock_bh(&sk->sk_callback_lock);
1524 	release_sock(sk);
1525 
1526 	if (tunnel->fd < 0)
1527 		sock_release(sock);
1528 	else
1529 		sockfd_put(sock);
1530 err:
1531 	l2tp_tunnel_remove(net, tunnel);
1532 	return ret;
1533 }
1534 EXPORT_SYMBOL_GPL(l2tp_tunnel_register);
1535 
1536 /* This function is used by the netlink TUNNEL_DELETE command.
1537  */
1538 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1539 {
1540 	if (!test_and_set_bit(0, &tunnel->dead)) {
1541 		trace_delete_tunnel(tunnel);
1542 		l2tp_tunnel_inc_refcount(tunnel);
1543 		queue_work(l2tp_wq, &tunnel->del_work);
1544 	}
1545 }
1546 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1547 
1548 void l2tp_session_delete(struct l2tp_session *session)
1549 {
1550 	if (test_and_set_bit(0, &session->dead))
1551 		return;
1552 
1553 	trace_delete_session(session);
1554 	l2tp_session_unhash(session);
1555 	l2tp_session_queue_purge(session);
1556 	if (session->session_close)
1557 		(*session->session_close)(session);
1558 
1559 	l2tp_session_dec_refcount(session);
1560 }
1561 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1562 
1563 /* We come here whenever a session's send_seq, cookie_len or
1564  * l2specific_type parameters are set.
1565  */
1566 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1567 {
1568 	if (version == L2TP_HDR_VER_2) {
1569 		session->hdr_len = 6;
1570 		if (session->send_seq)
1571 			session->hdr_len += 4;
1572 	} else {
1573 		session->hdr_len = 4 + session->cookie_len;
1574 		session->hdr_len += l2tp_get_l2specific_len(session);
1575 		if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1576 			session->hdr_len += 4;
1577 	}
1578 }
1579 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1580 
1581 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id,
1582 					 u32 peer_session_id, struct l2tp_session_cfg *cfg)
1583 {
1584 	struct l2tp_session *session;
1585 
1586 	session = kzalloc(sizeof(*session) + priv_size, GFP_KERNEL);
1587 	if (session) {
1588 		session->magic = L2TP_SESSION_MAGIC;
1589 		session->tunnel = tunnel;
1590 
1591 		session->session_id = session_id;
1592 		session->peer_session_id = peer_session_id;
1593 		session->nr = 0;
1594 		if (tunnel->version == L2TP_HDR_VER_2)
1595 			session->nr_max = 0xffff;
1596 		else
1597 			session->nr_max = 0xffffff;
1598 		session->nr_window_size = session->nr_max / 2;
1599 		session->nr_oos_count_max = 4;
1600 
1601 		/* Use NR of first received packet */
1602 		session->reorder_skip = 1;
1603 
1604 		sprintf(&session->name[0], "sess %u/%u",
1605 			tunnel->tunnel_id, session->session_id);
1606 
1607 		skb_queue_head_init(&session->reorder_q);
1608 
1609 		INIT_HLIST_NODE(&session->hlist);
1610 		INIT_HLIST_NODE(&session->global_hlist);
1611 
1612 		if (cfg) {
1613 			session->pwtype = cfg->pw_type;
1614 			session->send_seq = cfg->send_seq;
1615 			session->recv_seq = cfg->recv_seq;
1616 			session->lns_mode = cfg->lns_mode;
1617 			session->reorder_timeout = cfg->reorder_timeout;
1618 			session->l2specific_type = cfg->l2specific_type;
1619 			session->cookie_len = cfg->cookie_len;
1620 			memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1621 			session->peer_cookie_len = cfg->peer_cookie_len;
1622 			memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1623 		}
1624 
1625 		l2tp_session_set_header_len(session, tunnel->version);
1626 
1627 		refcount_set(&session->ref_count, 1);
1628 
1629 		return session;
1630 	}
1631 
1632 	return ERR_PTR(-ENOMEM);
1633 }
1634 EXPORT_SYMBOL_GPL(l2tp_session_create);
1635 
1636 /*****************************************************************************
1637  * Init and cleanup
1638  *****************************************************************************/
1639 
1640 static __net_init int l2tp_init_net(struct net *net)
1641 {
1642 	struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1643 	int hash;
1644 
1645 	idr_init(&pn->l2tp_tunnel_idr);
1646 	spin_lock_init(&pn->l2tp_tunnel_idr_lock);
1647 
1648 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1649 		INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1650 
1651 	spin_lock_init(&pn->l2tp_session_hlist_lock);
1652 
1653 	return 0;
1654 }
1655 
1656 static __net_exit void l2tp_exit_net(struct net *net)
1657 {
1658 	struct l2tp_net *pn = l2tp_pernet(net);
1659 	struct l2tp_tunnel *tunnel = NULL;
1660 	unsigned long tunnel_id, tmp;
1661 	int hash;
1662 
1663 	rcu_read_lock_bh();
1664 	idr_for_each_entry_ul(&pn->l2tp_tunnel_idr, tunnel, tmp, tunnel_id) {
1665 		if (tunnel)
1666 			l2tp_tunnel_delete(tunnel);
1667 	}
1668 	rcu_read_unlock_bh();
1669 
1670 	if (l2tp_wq)
1671 		flush_workqueue(l2tp_wq);
1672 	rcu_barrier();
1673 
1674 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1675 		WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1676 	idr_destroy(&pn->l2tp_tunnel_idr);
1677 }
1678 
1679 static struct pernet_operations l2tp_net_ops = {
1680 	.init = l2tp_init_net,
1681 	.exit = l2tp_exit_net,
1682 	.id   = &l2tp_net_id,
1683 	.size = sizeof(struct l2tp_net),
1684 };
1685 
1686 static int __init l2tp_init(void)
1687 {
1688 	int rc = 0;
1689 
1690 	rc = register_pernet_device(&l2tp_net_ops);
1691 	if (rc)
1692 		goto out;
1693 
1694 	l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1695 	if (!l2tp_wq) {
1696 		pr_err("alloc_workqueue failed\n");
1697 		unregister_pernet_device(&l2tp_net_ops);
1698 		rc = -ENOMEM;
1699 		goto out;
1700 	}
1701 
1702 	pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1703 
1704 out:
1705 	return rc;
1706 }
1707 
1708 static void __exit l2tp_exit(void)
1709 {
1710 	unregister_pernet_device(&l2tp_net_ops);
1711 	if (l2tp_wq) {
1712 		destroy_workqueue(l2tp_wq);
1713 		l2tp_wq = NULL;
1714 	}
1715 }
1716 
1717 module_init(l2tp_init);
1718 module_exit(l2tp_exit);
1719 
1720 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1721 MODULE_DESCRIPTION("L2TP core");
1722 MODULE_LICENSE("GPL");
1723 MODULE_VERSION(L2TP_DRV_VERSION);
1724