Merge tag 'io_uring-5.9-2020-08-15' of git://git.kernel.dk/linux-block
[linux-2.6-microblaze.git] / net / l2tp / l2tp_core.c
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
65 #define L2TP_DRV_VERSION        "V2.0"
66
67 /* L2TP header constants */
68 #define L2TP_HDRFLAG_T     0x8000
69 #define L2TP_HDRFLAG_L     0x4000
70 #define L2TP_HDRFLAG_S     0x0800
71 #define L2TP_HDRFLAG_O     0x0200
72 #define L2TP_HDRFLAG_P     0x0100
73
74 #define L2TP_HDR_VER_MASK  0x000F
75 #define L2TP_HDR_VER_2     0x0002
76 #define L2TP_HDR_VER_3     0x0003
77
78 /* L2TPv3 default L2-specific sublayer */
79 #define L2TP_SLFLAG_S      0x40000000
80 #define L2TP_SL_SEQ_MASK   0x00ffffff
81
82 #define L2TP_HDR_SIZE_MAX               14
83
84 /* Default trace flags */
85 #define L2TP_DEFAULT_DEBUG_FLAGS        0
86
87 /* Private data stored for received packets in the skb.
88  */
89 struct l2tp_skb_cb {
90         u32                     ns;
91         u16                     has_seq;
92         u16                     length;
93         unsigned long           expires;
94 };
95
96 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *)&(skb)->cb[sizeof(struct inet_skb_parm)])
97
98 static struct workqueue_struct *l2tp_wq;
99
100 /* per-net private data for this module */
101 static unsigned int l2tp_net_id;
102 struct l2tp_net {
103         struct list_head l2tp_tunnel_list;
104         /* Lock for write access to l2tp_tunnel_list */
105         spinlock_t l2tp_tunnel_list_lock;
106         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
107         /* Lock for write access to l2tp_session_hlist */
108         spinlock_t l2tp_session_hlist_lock;
109 };
110
111 #if IS_ENABLED(CONFIG_IPV6)
112 static bool l2tp_sk_is_v6(struct sock *sk)
113 {
114         return sk->sk_family == PF_INET6 &&
115                !ipv6_addr_v4mapped(&sk->sk_v6_daddr);
116 }
117 #endif
118
119 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
120 {
121         return sk->sk_user_data;
122 }
123
124 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
125 {
126         return net_generic(net, l2tp_net_id);
127 }
128
129 /* Session hash global list for L2TPv3.
130  * The session_id SHOULD be random according to RFC3931, but several
131  * L2TP implementations use incrementing session_ids.  So we do a real
132  * hash on the session_id, rather than a simple bitmask.
133  */
134 static inline struct hlist_head *
135 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
136 {
137         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
138 }
139
140 /* Session hash list.
141  * The session_id SHOULD be random according to RFC2661, but several
142  * L2TP implementations (Cisco and Microsoft) use incrementing
143  * session_ids.  So we do a real hash on the session_id, rather than a
144  * simple bitmask.
145  */
146 static inline struct hlist_head *
147 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
148 {
149         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
150 }
151
152 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
153 {
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         struct l2tp_tunnel *tunnel = session->tunnel;
161
162         if (tunnel) {
163                 if (WARN_ON(tunnel->magic != L2TP_TUNNEL_MAGIC))
164                         goto out;
165                 l2tp_tunnel_dec_refcount(tunnel);
166         }
167
168 out:
169         kfree(session);
170 }
171
172 void l2tp_tunnel_inc_refcount(struct l2tp_tunnel *tunnel)
173 {
174         refcount_inc(&tunnel->ref_count);
175 }
176 EXPORT_SYMBOL_GPL(l2tp_tunnel_inc_refcount);
177
178 void l2tp_tunnel_dec_refcount(struct l2tp_tunnel *tunnel)
179 {
180         if (refcount_dec_and_test(&tunnel->ref_count))
181                 l2tp_tunnel_free(tunnel);
182 }
183 EXPORT_SYMBOL_GPL(l2tp_tunnel_dec_refcount);
184
185 void l2tp_session_inc_refcount(struct l2tp_session *session)
186 {
187         refcount_inc(&session->ref_count);
188 }
189 EXPORT_SYMBOL_GPL(l2tp_session_inc_refcount);
190
191 void l2tp_session_dec_refcount(struct l2tp_session *session)
192 {
193         if (refcount_dec_and_test(&session->ref_count))
194                 l2tp_session_free(session);
195 }
196 EXPORT_SYMBOL_GPL(l2tp_session_dec_refcount);
197
198 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
199 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
200 {
201         const struct l2tp_net *pn = l2tp_pernet(net);
202         struct l2tp_tunnel *tunnel;
203
204         rcu_read_lock_bh();
205         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
206                 if (tunnel->tunnel_id == tunnel_id &&
207                     refcount_inc_not_zero(&tunnel->ref_count)) {
208                         rcu_read_unlock_bh();
209
210                         return tunnel;
211                 }
212         }
213         rcu_read_unlock_bh();
214
215         return NULL;
216 }
217 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
218
219 struct l2tp_tunnel *l2tp_tunnel_get_nth(const struct net *net, int nth)
220 {
221         const struct l2tp_net *pn = l2tp_pernet(net);
222         struct l2tp_tunnel *tunnel;
223         int count = 0;
224
225         rcu_read_lock_bh();
226         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
227                 if (++count > nth &&
228                     refcount_inc_not_zero(&tunnel->ref_count)) {
229                         rcu_read_unlock_bh();
230                         return tunnel;
231                 }
232         }
233         rcu_read_unlock_bh();
234
235         return NULL;
236 }
237 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_nth);
238
239 struct l2tp_session *l2tp_tunnel_get_session(struct l2tp_tunnel *tunnel,
240                                              u32 session_id)
241 {
242         struct hlist_head *session_list;
243         struct l2tp_session *session;
244
245         session_list = l2tp_session_id_hash(tunnel, session_id);
246
247         read_lock_bh(&tunnel->hlist_lock);
248         hlist_for_each_entry(session, session_list, hlist)
249                 if (session->session_id == session_id) {
250                         l2tp_session_inc_refcount(session);
251                         read_unlock_bh(&tunnel->hlist_lock);
252
253                         return session;
254                 }
255         read_unlock_bh(&tunnel->hlist_lock);
256
257         return NULL;
258 }
259 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_session);
260
261 struct l2tp_session *l2tp_session_get(const struct net *net, u32 session_id)
262 {
263         struct hlist_head *session_list;
264         struct l2tp_session *session;
265
266         session_list = l2tp_session_id_hash_2(l2tp_pernet(net), session_id);
267
268         rcu_read_lock_bh();
269         hlist_for_each_entry_rcu(session, session_list, global_hlist)
270                 if (session->session_id == session_id) {
271                         l2tp_session_inc_refcount(session);
272                         rcu_read_unlock_bh();
273
274                         return session;
275                 }
276         rcu_read_unlock_bh();
277
278         return NULL;
279 }
280 EXPORT_SYMBOL_GPL(l2tp_session_get);
281
282 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth)
283 {
284         int hash;
285         struct l2tp_session *session;
286         int count = 0;
287
288         read_lock_bh(&tunnel->hlist_lock);
289         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
290                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
291                         if (++count > nth) {
292                                 l2tp_session_inc_refcount(session);
293                                 read_unlock_bh(&tunnel->hlist_lock);
294                                 return session;
295                         }
296                 }
297         }
298
299         read_unlock_bh(&tunnel->hlist_lock);
300
301         return NULL;
302 }
303 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
304
305 /* Lookup a session by interface name.
306  * This is very inefficient but is only used by management interfaces.
307  */
308 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
309                                                 const char *ifname)
310 {
311         struct l2tp_net *pn = l2tp_pernet(net);
312         int hash;
313         struct l2tp_session *session;
314
315         rcu_read_lock_bh();
316         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
317                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
318                         if (!strcmp(session->ifname, ifname)) {
319                                 l2tp_session_inc_refcount(session);
320                                 rcu_read_unlock_bh();
321
322                                 return session;
323                         }
324                 }
325         }
326
327         rcu_read_unlock_bh();
328
329         return NULL;
330 }
331 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
332
333 int l2tp_session_register(struct l2tp_session *session,
334                           struct l2tp_tunnel *tunnel)
335 {
336         struct l2tp_session *session_walk;
337         struct hlist_head *g_head;
338         struct hlist_head *head;
339         struct l2tp_net *pn;
340         int err;
341
342         head = l2tp_session_id_hash(tunnel, session->session_id);
343
344         write_lock_bh(&tunnel->hlist_lock);
345         if (!tunnel->acpt_newsess) {
346                 err = -ENODEV;
347                 goto err_tlock;
348         }
349
350         hlist_for_each_entry(session_walk, head, hlist)
351                 if (session_walk->session_id == session->session_id) {
352                         err = -EEXIST;
353                         goto err_tlock;
354                 }
355
356         if (tunnel->version == L2TP_HDR_VER_3) {
357                 pn = l2tp_pernet(tunnel->l2tp_net);
358                 g_head = l2tp_session_id_hash_2(pn, session->session_id);
359
360                 spin_lock_bh(&pn->l2tp_session_hlist_lock);
361
362                 /* IP encap expects session IDs to be globally unique, while
363                  * UDP encap doesn't.
364                  */
365                 hlist_for_each_entry(session_walk, g_head, global_hlist)
366                         if (session_walk->session_id == session->session_id &&
367                             (session_walk->tunnel->encap == L2TP_ENCAPTYPE_IP ||
368                              tunnel->encap == L2TP_ENCAPTYPE_IP)) {
369                                 err = -EEXIST;
370                                 goto err_tlock_pnlock;
371                         }
372
373                 l2tp_tunnel_inc_refcount(tunnel);
374                 hlist_add_head_rcu(&session->global_hlist, g_head);
375
376                 spin_unlock_bh(&pn->l2tp_session_hlist_lock);
377         } else {
378                 l2tp_tunnel_inc_refcount(tunnel);
379         }
380
381         hlist_add_head(&session->hlist, head);
382         write_unlock_bh(&tunnel->hlist_lock);
383
384         return 0;
385
386 err_tlock_pnlock:
387         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
388 err_tlock:
389         write_unlock_bh(&tunnel->hlist_lock);
390
391         return err;
392 }
393 EXPORT_SYMBOL_GPL(l2tp_session_register);
394
395 /*****************************************************************************
396  * Receive data handling
397  *****************************************************************************/
398
399 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
400  * number.
401  */
402 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
403 {
404         struct sk_buff *skbp;
405         struct sk_buff *tmp;
406         u32 ns = L2TP_SKB_CB(skb)->ns;
407
408         spin_lock_bh(&session->reorder_q.lock);
409         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
410                 if (L2TP_SKB_CB(skbp)->ns > ns) {
411                         __skb_queue_before(&session->reorder_q, skbp, skb);
412                         l2tp_dbg(session, L2TP_MSG_SEQ,
413                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
414                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
415                                  skb_queue_len(&session->reorder_q));
416                         atomic_long_inc(&session->stats.rx_oos_packets);
417                         goto out;
418                 }
419         }
420
421         __skb_queue_tail(&session->reorder_q, skb);
422
423 out:
424         spin_unlock_bh(&session->reorder_q.lock);
425 }
426
427 /* Dequeue a single skb.
428  */
429 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
430 {
431         struct l2tp_tunnel *tunnel = session->tunnel;
432         int length = L2TP_SKB_CB(skb)->length;
433
434         /* We're about to requeue the skb, so return resources
435          * to its current owner (a socket receive buffer).
436          */
437         skb_orphan(skb);
438
439         atomic_long_inc(&tunnel->stats.rx_packets);
440         atomic_long_add(length, &tunnel->stats.rx_bytes);
441         atomic_long_inc(&session->stats.rx_packets);
442         atomic_long_add(length, &session->stats.rx_bytes);
443
444         if (L2TP_SKB_CB(skb)->has_seq) {
445                 /* Bump our Nr */
446                 session->nr++;
447                 session->nr &= session->nr_max;
448
449                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
450                          session->name, session->nr);
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                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
476                         atomic_long_inc(&session->stats.rx_seq_discards);
477                         atomic_long_inc(&session->stats.rx_errors);
478                         l2tp_dbg(session, L2TP_MSG_SEQ,
479                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
480                                  session->name, L2TP_SKB_CB(skb)->ns,
481                                  L2TP_SKB_CB(skb)->length, session->nr,
482                                  skb_queue_len(&session->reorder_q));
483                         session->reorder_skip = 1;
484                         __skb_unlink(skb, &session->reorder_q);
485                         kfree_skb(skb);
486                         continue;
487                 }
488
489                 if (L2TP_SKB_CB(skb)->has_seq) {
490                         if (session->reorder_skip) {
491                                 l2tp_dbg(session, L2TP_MSG_SEQ,
492                                          "%s: advancing nr to next pkt: %u -> %u",
493                                          session->name, session->nr,
494                                          L2TP_SKB_CB(skb)->ns);
495                                 session->reorder_skip = 0;
496                                 session->nr = L2TP_SKB_CB(skb)->ns;
497                         }
498                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
499                                 l2tp_dbg(session, L2TP_MSG_SEQ,
500                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
501                                          session->name, L2TP_SKB_CB(skb)->ns,
502                                          L2TP_SKB_CB(skb)->length, session->nr,
503                                          skb_queue_len(&session->reorder_q));
504                                 goto out;
505                         }
506                 }
507                 __skb_unlink(skb, &session->reorder_q);
508
509                 /* Process the skb. We release the queue lock while we
510                  * do so to let other contexts process the queue.
511                  */
512                 spin_unlock_bh(&session->reorder_q.lock);
513                 l2tp_recv_dequeue_skb(session, skb);
514                 goto start;
515         }
516
517 out:
518         spin_unlock_bh(&session->reorder_q.lock);
519 }
520
521 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
522 {
523         u32 nws;
524
525         if (nr >= session->nr)
526                 nws = nr - session->nr;
527         else
528                 nws = (session->nr_max + 1) - (session->nr - nr);
529
530         return nws < session->nr_window_size;
531 }
532
533 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
534  * acceptable, else non-zero.
535  */
536 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
537 {
538         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
539                 /* Packet sequence number is outside allowed window.
540                  * Discard it.
541                  */
542                 l2tp_dbg(session, L2TP_MSG_SEQ,
543                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
544                          session->name, L2TP_SKB_CB(skb)->ns,
545                          L2TP_SKB_CB(skb)->length, session->nr);
546                 goto discard;
547         }
548
549         if (session->reorder_timeout != 0) {
550                 /* Packet reordering enabled. Add skb to session's
551                  * reorder queue, in order of ns.
552                  */
553                 l2tp_recv_queue_skb(session, skb);
554                 goto out;
555         }
556
557         /* Packet reordering disabled. Discard out-of-sequence packets, while
558          * tracking the number if in-sequence packets after the first OOS packet
559          * is seen. After nr_oos_count_max in-sequence packets, reset the
560          * sequence number to re-enable packet reception.
561          */
562         if (L2TP_SKB_CB(skb)->ns == session->nr) {
563                 skb_queue_tail(&session->reorder_q, skb);
564         } else {
565                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
566                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
567
568                 if (nr_oos == nr_next)
569                         session->nr_oos_count++;
570                 else
571                         session->nr_oos_count = 0;
572
573                 session->nr_oos = nr_oos;
574                 if (session->nr_oos_count > session->nr_oos_count_max) {
575                         session->reorder_skip = 1;
576                         l2tp_dbg(session, L2TP_MSG_SEQ,
577                                  "%s: %d oos packets received. Resetting sequence numbers\n",
578                                  session->name, session->nr_oos_count);
579                 }
580                 if (!session->reorder_skip) {
581                         atomic_long_inc(&session->stats.rx_seq_discards);
582                         l2tp_dbg(session, L2TP_MSG_SEQ,
583                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
584                                  session->name, L2TP_SKB_CB(skb)->ns,
585                                  L2TP_SKB_CB(skb)->length, session->nr,
586                                  skb_queue_len(&session->reorder_q));
587                         goto discard;
588                 }
589                 skb_queue_tail(&session->reorder_q, skb);
590         }
591
592 out:
593         return 0;
594
595 discard:
596         return 1;
597 }
598
599 /* Do receive processing of L2TP data frames. We handle both L2TPv2
600  * and L2TPv3 data frames here.
601  *
602  * L2TPv2 Data Message Header
603  *
604  *  0                   1                   2                   3
605  *  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
606  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
607  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
608  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
609  * |           Tunnel ID           |           Session ID          |
610  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
611  * |             Ns (opt)          |             Nr (opt)          |
612  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
613  * |      Offset Size (opt)        |    Offset pad... (opt)
614  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
615  *
616  * Data frames are marked by T=0. All other fields are the same as
617  * those in L2TP control frames.
618  *
619  * L2TPv3 Data Message Header
620  *
621  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
622  * |                      L2TP Session Header                      |
623  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
624  * |                      L2-Specific Sublayer                     |
625  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
626  * |                        Tunnel Payload                      ...
627  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
628  *
629  * L2TPv3 Session Header Over IP
630  *
631  *  0                   1                   2                   3
632  *  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
633  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
634  * |                           Session ID                          |
635  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
636  * |               Cookie (optional, maximum 64 bits)...
637  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
638  *                                                                 |
639  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
640  *
641  * L2TPv3 L2-Specific Sublayer Format
642  *
643  *  0                   1                   2                   3
644  *  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
645  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
646  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
647  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
648  *
649  * Cookie value and sublayer format are negotiated with the peer when
650  * the session is set up. Unlike L2TPv2, we do not need to parse the
651  * packet header to determine if optional fields are present.
652  *
653  * Caller must already have parsed the frame and determined that it is
654  * a data (not control) frame before coming here. Fields up to the
655  * session-id have already been parsed and ptr points to the data
656  * after the session-id.
657  */
658 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
659                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
660                       int length)
661 {
662         struct l2tp_tunnel *tunnel = session->tunnel;
663         u32 ns = 0, nr = 0;
664         int offset;
665
666         /* Parse and check optional cookie */
667         if (session->peer_cookie_len > 0) {
668                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
669                         l2tp_info(tunnel, L2TP_MSG_DATA,
670                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
671                                   tunnel->name, tunnel->tunnel_id,
672                                   session->session_id);
673                         atomic_long_inc(&session->stats.rx_cookie_discards);
674                         goto discard;
675                 }
676                 ptr += session->peer_cookie_len;
677         }
678
679         /* Handle the optional sequence numbers. Sequence numbers are
680          * in different places for L2TPv2 and L2TPv3.
681          *
682          * If we are the LAC, enable/disable sequence numbers under
683          * the control of the LNS.  If no sequence numbers present but
684          * we were expecting them, discard frame.
685          */
686         L2TP_SKB_CB(skb)->has_seq = 0;
687         if (tunnel->version == L2TP_HDR_VER_2) {
688                 if (hdrflags & L2TP_HDRFLAG_S) {
689                         ns = ntohs(*(__be16 *)ptr);
690                         ptr += 2;
691                         nr = ntohs(*(__be16 *)ptr);
692                         ptr += 2;
693
694                         /* Store L2TP info in the skb */
695                         L2TP_SKB_CB(skb)->ns = ns;
696                         L2TP_SKB_CB(skb)->has_seq = 1;
697
698                         l2tp_dbg(session, L2TP_MSG_SEQ,
699                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
700                                  session->name, ns, nr, session->nr);
701                 }
702         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
703                 u32 l2h = ntohl(*(__be32 *)ptr);
704
705                 if (l2h & 0x40000000) {
706                         ns = l2h & 0x00ffffff;
707
708                         /* Store L2TP info in the skb */
709                         L2TP_SKB_CB(skb)->ns = ns;
710                         L2TP_SKB_CB(skb)->has_seq = 1;
711
712                         l2tp_dbg(session, L2TP_MSG_SEQ,
713                                  "%s: recv data ns=%u, session nr=%u\n",
714                                  session->name, ns, session->nr);
715                 }
716                 ptr += 4;
717         }
718
719         if (L2TP_SKB_CB(skb)->has_seq) {
720                 /* Received a packet with sequence numbers. If we're the LAC,
721                  * check if we sre sending sequence numbers and if not,
722                  * configure it so.
723                  */
724                 if (!session->lns_mode && !session->send_seq) {
725                         l2tp_info(session, L2TP_MSG_SEQ,
726                                   "%s: requested to enable seq numbers by LNS\n",
727                                   session->name);
728                         session->send_seq = 1;
729                         l2tp_session_set_header_len(session, tunnel->version);
730                 }
731         } else {
732                 /* No sequence numbers.
733                  * If user has configured mandatory sequence numbers, discard.
734                  */
735                 if (session->recv_seq) {
736                         l2tp_warn(session, L2TP_MSG_SEQ,
737                                   "%s: recv data has no seq numbers when required. Discarding.\n",
738                                   session->name);
739                         atomic_long_inc(&session->stats.rx_seq_discards);
740                         goto discard;
741                 }
742
743                 /* If we're the LAC and we're sending sequence numbers, the
744                  * LNS has requested that we no longer send sequence numbers.
745                  * If we're the LNS and we're sending sequence numbers, the
746                  * LAC is broken. Discard the frame.
747                  */
748                 if (!session->lns_mode && session->send_seq) {
749                         l2tp_info(session, L2TP_MSG_SEQ,
750                                   "%s: requested to disable seq numbers by LNS\n",
751                                   session->name);
752                         session->send_seq = 0;
753                         l2tp_session_set_header_len(session, tunnel->version);
754                 } else if (session->send_seq) {
755                         l2tp_warn(session, L2TP_MSG_SEQ,
756                                   "%s: recv data has no seq numbers when required. Discarding.\n",
757                                   session->name);
758                         atomic_long_inc(&session->stats.rx_seq_discards);
759                         goto discard;
760                 }
761         }
762
763         /* Session data offset is defined only for L2TPv2 and is
764          * indicated by an optional 16-bit value in the header.
765          */
766         if (tunnel->version == L2TP_HDR_VER_2) {
767                 /* If offset bit set, skip it. */
768                 if (hdrflags & L2TP_HDRFLAG_O) {
769                         offset = ntohs(*(__be16 *)ptr);
770                         ptr += 2 + offset;
771                 }
772         }
773
774         offset = ptr - optr;
775         if (!pskb_may_pull(skb, offset))
776                 goto discard;
777
778         __skb_pull(skb, offset);
779
780         /* Prepare skb for adding to the session's reorder_q.  Hold
781          * packets for max reorder_timeout or 1 second if not
782          * reordering.
783          */
784         L2TP_SKB_CB(skb)->length = length;
785         L2TP_SKB_CB(skb)->expires = jiffies +
786                 (session->reorder_timeout ? session->reorder_timeout : HZ);
787
788         /* Add packet to the session's receive queue. Reordering is done here, if
789          * enabled. Saved L2TP protocol info is stored in skb->sb[].
790          */
791         if (L2TP_SKB_CB(skb)->has_seq) {
792                 if (l2tp_recv_data_seq(session, skb))
793                         goto discard;
794         } else {
795                 /* No sequence numbers. Add the skb to the tail of the
796                  * reorder queue. This ensures that it will be
797                  * delivered after all previous sequenced skbs.
798                  */
799                 skb_queue_tail(&session->reorder_q, skb);
800         }
801
802         /* Try to dequeue as many skbs from reorder_q as we can. */
803         l2tp_recv_dequeue(session);
804
805         return;
806
807 discard:
808         atomic_long_inc(&session->stats.rx_errors);
809         kfree_skb(skb);
810 }
811 EXPORT_SYMBOL_GPL(l2tp_recv_common);
812
813 /* Drop skbs from the session's reorder_q
814  */
815 static void l2tp_session_queue_purge(struct l2tp_session *session)
816 {
817         struct sk_buff *skb = NULL;
818
819         if (WARN_ON(session->magic != L2TP_SESSION_MAGIC))
820                 return;
821
822         while ((skb = skb_dequeue(&session->reorder_q))) {
823                 atomic_long_inc(&session->stats.rx_errors);
824                 kfree_skb(skb);
825         }
826 }
827
828 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
829  * here. The skb is not on a list when we get here.
830  * Returns 0 if the packet was a data packet and was successfully passed on.
831  * Returns 1 if the packet was not a good data packet and could not be
832  * forwarded.  All such packets are passed up to userspace to deal with.
833  */
834 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb)
835 {
836         struct l2tp_session *session = NULL;
837         unsigned char *ptr, *optr;
838         u16 hdrflags;
839         u32 tunnel_id, session_id;
840         u16 version;
841         int length;
842
843         /* UDP has verifed checksum */
844
845         /* UDP always verifies the packet length. */
846         __skb_pull(skb, sizeof(struct udphdr));
847
848         /* Short packet? */
849         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_MAX)) {
850                 l2tp_info(tunnel, L2TP_MSG_DATA,
851                           "%s: recv short packet (len=%d)\n",
852                           tunnel->name, skb->len);
853                 goto error;
854         }
855
856         /* Trace packet contents, if enabled */
857         if (tunnel->debug & L2TP_MSG_DATA) {
858                 length = min(32u, skb->len);
859                 if (!pskb_may_pull(skb, length))
860                         goto error;
861
862                 pr_debug("%s: recv\n", tunnel->name);
863                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
864         }
865
866         /* Point to L2TP header */
867         optr = skb->data;
868         ptr = skb->data;
869
870         /* Get L2TP header flags */
871         hdrflags = ntohs(*(__be16 *)ptr);
872
873         /* Check protocol version */
874         version = hdrflags & L2TP_HDR_VER_MASK;
875         if (version != tunnel->version) {
876                 l2tp_info(tunnel, L2TP_MSG_DATA,
877                           "%s: recv protocol version mismatch: got %d expected %d\n",
878                           tunnel->name, version, tunnel->version);
879                 goto error;
880         }
881
882         /* Get length of L2TP packet */
883         length = skb->len;
884
885         /* If type is control packet, it is handled by userspace. */
886         if (hdrflags & L2TP_HDRFLAG_T) {
887                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
888                          "%s: recv control packet, len=%d\n",
889                          tunnel->name, length);
890                 goto error;
891         }
892
893         /* Skip flags */
894         ptr += 2;
895
896         if (tunnel->version == L2TP_HDR_VER_2) {
897                 /* If length is present, skip it */
898                 if (hdrflags & L2TP_HDRFLAG_L)
899                         ptr += 2;
900
901                 /* Extract tunnel and session ID */
902                 tunnel_id = ntohs(*(__be16 *)ptr);
903                 ptr += 2;
904                 session_id = ntohs(*(__be16 *)ptr);
905                 ptr += 2;
906         } else {
907                 ptr += 2;       /* skip reserved bits */
908                 tunnel_id = tunnel->tunnel_id;
909                 session_id = ntohl(*(__be32 *)ptr);
910                 ptr += 4;
911         }
912
913         /* Find the session context */
914         session = l2tp_tunnel_get_session(tunnel, session_id);
915         if (!session || !session->recv_skb) {
916                 if (session)
917                         l2tp_session_dec_refcount(session);
918
919                 /* Not found? Pass to userspace to deal with */
920                 l2tp_info(tunnel, L2TP_MSG_DATA,
921                           "%s: no session found (%u/%u). Passing up.\n",
922                           tunnel->name, tunnel_id, session_id);
923                 goto error;
924         }
925
926         if (tunnel->version == L2TP_HDR_VER_3 &&
927             l2tp_v3_ensure_opt_in_linear(session, skb, &ptr, &optr))
928                 goto error;
929
930         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length);
931         l2tp_session_dec_refcount(session);
932
933         return 0;
934
935 error:
936         /* Put UDP header back */
937         __skb_push(skb, sizeof(struct udphdr));
938
939         return 1;
940 }
941
942 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
943  * Return codes:
944  * 0 : success.
945  * <0: error
946  * >0: skb should be passed up to userspace as UDP.
947  */
948 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
949 {
950         struct l2tp_tunnel *tunnel;
951
952         tunnel = rcu_dereference_sk_user_data(sk);
953         if (!tunnel)
954                 goto pass_up;
955
956         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
957                  tunnel->name, skb->len);
958
959         if (l2tp_udp_recv_core(tunnel, skb))
960                 goto pass_up;
961
962         return 0;
963
964 pass_up:
965         return 1;
966 }
967 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
968
969 /************************************************************************
970  * Transmit handling
971  ***********************************************************************/
972
973 /* Build an L2TP header for the session into the buffer provided.
974  */
975 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
976 {
977         struct l2tp_tunnel *tunnel = session->tunnel;
978         __be16 *bufp = buf;
979         __be16 *optr = buf;
980         u16 flags = L2TP_HDR_VER_2;
981         u32 tunnel_id = tunnel->peer_tunnel_id;
982         u32 session_id = session->peer_session_id;
983
984         if (session->send_seq)
985                 flags |= L2TP_HDRFLAG_S;
986
987         /* Setup L2TP header. */
988         *bufp++ = htons(flags);
989         *bufp++ = htons(tunnel_id);
990         *bufp++ = htons(session_id);
991         if (session->send_seq) {
992                 *bufp++ = htons(session->ns);
993                 *bufp++ = 0;
994                 session->ns++;
995                 session->ns &= 0xffff;
996                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
997                          session->name, session->ns);
998         }
999
1000         return bufp - optr;
1001 }
1002
1003 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1004 {
1005         struct l2tp_tunnel *tunnel = session->tunnel;
1006         char *bufp = buf;
1007         char *optr = bufp;
1008
1009         /* Setup L2TP header. The header differs slightly for UDP and
1010          * IP encapsulations. For UDP, there is 4 bytes of flags.
1011          */
1012         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1013                 u16 flags = L2TP_HDR_VER_3;
1014                 *((__be16 *)bufp) = htons(flags);
1015                 bufp += 2;
1016                 *((__be16 *)bufp) = 0;
1017                 bufp += 2;
1018         }
1019
1020         *((__be32 *)bufp) = htonl(session->peer_session_id);
1021         bufp += 4;
1022         if (session->cookie_len) {
1023                 memcpy(bufp, &session->cookie[0], session->cookie_len);
1024                 bufp += session->cookie_len;
1025         }
1026         if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1027                 u32 l2h = 0;
1028
1029                 if (session->send_seq) {
1030                         l2h = 0x40000000 | session->ns;
1031                         session->ns++;
1032                         session->ns &= 0xffffff;
1033                         l2tp_dbg(session, L2TP_MSG_SEQ,
1034                                  "%s: updated ns to %u\n",
1035                                  session->name, session->ns);
1036                 }
1037
1038                 *((__be32 *)bufp) = htonl(l2h);
1039                 bufp += 4;
1040         }
1041
1042         return bufp - optr;
1043 }
1044
1045 static void l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1046                            struct flowi *fl, size_t data_len)
1047 {
1048         struct l2tp_tunnel *tunnel = session->tunnel;
1049         unsigned int len = skb->len;
1050         int error;
1051
1052         /* Debug */
1053         if (session->send_seq)
1054                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes, ns=%u\n",
1055                          session->name, data_len, session->ns - 1);
1056         else
1057                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes\n",
1058                          session->name, data_len);
1059
1060         if (session->debug & L2TP_MSG_DATA) {
1061                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1062                 unsigned char *datap = skb->data + uhlen;
1063
1064                 pr_debug("%s: xmit\n", session->name);
1065                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1066                                      datap, min_t(size_t, 32, len - uhlen));
1067         }
1068
1069         /* Queue the packet to IP for output */
1070         skb->ignore_df = 1;
1071         skb_dst_drop(skb);
1072 #if IS_ENABLED(CONFIG_IPV6)
1073         if (l2tp_sk_is_v6(tunnel->sock))
1074                 error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1075         else
1076 #endif
1077                 error = ip_queue_xmit(tunnel->sock, skb, fl);
1078
1079         /* Update stats */
1080         if (error >= 0) {
1081                 atomic_long_inc(&tunnel->stats.tx_packets);
1082                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1083                 atomic_long_inc(&session->stats.tx_packets);
1084                 atomic_long_add(len, &session->stats.tx_bytes);
1085         } else {
1086                 atomic_long_inc(&tunnel->stats.tx_errors);
1087                 atomic_long_inc(&session->stats.tx_errors);
1088         }
1089 }
1090
1091 /* If caller requires the skb to have a ppp header, the header must be
1092  * inserted in the skb data before calling this function.
1093  */
1094 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1095 {
1096         int data_len = skb->len;
1097         struct l2tp_tunnel *tunnel = session->tunnel;
1098         struct sock *sk = tunnel->sock;
1099         struct flowi *fl;
1100         struct udphdr *uh;
1101         struct inet_sock *inet;
1102         int headroom;
1103         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1104         int udp_len;
1105         int ret = NET_XMIT_SUCCESS;
1106
1107         /* Check that there's enough headroom in the skb to insert IP,
1108          * UDP and L2TP headers. If not enough, expand it to
1109          * make room. Adjust truesize.
1110          */
1111         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1112                 uhlen + hdr_len;
1113         if (skb_cow_head(skb, headroom)) {
1114                 kfree_skb(skb);
1115                 return NET_XMIT_DROP;
1116         }
1117
1118         /* Setup L2TP header */
1119         if (tunnel->version == L2TP_HDR_VER_2)
1120                 l2tp_build_l2tpv2_header(session, __skb_push(skb, hdr_len));
1121         else
1122                 l2tp_build_l2tpv3_header(session, __skb_push(skb, hdr_len));
1123
1124         /* Reset skb netfilter state */
1125         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1126         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1127                               IPSKB_REROUTED);
1128         nf_reset_ct(skb);
1129
1130         bh_lock_sock(sk);
1131         if (sock_owned_by_user(sk)) {
1132                 kfree_skb(skb);
1133                 ret = NET_XMIT_DROP;
1134                 goto out_unlock;
1135         }
1136
1137         /* The user-space may change the connection status for the user-space
1138          * provided socket at run time: we must check it under the socket lock
1139          */
1140         if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1141                 kfree_skb(skb);
1142                 ret = NET_XMIT_DROP;
1143                 goto out_unlock;
1144         }
1145
1146         inet = inet_sk(sk);
1147         fl = &inet->cork.fl;
1148         switch (tunnel->encap) {
1149         case L2TP_ENCAPTYPE_UDP:
1150                 /* Setup UDP header */
1151                 __skb_push(skb, sizeof(*uh));
1152                 skb_reset_transport_header(skb);
1153                 uh = udp_hdr(skb);
1154                 uh->source = inet->inet_sport;
1155                 uh->dest = inet->inet_dport;
1156                 udp_len = uhlen + hdr_len + data_len;
1157                 uh->len = htons(udp_len);
1158
1159                 /* Calculate UDP checksum if configured to do so */
1160 #if IS_ENABLED(CONFIG_IPV6)
1161                 if (l2tp_sk_is_v6(sk))
1162                         udp6_set_csum(udp_get_no_check6_tx(sk),
1163                                       skb, &inet6_sk(sk)->saddr,
1164                                       &sk->sk_v6_daddr, udp_len);
1165                 else
1166 #endif
1167                         udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1168                                      inet->inet_daddr, udp_len);
1169                 break;
1170
1171         case L2TP_ENCAPTYPE_IP:
1172                 break;
1173         }
1174
1175         l2tp_xmit_core(session, skb, fl, data_len);
1176 out_unlock:
1177         bh_unlock_sock(sk);
1178
1179         return ret;
1180 }
1181 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1182
1183 /*****************************************************************************
1184  * Tinnel and session create/destroy.
1185  *****************************************************************************/
1186
1187 /* Tunnel socket destruct hook.
1188  * The tunnel context is deleted only when all session sockets have been
1189  * closed.
1190  */
1191 static void l2tp_tunnel_destruct(struct sock *sk)
1192 {
1193         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1194
1195         if (!tunnel)
1196                 goto end;
1197
1198         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1199
1200         /* Disable udp encapsulation */
1201         switch (tunnel->encap) {
1202         case L2TP_ENCAPTYPE_UDP:
1203                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1204                 (udp_sk(sk))->encap_type = 0;
1205                 (udp_sk(sk))->encap_rcv = NULL;
1206                 (udp_sk(sk))->encap_destroy = NULL;
1207                 break;
1208         case L2TP_ENCAPTYPE_IP:
1209                 break;
1210         }
1211
1212         /* Remove hooks into tunnel socket */
1213         sk->sk_destruct = tunnel->old_sk_destruct;
1214         sk->sk_user_data = NULL;
1215
1216         /* Call the original destructor */
1217         if (sk->sk_destruct)
1218                 (*sk->sk_destruct)(sk);
1219
1220         kfree_rcu(tunnel, rcu);
1221 end:
1222         return;
1223 }
1224
1225 /* Remove an l2tp session from l2tp_core's hash lists. */
1226 static void l2tp_session_unhash(struct l2tp_session *session)
1227 {
1228         struct l2tp_tunnel *tunnel = session->tunnel;
1229
1230         /* Remove the session from core hashes */
1231         if (tunnel) {
1232                 /* Remove from the per-tunnel hash */
1233                 write_lock_bh(&tunnel->hlist_lock);
1234                 hlist_del_init(&session->hlist);
1235                 write_unlock_bh(&tunnel->hlist_lock);
1236
1237                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1238                 if (tunnel->version != L2TP_HDR_VER_2) {
1239                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1240
1241                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1242                         hlist_del_init_rcu(&session->global_hlist);
1243                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1244                         synchronize_rcu();
1245                 }
1246         }
1247 }
1248
1249 /* When the tunnel is closed, all the attached sessions need to go too.
1250  */
1251 static void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1252 {
1253         int hash;
1254         struct hlist_node *walk;
1255         struct hlist_node *tmp;
1256         struct l2tp_session *session;
1257
1258         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1259                   tunnel->name);
1260
1261         write_lock_bh(&tunnel->hlist_lock);
1262         tunnel->acpt_newsess = false;
1263         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1264 again:
1265                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1266                         session = hlist_entry(walk, struct l2tp_session, hlist);
1267
1268                         l2tp_info(session, L2TP_MSG_CONTROL,
1269                                   "%s: closing session\n", session->name);
1270
1271                         hlist_del_init(&session->hlist);
1272
1273                         if (test_and_set_bit(0, &session->dead))
1274                                 goto again;
1275
1276                         write_unlock_bh(&tunnel->hlist_lock);
1277
1278                         l2tp_session_unhash(session);
1279                         l2tp_session_queue_purge(session);
1280
1281                         if (session->session_close)
1282                                 (*session->session_close)(session);
1283
1284                         l2tp_session_dec_refcount(session);
1285
1286                         write_lock_bh(&tunnel->hlist_lock);
1287
1288                         /* Now restart from the beginning of this hash
1289                          * chain.  We always remove a session from the
1290                          * list so we are guaranteed to make forward
1291                          * progress.
1292                          */
1293                         goto again;
1294                 }
1295         }
1296         write_unlock_bh(&tunnel->hlist_lock);
1297 }
1298
1299 /* Tunnel socket destroy hook for UDP encapsulation */
1300 static void l2tp_udp_encap_destroy(struct sock *sk)
1301 {
1302         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1303
1304         if (tunnel)
1305                 l2tp_tunnel_delete(tunnel);
1306 }
1307
1308 /* Workqueue tunnel deletion function */
1309 static void l2tp_tunnel_del_work(struct work_struct *work)
1310 {
1311         struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1312                                                   del_work);
1313         struct sock *sk = tunnel->sock;
1314         struct socket *sock = sk->sk_socket;
1315         struct l2tp_net *pn;
1316
1317         l2tp_tunnel_closeall(tunnel);
1318
1319         /* If the tunnel socket was created within the kernel, use
1320          * the sk API to release it here.
1321          */
1322         if (tunnel->fd < 0) {
1323                 if (sock) {
1324                         kernel_sock_shutdown(sock, SHUT_RDWR);
1325                         sock_release(sock);
1326                 }
1327         }
1328
1329         /* Remove the tunnel struct from the tunnel list */
1330         pn = l2tp_pernet(tunnel->l2tp_net);
1331         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1332         list_del_rcu(&tunnel->list);
1333         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1334
1335         /* drop initial ref */
1336         l2tp_tunnel_dec_refcount(tunnel);
1337
1338         /* drop workqueue ref */
1339         l2tp_tunnel_dec_refcount(tunnel);
1340 }
1341
1342 /* Create a socket for the tunnel, if one isn't set up by
1343  * userspace. This is used for static tunnels where there is no
1344  * managing L2TP daemon.
1345  *
1346  * Since we don't want these sockets to keep a namespace alive by
1347  * themselves, we drop the socket's namespace refcount after creation.
1348  * These sockets are freed when the namespace exits using the pernet
1349  * exit hook.
1350  */
1351 static int l2tp_tunnel_sock_create(struct net *net,
1352                                    u32 tunnel_id,
1353                                    u32 peer_tunnel_id,
1354                                    struct l2tp_tunnel_cfg *cfg,
1355                                    struct socket **sockp)
1356 {
1357         int err = -EINVAL;
1358         struct socket *sock = NULL;
1359         struct udp_port_cfg udp_conf;
1360
1361         switch (cfg->encap) {
1362         case L2TP_ENCAPTYPE_UDP:
1363                 memset(&udp_conf, 0, sizeof(udp_conf));
1364
1365 #if IS_ENABLED(CONFIG_IPV6)
1366                 if (cfg->local_ip6 && cfg->peer_ip6) {
1367                         udp_conf.family = AF_INET6;
1368                         memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1369                                sizeof(udp_conf.local_ip6));
1370                         memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1371                                sizeof(udp_conf.peer_ip6));
1372                         udp_conf.use_udp6_tx_checksums =
1373                           !cfg->udp6_zero_tx_checksums;
1374                         udp_conf.use_udp6_rx_checksums =
1375                           !cfg->udp6_zero_rx_checksums;
1376                 } else
1377 #endif
1378                 {
1379                         udp_conf.family = AF_INET;
1380                         udp_conf.local_ip = cfg->local_ip;
1381                         udp_conf.peer_ip = cfg->peer_ip;
1382                         udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1383                 }
1384
1385                 udp_conf.local_udp_port = htons(cfg->local_udp_port);
1386                 udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1387
1388                 err = udp_sock_create(net, &udp_conf, &sock);
1389                 if (err < 0)
1390                         goto out;
1391
1392                 break;
1393
1394         case L2TP_ENCAPTYPE_IP:
1395 #if IS_ENABLED(CONFIG_IPV6)
1396                 if (cfg->local_ip6 && cfg->peer_ip6) {
1397                         struct sockaddr_l2tpip6 ip6_addr = {0};
1398
1399                         err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1400                                                IPPROTO_L2TP, &sock);
1401                         if (err < 0)
1402                                 goto out;
1403
1404                         ip6_addr.l2tp_family = AF_INET6;
1405                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1406                                sizeof(ip6_addr.l2tp_addr));
1407                         ip6_addr.l2tp_conn_id = tunnel_id;
1408                         err = kernel_bind(sock, (struct sockaddr *)&ip6_addr,
1409                                           sizeof(ip6_addr));
1410                         if (err < 0)
1411                                 goto out;
1412
1413                         ip6_addr.l2tp_family = AF_INET6;
1414                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1415                                sizeof(ip6_addr.l2tp_addr));
1416                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1417                         err = kernel_connect(sock,
1418                                              (struct sockaddr *)&ip6_addr,
1419                                              sizeof(ip6_addr), 0);
1420                         if (err < 0)
1421                                 goto out;
1422                 } else
1423 #endif
1424                 {
1425                         struct sockaddr_l2tpip ip_addr = {0};
1426
1427                         err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1428                                                IPPROTO_L2TP, &sock);
1429                         if (err < 0)
1430                                 goto out;
1431
1432                         ip_addr.l2tp_family = AF_INET;
1433                         ip_addr.l2tp_addr = cfg->local_ip;
1434                         ip_addr.l2tp_conn_id = tunnel_id;
1435                         err = kernel_bind(sock, (struct sockaddr *)&ip_addr,
1436                                           sizeof(ip_addr));
1437                         if (err < 0)
1438                                 goto out;
1439
1440                         ip_addr.l2tp_family = AF_INET;
1441                         ip_addr.l2tp_addr = cfg->peer_ip;
1442                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1443                         err = kernel_connect(sock, (struct sockaddr *)&ip_addr,
1444                                              sizeof(ip_addr), 0);
1445                         if (err < 0)
1446                                 goto out;
1447                 }
1448                 break;
1449
1450         default:
1451                 goto out;
1452         }
1453
1454 out:
1455         *sockp = sock;
1456         if (err < 0 && sock) {
1457                 kernel_sock_shutdown(sock, SHUT_RDWR);
1458                 sock_release(sock);
1459                 *sockp = NULL;
1460         }
1461
1462         return err;
1463 }
1464
1465 static struct lock_class_key l2tp_socket_class;
1466
1467 int l2tp_tunnel_create(struct net *net, int fd, int version, u32 tunnel_id, u32 peer_tunnel_id,
1468                        struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1469 {
1470         struct l2tp_tunnel *tunnel = NULL;
1471         int err;
1472         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1473
1474         if (cfg)
1475                 encap = cfg->encap;
1476
1477         tunnel = kzalloc(sizeof(*tunnel), GFP_KERNEL);
1478         if (!tunnel) {
1479                 err = -ENOMEM;
1480                 goto err;
1481         }
1482
1483         tunnel->version = version;
1484         tunnel->tunnel_id = tunnel_id;
1485         tunnel->peer_tunnel_id = peer_tunnel_id;
1486         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1487
1488         tunnel->magic = L2TP_TUNNEL_MAGIC;
1489         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1490         rwlock_init(&tunnel->hlist_lock);
1491         tunnel->acpt_newsess = true;
1492
1493         if (cfg)
1494                 tunnel->debug = cfg->debug;
1495
1496         tunnel->encap = encap;
1497
1498         refcount_set(&tunnel->ref_count, 1);
1499         tunnel->fd = fd;
1500
1501         /* Init delete workqueue struct */
1502         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1503
1504         INIT_LIST_HEAD(&tunnel->list);
1505
1506         err = 0;
1507 err:
1508         if (tunnelp)
1509                 *tunnelp = tunnel;
1510
1511         return err;
1512 }
1513 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1514
1515 static int l2tp_validate_socket(const struct sock *sk, const struct net *net,
1516                                 enum l2tp_encap_type encap)
1517 {
1518         if (!net_eq(sock_net(sk), net))
1519                 return -EINVAL;
1520
1521         if (sk->sk_type != SOCK_DGRAM)
1522                 return -EPROTONOSUPPORT;
1523
1524         if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
1525                 return -EPROTONOSUPPORT;
1526
1527         if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) ||
1528             (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP))
1529                 return -EPROTONOSUPPORT;
1530
1531         if (sk->sk_user_data)
1532                 return -EBUSY;
1533
1534         return 0;
1535 }
1536
1537 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net,
1538                          struct l2tp_tunnel_cfg *cfg)
1539 {
1540         struct l2tp_tunnel *tunnel_walk;
1541         struct l2tp_net *pn;
1542         struct socket *sock;
1543         struct sock *sk;
1544         int ret;
1545
1546         if (tunnel->fd < 0) {
1547                 ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id,
1548                                               tunnel->peer_tunnel_id, cfg,
1549                                               &sock);
1550                 if (ret < 0)
1551                         goto err;
1552         } else {
1553                 sock = sockfd_lookup(tunnel->fd, &ret);
1554                 if (!sock)
1555                         goto err;
1556
1557                 ret = l2tp_validate_socket(sock->sk, net, tunnel->encap);
1558                 if (ret < 0)
1559                         goto err_sock;
1560         }
1561
1562         tunnel->l2tp_net = net;
1563         pn = l2tp_pernet(net);
1564
1565         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1566         list_for_each_entry(tunnel_walk, &pn->l2tp_tunnel_list, list) {
1567                 if (tunnel_walk->tunnel_id == tunnel->tunnel_id) {
1568                         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1569
1570                         ret = -EEXIST;
1571                         goto err_sock;
1572                 }
1573         }
1574         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1575         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1576
1577         sk = sock->sk;
1578         sock_hold(sk);
1579         tunnel->sock = sk;
1580
1581         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1582                 struct udp_tunnel_sock_cfg udp_cfg = {
1583                         .sk_user_data = tunnel,
1584                         .encap_type = UDP_ENCAP_L2TPINUDP,
1585                         .encap_rcv = l2tp_udp_encap_recv,
1586                         .encap_destroy = l2tp_udp_encap_destroy,
1587                 };
1588
1589                 setup_udp_tunnel_sock(net, sock, &udp_cfg);
1590         } else {
1591                 sk->sk_user_data = tunnel;
1592         }
1593
1594         tunnel->old_sk_destruct = sk->sk_destruct;
1595         sk->sk_destruct = &l2tp_tunnel_destruct;
1596         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class,
1597                                    "l2tp_sock");
1598         sk->sk_allocation = GFP_ATOMIC;
1599
1600         if (tunnel->fd >= 0)
1601                 sockfd_put(sock);
1602
1603         return 0;
1604
1605 err_sock:
1606         if (tunnel->fd < 0)
1607                 sock_release(sock);
1608         else
1609                 sockfd_put(sock);
1610 err:
1611         return ret;
1612 }
1613 EXPORT_SYMBOL_GPL(l2tp_tunnel_register);
1614
1615 /* This function is used by the netlink TUNNEL_DELETE command.
1616  */
1617 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1618 {
1619         if (!test_and_set_bit(0, &tunnel->dead)) {
1620                 l2tp_tunnel_inc_refcount(tunnel);
1621                 queue_work(l2tp_wq, &tunnel->del_work);
1622         }
1623 }
1624 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1625
1626 void l2tp_session_delete(struct l2tp_session *session)
1627 {
1628         if (test_and_set_bit(0, &session->dead))
1629                 return;
1630
1631         l2tp_session_unhash(session);
1632         l2tp_session_queue_purge(session);
1633         if (session->session_close)
1634                 (*session->session_close)(session);
1635
1636         l2tp_session_dec_refcount(session);
1637 }
1638 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1639
1640 /* We come here whenever a session's send_seq, cookie_len or
1641  * l2specific_type parameters are set.
1642  */
1643 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1644 {
1645         if (version == L2TP_HDR_VER_2) {
1646                 session->hdr_len = 6;
1647                 if (session->send_seq)
1648                         session->hdr_len += 4;
1649         } else {
1650                 session->hdr_len = 4 + session->cookie_len;
1651                 session->hdr_len += l2tp_get_l2specific_len(session);
1652                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1653                         session->hdr_len += 4;
1654         }
1655 }
1656 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1657
1658 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id,
1659                                          u32 peer_session_id, struct l2tp_session_cfg *cfg)
1660 {
1661         struct l2tp_session *session;
1662
1663         session = kzalloc(sizeof(*session) + priv_size, GFP_KERNEL);
1664         if (session) {
1665                 session->magic = L2TP_SESSION_MAGIC;
1666                 session->tunnel = tunnel;
1667
1668                 session->session_id = session_id;
1669                 session->peer_session_id = peer_session_id;
1670                 session->nr = 0;
1671                 if (tunnel->version == L2TP_HDR_VER_2)
1672                         session->nr_max = 0xffff;
1673                 else
1674                         session->nr_max = 0xffffff;
1675                 session->nr_window_size = session->nr_max / 2;
1676                 session->nr_oos_count_max = 4;
1677
1678                 /* Use NR of first received packet */
1679                 session->reorder_skip = 1;
1680
1681                 sprintf(&session->name[0], "sess %u/%u",
1682                         tunnel->tunnel_id, session->session_id);
1683
1684                 skb_queue_head_init(&session->reorder_q);
1685
1686                 INIT_HLIST_NODE(&session->hlist);
1687                 INIT_HLIST_NODE(&session->global_hlist);
1688
1689                 /* Inherit debug options from tunnel */
1690                 session->debug = tunnel->debug;
1691
1692                 if (cfg) {
1693                         session->pwtype = cfg->pw_type;
1694                         session->debug = cfg->debug;
1695                         session->send_seq = cfg->send_seq;
1696                         session->recv_seq = cfg->recv_seq;
1697                         session->lns_mode = cfg->lns_mode;
1698                         session->reorder_timeout = cfg->reorder_timeout;
1699                         session->l2specific_type = cfg->l2specific_type;
1700                         session->cookie_len = cfg->cookie_len;
1701                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1702                         session->peer_cookie_len = cfg->peer_cookie_len;
1703                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1704                 }
1705
1706                 l2tp_session_set_header_len(session, tunnel->version);
1707
1708                 refcount_set(&session->ref_count, 1);
1709
1710                 return session;
1711         }
1712
1713         return ERR_PTR(-ENOMEM);
1714 }
1715 EXPORT_SYMBOL_GPL(l2tp_session_create);
1716
1717 /*****************************************************************************
1718  * Init and cleanup
1719  *****************************************************************************/
1720
1721 static __net_init int l2tp_init_net(struct net *net)
1722 {
1723         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1724         int hash;
1725
1726         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1727         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1728
1729         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1730                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1731
1732         spin_lock_init(&pn->l2tp_session_hlist_lock);
1733
1734         return 0;
1735 }
1736
1737 static __net_exit void l2tp_exit_net(struct net *net)
1738 {
1739         struct l2tp_net *pn = l2tp_pernet(net);
1740         struct l2tp_tunnel *tunnel = NULL;
1741         int hash;
1742
1743         rcu_read_lock_bh();
1744         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1745                 l2tp_tunnel_delete(tunnel);
1746         }
1747         rcu_read_unlock_bh();
1748
1749         if (l2tp_wq)
1750                 flush_workqueue(l2tp_wq);
1751         rcu_barrier();
1752
1753         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1754                 WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1755 }
1756
1757 static struct pernet_operations l2tp_net_ops = {
1758         .init = l2tp_init_net,
1759         .exit = l2tp_exit_net,
1760         .id   = &l2tp_net_id,
1761         .size = sizeof(struct l2tp_net),
1762 };
1763
1764 static int __init l2tp_init(void)
1765 {
1766         int rc = 0;
1767
1768         rc = register_pernet_device(&l2tp_net_ops);
1769         if (rc)
1770                 goto out;
1771
1772         l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1773         if (!l2tp_wq) {
1774                 pr_err("alloc_workqueue failed\n");
1775                 unregister_pernet_device(&l2tp_net_ops);
1776                 rc = -ENOMEM;
1777                 goto out;
1778         }
1779
1780         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1781
1782 out:
1783         return rc;
1784 }
1785
1786 static void __exit l2tp_exit(void)
1787 {
1788         unregister_pernet_device(&l2tp_net_ops);
1789         if (l2tp_wq) {
1790                 destroy_workqueue(l2tp_wq);
1791                 l2tp_wq = NULL;
1792         }
1793 }
1794
1795 module_init(l2tp_init);
1796 module_exit(l2tp_exit);
1797
1798 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1799 MODULE_DESCRIPTION("L2TP core");
1800 MODULE_LICENSE("GPL");
1801 MODULE_VERSION(L2TP_DRV_VERSION);