xref: /linux/net/l2tp/l2tp_core.c (revision b4ba157044ea433a66126603ad7140e12dbc794b)
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 	struct l2tp_tunnel *orig_tunnel = tunnel;
798 	unsigned char *ptr, *optr;
799 	u16 hdrflags;
800 	u32 tunnel_id, session_id;
801 	u16 version;
802 	int length;
803 
804 	/* UDP has verified checksum */
805 
806 	/* UDP always verifies the packet length. */
807 	__skb_pull(skb, sizeof(struct udphdr));
808 
809 	/* Short packet? */
810 	if (!pskb_may_pull(skb, L2TP_HDR_SIZE_MAX)) {
811 		pr_debug_ratelimited("%s: recv short packet (len=%d)\n",
812 				     tunnel->name, skb->len);
813 		goto invalid;
814 	}
815 
816 	/* Point to L2TP header */
817 	optr = skb->data;
818 	ptr = skb->data;
819 
820 	/* Get L2TP header flags */
821 	hdrflags = ntohs(*(__be16 *)ptr);
822 
823 	/* Get protocol version */
824 	version = hdrflags & L2TP_HDR_VER_MASK;
825 
826 	/* Get length of L2TP packet */
827 	length = skb->len;
828 
829 	/* If type is control packet, it is handled by userspace. */
830 	if (hdrflags & L2TP_HDRFLAG_T)
831 		goto pass;
832 
833 	/* Skip flags */
834 	ptr += 2;
835 
836 	if (version == L2TP_HDR_VER_2) {
837 		/* If length is present, skip it */
838 		if (hdrflags & L2TP_HDRFLAG_L)
839 			ptr += 2;
840 
841 		/* Extract tunnel and session ID */
842 		tunnel_id = ntohs(*(__be16 *)ptr);
843 		ptr += 2;
844 
845 		if (tunnel_id != tunnel->tunnel_id) {
846 			/* We are receiving trafic for another tunnel, probably
847 			 * because we have several tunnels between the same
848 			 * IP/port quadruple, look it up.
849 			 */
850 			struct l2tp_tunnel *alt_tunnel;
851 
852 			alt_tunnel = l2tp_tunnel_get(tunnel->l2tp_net, tunnel_id);
853 			if (!alt_tunnel)
854 				goto pass;
855 			tunnel = alt_tunnel;
856 		}
857 
858 		session_id = ntohs(*(__be16 *)ptr);
859 		ptr += 2;
860 	} else {
861 		ptr += 2;	/* skip reserved bits */
862 		tunnel_id = tunnel->tunnel_id;
863 		session_id = ntohl(*(__be32 *)ptr);
864 		ptr += 4;
865 	}
866 
867 	/* Check protocol version */
868 	if (version != tunnel->version) {
869 		pr_debug_ratelimited("%s: recv protocol version mismatch: got %d expected %d\n",
870 				     tunnel->name, version, tunnel->version);
871 		goto invalid;
872 	}
873 
874 	/* Find the session context */
875 	session = l2tp_tunnel_get_session(tunnel, session_id);
876 	if (!session || !session->recv_skb) {
877 		if (session)
878 			l2tp_session_dec_refcount(session);
879 
880 		/* Not found? Pass to userspace to deal with */
881 		pr_debug_ratelimited("%s: no session found (%u/%u). Passing up.\n",
882 				     tunnel->name, tunnel_id, session_id);
883 		goto pass;
884 	}
885 
886 	if (tunnel->version == L2TP_HDR_VER_3 &&
887 	    l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr)) {
888 		l2tp_session_dec_refcount(session);
889 		goto invalid;
890 	}
891 
892 	l2tp_recv_common(session, skb, ptr, optr, hdrflags, length);
893 	l2tp_session_dec_refcount(session);
894 
895 	if (tunnel != orig_tunnel)
896 		l2tp_tunnel_dec_refcount(tunnel);
897 
898 	return 0;
899 
900 invalid:
901 	atomic_long_inc(&tunnel->stats.rx_invalid);
902 
903 pass:
904 	/* Put UDP header back */
905 	__skb_push(skb, sizeof(struct udphdr));
906 
907 	if (tunnel != orig_tunnel)
908 		l2tp_tunnel_dec_refcount(tunnel);
909 
910 	return 1;
911 }
912 
913 /* UDP encapsulation receive and error receive handlers.
914  * See net/ipv4/udp.c for details.
915  *
916  * Note that these functions are called from inside an
917  * RCU-protected region, but without the socket being locked.
918  *
919  * Hence we use rcu_dereference_sk_user_data to access the
920  * tunnel data structure rather the usual l2tp_sk_to_tunnel
921  * accessor function.
922  */
923 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
924 {
925 	struct l2tp_tunnel *tunnel;
926 
927 	tunnel = rcu_dereference_sk_user_data(sk);
928 	if (!tunnel)
929 		goto pass_up;
930 	if (WARN_ON(tunnel->magic != L2TP_TUNNEL_MAGIC))
931 		goto pass_up;
932 
933 	if (l2tp_udp_recv_core(tunnel, skb))
934 		goto pass_up;
935 
936 	return 0;
937 
938 pass_up:
939 	return 1;
940 }
941 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
942 
943 static void l2tp_udp_encap_err_recv(struct sock *sk, struct sk_buff *skb, int err,
944 				    __be16 port, u32 info, u8 *payload)
945 {
946 	struct l2tp_tunnel *tunnel;
947 
948 	tunnel = rcu_dereference_sk_user_data(sk);
949 	if (!tunnel || tunnel->fd < 0)
950 		return;
951 
952 	sk->sk_err = err;
953 	sk_error_report(sk);
954 
955 	if (ip_hdr(skb)->version == IPVERSION) {
956 		if (inet_test_bit(RECVERR, sk))
957 			return ip_icmp_error(sk, skb, err, port, info, payload);
958 #if IS_ENABLED(CONFIG_IPV6)
959 	} else {
960 		if (inet6_test_bit(RECVERR6, sk))
961 			return ipv6_icmp_error(sk, skb, err, port, info, payload);
962 #endif
963 	}
964 }
965 
966 /************************************************************************
967  * Transmit handling
968  ***********************************************************************/
969 
970 /* Build an L2TP header for the session into the buffer provided.
971  */
972 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
973 {
974 	struct l2tp_tunnel *tunnel = session->tunnel;
975 	__be16 *bufp = buf;
976 	__be16 *optr = buf;
977 	u16 flags = L2TP_HDR_VER_2;
978 	u32 tunnel_id = tunnel->peer_tunnel_id;
979 	u32 session_id = session->peer_session_id;
980 
981 	if (session->send_seq)
982 		flags |= L2TP_HDRFLAG_S;
983 
984 	/* Setup L2TP header. */
985 	*bufp++ = htons(flags);
986 	*bufp++ = htons(tunnel_id);
987 	*bufp++ = htons(session_id);
988 	if (session->send_seq) {
989 		*bufp++ = htons(session->ns);
990 		*bufp++ = 0;
991 		session->ns++;
992 		session->ns &= 0xffff;
993 		trace_session_seqnum_update(session);
994 	}
995 
996 	return bufp - optr;
997 }
998 
999 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1000 {
1001 	struct l2tp_tunnel *tunnel = session->tunnel;
1002 	char *bufp = buf;
1003 	char *optr = bufp;
1004 
1005 	/* Setup L2TP header. The header differs slightly for UDP and
1006 	 * IP encapsulations. For UDP, there is 4 bytes of flags.
1007 	 */
1008 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1009 		u16 flags = L2TP_HDR_VER_3;
1010 		*((__be16 *)bufp) = htons(flags);
1011 		bufp += 2;
1012 		*((__be16 *)bufp) = 0;
1013 		bufp += 2;
1014 	}
1015 
1016 	*((__be32 *)bufp) = htonl(session->peer_session_id);
1017 	bufp += 4;
1018 	if (session->cookie_len) {
1019 		memcpy(bufp, &session->cookie[0], session->cookie_len);
1020 		bufp += session->cookie_len;
1021 	}
1022 	if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1023 		u32 l2h = 0;
1024 
1025 		if (session->send_seq) {
1026 			l2h = 0x40000000 | session->ns;
1027 			session->ns++;
1028 			session->ns &= 0xffffff;
1029 			trace_session_seqnum_update(session);
1030 		}
1031 
1032 		*((__be32 *)bufp) = htonl(l2h);
1033 		bufp += 4;
1034 	}
1035 
1036 	return bufp - optr;
1037 }
1038 
1039 /* Queue the packet to IP for output: tunnel socket lock must be held */
1040 static int l2tp_xmit_queue(struct l2tp_tunnel *tunnel, struct sk_buff *skb, struct flowi *fl)
1041 {
1042 	int err;
1043 
1044 	skb->ignore_df = 1;
1045 	skb_dst_drop(skb);
1046 #if IS_ENABLED(CONFIG_IPV6)
1047 	if (l2tp_sk_is_v6(tunnel->sock))
1048 		err = inet6_csk_xmit(tunnel->sock, skb, NULL);
1049 	else
1050 #endif
1051 		err = ip_queue_xmit(tunnel->sock, skb, fl);
1052 
1053 	return err >= 0 ? NET_XMIT_SUCCESS : NET_XMIT_DROP;
1054 }
1055 
1056 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, unsigned int *len)
1057 {
1058 	struct l2tp_tunnel *tunnel = session->tunnel;
1059 	unsigned int data_len = skb->len;
1060 	struct sock *sk = tunnel->sock;
1061 	int headroom, uhlen, udp_len;
1062 	int ret = NET_XMIT_SUCCESS;
1063 	struct inet_sock *inet;
1064 	struct udphdr *uh;
1065 
1066 	/* Check that there's enough headroom in the skb to insert IP,
1067 	 * UDP and L2TP headers. If not enough, expand it to
1068 	 * make room. Adjust truesize.
1069 	 */
1070 	uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(*uh) : 0;
1071 	headroom = NET_SKB_PAD + sizeof(struct iphdr) + uhlen + session->hdr_len;
1072 	if (skb_cow_head(skb, headroom)) {
1073 		kfree_skb(skb);
1074 		return NET_XMIT_DROP;
1075 	}
1076 
1077 	/* Setup L2TP header */
1078 	if (tunnel->version == L2TP_HDR_VER_2)
1079 		l2tp_build_l2tpv2_header(session, __skb_push(skb, session->hdr_len));
1080 	else
1081 		l2tp_build_l2tpv3_header(session, __skb_push(skb, session->hdr_len));
1082 
1083 	/* Reset skb netfilter state */
1084 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1085 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | IPSKB_REROUTED);
1086 	nf_reset_ct(skb);
1087 
1088 	bh_lock_sock_nested(sk);
1089 	if (sock_owned_by_user(sk)) {
1090 		kfree_skb(skb);
1091 		ret = NET_XMIT_DROP;
1092 		goto out_unlock;
1093 	}
1094 
1095 	/* The user-space may change the connection status for the user-space
1096 	 * provided socket at run time: we must check it under the socket lock
1097 	 */
1098 	if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1099 		kfree_skb(skb);
1100 		ret = NET_XMIT_DROP;
1101 		goto out_unlock;
1102 	}
1103 
1104 	/* Report transmitted length before we add encap header, which keeps
1105 	 * statistics consistent for both UDP and IP encap tx/rx paths.
1106 	 */
1107 	*len = skb->len;
1108 
1109 	inet = inet_sk(sk);
1110 	switch (tunnel->encap) {
1111 	case L2TP_ENCAPTYPE_UDP:
1112 		/* Setup UDP header */
1113 		__skb_push(skb, sizeof(*uh));
1114 		skb_reset_transport_header(skb);
1115 		uh = udp_hdr(skb);
1116 		uh->source = inet->inet_sport;
1117 		uh->dest = inet->inet_dport;
1118 		udp_len = uhlen + session->hdr_len + data_len;
1119 		uh->len = htons(udp_len);
1120 
1121 		/* Calculate UDP checksum if configured to do so */
1122 #if IS_ENABLED(CONFIG_IPV6)
1123 		if (l2tp_sk_is_v6(sk))
1124 			udp6_set_csum(udp_get_no_check6_tx(sk),
1125 				      skb, &inet6_sk(sk)->saddr,
1126 				      &sk->sk_v6_daddr, udp_len);
1127 		else
1128 #endif
1129 			udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1130 				     inet->inet_daddr, udp_len);
1131 		break;
1132 
1133 	case L2TP_ENCAPTYPE_IP:
1134 		break;
1135 	}
1136 
1137 	ret = l2tp_xmit_queue(tunnel, skb, &inet->cork.fl);
1138 
1139 out_unlock:
1140 	bh_unlock_sock(sk);
1141 
1142 	return ret;
1143 }
1144 
1145 /* If caller requires the skb to have a ppp header, the header must be
1146  * inserted in the skb data before calling this function.
1147  */
1148 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb)
1149 {
1150 	unsigned int len = 0;
1151 	int ret;
1152 
1153 	ret = l2tp_xmit_core(session, skb, &len);
1154 	if (ret == NET_XMIT_SUCCESS) {
1155 		atomic_long_inc(&session->tunnel->stats.tx_packets);
1156 		atomic_long_add(len, &session->tunnel->stats.tx_bytes);
1157 		atomic_long_inc(&session->stats.tx_packets);
1158 		atomic_long_add(len, &session->stats.tx_bytes);
1159 	} else {
1160 		atomic_long_inc(&session->tunnel->stats.tx_errors);
1161 		atomic_long_inc(&session->stats.tx_errors);
1162 	}
1163 	return ret;
1164 }
1165 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1166 
1167 /*****************************************************************************
1168  * Tinnel and session create/destroy.
1169  *****************************************************************************/
1170 
1171 /* Tunnel socket destruct hook.
1172  * The tunnel context is deleted only when all session sockets have been
1173  * closed.
1174  */
1175 static void l2tp_tunnel_destruct(struct sock *sk)
1176 {
1177 	struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
1178 
1179 	if (!tunnel)
1180 		goto end;
1181 
1182 	/* Disable udp encapsulation */
1183 	switch (tunnel->encap) {
1184 	case L2TP_ENCAPTYPE_UDP:
1185 		/* No longer an encapsulation socket. See net/ipv4/udp.c */
1186 		WRITE_ONCE(udp_sk(sk)->encap_type, 0);
1187 		udp_sk(sk)->encap_rcv = NULL;
1188 		udp_sk(sk)->encap_destroy = NULL;
1189 		break;
1190 	case L2TP_ENCAPTYPE_IP:
1191 		break;
1192 	}
1193 
1194 	/* Remove hooks into tunnel socket */
1195 	write_lock_bh(&sk->sk_callback_lock);
1196 	sk->sk_destruct = tunnel->old_sk_destruct;
1197 	sk->sk_user_data = NULL;
1198 	write_unlock_bh(&sk->sk_callback_lock);
1199 
1200 	/* Call the original destructor */
1201 	if (sk->sk_destruct)
1202 		(*sk->sk_destruct)(sk);
1203 
1204 	kfree_rcu(tunnel, rcu);
1205 end:
1206 	return;
1207 }
1208 
1209 /* Remove an l2tp session from l2tp_core's hash lists. */
1210 static void l2tp_session_unhash(struct l2tp_session *session)
1211 {
1212 	struct l2tp_tunnel *tunnel = session->tunnel;
1213 
1214 	/* Remove the session from core hashes */
1215 	if (tunnel) {
1216 		/* Remove from the per-tunnel hash */
1217 		spin_lock_bh(&tunnel->hlist_lock);
1218 		hlist_del_init_rcu(&session->hlist);
1219 		spin_unlock_bh(&tunnel->hlist_lock);
1220 
1221 		/* For L2TPv3 we have a per-net hash: remove from there, too */
1222 		if (tunnel->version != L2TP_HDR_VER_2) {
1223 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1224 
1225 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1226 			hlist_del_init_rcu(&session->global_hlist);
1227 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1228 		}
1229 
1230 		synchronize_rcu();
1231 	}
1232 }
1233 
1234 /* When the tunnel is closed, all the attached sessions need to go too.
1235  */
1236 static void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1237 {
1238 	struct l2tp_session *session;
1239 	int hash;
1240 
1241 	spin_lock_bh(&tunnel->hlist_lock);
1242 	tunnel->acpt_newsess = false;
1243 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1244 again:
1245 		hlist_for_each_entry_rcu(session, &tunnel->session_hlist[hash], hlist) {
1246 			hlist_del_init_rcu(&session->hlist);
1247 
1248 			spin_unlock_bh(&tunnel->hlist_lock);
1249 			l2tp_session_delete(session);
1250 			spin_lock_bh(&tunnel->hlist_lock);
1251 
1252 			/* Now restart from the beginning of this hash
1253 			 * chain.  We always remove a session from the
1254 			 * list so we are guaranteed to make forward
1255 			 * progress.
1256 			 */
1257 			goto again;
1258 		}
1259 	}
1260 	spin_unlock_bh(&tunnel->hlist_lock);
1261 }
1262 
1263 /* Tunnel socket destroy hook for UDP encapsulation */
1264 static void l2tp_udp_encap_destroy(struct sock *sk)
1265 {
1266 	struct l2tp_tunnel *tunnel = l2tp_sk_to_tunnel(sk);
1267 
1268 	if (tunnel)
1269 		l2tp_tunnel_delete(tunnel);
1270 }
1271 
1272 static void l2tp_tunnel_remove(struct net *net, struct l2tp_tunnel *tunnel)
1273 {
1274 	struct l2tp_net *pn = l2tp_pernet(net);
1275 
1276 	spin_lock_bh(&pn->l2tp_tunnel_idr_lock);
1277 	idr_remove(&pn->l2tp_tunnel_idr, tunnel->tunnel_id);
1278 	spin_unlock_bh(&pn->l2tp_tunnel_idr_lock);
1279 }
1280 
1281 /* Workqueue tunnel deletion function */
1282 static void l2tp_tunnel_del_work(struct work_struct *work)
1283 {
1284 	struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1285 						  del_work);
1286 	struct sock *sk = tunnel->sock;
1287 	struct socket *sock = sk->sk_socket;
1288 
1289 	l2tp_tunnel_closeall(tunnel);
1290 
1291 	/* If the tunnel socket was created within the kernel, use
1292 	 * the sk API to release it here.
1293 	 */
1294 	if (tunnel->fd < 0) {
1295 		if (sock) {
1296 			kernel_sock_shutdown(sock, SHUT_RDWR);
1297 			sock_release(sock);
1298 		}
1299 	}
1300 
1301 	l2tp_tunnel_remove(tunnel->l2tp_net, tunnel);
1302 	/* drop initial ref */
1303 	l2tp_tunnel_dec_refcount(tunnel);
1304 
1305 	/* drop workqueue ref */
1306 	l2tp_tunnel_dec_refcount(tunnel);
1307 }
1308 
1309 /* Create a socket for the tunnel, if one isn't set up by
1310  * userspace. This is used for static tunnels where there is no
1311  * managing L2TP daemon.
1312  *
1313  * Since we don't want these sockets to keep a namespace alive by
1314  * themselves, we drop the socket's namespace refcount after creation.
1315  * These sockets are freed when the namespace exits using the pernet
1316  * exit hook.
1317  */
1318 static int l2tp_tunnel_sock_create(struct net *net,
1319 				   u32 tunnel_id,
1320 				   u32 peer_tunnel_id,
1321 				   struct l2tp_tunnel_cfg *cfg,
1322 				   struct socket **sockp)
1323 {
1324 	int err = -EINVAL;
1325 	struct socket *sock = NULL;
1326 	struct udp_port_cfg udp_conf;
1327 
1328 	switch (cfg->encap) {
1329 	case L2TP_ENCAPTYPE_UDP:
1330 		memset(&udp_conf, 0, sizeof(udp_conf));
1331 
1332 #if IS_ENABLED(CONFIG_IPV6)
1333 		if (cfg->local_ip6 && cfg->peer_ip6) {
1334 			udp_conf.family = AF_INET6;
1335 			memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1336 			       sizeof(udp_conf.local_ip6));
1337 			memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1338 			       sizeof(udp_conf.peer_ip6));
1339 			udp_conf.use_udp6_tx_checksums =
1340 			  !cfg->udp6_zero_tx_checksums;
1341 			udp_conf.use_udp6_rx_checksums =
1342 			  !cfg->udp6_zero_rx_checksums;
1343 		} else
1344 #endif
1345 		{
1346 			udp_conf.family = AF_INET;
1347 			udp_conf.local_ip = cfg->local_ip;
1348 			udp_conf.peer_ip = cfg->peer_ip;
1349 			udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1350 		}
1351 
1352 		udp_conf.local_udp_port = htons(cfg->local_udp_port);
1353 		udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1354 
1355 		err = udp_sock_create(net, &udp_conf, &sock);
1356 		if (err < 0)
1357 			goto out;
1358 
1359 		break;
1360 
1361 	case L2TP_ENCAPTYPE_IP:
1362 #if IS_ENABLED(CONFIG_IPV6)
1363 		if (cfg->local_ip6 && cfg->peer_ip6) {
1364 			struct sockaddr_l2tpip6 ip6_addr = {0};
1365 
1366 			err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1367 					       IPPROTO_L2TP, &sock);
1368 			if (err < 0)
1369 				goto out;
1370 
1371 			ip6_addr.l2tp_family = AF_INET6;
1372 			memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1373 			       sizeof(ip6_addr.l2tp_addr));
1374 			ip6_addr.l2tp_conn_id = tunnel_id;
1375 			err = kernel_bind(sock, (struct sockaddr *)&ip6_addr,
1376 					  sizeof(ip6_addr));
1377 			if (err < 0)
1378 				goto out;
1379 
1380 			ip6_addr.l2tp_family = AF_INET6;
1381 			memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1382 			       sizeof(ip6_addr.l2tp_addr));
1383 			ip6_addr.l2tp_conn_id = peer_tunnel_id;
1384 			err = kernel_connect(sock,
1385 					     (struct sockaddr *)&ip6_addr,
1386 					     sizeof(ip6_addr), 0);
1387 			if (err < 0)
1388 				goto out;
1389 		} else
1390 #endif
1391 		{
1392 			struct sockaddr_l2tpip ip_addr = {0};
1393 
1394 			err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1395 					       IPPROTO_L2TP, &sock);
1396 			if (err < 0)
1397 				goto out;
1398 
1399 			ip_addr.l2tp_family = AF_INET;
1400 			ip_addr.l2tp_addr = cfg->local_ip;
1401 			ip_addr.l2tp_conn_id = tunnel_id;
1402 			err = kernel_bind(sock, (struct sockaddr *)&ip_addr,
1403 					  sizeof(ip_addr));
1404 			if (err < 0)
1405 				goto out;
1406 
1407 			ip_addr.l2tp_family = AF_INET;
1408 			ip_addr.l2tp_addr = cfg->peer_ip;
1409 			ip_addr.l2tp_conn_id = peer_tunnel_id;
1410 			err = kernel_connect(sock, (struct sockaddr *)&ip_addr,
1411 					     sizeof(ip_addr), 0);
1412 			if (err < 0)
1413 				goto out;
1414 		}
1415 		break;
1416 
1417 	default:
1418 		goto out;
1419 	}
1420 
1421 out:
1422 	*sockp = sock;
1423 	if (err < 0 && sock) {
1424 		kernel_sock_shutdown(sock, SHUT_RDWR);
1425 		sock_release(sock);
1426 		*sockp = NULL;
1427 	}
1428 
1429 	return err;
1430 }
1431 
1432 int l2tp_tunnel_create(int fd, int version, u32 tunnel_id, u32 peer_tunnel_id,
1433 		       struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1434 {
1435 	struct l2tp_tunnel *tunnel = NULL;
1436 	int err;
1437 	enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1438 
1439 	if (cfg)
1440 		encap = cfg->encap;
1441 
1442 	tunnel = kzalloc(sizeof(*tunnel), GFP_KERNEL);
1443 	if (!tunnel) {
1444 		err = -ENOMEM;
1445 		goto err;
1446 	}
1447 
1448 	tunnel->version = version;
1449 	tunnel->tunnel_id = tunnel_id;
1450 	tunnel->peer_tunnel_id = peer_tunnel_id;
1451 
1452 	tunnel->magic = L2TP_TUNNEL_MAGIC;
1453 	sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1454 	spin_lock_init(&tunnel->hlist_lock);
1455 	tunnel->acpt_newsess = true;
1456 
1457 	tunnel->encap = encap;
1458 
1459 	refcount_set(&tunnel->ref_count, 1);
1460 	tunnel->fd = fd;
1461 
1462 	/* Init delete workqueue struct */
1463 	INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1464 
1465 	INIT_LIST_HEAD(&tunnel->list);
1466 
1467 	err = 0;
1468 err:
1469 	if (tunnelp)
1470 		*tunnelp = tunnel;
1471 
1472 	return err;
1473 }
1474 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1475 
1476 static int l2tp_validate_socket(const struct sock *sk, const struct net *net,
1477 				enum l2tp_encap_type encap)
1478 {
1479 	if (!net_eq(sock_net(sk), net))
1480 		return -EINVAL;
1481 
1482 	if (sk->sk_type != SOCK_DGRAM)
1483 		return -EPROTONOSUPPORT;
1484 
1485 	if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
1486 		return -EPROTONOSUPPORT;
1487 
1488 	if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) ||
1489 	    (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP))
1490 		return -EPROTONOSUPPORT;
1491 
1492 	if (sk->sk_user_data)
1493 		return -EBUSY;
1494 
1495 	return 0;
1496 }
1497 
1498 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net,
1499 			 struct l2tp_tunnel_cfg *cfg)
1500 {
1501 	struct l2tp_net *pn = l2tp_pernet(net);
1502 	u32 tunnel_id = tunnel->tunnel_id;
1503 	struct socket *sock;
1504 	struct sock *sk;
1505 	int ret;
1506 
1507 	spin_lock_bh(&pn->l2tp_tunnel_idr_lock);
1508 	ret = idr_alloc_u32(&pn->l2tp_tunnel_idr, NULL, &tunnel_id, tunnel_id,
1509 			    GFP_ATOMIC);
1510 	spin_unlock_bh(&pn->l2tp_tunnel_idr_lock);
1511 	if (ret)
1512 		return ret == -ENOSPC ? -EEXIST : ret;
1513 
1514 	if (tunnel->fd < 0) {
1515 		ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id,
1516 					      tunnel->peer_tunnel_id, cfg,
1517 					      &sock);
1518 		if (ret < 0)
1519 			goto err;
1520 	} else {
1521 		sock = sockfd_lookup(tunnel->fd, &ret);
1522 		if (!sock)
1523 			goto err;
1524 	}
1525 
1526 	sk = sock->sk;
1527 	lock_sock(sk);
1528 	write_lock_bh(&sk->sk_callback_lock);
1529 	ret = l2tp_validate_socket(sk, net, tunnel->encap);
1530 	if (ret < 0)
1531 		goto err_inval_sock;
1532 	rcu_assign_sk_user_data(sk, tunnel);
1533 	write_unlock_bh(&sk->sk_callback_lock);
1534 
1535 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1536 		struct udp_tunnel_sock_cfg udp_cfg = {
1537 			.sk_user_data = tunnel,
1538 			.encap_type = UDP_ENCAP_L2TPINUDP,
1539 			.encap_rcv = l2tp_udp_encap_recv,
1540 			.encap_err_rcv = l2tp_udp_encap_err_recv,
1541 			.encap_destroy = l2tp_udp_encap_destroy,
1542 		};
1543 
1544 		setup_udp_tunnel_sock(net, sock, &udp_cfg);
1545 	}
1546 
1547 	tunnel->old_sk_destruct = sk->sk_destruct;
1548 	sk->sk_destruct = &l2tp_tunnel_destruct;
1549 	sk->sk_allocation = GFP_ATOMIC;
1550 	release_sock(sk);
1551 
1552 	sock_hold(sk);
1553 	tunnel->sock = sk;
1554 	tunnel->l2tp_net = net;
1555 
1556 	spin_lock_bh(&pn->l2tp_tunnel_idr_lock);
1557 	idr_replace(&pn->l2tp_tunnel_idr, tunnel, tunnel->tunnel_id);
1558 	spin_unlock_bh(&pn->l2tp_tunnel_idr_lock);
1559 
1560 	trace_register_tunnel(tunnel);
1561 
1562 	if (tunnel->fd >= 0)
1563 		sockfd_put(sock);
1564 
1565 	return 0;
1566 
1567 err_inval_sock:
1568 	write_unlock_bh(&sk->sk_callback_lock);
1569 	release_sock(sk);
1570 
1571 	if (tunnel->fd < 0)
1572 		sock_release(sock);
1573 	else
1574 		sockfd_put(sock);
1575 err:
1576 	l2tp_tunnel_remove(net, tunnel);
1577 	return ret;
1578 }
1579 EXPORT_SYMBOL_GPL(l2tp_tunnel_register);
1580 
1581 /* This function is used by the netlink TUNNEL_DELETE command.
1582  */
1583 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1584 {
1585 	if (!test_and_set_bit(0, &tunnel->dead)) {
1586 		trace_delete_tunnel(tunnel);
1587 		l2tp_tunnel_inc_refcount(tunnel);
1588 		queue_work(l2tp_wq, &tunnel->del_work);
1589 	}
1590 }
1591 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1592 
1593 void l2tp_session_delete(struct l2tp_session *session)
1594 {
1595 	if (test_and_set_bit(0, &session->dead))
1596 		return;
1597 
1598 	trace_delete_session(session);
1599 	l2tp_session_unhash(session);
1600 	l2tp_session_queue_purge(session);
1601 	if (session->session_close)
1602 		(*session->session_close)(session);
1603 
1604 	l2tp_session_dec_refcount(session);
1605 }
1606 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1607 
1608 /* We come here whenever a session's send_seq, cookie_len or
1609  * l2specific_type parameters are set.
1610  */
1611 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1612 {
1613 	if (version == L2TP_HDR_VER_2) {
1614 		session->hdr_len = 6;
1615 		if (session->send_seq)
1616 			session->hdr_len += 4;
1617 	} else {
1618 		session->hdr_len = 4 + session->cookie_len;
1619 		session->hdr_len += l2tp_get_l2specific_len(session);
1620 		if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1621 			session->hdr_len += 4;
1622 	}
1623 }
1624 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1625 
1626 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id,
1627 					 u32 peer_session_id, struct l2tp_session_cfg *cfg)
1628 {
1629 	struct l2tp_session *session;
1630 
1631 	session = kzalloc(sizeof(*session) + priv_size, GFP_KERNEL);
1632 	if (session) {
1633 		session->magic = L2TP_SESSION_MAGIC;
1634 		session->tunnel = tunnel;
1635 
1636 		session->session_id = session_id;
1637 		session->peer_session_id = peer_session_id;
1638 		session->nr = 0;
1639 		if (tunnel->version == L2TP_HDR_VER_2)
1640 			session->nr_max = 0xffff;
1641 		else
1642 			session->nr_max = 0xffffff;
1643 		session->nr_window_size = session->nr_max / 2;
1644 		session->nr_oos_count_max = 4;
1645 
1646 		/* Use NR of first received packet */
1647 		session->reorder_skip = 1;
1648 
1649 		sprintf(&session->name[0], "sess %u/%u",
1650 			tunnel->tunnel_id, session->session_id);
1651 
1652 		skb_queue_head_init(&session->reorder_q);
1653 
1654 		INIT_HLIST_NODE(&session->hlist);
1655 		INIT_HLIST_NODE(&session->global_hlist);
1656 
1657 		if (cfg) {
1658 			session->pwtype = cfg->pw_type;
1659 			session->send_seq = cfg->send_seq;
1660 			session->recv_seq = cfg->recv_seq;
1661 			session->lns_mode = cfg->lns_mode;
1662 			session->reorder_timeout = cfg->reorder_timeout;
1663 			session->l2specific_type = cfg->l2specific_type;
1664 			session->cookie_len = cfg->cookie_len;
1665 			memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1666 			session->peer_cookie_len = cfg->peer_cookie_len;
1667 			memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1668 		}
1669 
1670 		l2tp_session_set_header_len(session, tunnel->version);
1671 
1672 		refcount_set(&session->ref_count, 1);
1673 
1674 		return session;
1675 	}
1676 
1677 	return ERR_PTR(-ENOMEM);
1678 }
1679 EXPORT_SYMBOL_GPL(l2tp_session_create);
1680 
1681 /*****************************************************************************
1682  * Init and cleanup
1683  *****************************************************************************/
1684 
1685 static __net_init int l2tp_init_net(struct net *net)
1686 {
1687 	struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1688 	int hash;
1689 
1690 	idr_init(&pn->l2tp_tunnel_idr);
1691 	spin_lock_init(&pn->l2tp_tunnel_idr_lock);
1692 
1693 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1694 		INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1695 
1696 	spin_lock_init(&pn->l2tp_session_hlist_lock);
1697 
1698 	return 0;
1699 }
1700 
1701 static __net_exit void l2tp_exit_net(struct net *net)
1702 {
1703 	struct l2tp_net *pn = l2tp_pernet(net);
1704 	struct l2tp_tunnel *tunnel = NULL;
1705 	unsigned long tunnel_id, tmp;
1706 	int hash;
1707 
1708 	rcu_read_lock_bh();
1709 	idr_for_each_entry_ul(&pn->l2tp_tunnel_idr, tunnel, tmp, tunnel_id) {
1710 		if (tunnel)
1711 			l2tp_tunnel_delete(tunnel);
1712 	}
1713 	rcu_read_unlock_bh();
1714 
1715 	if (l2tp_wq)
1716 		flush_workqueue(l2tp_wq);
1717 	rcu_barrier();
1718 
1719 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1720 		WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1721 	idr_destroy(&pn->l2tp_tunnel_idr);
1722 }
1723 
1724 static struct pernet_operations l2tp_net_ops = {
1725 	.init = l2tp_init_net,
1726 	.exit = l2tp_exit_net,
1727 	.id   = &l2tp_net_id,
1728 	.size = sizeof(struct l2tp_net),
1729 };
1730 
1731 static int __init l2tp_init(void)
1732 {
1733 	int rc = 0;
1734 
1735 	rc = register_pernet_device(&l2tp_net_ops);
1736 	if (rc)
1737 		goto out;
1738 
1739 	l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1740 	if (!l2tp_wq) {
1741 		pr_err("alloc_workqueue failed\n");
1742 		unregister_pernet_device(&l2tp_net_ops);
1743 		rc = -ENOMEM;
1744 		goto out;
1745 	}
1746 
1747 	pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1748 
1749 out:
1750 	return rc;
1751 }
1752 
1753 static void __exit l2tp_exit(void)
1754 {
1755 	unregister_pernet_device(&l2tp_net_ops);
1756 	if (l2tp_wq) {
1757 		destroy_workqueue(l2tp_wq);
1758 		l2tp_wq = NULL;
1759 	}
1760 }
1761 
1762 module_init(l2tp_init);
1763 module_exit(l2tp_exit);
1764 
1765 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1766 MODULE_DESCRIPTION("L2TP core");
1767 MODULE_LICENSE("GPL");
1768 MODULE_VERSION(L2TP_DRV_VERSION);
1769