Merge tag 'regmap-v4.17' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie...
[linux] / net / l2tp / l2tp_core.c
1 /*
2  * L2TP core.
3  *
4  * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
5  *
6  * This file contains some code of the original L2TPv2 pppol2tp
7  * driver, which has the following copyright:
8  *
9  * Authors:     Martijn van Oosterhout <kleptog@svana.org>
10  *              James Chapman (jchapman@katalix.com)
11  * Contributors:
12  *              Michal Ostrowski <mostrows@speakeasy.net>
13  *              Arnaldo Carvalho de Melo <acme@xconectiva.com.br>
14  *              David S. Miller (davem@redhat.com)
15  *
16  * This program is free software; you can redistribute it and/or modify
17  * it under the terms of the GNU General Public License version 2 as
18  * published by the Free Software Foundation.
19  */
20
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22
23 #include <linux/module.h>
24 #include <linux/string.h>
25 #include <linux/list.h>
26 #include <linux/rculist.h>
27 #include <linux/uaccess.h>
28
29 #include <linux/kernel.h>
30 #include <linux/spinlock.h>
31 #include <linux/kthread.h>
32 #include <linux/sched.h>
33 #include <linux/slab.h>
34 #include <linux/errno.h>
35 #include <linux/jiffies.h>
36
37 #include <linux/netdevice.h>
38 #include <linux/net.h>
39 #include <linux/inetdevice.h>
40 #include <linux/skbuff.h>
41 #include <linux/init.h>
42 #include <linux/in.h>
43 #include <linux/ip.h>
44 #include <linux/udp.h>
45 #include <linux/l2tp.h>
46 #include <linux/hash.h>
47 #include <linux/sort.h>
48 #include <linux/file.h>
49 #include <linux/nsproxy.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/dst.h>
53 #include <net/ip.h>
54 #include <net/udp.h>
55 #include <net/udp_tunnel.h>
56 #include <net/inet_common.h>
57 #include <net/xfrm.h>
58 #include <net/protocol.h>
59 #include <net/inet6_connection_sock.h>
60 #include <net/inet_ecn.h>
61 #include <net/ip6_route.h>
62 #include <net/ip6_checksum.h>
63
64 #include <asm/byteorder.h>
65 #include <linux/atomic.h>
66
67 #include "l2tp_core.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_SEQ               10
87 #define L2TP_HDR_SIZE_NOSEQ             6
88
89 /* Default trace flags */
90 #define L2TP_DEFAULT_DEBUG_FLAGS        0
91
92 /* Private data stored for received packets in the skb.
93  */
94 struct l2tp_skb_cb {
95         u32                     ns;
96         u16                     has_seq;
97         u16                     length;
98         unsigned long           expires;
99 };
100
101 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
102
103 static struct workqueue_struct *l2tp_wq;
104
105 /* per-net private data for this module */
106 static unsigned int l2tp_net_id;
107 struct l2tp_net {
108         struct list_head l2tp_tunnel_list;
109         spinlock_t l2tp_tunnel_list_lock;
110         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
111         spinlock_t l2tp_session_hlist_lock;
112 };
113
114 #if IS_ENABLED(CONFIG_IPV6)
115 static bool l2tp_sk_is_v6(struct sock *sk)
116 {
117         return sk->sk_family == PF_INET6 &&
118                !ipv6_addr_v4mapped(&sk->sk_v6_daddr);
119 }
120 #endif
121
122 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
123 {
124         return sk->sk_user_data;
125 }
126
127 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
128 {
129         BUG_ON(!net);
130
131         return net_generic(net, l2tp_net_id);
132 }
133
134 /* Session hash global list for L2TPv3.
135  * The session_id SHOULD be random according to RFC3931, but several
136  * L2TP implementations use incrementing session_ids.  So we do a real
137  * hash on the session_id, rather than a simple bitmask.
138  */
139 static inline struct hlist_head *
140 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
141 {
142         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
143
144 }
145
146 /* Session hash list.
147  * The session_id SHOULD be random according to RFC2661, but several
148  * L2TP implementations (Cisco and Microsoft) use incrementing
149  * session_ids.  So we do a real hash on the session_id, rather than a
150  * simple bitmask.
151  */
152 static inline struct hlist_head *
153 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
154 {
155         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
156 }
157
158 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
159 {
160         sock_put(tunnel->sock);
161         /* the tunnel is freed in the socket destructor */
162 }
163 EXPORT_SYMBOL(l2tp_tunnel_free);
164
165 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
166 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
167 {
168         const struct l2tp_net *pn = l2tp_pernet(net);
169         struct l2tp_tunnel *tunnel;
170
171         rcu_read_lock_bh();
172         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
173                 if (tunnel->tunnel_id == tunnel_id) {
174                         l2tp_tunnel_inc_refcount(tunnel);
175                         rcu_read_unlock_bh();
176
177                         return tunnel;
178                 }
179         }
180         rcu_read_unlock_bh();
181
182         return NULL;
183 }
184 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
185
186 /* Lookup a session. A new reference is held on the returned session. */
187 struct l2tp_session *l2tp_session_get(const struct net *net,
188                                       struct l2tp_tunnel *tunnel,
189                                       u32 session_id)
190 {
191         struct hlist_head *session_list;
192         struct l2tp_session *session;
193
194         if (!tunnel) {
195                 struct l2tp_net *pn = l2tp_pernet(net);
196
197                 session_list = l2tp_session_id_hash_2(pn, session_id);
198
199                 rcu_read_lock_bh();
200                 hlist_for_each_entry_rcu(session, session_list, global_hlist) {
201                         if (session->session_id == session_id) {
202                                 l2tp_session_inc_refcount(session);
203                                 rcu_read_unlock_bh();
204
205                                 return session;
206                         }
207                 }
208                 rcu_read_unlock_bh();
209
210                 return NULL;
211         }
212
213         session_list = l2tp_session_id_hash(tunnel, session_id);
214         read_lock_bh(&tunnel->hlist_lock);
215         hlist_for_each_entry(session, session_list, hlist) {
216                 if (session->session_id == session_id) {
217                         l2tp_session_inc_refcount(session);
218                         read_unlock_bh(&tunnel->hlist_lock);
219
220                         return session;
221                 }
222         }
223         read_unlock_bh(&tunnel->hlist_lock);
224
225         return NULL;
226 }
227 EXPORT_SYMBOL_GPL(l2tp_session_get);
228
229 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth)
230 {
231         int hash;
232         struct l2tp_session *session;
233         int count = 0;
234
235         read_lock_bh(&tunnel->hlist_lock);
236         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
237                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
238                         if (++count > nth) {
239                                 l2tp_session_inc_refcount(session);
240                                 read_unlock_bh(&tunnel->hlist_lock);
241                                 return session;
242                         }
243                 }
244         }
245
246         read_unlock_bh(&tunnel->hlist_lock);
247
248         return NULL;
249 }
250 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
251
252 /* Lookup a session by interface name.
253  * This is very inefficient but is only used by management interfaces.
254  */
255 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
256                                                 const char *ifname)
257 {
258         struct l2tp_net *pn = l2tp_pernet(net);
259         int hash;
260         struct l2tp_session *session;
261
262         rcu_read_lock_bh();
263         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
264                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
265                         if (!strcmp(session->ifname, ifname)) {
266                                 l2tp_session_inc_refcount(session);
267                                 rcu_read_unlock_bh();
268
269                                 return session;
270                         }
271                 }
272         }
273
274         rcu_read_unlock_bh();
275
276         return NULL;
277 }
278 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
279
280 int l2tp_session_register(struct l2tp_session *session,
281                           struct l2tp_tunnel *tunnel)
282 {
283         struct l2tp_session *session_walk;
284         struct hlist_head *g_head;
285         struct hlist_head *head;
286         struct l2tp_net *pn;
287         int err;
288
289         head = l2tp_session_id_hash(tunnel, session->session_id);
290
291         write_lock_bh(&tunnel->hlist_lock);
292         if (!tunnel->acpt_newsess) {
293                 err = -ENODEV;
294                 goto err_tlock;
295         }
296
297         hlist_for_each_entry(session_walk, head, hlist)
298                 if (session_walk->session_id == session->session_id) {
299                         err = -EEXIST;
300                         goto err_tlock;
301                 }
302
303         if (tunnel->version == L2TP_HDR_VER_3) {
304                 pn = l2tp_pernet(tunnel->l2tp_net);
305                 g_head = l2tp_session_id_hash_2(l2tp_pernet(tunnel->l2tp_net),
306                                                 session->session_id);
307
308                 spin_lock_bh(&pn->l2tp_session_hlist_lock);
309
310                 hlist_for_each_entry(session_walk, g_head, global_hlist)
311                         if (session_walk->session_id == session->session_id) {
312                                 err = -EEXIST;
313                                 goto err_tlock_pnlock;
314                         }
315
316                 l2tp_tunnel_inc_refcount(tunnel);
317                 hlist_add_head_rcu(&session->global_hlist, g_head);
318
319                 spin_unlock_bh(&pn->l2tp_session_hlist_lock);
320         } else {
321                 l2tp_tunnel_inc_refcount(tunnel);
322         }
323
324         hlist_add_head(&session->hlist, head);
325         write_unlock_bh(&tunnel->hlist_lock);
326
327         return 0;
328
329 err_tlock_pnlock:
330         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
331 err_tlock:
332         write_unlock_bh(&tunnel->hlist_lock);
333
334         return err;
335 }
336 EXPORT_SYMBOL_GPL(l2tp_session_register);
337
338 /* Lookup a tunnel by id
339  */
340 struct l2tp_tunnel *l2tp_tunnel_find(const struct net *net, u32 tunnel_id)
341 {
342         struct l2tp_tunnel *tunnel;
343         struct l2tp_net *pn = l2tp_pernet(net);
344
345         rcu_read_lock_bh();
346         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
347                 if (tunnel->tunnel_id == tunnel_id) {
348                         rcu_read_unlock_bh();
349                         return tunnel;
350                 }
351         }
352         rcu_read_unlock_bh();
353
354         return NULL;
355 }
356 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
357
358 struct l2tp_tunnel *l2tp_tunnel_find_nth(const struct net *net, int nth)
359 {
360         struct l2tp_net *pn = l2tp_pernet(net);
361         struct l2tp_tunnel *tunnel;
362         int count = 0;
363
364         rcu_read_lock_bh();
365         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
366                 if (++count > nth) {
367                         rcu_read_unlock_bh();
368                         return tunnel;
369                 }
370         }
371
372         rcu_read_unlock_bh();
373
374         return NULL;
375 }
376 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
377
378 /*****************************************************************************
379  * Receive data handling
380  *****************************************************************************/
381
382 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
383  * number.
384  */
385 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
386 {
387         struct sk_buff *skbp;
388         struct sk_buff *tmp;
389         u32 ns = L2TP_SKB_CB(skb)->ns;
390
391         spin_lock_bh(&session->reorder_q.lock);
392         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
393                 if (L2TP_SKB_CB(skbp)->ns > ns) {
394                         __skb_queue_before(&session->reorder_q, skbp, skb);
395                         l2tp_dbg(session, L2TP_MSG_SEQ,
396                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
397                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
398                                  skb_queue_len(&session->reorder_q));
399                         atomic_long_inc(&session->stats.rx_oos_packets);
400                         goto out;
401                 }
402         }
403
404         __skb_queue_tail(&session->reorder_q, skb);
405
406 out:
407         spin_unlock_bh(&session->reorder_q.lock);
408 }
409
410 /* Dequeue a single skb.
411  */
412 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
413 {
414         struct l2tp_tunnel *tunnel = session->tunnel;
415         int length = L2TP_SKB_CB(skb)->length;
416
417         /* We're about to requeue the skb, so return resources
418          * to its current owner (a socket receive buffer).
419          */
420         skb_orphan(skb);
421
422         atomic_long_inc(&tunnel->stats.rx_packets);
423         atomic_long_add(length, &tunnel->stats.rx_bytes);
424         atomic_long_inc(&session->stats.rx_packets);
425         atomic_long_add(length, &session->stats.rx_bytes);
426
427         if (L2TP_SKB_CB(skb)->has_seq) {
428                 /* Bump our Nr */
429                 session->nr++;
430                 session->nr &= session->nr_max;
431
432                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
433                          session->name, session->nr);
434         }
435
436         /* call private receive handler */
437         if (session->recv_skb != NULL)
438                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
439         else
440                 kfree_skb(skb);
441 }
442
443 /* Dequeue skbs from the session's reorder_q, subject to packet order.
444  * Skbs that have been in the queue for too long are simply discarded.
445  */
446 static void l2tp_recv_dequeue(struct l2tp_session *session)
447 {
448         struct sk_buff *skb;
449         struct sk_buff *tmp;
450
451         /* If the pkt at the head of the queue has the nr that we
452          * expect to send up next, dequeue it and any other
453          * in-sequence packets behind it.
454          */
455 start:
456         spin_lock_bh(&session->reorder_q.lock);
457         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
458                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
459                         atomic_long_inc(&session->stats.rx_seq_discards);
460                         atomic_long_inc(&session->stats.rx_errors);
461                         l2tp_dbg(session, L2TP_MSG_SEQ,
462                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
463                                  session->name, L2TP_SKB_CB(skb)->ns,
464                                  L2TP_SKB_CB(skb)->length, session->nr,
465                                  skb_queue_len(&session->reorder_q));
466                         session->reorder_skip = 1;
467                         __skb_unlink(skb, &session->reorder_q);
468                         kfree_skb(skb);
469                         continue;
470                 }
471
472                 if (L2TP_SKB_CB(skb)->has_seq) {
473                         if (session->reorder_skip) {
474                                 l2tp_dbg(session, L2TP_MSG_SEQ,
475                                          "%s: advancing nr to next pkt: %u -> %u",
476                                          session->name, session->nr,
477                                          L2TP_SKB_CB(skb)->ns);
478                                 session->reorder_skip = 0;
479                                 session->nr = L2TP_SKB_CB(skb)->ns;
480                         }
481                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
482                                 l2tp_dbg(session, L2TP_MSG_SEQ,
483                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
484                                          session->name, L2TP_SKB_CB(skb)->ns,
485                                          L2TP_SKB_CB(skb)->length, session->nr,
486                                          skb_queue_len(&session->reorder_q));
487                                 goto out;
488                         }
489                 }
490                 __skb_unlink(skb, &session->reorder_q);
491
492                 /* Process the skb. We release the queue lock while we
493                  * do so to let other contexts process the queue.
494                  */
495                 spin_unlock_bh(&session->reorder_q.lock);
496                 l2tp_recv_dequeue_skb(session, skb);
497                 goto start;
498         }
499
500 out:
501         spin_unlock_bh(&session->reorder_q.lock);
502 }
503
504 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
505 {
506         u32 nws;
507
508         if (nr >= session->nr)
509                 nws = nr - session->nr;
510         else
511                 nws = (session->nr_max + 1) - (session->nr - nr);
512
513         return nws < session->nr_window_size;
514 }
515
516 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
517  * acceptable, else non-zero.
518  */
519 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
520 {
521         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
522                 /* Packet sequence number is outside allowed window.
523                  * Discard it.
524                  */
525                 l2tp_dbg(session, L2TP_MSG_SEQ,
526                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
527                          session->name, L2TP_SKB_CB(skb)->ns,
528                          L2TP_SKB_CB(skb)->length, session->nr);
529                 goto discard;
530         }
531
532         if (session->reorder_timeout != 0) {
533                 /* Packet reordering enabled. Add skb to session's
534                  * reorder queue, in order of ns.
535                  */
536                 l2tp_recv_queue_skb(session, skb);
537                 goto out;
538         }
539
540         /* Packet reordering disabled. Discard out-of-sequence packets, while
541          * tracking the number if in-sequence packets after the first OOS packet
542          * is seen. After nr_oos_count_max in-sequence packets, reset the
543          * sequence number to re-enable packet reception.
544          */
545         if (L2TP_SKB_CB(skb)->ns == session->nr) {
546                 skb_queue_tail(&session->reorder_q, skb);
547         } else {
548                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
549                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
550
551                 if (nr_oos == nr_next)
552                         session->nr_oos_count++;
553                 else
554                         session->nr_oos_count = 0;
555
556                 session->nr_oos = nr_oos;
557                 if (session->nr_oos_count > session->nr_oos_count_max) {
558                         session->reorder_skip = 1;
559                         l2tp_dbg(session, L2TP_MSG_SEQ,
560                                  "%s: %d oos packets received. Resetting sequence numbers\n",
561                                  session->name, session->nr_oos_count);
562                 }
563                 if (!session->reorder_skip) {
564                         atomic_long_inc(&session->stats.rx_seq_discards);
565                         l2tp_dbg(session, L2TP_MSG_SEQ,
566                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
567                                  session->name, L2TP_SKB_CB(skb)->ns,
568                                  L2TP_SKB_CB(skb)->length, session->nr,
569                                  skb_queue_len(&session->reorder_q));
570                         goto discard;
571                 }
572                 skb_queue_tail(&session->reorder_q, skb);
573         }
574
575 out:
576         return 0;
577
578 discard:
579         return 1;
580 }
581
582 /* Do receive processing of L2TP data frames. We handle both L2TPv2
583  * and L2TPv3 data frames here.
584  *
585  * L2TPv2 Data Message Header
586  *
587  *  0                   1                   2                   3
588  *  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
589  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
590  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
591  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
592  * |           Tunnel ID           |           Session ID          |
593  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
594  * |             Ns (opt)          |             Nr (opt)          |
595  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
596  * |      Offset Size (opt)        |    Offset pad... (opt)
597  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
598  *
599  * Data frames are marked by T=0. All other fields are the same as
600  * those in L2TP control frames.
601  *
602  * L2TPv3 Data Message Header
603  *
604  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
605  * |                      L2TP Session Header                      |
606  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
607  * |                      L2-Specific Sublayer                     |
608  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
609  * |                        Tunnel Payload                      ...
610  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
611  *
612  * L2TPv3 Session Header Over IP
613  *
614  *  0                   1                   2                   3
615  *  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
616  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
617  * |                           Session ID                          |
618  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
619  * |               Cookie (optional, maximum 64 bits)...
620  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
621  *                                                                 |
622  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
623  *
624  * L2TPv3 L2-Specific Sublayer Format
625  *
626  *  0                   1                   2                   3
627  *  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
628  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
629  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
630  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
631  *
632  * Cookie value and sublayer format are negotiated with the peer when
633  * the session is set up. Unlike L2TPv2, we do not need to parse the
634  * packet header to determine if optional fields are present.
635  *
636  * Caller must already have parsed the frame and determined that it is
637  * a data (not control) frame before coming here. Fields up to the
638  * session-id have already been parsed and ptr points to the data
639  * after the session-id.
640  */
641 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
642                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
643                       int length, int (*payload_hook)(struct sk_buff *skb))
644 {
645         struct l2tp_tunnel *tunnel = session->tunnel;
646         int offset;
647         u32 ns, nr;
648
649         /* Parse and check optional cookie */
650         if (session->peer_cookie_len > 0) {
651                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
652                         l2tp_info(tunnel, L2TP_MSG_DATA,
653                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
654                                   tunnel->name, tunnel->tunnel_id,
655                                   session->session_id);
656                         atomic_long_inc(&session->stats.rx_cookie_discards);
657                         goto discard;
658                 }
659                 ptr += session->peer_cookie_len;
660         }
661
662         /* Handle the optional sequence numbers. Sequence numbers are
663          * in different places for L2TPv2 and L2TPv3.
664          *
665          * If we are the LAC, enable/disable sequence numbers under
666          * the control of the LNS.  If no sequence numbers present but
667          * we were expecting them, discard frame.
668          */
669         ns = nr = 0;
670         L2TP_SKB_CB(skb)->has_seq = 0;
671         if (tunnel->version == L2TP_HDR_VER_2) {
672                 if (hdrflags & L2TP_HDRFLAG_S) {
673                         ns = ntohs(*(__be16 *) ptr);
674                         ptr += 2;
675                         nr = ntohs(*(__be16 *) ptr);
676                         ptr += 2;
677
678                         /* Store L2TP info in the skb */
679                         L2TP_SKB_CB(skb)->ns = ns;
680                         L2TP_SKB_CB(skb)->has_seq = 1;
681
682                         l2tp_dbg(session, L2TP_MSG_SEQ,
683                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
684                                  session->name, ns, nr, session->nr);
685                 }
686         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
687                 u32 l2h = ntohl(*(__be32 *) ptr);
688
689                 if (l2h & 0x40000000) {
690                         ns = l2h & 0x00ffffff;
691
692                         /* Store L2TP info in the skb */
693                         L2TP_SKB_CB(skb)->ns = ns;
694                         L2TP_SKB_CB(skb)->has_seq = 1;
695
696                         l2tp_dbg(session, L2TP_MSG_SEQ,
697                                  "%s: recv data ns=%u, session nr=%u\n",
698                                  session->name, ns, session->nr);
699                 }
700                 ptr += 4;
701         }
702
703         if (L2TP_SKB_CB(skb)->has_seq) {
704                 /* Received a packet with sequence numbers. If we're the LNS,
705                  * check if we sre sending sequence numbers and if not,
706                  * configure it so.
707                  */
708                 if ((!session->lns_mode) && (!session->send_seq)) {
709                         l2tp_info(session, L2TP_MSG_SEQ,
710                                   "%s: requested to enable seq numbers by LNS\n",
711                                   session->name);
712                         session->send_seq = 1;
713                         l2tp_session_set_header_len(session, tunnel->version);
714                 }
715         } else {
716                 /* No sequence numbers.
717                  * If user has configured mandatory sequence numbers, discard.
718                  */
719                 if (session->recv_seq) {
720                         l2tp_warn(session, L2TP_MSG_SEQ,
721                                   "%s: recv data has no seq numbers when required. Discarding.\n",
722                                   session->name);
723                         atomic_long_inc(&session->stats.rx_seq_discards);
724                         goto discard;
725                 }
726
727                 /* If we're the LAC and we're sending sequence numbers, the
728                  * LNS has requested that we no longer send sequence numbers.
729                  * If we're the LNS and we're sending sequence numbers, the
730                  * LAC is broken. Discard the frame.
731                  */
732                 if ((!session->lns_mode) && (session->send_seq)) {
733                         l2tp_info(session, L2TP_MSG_SEQ,
734                                   "%s: requested to disable seq numbers by LNS\n",
735                                   session->name);
736                         session->send_seq = 0;
737                         l2tp_session_set_header_len(session, tunnel->version);
738                 } else if (session->send_seq) {
739                         l2tp_warn(session, L2TP_MSG_SEQ,
740                                   "%s: recv data has no seq numbers when required. Discarding.\n",
741                                   session->name);
742                         atomic_long_inc(&session->stats.rx_seq_discards);
743                         goto discard;
744                 }
745         }
746
747         /* Session data offset is defined only for L2TPv2 and is
748          * indicated by an optional 16-bit value in the header.
749          */
750         if (tunnel->version == L2TP_HDR_VER_2) {
751                 /* If offset bit set, skip it. */
752                 if (hdrflags & L2TP_HDRFLAG_O) {
753                         offset = ntohs(*(__be16 *)ptr);
754                         ptr += 2 + offset;
755                 }
756         }
757
758         offset = ptr - optr;
759         if (!pskb_may_pull(skb, offset))
760                 goto discard;
761
762         __skb_pull(skb, offset);
763
764         /* If caller wants to process the payload before we queue the
765          * packet, do so now.
766          */
767         if (payload_hook)
768                 if ((*payload_hook)(skb))
769                         goto discard;
770
771         /* Prepare skb for adding to the session's reorder_q.  Hold
772          * packets for max reorder_timeout or 1 second if not
773          * reordering.
774          */
775         L2TP_SKB_CB(skb)->length = length;
776         L2TP_SKB_CB(skb)->expires = jiffies +
777                 (session->reorder_timeout ? session->reorder_timeout : HZ);
778
779         /* Add packet to the session's receive queue. Reordering is done here, if
780          * enabled. Saved L2TP protocol info is stored in skb->sb[].
781          */
782         if (L2TP_SKB_CB(skb)->has_seq) {
783                 if (l2tp_recv_data_seq(session, skb))
784                         goto discard;
785         } else {
786                 /* No sequence numbers. Add the skb to the tail of the
787                  * reorder queue. This ensures that it will be
788                  * delivered after all previous sequenced skbs.
789                  */
790                 skb_queue_tail(&session->reorder_q, skb);
791         }
792
793         /* Try to dequeue as many skbs from reorder_q as we can. */
794         l2tp_recv_dequeue(session);
795
796         return;
797
798 discard:
799         atomic_long_inc(&session->stats.rx_errors);
800         kfree_skb(skb);
801 }
802 EXPORT_SYMBOL(l2tp_recv_common);
803
804 /* Drop skbs from the session's reorder_q
805  */
806 int l2tp_session_queue_purge(struct l2tp_session *session)
807 {
808         struct sk_buff *skb = NULL;
809         BUG_ON(!session);
810         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
811         while ((skb = skb_dequeue(&session->reorder_q))) {
812                 atomic_long_inc(&session->stats.rx_errors);
813                 kfree_skb(skb);
814         }
815         return 0;
816 }
817 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
818
819 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
820  * here. The skb is not on a list when we get here.
821  * Returns 0 if the packet was a data packet and was successfully passed on.
822  * Returns 1 if the packet was not a good data packet and could not be
823  * forwarded.  All such packets are passed up to userspace to deal with.
824  */
825 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
826                               int (*payload_hook)(struct sk_buff *skb))
827 {
828         struct l2tp_session *session = NULL;
829         unsigned char *ptr, *optr;
830         u16 hdrflags;
831         u32 tunnel_id, session_id;
832         u16 version;
833         int length;
834
835         /* UDP has verifed checksum */
836
837         /* UDP always verifies the packet length. */
838         __skb_pull(skb, sizeof(struct udphdr));
839
840         /* Short packet? */
841         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
842                 l2tp_info(tunnel, L2TP_MSG_DATA,
843                           "%s: recv short packet (len=%d)\n",
844                           tunnel->name, skb->len);
845                 goto error;
846         }
847
848         /* Trace packet contents, if enabled */
849         if (tunnel->debug & L2TP_MSG_DATA) {
850                 length = min(32u, skb->len);
851                 if (!pskb_may_pull(skb, length))
852                         goto error;
853
854                 pr_debug("%s: recv\n", tunnel->name);
855                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
856         }
857
858         /* Point to L2TP header */
859         optr = ptr = skb->data;
860
861         /* Get L2TP header flags */
862         hdrflags = ntohs(*(__be16 *) ptr);
863
864         /* Check protocol version */
865         version = hdrflags & L2TP_HDR_VER_MASK;
866         if (version != tunnel->version) {
867                 l2tp_info(tunnel, L2TP_MSG_DATA,
868                           "%s: recv protocol version mismatch: got %d expected %d\n",
869                           tunnel->name, version, tunnel->version);
870                 goto error;
871         }
872
873         /* Get length of L2TP packet */
874         length = skb->len;
875
876         /* If type is control packet, it is handled by userspace. */
877         if (hdrflags & L2TP_HDRFLAG_T) {
878                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
879                          "%s: recv control packet, len=%d\n",
880                          tunnel->name, length);
881                 goto error;
882         }
883
884         /* Skip flags */
885         ptr += 2;
886
887         if (tunnel->version == L2TP_HDR_VER_2) {
888                 /* If length is present, skip it */
889                 if (hdrflags & L2TP_HDRFLAG_L)
890                         ptr += 2;
891
892                 /* Extract tunnel and session ID */
893                 tunnel_id = ntohs(*(__be16 *) ptr);
894                 ptr += 2;
895                 session_id = ntohs(*(__be16 *) ptr);
896                 ptr += 2;
897         } else {
898                 ptr += 2;       /* skip reserved bits */
899                 tunnel_id = tunnel->tunnel_id;
900                 session_id = ntohl(*(__be32 *) ptr);
901                 ptr += 4;
902         }
903
904         /* Find the session context */
905         session = l2tp_session_get(tunnel->l2tp_net, tunnel, session_id);
906         if (!session || !session->recv_skb) {
907                 if (session)
908                         l2tp_session_dec_refcount(session);
909
910                 /* Not found? Pass to userspace to deal with */
911                 l2tp_info(tunnel, L2TP_MSG_DATA,
912                           "%s: no session found (%u/%u). Passing up.\n",
913                           tunnel->name, tunnel_id, session_id);
914                 goto error;
915         }
916
917         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
918         l2tp_session_dec_refcount(session);
919
920         return 0;
921
922 error:
923         /* Put UDP header back */
924         __skb_push(skb, sizeof(struct udphdr));
925
926         return 1;
927 }
928
929 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
930  * Return codes:
931  * 0 : success.
932  * <0: error
933  * >0: skb should be passed up to userspace as UDP.
934  */
935 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
936 {
937         struct l2tp_tunnel *tunnel;
938
939         tunnel = l2tp_tunnel(sk);
940         if (tunnel == NULL)
941                 goto pass_up;
942
943         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
944                  tunnel->name, skb->len);
945
946         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
947                 goto pass_up;
948
949         return 0;
950
951 pass_up:
952         return 1;
953 }
954 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
955
956 /************************************************************************
957  * Transmit handling
958  ***********************************************************************/
959
960 /* Build an L2TP header for the session into the buffer provided.
961  */
962 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
963 {
964         struct l2tp_tunnel *tunnel = session->tunnel;
965         __be16 *bufp = buf;
966         __be16 *optr = buf;
967         u16 flags = L2TP_HDR_VER_2;
968         u32 tunnel_id = tunnel->peer_tunnel_id;
969         u32 session_id = session->peer_session_id;
970
971         if (session->send_seq)
972                 flags |= L2TP_HDRFLAG_S;
973
974         /* Setup L2TP header. */
975         *bufp++ = htons(flags);
976         *bufp++ = htons(tunnel_id);
977         *bufp++ = htons(session_id);
978         if (session->send_seq) {
979                 *bufp++ = htons(session->ns);
980                 *bufp++ = 0;
981                 session->ns++;
982                 session->ns &= 0xffff;
983                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
984                          session->name, session->ns);
985         }
986
987         return bufp - optr;
988 }
989
990 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
991 {
992         struct l2tp_tunnel *tunnel = session->tunnel;
993         char *bufp = buf;
994         char *optr = bufp;
995
996         /* Setup L2TP header. The header differs slightly for UDP and
997          * IP encapsulations. For UDP, there is 4 bytes of flags.
998          */
999         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1000                 u16 flags = L2TP_HDR_VER_3;
1001                 *((__be16 *) bufp) = htons(flags);
1002                 bufp += 2;
1003                 *((__be16 *) bufp) = 0;
1004                 bufp += 2;
1005         }
1006
1007         *((__be32 *) bufp) = htonl(session->peer_session_id);
1008         bufp += 4;
1009         if (session->cookie_len) {
1010                 memcpy(bufp, &session->cookie[0], session->cookie_len);
1011                 bufp += session->cookie_len;
1012         }
1013         if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1014                 u32 l2h = 0;
1015
1016                 if (session->send_seq) {
1017                         l2h = 0x40000000 | session->ns;
1018                         session->ns++;
1019                         session->ns &= 0xffffff;
1020                         l2tp_dbg(session, L2TP_MSG_SEQ,
1021                                  "%s: updated ns to %u\n",
1022                                  session->name, session->ns);
1023                 }
1024
1025                 *((__be32 *)bufp) = htonl(l2h);
1026                 bufp += 4;
1027         }
1028
1029         return bufp - optr;
1030 }
1031
1032 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1033                           struct flowi *fl, size_t data_len)
1034 {
1035         struct l2tp_tunnel *tunnel = session->tunnel;
1036         unsigned int len = skb->len;
1037         int error;
1038
1039         /* Debug */
1040         if (session->send_seq)
1041                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes, ns=%u\n",
1042                          session->name, data_len, session->ns - 1);
1043         else
1044                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes\n",
1045                          session->name, data_len);
1046
1047         if (session->debug & L2TP_MSG_DATA) {
1048                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1049                 unsigned char *datap = skb->data + uhlen;
1050
1051                 pr_debug("%s: xmit\n", session->name);
1052                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1053                                      datap, min_t(size_t, 32, len - uhlen));
1054         }
1055
1056         /* Queue the packet to IP for output */
1057         skb->ignore_df = 1;
1058 #if IS_ENABLED(CONFIG_IPV6)
1059         if (l2tp_sk_is_v6(tunnel->sock))
1060                 error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1061         else
1062 #endif
1063                 error = ip_queue_xmit(tunnel->sock, skb, fl);
1064
1065         /* Update stats */
1066         if (error >= 0) {
1067                 atomic_long_inc(&tunnel->stats.tx_packets);
1068                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1069                 atomic_long_inc(&session->stats.tx_packets);
1070                 atomic_long_add(len, &session->stats.tx_bytes);
1071         } else {
1072                 atomic_long_inc(&tunnel->stats.tx_errors);
1073                 atomic_long_inc(&session->stats.tx_errors);
1074         }
1075
1076         return 0;
1077 }
1078
1079 /* If caller requires the skb to have a ppp header, the header must be
1080  * inserted in the skb data before calling this function.
1081  */
1082 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1083 {
1084         int data_len = skb->len;
1085         struct l2tp_tunnel *tunnel = session->tunnel;
1086         struct sock *sk = tunnel->sock;
1087         struct flowi *fl;
1088         struct udphdr *uh;
1089         struct inet_sock *inet;
1090         int headroom;
1091         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1092         int udp_len;
1093         int ret = NET_XMIT_SUCCESS;
1094
1095         /* Check that there's enough headroom in the skb to insert IP,
1096          * UDP and L2TP headers. If not enough, expand it to
1097          * make room. Adjust truesize.
1098          */
1099         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1100                 uhlen + hdr_len;
1101         if (skb_cow_head(skb, headroom)) {
1102                 kfree_skb(skb);
1103                 return NET_XMIT_DROP;
1104         }
1105
1106         /* Setup L2TP header */
1107         session->build_header(session, __skb_push(skb, hdr_len));
1108
1109         /* Reset skb netfilter state */
1110         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1111         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1112                               IPSKB_REROUTED);
1113         nf_reset(skb);
1114
1115         bh_lock_sock(sk);
1116         if (sock_owned_by_user(sk)) {
1117                 kfree_skb(skb);
1118                 ret = NET_XMIT_DROP;
1119                 goto out_unlock;
1120         }
1121
1122         /* The user-space may change the connection status for the user-space
1123          * provided socket at run time: we must check it under the socket lock
1124          */
1125         if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1126                 kfree_skb(skb);
1127                 ret = NET_XMIT_DROP;
1128                 goto out_unlock;
1129         }
1130
1131         /* Get routing info from the tunnel socket */
1132         skb_dst_drop(skb);
1133         skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1134
1135         inet = inet_sk(sk);
1136         fl = &inet->cork.fl;
1137         switch (tunnel->encap) {
1138         case L2TP_ENCAPTYPE_UDP:
1139                 /* Setup UDP header */
1140                 __skb_push(skb, sizeof(*uh));
1141                 skb_reset_transport_header(skb);
1142                 uh = udp_hdr(skb);
1143                 uh->source = inet->inet_sport;
1144                 uh->dest = inet->inet_dport;
1145                 udp_len = uhlen + hdr_len + data_len;
1146                 uh->len = htons(udp_len);
1147
1148                 /* Calculate UDP checksum if configured to do so */
1149 #if IS_ENABLED(CONFIG_IPV6)
1150                 if (l2tp_sk_is_v6(sk))
1151                         udp6_set_csum(udp_get_no_check6_tx(sk),
1152                                       skb, &inet6_sk(sk)->saddr,
1153                                       &sk->sk_v6_daddr, udp_len);
1154                 else
1155 #endif
1156                 udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1157                              inet->inet_daddr, udp_len);
1158                 break;
1159
1160         case L2TP_ENCAPTYPE_IP:
1161                 break;
1162         }
1163
1164         l2tp_xmit_core(session, skb, fl, data_len);
1165 out_unlock:
1166         bh_unlock_sock(sk);
1167
1168         return ret;
1169 }
1170 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1171
1172 /*****************************************************************************
1173  * Tinnel and session create/destroy.
1174  *****************************************************************************/
1175
1176 /* Tunnel socket destruct hook.
1177  * The tunnel context is deleted only when all session sockets have been
1178  * closed.
1179  */
1180 static void l2tp_tunnel_destruct(struct sock *sk)
1181 {
1182         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1183
1184         if (tunnel == NULL)
1185                 goto end;
1186
1187         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1188
1189         /* Disable udp encapsulation */
1190         switch (tunnel->encap) {
1191         case L2TP_ENCAPTYPE_UDP:
1192                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1193                 (udp_sk(sk))->encap_type = 0;
1194                 (udp_sk(sk))->encap_rcv = NULL;
1195                 (udp_sk(sk))->encap_destroy = NULL;
1196                 break;
1197         case L2TP_ENCAPTYPE_IP:
1198                 break;
1199         }
1200
1201         /* Remove hooks into tunnel socket */
1202         sk->sk_destruct = tunnel->old_sk_destruct;
1203         sk->sk_user_data = NULL;
1204
1205         /* Call the original destructor */
1206         if (sk->sk_destruct)
1207                 (*sk->sk_destruct)(sk);
1208
1209         kfree_rcu(tunnel, rcu);
1210 end:
1211         return;
1212 }
1213
1214 /* When the tunnel is closed, all the attached sessions need to go too.
1215  */
1216 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1217 {
1218         int hash;
1219         struct hlist_node *walk;
1220         struct hlist_node *tmp;
1221         struct l2tp_session *session;
1222
1223         BUG_ON(tunnel == NULL);
1224
1225         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1226                   tunnel->name);
1227
1228         write_lock_bh(&tunnel->hlist_lock);
1229         tunnel->acpt_newsess = false;
1230         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1231 again:
1232                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1233                         session = hlist_entry(walk, struct l2tp_session, hlist);
1234
1235                         l2tp_info(session, L2TP_MSG_CONTROL,
1236                                   "%s: closing session\n", session->name);
1237
1238                         hlist_del_init(&session->hlist);
1239
1240                         if (test_and_set_bit(0, &session->dead))
1241                                 goto again;
1242
1243                         write_unlock_bh(&tunnel->hlist_lock);
1244
1245                         __l2tp_session_unhash(session);
1246                         l2tp_session_queue_purge(session);
1247
1248                         if (session->session_close != NULL)
1249                                 (*session->session_close)(session);
1250
1251                         l2tp_session_dec_refcount(session);
1252
1253                         write_lock_bh(&tunnel->hlist_lock);
1254
1255                         /* Now restart from the beginning of this hash
1256                          * chain.  We always remove a session from the
1257                          * list so we are guaranteed to make forward
1258                          * progress.
1259                          */
1260                         goto again;
1261                 }
1262         }
1263         write_unlock_bh(&tunnel->hlist_lock);
1264 }
1265 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1266
1267 /* Tunnel socket destroy hook for UDP encapsulation */
1268 static void l2tp_udp_encap_destroy(struct sock *sk)
1269 {
1270         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1271
1272         if (tunnel)
1273                 l2tp_tunnel_delete(tunnel);
1274 }
1275
1276 /* Workqueue tunnel deletion function */
1277 static void l2tp_tunnel_del_work(struct work_struct *work)
1278 {
1279         struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1280                                                   del_work);
1281         struct sock *sk = tunnel->sock;
1282         struct socket *sock = sk->sk_socket;
1283         struct l2tp_net *pn;
1284
1285         l2tp_tunnel_closeall(tunnel);
1286
1287         /* If the tunnel socket was created within the kernel, use
1288          * the sk API to release it here.
1289          */
1290         if (tunnel->fd < 0) {
1291                 if (sock) {
1292                         kernel_sock_shutdown(sock, SHUT_RDWR);
1293                         sock_release(sock);
1294                 }
1295         }
1296
1297         /* Remove the tunnel struct from the tunnel list */
1298         pn = l2tp_pernet(tunnel->l2tp_net);
1299         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1300         list_del_rcu(&tunnel->list);
1301         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1302
1303         /* drop initial ref */
1304         l2tp_tunnel_dec_refcount(tunnel);
1305
1306         /* drop workqueue ref */
1307         l2tp_tunnel_dec_refcount(tunnel);
1308 }
1309
1310 /* Create a socket for the tunnel, if one isn't set up by
1311  * userspace. This is used for static tunnels where there is no
1312  * managing L2TP daemon.
1313  *
1314  * Since we don't want these sockets to keep a namespace alive by
1315  * themselves, we drop the socket's namespace refcount after creation.
1316  * These sockets are freed when the namespace exits using the pernet
1317  * exit hook.
1318  */
1319 static int l2tp_tunnel_sock_create(struct net *net,
1320                                 u32 tunnel_id,
1321                                 u32 peer_tunnel_id,
1322                                 struct l2tp_tunnel_cfg *cfg,
1323                                 struct socket **sockp)
1324 {
1325         int err = -EINVAL;
1326         struct socket *sock = NULL;
1327         struct udp_port_cfg udp_conf;
1328
1329         switch (cfg->encap) {
1330         case L2TP_ENCAPTYPE_UDP:
1331                 memset(&udp_conf, 0, sizeof(udp_conf));
1332
1333 #if IS_ENABLED(CONFIG_IPV6)
1334                 if (cfg->local_ip6 && cfg->peer_ip6) {
1335                         udp_conf.family = AF_INET6;
1336                         memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1337                                sizeof(udp_conf.local_ip6));
1338                         memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1339                                sizeof(udp_conf.peer_ip6));
1340                         udp_conf.use_udp6_tx_checksums =
1341                           ! cfg->udp6_zero_tx_checksums;
1342                         udp_conf.use_udp6_rx_checksums =
1343                           ! cfg->udp6_zero_rx_checksums;
1344                 } else
1345 #endif
1346                 {
1347                         udp_conf.family = AF_INET;
1348                         udp_conf.local_ip = cfg->local_ip;
1349                         udp_conf.peer_ip = cfg->peer_ip;
1350                         udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1351                 }
1352
1353                 udp_conf.local_udp_port = htons(cfg->local_udp_port);
1354                 udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1355
1356                 err = udp_sock_create(net, &udp_conf, &sock);
1357                 if (err < 0)
1358                         goto out;
1359
1360                 break;
1361
1362         case L2TP_ENCAPTYPE_IP:
1363 #if IS_ENABLED(CONFIG_IPV6)
1364                 if (cfg->local_ip6 && cfg->peer_ip6) {
1365                         struct sockaddr_l2tpip6 ip6_addr = {0};
1366
1367                         err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1368                                           IPPROTO_L2TP, &sock);
1369                         if (err < 0)
1370                                 goto out;
1371
1372                         ip6_addr.l2tp_family = AF_INET6;
1373                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1374                                sizeof(ip6_addr.l2tp_addr));
1375                         ip6_addr.l2tp_conn_id = tunnel_id;
1376                         err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1377                                           sizeof(ip6_addr));
1378                         if (err < 0)
1379                                 goto out;
1380
1381                         ip6_addr.l2tp_family = AF_INET6;
1382                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1383                                sizeof(ip6_addr.l2tp_addr));
1384                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1385                         err = kernel_connect(sock,
1386                                              (struct sockaddr *) &ip6_addr,
1387                                              sizeof(ip6_addr), 0);
1388                         if (err < 0)
1389                                 goto out;
1390                 } else
1391 #endif
1392                 {
1393                         struct sockaddr_l2tpip ip_addr = {0};
1394
1395                         err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1396                                           IPPROTO_L2TP, &sock);
1397                         if (err < 0)
1398                                 goto out;
1399
1400                         ip_addr.l2tp_family = AF_INET;
1401                         ip_addr.l2tp_addr = cfg->local_ip;
1402                         ip_addr.l2tp_conn_id = tunnel_id;
1403                         err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1404                                           sizeof(ip_addr));
1405                         if (err < 0)
1406                                 goto out;
1407
1408                         ip_addr.l2tp_family = AF_INET;
1409                         ip_addr.l2tp_addr = cfg->peer_ip;
1410                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1411                         err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1412                                              sizeof(ip_addr), 0);
1413                         if (err < 0)
1414                                 goto out;
1415                 }
1416                 break;
1417
1418         default:
1419                 goto out;
1420         }
1421
1422 out:
1423         *sockp = sock;
1424         if ((err < 0) && sock) {
1425                 kernel_sock_shutdown(sock, SHUT_RDWR);
1426                 sock_release(sock);
1427                 *sockp = NULL;
1428         }
1429
1430         return err;
1431 }
1432
1433 static struct lock_class_key l2tp_socket_class;
1434
1435 int l2tp_tunnel_create(struct net *net, int fd, int version, u32 tunnel_id, u32 peer_tunnel_id, struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1436 {
1437         struct l2tp_tunnel *tunnel = NULL;
1438         int err;
1439         struct socket *sock = NULL;
1440         struct sock *sk = NULL;
1441         struct l2tp_net *pn;
1442         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1443
1444         /* Get the tunnel socket from the fd, which was opened by
1445          * the userspace L2TP daemon. If not specified, create a
1446          * kernel socket.
1447          */
1448         if (fd < 0) {
1449                 err = l2tp_tunnel_sock_create(net, tunnel_id, peer_tunnel_id,
1450                                 cfg, &sock);
1451                 if (err < 0)
1452                         goto err;
1453         } else {
1454                 sock = sockfd_lookup(fd, &err);
1455                 if (!sock) {
1456                         pr_err("tunl %u: sockfd_lookup(fd=%d) returned %d\n",
1457                                tunnel_id, fd, err);
1458                         err = -EBADF;
1459                         goto err;
1460                 }
1461
1462                 /* Reject namespace mismatches */
1463                 if (!net_eq(sock_net(sock->sk), net)) {
1464                         pr_err("tunl %u: netns mismatch\n", tunnel_id);
1465                         err = -EINVAL;
1466                         goto err;
1467                 }
1468         }
1469
1470         sk = sock->sk;
1471
1472         if (cfg != NULL)
1473                 encap = cfg->encap;
1474
1475         /* Quick sanity checks */
1476         err = -EPROTONOSUPPORT;
1477         if (sk->sk_type != SOCK_DGRAM) {
1478                 pr_debug("tunl %hu: fd %d wrong socket type\n",
1479                          tunnel_id, fd);
1480                 goto err;
1481         }
1482         switch (encap) {
1483         case L2TP_ENCAPTYPE_UDP:
1484                 if (sk->sk_protocol != IPPROTO_UDP) {
1485                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1486                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1487                         goto err;
1488                 }
1489                 break;
1490         case L2TP_ENCAPTYPE_IP:
1491                 if (sk->sk_protocol != IPPROTO_L2TP) {
1492                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1493                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1494                         goto err;
1495                 }
1496                 break;
1497         }
1498
1499         /* Check if this socket has already been prepped */
1500         tunnel = l2tp_tunnel(sk);
1501         if (tunnel != NULL) {
1502                 /* This socket has already been prepped */
1503                 err = -EBUSY;
1504                 goto err;
1505         }
1506
1507         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1508         if (tunnel == NULL) {
1509                 err = -ENOMEM;
1510                 goto err;
1511         }
1512
1513         tunnel->version = version;
1514         tunnel->tunnel_id = tunnel_id;
1515         tunnel->peer_tunnel_id = peer_tunnel_id;
1516         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1517
1518         tunnel->magic = L2TP_TUNNEL_MAGIC;
1519         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1520         rwlock_init(&tunnel->hlist_lock);
1521         tunnel->acpt_newsess = true;
1522
1523         /* The net we belong to */
1524         tunnel->l2tp_net = net;
1525         pn = l2tp_pernet(net);
1526
1527         if (cfg != NULL)
1528                 tunnel->debug = cfg->debug;
1529
1530         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1531         tunnel->encap = encap;
1532         if (encap == L2TP_ENCAPTYPE_UDP) {
1533                 struct udp_tunnel_sock_cfg udp_cfg = { };
1534
1535                 udp_cfg.sk_user_data = tunnel;
1536                 udp_cfg.encap_type = UDP_ENCAP_L2TPINUDP;
1537                 udp_cfg.encap_rcv = l2tp_udp_encap_recv;
1538                 udp_cfg.encap_destroy = l2tp_udp_encap_destroy;
1539
1540                 setup_udp_tunnel_sock(net, sock, &udp_cfg);
1541         } else {
1542                 sk->sk_user_data = tunnel;
1543         }
1544
1545         /* Bump the reference count. The tunnel context is deleted
1546          * only when this drops to zero. A reference is also held on
1547          * the tunnel socket to ensure that it is not released while
1548          * the tunnel is extant. Must be done before sk_destruct is
1549          * set.
1550          */
1551         refcount_set(&tunnel->ref_count, 1);
1552         sock_hold(sk);
1553         tunnel->sock = sk;
1554         tunnel->fd = fd;
1555
1556         /* Hook on the tunnel socket destructor so that we can cleanup
1557          * if the tunnel socket goes away.
1558          */
1559         tunnel->old_sk_destruct = sk->sk_destruct;
1560         sk->sk_destruct = &l2tp_tunnel_destruct;
1561         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class, "l2tp_sock");
1562
1563         sk->sk_allocation = GFP_ATOMIC;
1564
1565         /* Init delete workqueue struct */
1566         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1567
1568         /* Add tunnel to our list */
1569         INIT_LIST_HEAD(&tunnel->list);
1570         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1571         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1572         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1573
1574         err = 0;
1575 err:
1576         if (tunnelp)
1577                 *tunnelp = tunnel;
1578
1579         /* If tunnel's socket was created by the kernel, it doesn't
1580          *  have a file.
1581          */
1582         if (sock && sock->file)
1583                 sockfd_put(sock);
1584
1585         return err;
1586 }
1587 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1588
1589 /* This function is used by the netlink TUNNEL_DELETE command.
1590  */
1591 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1592 {
1593         if (!test_and_set_bit(0, &tunnel->dead)) {
1594                 l2tp_tunnel_inc_refcount(tunnel);
1595                 queue_work(l2tp_wq, &tunnel->del_work);
1596         }
1597 }
1598 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1599
1600 /* Really kill the session.
1601  */
1602 void l2tp_session_free(struct l2tp_session *session)
1603 {
1604         struct l2tp_tunnel *tunnel = session->tunnel;
1605
1606         BUG_ON(refcount_read(&session->ref_count) != 0);
1607
1608         if (tunnel) {
1609                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1610                 l2tp_tunnel_dec_refcount(tunnel);
1611         }
1612
1613         kfree(session);
1614 }
1615 EXPORT_SYMBOL_GPL(l2tp_session_free);
1616
1617 /* Remove an l2tp session from l2tp_core's hash lists.
1618  * Provides a tidyup interface for pseudowire code which can't just route all
1619  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1620  * callback.
1621  */
1622 void __l2tp_session_unhash(struct l2tp_session *session)
1623 {
1624         struct l2tp_tunnel *tunnel = session->tunnel;
1625
1626         /* Remove the session from core hashes */
1627         if (tunnel) {
1628                 /* Remove from the per-tunnel hash */
1629                 write_lock_bh(&tunnel->hlist_lock);
1630                 hlist_del_init(&session->hlist);
1631                 write_unlock_bh(&tunnel->hlist_lock);
1632
1633                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1634                 if (tunnel->version != L2TP_HDR_VER_2) {
1635                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1636                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1637                         hlist_del_init_rcu(&session->global_hlist);
1638                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1639                         synchronize_rcu();
1640                 }
1641         }
1642 }
1643 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1644
1645 /* This function is used by the netlink SESSION_DELETE command and by
1646    pseudowire modules.
1647  */
1648 int l2tp_session_delete(struct l2tp_session *session)
1649 {
1650         if (test_and_set_bit(0, &session->dead))
1651                 return 0;
1652
1653         __l2tp_session_unhash(session);
1654         l2tp_session_queue_purge(session);
1655         if (session->session_close != NULL)
1656                 (*session->session_close)(session);
1657
1658         l2tp_session_dec_refcount(session);
1659
1660         return 0;
1661 }
1662 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1663
1664 /* We come here whenever a session's send_seq, cookie_len or
1665  * l2specific_type parameters are set.
1666  */
1667 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1668 {
1669         if (version == L2TP_HDR_VER_2) {
1670                 session->hdr_len = 6;
1671                 if (session->send_seq)
1672                         session->hdr_len += 4;
1673         } else {
1674                 session->hdr_len = 4 + session->cookie_len;
1675                 session->hdr_len += l2tp_get_l2specific_len(session);
1676                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1677                         session->hdr_len += 4;
1678         }
1679
1680 }
1681 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1682
1683 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id, u32 peer_session_id, struct l2tp_session_cfg *cfg)
1684 {
1685         struct l2tp_session *session;
1686
1687         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1688         if (session != NULL) {
1689                 session->magic = L2TP_SESSION_MAGIC;
1690                 session->tunnel = tunnel;
1691
1692                 session->session_id = session_id;
1693                 session->peer_session_id = peer_session_id;
1694                 session->nr = 0;
1695                 if (tunnel->version == L2TP_HDR_VER_2)
1696                         session->nr_max = 0xffff;
1697                 else
1698                         session->nr_max = 0xffffff;
1699                 session->nr_window_size = session->nr_max / 2;
1700                 session->nr_oos_count_max = 4;
1701
1702                 /* Use NR of first received packet */
1703                 session->reorder_skip = 1;
1704
1705                 sprintf(&session->name[0], "sess %u/%u",
1706                         tunnel->tunnel_id, session->session_id);
1707
1708                 skb_queue_head_init(&session->reorder_q);
1709
1710                 INIT_HLIST_NODE(&session->hlist);
1711                 INIT_HLIST_NODE(&session->global_hlist);
1712
1713                 /* Inherit debug options from tunnel */
1714                 session->debug = tunnel->debug;
1715
1716                 if (cfg) {
1717                         session->pwtype = cfg->pw_type;
1718                         session->debug = cfg->debug;
1719                         session->mtu = cfg->mtu;
1720                         session->mru = cfg->mru;
1721                         session->send_seq = cfg->send_seq;
1722                         session->recv_seq = cfg->recv_seq;
1723                         session->lns_mode = cfg->lns_mode;
1724                         session->reorder_timeout = cfg->reorder_timeout;
1725                         session->l2specific_type = cfg->l2specific_type;
1726                         session->cookie_len = cfg->cookie_len;
1727                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1728                         session->peer_cookie_len = cfg->peer_cookie_len;
1729                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1730                 }
1731
1732                 if (tunnel->version == L2TP_HDR_VER_2)
1733                         session->build_header = l2tp_build_l2tpv2_header;
1734                 else
1735                         session->build_header = l2tp_build_l2tpv3_header;
1736
1737                 l2tp_session_set_header_len(session, tunnel->version);
1738
1739                 refcount_set(&session->ref_count, 1);
1740
1741                 return session;
1742         }
1743
1744         return ERR_PTR(-ENOMEM);
1745 }
1746 EXPORT_SYMBOL_GPL(l2tp_session_create);
1747
1748 /*****************************************************************************
1749  * Init and cleanup
1750  *****************************************************************************/
1751
1752 static __net_init int l2tp_init_net(struct net *net)
1753 {
1754         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1755         int hash;
1756
1757         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1758         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1759
1760         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1761                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1762
1763         spin_lock_init(&pn->l2tp_session_hlist_lock);
1764
1765         return 0;
1766 }
1767
1768 static __net_exit void l2tp_exit_net(struct net *net)
1769 {
1770         struct l2tp_net *pn = l2tp_pernet(net);
1771         struct l2tp_tunnel *tunnel = NULL;
1772         int hash;
1773
1774         rcu_read_lock_bh();
1775         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1776                 l2tp_tunnel_delete(tunnel);
1777         }
1778         rcu_read_unlock_bh();
1779
1780         flush_workqueue(l2tp_wq);
1781         rcu_barrier();
1782
1783         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1784                 WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1785 }
1786
1787 static struct pernet_operations l2tp_net_ops = {
1788         .init = l2tp_init_net,
1789         .exit = l2tp_exit_net,
1790         .id   = &l2tp_net_id,
1791         .size = sizeof(struct l2tp_net),
1792 };
1793
1794 static int __init l2tp_init(void)
1795 {
1796         int rc = 0;
1797
1798         rc = register_pernet_device(&l2tp_net_ops);
1799         if (rc)
1800                 goto out;
1801
1802         l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1803         if (!l2tp_wq) {
1804                 pr_err("alloc_workqueue failed\n");
1805                 unregister_pernet_device(&l2tp_net_ops);
1806                 rc = -ENOMEM;
1807                 goto out;
1808         }
1809
1810         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1811
1812 out:
1813         return rc;
1814 }
1815
1816 static void __exit l2tp_exit(void)
1817 {
1818         unregister_pernet_device(&l2tp_net_ops);
1819         if (l2tp_wq) {
1820                 destroy_workqueue(l2tp_wq);
1821                 l2tp_wq = NULL;
1822         }
1823 }
1824
1825 module_init(l2tp_init);
1826 module_exit(l2tp_exit);
1827
1828 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1829 MODULE_DESCRIPTION("L2TP core");
1830 MODULE_LICENSE("GPL");
1831 MODULE_VERSION(L2TP_DRV_VERSION);
1832