]> git.karo-electronics.de Git - karo-tx-linux.git/blob - net/l2tp/l2tp_core.c
Merge tag 'batman-adv-fix-for-davem' of git://git.open-mesh.org/linux-merge
[karo-tx-linux.git] / 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/inet_common.h>
56 #include <net/xfrm.h>
57 #include <net/protocol.h>
58 #include <net/inet6_connection_sock.h>
59 #include <net/inet_ecn.h>
60 #include <net/ip6_route.h>
61 #include <net/ip6_checksum.h>
62
63 #include <asm/byteorder.h>
64 #include <linux/atomic.h>
65
66 #include "l2tp_core.h"
67
68 #define L2TP_DRV_VERSION        "V2.0"
69
70 /* L2TP header constants */
71 #define L2TP_HDRFLAG_T     0x8000
72 #define L2TP_HDRFLAG_L     0x4000
73 #define L2TP_HDRFLAG_S     0x0800
74 #define L2TP_HDRFLAG_O     0x0200
75 #define L2TP_HDRFLAG_P     0x0100
76
77 #define L2TP_HDR_VER_MASK  0x000F
78 #define L2TP_HDR_VER_2     0x0002
79 #define L2TP_HDR_VER_3     0x0003
80
81 /* L2TPv3 default L2-specific sublayer */
82 #define L2TP_SLFLAG_S      0x40000000
83 #define L2TP_SL_SEQ_MASK   0x00ffffff
84
85 #define L2TP_HDR_SIZE_SEQ               10
86 #define L2TP_HDR_SIZE_NOSEQ             6
87
88 /* Default trace flags */
89 #define L2TP_DEFAULT_DEBUG_FLAGS        0
90
91 /* Private data stored for received packets in the skb.
92  */
93 struct l2tp_skb_cb {
94         u32                     ns;
95         u16                     has_seq;
96         u16                     length;
97         unsigned long           expires;
98 };
99
100 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
101
102 static atomic_t l2tp_tunnel_count;
103 static atomic_t l2tp_session_count;
104 static struct workqueue_struct *l2tp_wq;
105
106 /* per-net private data for this module */
107 static unsigned int l2tp_net_id;
108 struct l2tp_net {
109         struct list_head l2tp_tunnel_list;
110         spinlock_t l2tp_tunnel_list_lock;
111         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
112         spinlock_t l2tp_session_hlist_lock;
113 };
114
115 static void l2tp_session_set_header_len(struct l2tp_session *session, int version);
116 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel);
117
118 static inline struct l2tp_net *l2tp_pernet(struct net *net)
119 {
120         BUG_ON(!net);
121
122         return net_generic(net, l2tp_net_id);
123 }
124
125 /* Tunnel reference counts. Incremented per session that is added to
126  * the tunnel.
127  */
128 static inline void l2tp_tunnel_inc_refcount_1(struct l2tp_tunnel *tunnel)
129 {
130         atomic_inc(&tunnel->ref_count);
131 }
132
133 static inline void l2tp_tunnel_dec_refcount_1(struct l2tp_tunnel *tunnel)
134 {
135         if (atomic_dec_and_test(&tunnel->ref_count))
136                 l2tp_tunnel_free(tunnel);
137 }
138 #ifdef L2TP_REFCNT_DEBUG
139 #define l2tp_tunnel_inc_refcount(_t)                                    \
140 do {                                                                    \
141         pr_debug("l2tp_tunnel_inc_refcount: %s:%d %s: cnt=%d\n",        \
142                  __func__, __LINE__, (_t)->name,                        \
143                  atomic_read(&_t->ref_count));                          \
144         l2tp_tunnel_inc_refcount_1(_t);                                 \
145 } while (0)
146 #define l2tp_tunnel_dec_refcount(_t)
147 do {                                                                    \
148         pr_debug("l2tp_tunnel_dec_refcount: %s:%d %s: cnt=%d\n",        \
149                  __func__, __LINE__, (_t)->name,                        \
150                  atomic_read(&_t->ref_count));                          \
151         l2tp_tunnel_dec_refcount_1(_t);                                 \
152 } while (0)
153 #else
154 #define l2tp_tunnel_inc_refcount(t) l2tp_tunnel_inc_refcount_1(t)
155 #define l2tp_tunnel_dec_refcount(t) l2tp_tunnel_dec_refcount_1(t)
156 #endif
157
158 /* Session hash global list for L2TPv3.
159  * The session_id SHOULD be random according to RFC3931, but several
160  * L2TP implementations use incrementing session_ids.  So we do a real
161  * hash on the session_id, rather than a simple bitmask.
162  */
163 static inline struct hlist_head *
164 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
165 {
166         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
167
168 }
169
170 /* Lookup the tunnel socket, possibly involving the fs code if the socket is
171  * owned by userspace.  A struct sock returned from this function must be
172  * released using l2tp_tunnel_sock_put once you're done with it.
173  */
174 struct sock *l2tp_tunnel_sock_lookup(struct l2tp_tunnel *tunnel)
175 {
176         int err = 0;
177         struct socket *sock = NULL;
178         struct sock *sk = NULL;
179
180         if (!tunnel)
181                 goto out;
182
183         if (tunnel->fd >= 0) {
184                 /* Socket is owned by userspace, who might be in the process
185                  * of closing it.  Look the socket up using the fd to ensure
186                  * consistency.
187                  */
188                 sock = sockfd_lookup(tunnel->fd, &err);
189                 if (sock)
190                         sk = sock->sk;
191         } else {
192                 /* Socket is owned by kernelspace */
193                 sk = tunnel->sock;
194                 sock_hold(sk);
195         }
196
197 out:
198         return sk;
199 }
200 EXPORT_SYMBOL_GPL(l2tp_tunnel_sock_lookup);
201
202 /* Drop a reference to a tunnel socket obtained via. l2tp_tunnel_sock_put */
203 void l2tp_tunnel_sock_put(struct sock *sk)
204 {
205         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
206         if (tunnel) {
207                 if (tunnel->fd >= 0) {
208                         /* Socket is owned by userspace */
209                         sockfd_put(sk->sk_socket);
210                 }
211                 sock_put(sk);
212         }
213         sock_put(sk);
214 }
215 EXPORT_SYMBOL_GPL(l2tp_tunnel_sock_put);
216
217 /* Lookup a session by id in the global session list
218  */
219 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
220 {
221         struct l2tp_net *pn = l2tp_pernet(net);
222         struct hlist_head *session_list =
223                 l2tp_session_id_hash_2(pn, session_id);
224         struct l2tp_session *session;
225
226         rcu_read_lock_bh();
227         hlist_for_each_entry_rcu(session, session_list, global_hlist) {
228                 if (session->session_id == session_id) {
229                         rcu_read_unlock_bh();
230                         return session;
231                 }
232         }
233         rcu_read_unlock_bh();
234
235         return NULL;
236 }
237
238 /* Session hash list.
239  * The session_id SHOULD be random according to RFC2661, but several
240  * L2TP implementations (Cisco and Microsoft) use incrementing
241  * session_ids.  So we do a real hash on the session_id, rather than a
242  * simple bitmask.
243  */
244 static inline struct hlist_head *
245 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
246 {
247         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
248 }
249
250 /* Lookup a session by id
251  */
252 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
253 {
254         struct hlist_head *session_list;
255         struct l2tp_session *session;
256
257         /* In L2TPv3, session_ids are unique over all tunnels and we
258          * sometimes need to look them up before we know the
259          * tunnel.
260          */
261         if (tunnel == NULL)
262                 return l2tp_session_find_2(net, session_id);
263
264         session_list = l2tp_session_id_hash(tunnel, session_id);
265         read_lock_bh(&tunnel->hlist_lock);
266         hlist_for_each_entry(session, session_list, hlist) {
267                 if (session->session_id == session_id) {
268                         read_unlock_bh(&tunnel->hlist_lock);
269                         return session;
270                 }
271         }
272         read_unlock_bh(&tunnel->hlist_lock);
273
274         return NULL;
275 }
276 EXPORT_SYMBOL_GPL(l2tp_session_find);
277
278 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
279 {
280         int hash;
281         struct l2tp_session *session;
282         int count = 0;
283
284         read_lock_bh(&tunnel->hlist_lock);
285         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
286                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
287                         if (++count > nth) {
288                                 read_unlock_bh(&tunnel->hlist_lock);
289                                 return session;
290                         }
291                 }
292         }
293
294         read_unlock_bh(&tunnel->hlist_lock);
295
296         return NULL;
297 }
298 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
299
300 /* Lookup a session by interface name.
301  * This is very inefficient but is only used by management interfaces.
302  */
303 struct l2tp_session *l2tp_session_find_by_ifname(struct net *net, char *ifname)
304 {
305         struct l2tp_net *pn = l2tp_pernet(net);
306         int hash;
307         struct l2tp_session *session;
308
309         rcu_read_lock_bh();
310         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
311                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
312                         if (!strcmp(session->ifname, ifname)) {
313                                 rcu_read_unlock_bh();
314                                 return session;
315                         }
316                 }
317         }
318
319         rcu_read_unlock_bh();
320
321         return NULL;
322 }
323 EXPORT_SYMBOL_GPL(l2tp_session_find_by_ifname);
324
325 /* Lookup a tunnel by id
326  */
327 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
328 {
329         struct l2tp_tunnel *tunnel;
330         struct l2tp_net *pn = l2tp_pernet(net);
331
332         rcu_read_lock_bh();
333         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
334                 if (tunnel->tunnel_id == tunnel_id) {
335                         rcu_read_unlock_bh();
336                         return tunnel;
337                 }
338         }
339         rcu_read_unlock_bh();
340
341         return NULL;
342 }
343 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
344
345 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
346 {
347         struct l2tp_net *pn = l2tp_pernet(net);
348         struct l2tp_tunnel *tunnel;
349         int count = 0;
350
351         rcu_read_lock_bh();
352         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
353                 if (++count > nth) {
354                         rcu_read_unlock_bh();
355                         return tunnel;
356                 }
357         }
358
359         rcu_read_unlock_bh();
360
361         return NULL;
362 }
363 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
364
365 /*****************************************************************************
366  * Receive data handling
367  *****************************************************************************/
368
369 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
370  * number.
371  */
372 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
373 {
374         struct sk_buff *skbp;
375         struct sk_buff *tmp;
376         u32 ns = L2TP_SKB_CB(skb)->ns;
377
378         spin_lock_bh(&session->reorder_q.lock);
379         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
380                 if (L2TP_SKB_CB(skbp)->ns > ns) {
381                         __skb_queue_before(&session->reorder_q, skbp, skb);
382                         l2tp_dbg(session, L2TP_MSG_SEQ,
383                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
384                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
385                                  skb_queue_len(&session->reorder_q));
386                         atomic_long_inc(&session->stats.rx_oos_packets);
387                         goto out;
388                 }
389         }
390
391         __skb_queue_tail(&session->reorder_q, skb);
392
393 out:
394         spin_unlock_bh(&session->reorder_q.lock);
395 }
396
397 /* Dequeue a single skb.
398  */
399 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
400 {
401         struct l2tp_tunnel *tunnel = session->tunnel;
402         int length = L2TP_SKB_CB(skb)->length;
403
404         /* We're about to requeue the skb, so return resources
405          * to its current owner (a socket receive buffer).
406          */
407         skb_orphan(skb);
408
409         atomic_long_inc(&tunnel->stats.rx_packets);
410         atomic_long_add(length, &tunnel->stats.rx_bytes);
411         atomic_long_inc(&session->stats.rx_packets);
412         atomic_long_add(length, &session->stats.rx_bytes);
413
414         if (L2TP_SKB_CB(skb)->has_seq) {
415                 /* Bump our Nr */
416                 session->nr++;
417                 session->nr &= session->nr_max;
418
419                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
420                          session->name, session->nr);
421         }
422
423         /* call private receive handler */
424         if (session->recv_skb != NULL)
425                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
426         else
427                 kfree_skb(skb);
428
429         if (session->deref)
430                 (*session->deref)(session);
431 }
432
433 /* Dequeue skbs from the session's reorder_q, subject to packet order.
434  * Skbs that have been in the queue for too long are simply discarded.
435  */
436 static void l2tp_recv_dequeue(struct l2tp_session *session)
437 {
438         struct sk_buff *skb;
439         struct sk_buff *tmp;
440
441         /* If the pkt at the head of the queue has the nr that we
442          * expect to send up next, dequeue it and any other
443          * in-sequence packets behind it.
444          */
445 start:
446         spin_lock_bh(&session->reorder_q.lock);
447         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
448                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
449                         atomic_long_inc(&session->stats.rx_seq_discards);
450                         atomic_long_inc(&session->stats.rx_errors);
451                         l2tp_dbg(session, L2TP_MSG_SEQ,
452                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
453                                  session->name, L2TP_SKB_CB(skb)->ns,
454                                  L2TP_SKB_CB(skb)->length, session->nr,
455                                  skb_queue_len(&session->reorder_q));
456                         session->reorder_skip = 1;
457                         __skb_unlink(skb, &session->reorder_q);
458                         kfree_skb(skb);
459                         if (session->deref)
460                                 (*session->deref)(session);
461                         continue;
462                 }
463
464                 if (L2TP_SKB_CB(skb)->has_seq) {
465                         if (session->reorder_skip) {
466                                 l2tp_dbg(session, L2TP_MSG_SEQ,
467                                          "%s: advancing nr to next pkt: %u -> %u",
468                                          session->name, session->nr,
469                                          L2TP_SKB_CB(skb)->ns);
470                                 session->reorder_skip = 0;
471                                 session->nr = L2TP_SKB_CB(skb)->ns;
472                         }
473                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
474                                 l2tp_dbg(session, L2TP_MSG_SEQ,
475                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
476                                          session->name, L2TP_SKB_CB(skb)->ns,
477                                          L2TP_SKB_CB(skb)->length, session->nr,
478                                          skb_queue_len(&session->reorder_q));
479                                 goto out;
480                         }
481                 }
482                 __skb_unlink(skb, &session->reorder_q);
483
484                 /* Process the skb. We release the queue lock while we
485                  * do so to let other contexts process the queue.
486                  */
487                 spin_unlock_bh(&session->reorder_q.lock);
488                 l2tp_recv_dequeue_skb(session, skb);
489                 goto start;
490         }
491
492 out:
493         spin_unlock_bh(&session->reorder_q.lock);
494 }
495
496 static inline int l2tp_verify_udp_checksum(struct sock *sk,
497                                            struct sk_buff *skb)
498 {
499         struct l2tp_tunnel *tunnel = (struct l2tp_tunnel *)sk->sk_user_data;
500         struct udphdr *uh = udp_hdr(skb);
501         u16 ulen = ntohs(uh->len);
502         __wsum psum;
503
504         if (sk->sk_no_check || skb_csum_unnecessary(skb))
505                 return 0;
506
507 #if IS_ENABLED(CONFIG_IPV6)
508         if (sk->sk_family == PF_INET6 && !tunnel->v4mapped) {
509                 if (!uh->check) {
510                         LIMIT_NETDEBUG(KERN_INFO "L2TP: IPv6: checksum is 0\n");
511                         return 1;
512                 }
513                 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
514                     !csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
515                                      &ipv6_hdr(skb)->daddr, ulen,
516                                      IPPROTO_UDP, skb->csum)) {
517                         skb->ip_summed = CHECKSUM_UNNECESSARY;
518                         return 0;
519                 }
520                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
521                                                          &ipv6_hdr(skb)->daddr,
522                                                          skb->len, IPPROTO_UDP,
523                                                          0));
524         } else
525 #endif
526         {
527                 struct inet_sock *inet;
528                 if (!uh->check)
529                         return 0;
530                 inet = inet_sk(sk);
531                 psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr,
532                                           ulen, IPPROTO_UDP, 0);
533
534                 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
535                     !csum_fold(csum_add(psum, skb->csum)))
536                         return 0;
537                 skb->csum = psum;
538         }
539
540         return __skb_checksum_complete(skb);
541 }
542
543 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
544 {
545         u32 nws;
546
547         if (nr >= session->nr)
548                 nws = nr - session->nr;
549         else
550                 nws = (session->nr_max + 1) - (session->nr - nr);
551
552         return nws < session->nr_window_size;
553 }
554
555 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
556  * acceptable, else non-zero.
557  */
558 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
559 {
560         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
561                 /* Packet sequence number is outside allowed window.
562                  * Discard it.
563                  */
564                 l2tp_dbg(session, L2TP_MSG_SEQ,
565                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
566                          session->name, L2TP_SKB_CB(skb)->ns,
567                          L2TP_SKB_CB(skb)->length, session->nr);
568                 goto discard;
569         }
570
571         if (session->reorder_timeout != 0) {
572                 /* Packet reordering enabled. Add skb to session's
573                  * reorder queue, in order of ns.
574                  */
575                 l2tp_recv_queue_skb(session, skb);
576                 goto out;
577         }
578
579         /* Packet reordering disabled. Discard out-of-sequence packets, while
580          * tracking the number if in-sequence packets after the first OOS packet
581          * is seen. After nr_oos_count_max in-sequence packets, reset the
582          * sequence number to re-enable packet reception.
583          */
584         if (L2TP_SKB_CB(skb)->ns == session->nr) {
585                 skb_queue_tail(&session->reorder_q, skb);
586         } else {
587                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
588                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
589
590                 if (nr_oos == nr_next)
591                         session->nr_oos_count++;
592                 else
593                         session->nr_oos_count = 0;
594
595                 session->nr_oos = nr_oos;
596                 if (session->nr_oos_count > session->nr_oos_count_max) {
597                         session->reorder_skip = 1;
598                         l2tp_dbg(session, L2TP_MSG_SEQ,
599                                  "%s: %d oos packets received. Resetting sequence numbers\n",
600                                  session->name, session->nr_oos_count);
601                 }
602                 if (!session->reorder_skip) {
603                         atomic_long_inc(&session->stats.rx_seq_discards);
604                         l2tp_dbg(session, L2TP_MSG_SEQ,
605                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
606                                  session->name, L2TP_SKB_CB(skb)->ns,
607                                  L2TP_SKB_CB(skb)->length, session->nr,
608                                  skb_queue_len(&session->reorder_q));
609                         goto discard;
610                 }
611                 skb_queue_tail(&session->reorder_q, skb);
612         }
613
614 out:
615         return 0;
616
617 discard:
618         return 1;
619 }
620
621 /* Do receive processing of L2TP data frames. We handle both L2TPv2
622  * and L2TPv3 data frames here.
623  *
624  * L2TPv2 Data Message Header
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  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
630  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
631  * |           Tunnel ID           |           Session ID          |
632  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
633  * |             Ns (opt)          |             Nr (opt)          |
634  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
635  * |      Offset Size (opt)        |    Offset pad... (opt)
636  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
637  *
638  * Data frames are marked by T=0. All other fields are the same as
639  * those in L2TP control frames.
640  *
641  * L2TPv3 Data Message Header
642  *
643  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
644  * |                      L2TP Session Header                      |
645  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
646  * |                      L2-Specific Sublayer                     |
647  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
648  * |                        Tunnel Payload                      ...
649  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
650  *
651  * L2TPv3 Session Header Over IP
652  *
653  *  0                   1                   2                   3
654  *  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
655  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
656  * |                           Session ID                          |
657  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
658  * |               Cookie (optional, maximum 64 bits)...
659  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
660  *                                                                 |
661  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
662  *
663  * L2TPv3 L2-Specific Sublayer Format
664  *
665  *  0                   1                   2                   3
666  *  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
667  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
668  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
669  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
670  *
671  * Cookie value, sublayer format and offset (pad) are negotiated with
672  * the peer when the session is set up. Unlike L2TPv2, we do not need
673  * to parse the packet header to determine if optional fields are
674  * present.
675  *
676  * Caller must already have parsed the frame and determined that it is
677  * a data (not control) frame before coming here. Fields up to the
678  * session-id have already been parsed and ptr points to the data
679  * after the session-id.
680  */
681 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
682                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
683                       int length, int (*payload_hook)(struct sk_buff *skb))
684 {
685         struct l2tp_tunnel *tunnel = session->tunnel;
686         int offset;
687         u32 ns, nr;
688
689         /* The ref count is increased since we now hold a pointer to
690          * the session. Take care to decrement the refcnt when exiting
691          * this function from now on...
692          */
693         l2tp_session_inc_refcount(session);
694         if (session->ref)
695                 (*session->ref)(session);
696
697         /* Parse and check optional cookie */
698         if (session->peer_cookie_len > 0) {
699                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
700                         l2tp_info(tunnel, L2TP_MSG_DATA,
701                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
702                                   tunnel->name, tunnel->tunnel_id,
703                                   session->session_id);
704                         atomic_long_inc(&session->stats.rx_cookie_discards);
705                         goto discard;
706                 }
707                 ptr += session->peer_cookie_len;
708         }
709
710         /* Handle the optional sequence numbers. Sequence numbers are
711          * in different places for L2TPv2 and L2TPv3.
712          *
713          * If we are the LAC, enable/disable sequence numbers under
714          * the control of the LNS.  If no sequence numbers present but
715          * we were expecting them, discard frame.
716          */
717         ns = nr = 0;
718         L2TP_SKB_CB(skb)->has_seq = 0;
719         if (tunnel->version == L2TP_HDR_VER_2) {
720                 if (hdrflags & L2TP_HDRFLAG_S) {
721                         ns = ntohs(*(__be16 *) ptr);
722                         ptr += 2;
723                         nr = ntohs(*(__be16 *) ptr);
724                         ptr += 2;
725
726                         /* Store L2TP info in the skb */
727                         L2TP_SKB_CB(skb)->ns = ns;
728                         L2TP_SKB_CB(skb)->has_seq = 1;
729
730                         l2tp_dbg(session, L2TP_MSG_SEQ,
731                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
732                                  session->name, ns, nr, session->nr);
733                 }
734         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
735                 u32 l2h = ntohl(*(__be32 *) ptr);
736
737                 if (l2h & 0x40000000) {
738                         ns = l2h & 0x00ffffff;
739
740                         /* Store L2TP info in the skb */
741                         L2TP_SKB_CB(skb)->ns = ns;
742                         L2TP_SKB_CB(skb)->has_seq = 1;
743
744                         l2tp_dbg(session, L2TP_MSG_SEQ,
745                                  "%s: recv data ns=%u, session nr=%u\n",
746                                  session->name, ns, session->nr);
747                 }
748         }
749
750         /* Advance past L2-specific header, if present */
751         ptr += session->l2specific_len;
752
753         if (L2TP_SKB_CB(skb)->has_seq) {
754                 /* Received a packet with sequence numbers. If we're the LNS,
755                  * check if we sre sending sequence numbers and if not,
756                  * configure it so.
757                  */
758                 if ((!session->lns_mode) && (!session->send_seq)) {
759                         l2tp_info(session, L2TP_MSG_SEQ,
760                                   "%s: requested to enable seq numbers by LNS\n",
761                                   session->name);
762                         session->send_seq = -1;
763                         l2tp_session_set_header_len(session, tunnel->version);
764                 }
765         } else {
766                 /* No sequence numbers.
767                  * If user has configured mandatory sequence numbers, discard.
768                  */
769                 if (session->recv_seq) {
770                         l2tp_warn(session, L2TP_MSG_SEQ,
771                                   "%s: recv data has no seq numbers when required. Discarding.\n",
772                                   session->name);
773                         atomic_long_inc(&session->stats.rx_seq_discards);
774                         goto discard;
775                 }
776
777                 /* If we're the LAC and we're sending sequence numbers, the
778                  * LNS has requested that we no longer send sequence numbers.
779                  * If we're the LNS and we're sending sequence numbers, the
780                  * LAC is broken. Discard the frame.
781                  */
782                 if ((!session->lns_mode) && (session->send_seq)) {
783                         l2tp_info(session, L2TP_MSG_SEQ,
784                                   "%s: requested to disable seq numbers by LNS\n",
785                                   session->name);
786                         session->send_seq = 0;
787                         l2tp_session_set_header_len(session, tunnel->version);
788                 } else if (session->send_seq) {
789                         l2tp_warn(session, L2TP_MSG_SEQ,
790                                   "%s: recv data has no seq numbers when required. Discarding.\n",
791                                   session->name);
792                         atomic_long_inc(&session->stats.rx_seq_discards);
793                         goto discard;
794                 }
795         }
796
797         /* Session data offset is handled differently for L2TPv2 and
798          * L2TPv3. For L2TPv2, there is an optional 16-bit value in
799          * the header. For L2TPv3, the offset is negotiated using AVPs
800          * in the session setup control protocol.
801          */
802         if (tunnel->version == L2TP_HDR_VER_2) {
803                 /* If offset bit set, skip it. */
804                 if (hdrflags & L2TP_HDRFLAG_O) {
805                         offset = ntohs(*(__be16 *)ptr);
806                         ptr += 2 + offset;
807                 }
808         } else
809                 ptr += session->offset;
810
811         offset = ptr - optr;
812         if (!pskb_may_pull(skb, offset))
813                 goto discard;
814
815         __skb_pull(skb, offset);
816
817         /* If caller wants to process the payload before we queue the
818          * packet, do so now.
819          */
820         if (payload_hook)
821                 if ((*payload_hook)(skb))
822                         goto discard;
823
824         /* Prepare skb for adding to the session's reorder_q.  Hold
825          * packets for max reorder_timeout or 1 second if not
826          * reordering.
827          */
828         L2TP_SKB_CB(skb)->length = length;
829         L2TP_SKB_CB(skb)->expires = jiffies +
830                 (session->reorder_timeout ? session->reorder_timeout : HZ);
831
832         /* Add packet to the session's receive queue. Reordering is done here, if
833          * enabled. Saved L2TP protocol info is stored in skb->sb[].
834          */
835         if (L2TP_SKB_CB(skb)->has_seq) {
836                 if (l2tp_recv_data_seq(session, skb))
837                         goto discard;
838         } else {
839                 /* No sequence numbers. Add the skb to the tail of the
840                  * reorder queue. This ensures that it will be
841                  * delivered after all previous sequenced skbs.
842                  */
843                 skb_queue_tail(&session->reorder_q, skb);
844         }
845
846         /* Try to dequeue as many skbs from reorder_q as we can. */
847         l2tp_recv_dequeue(session);
848
849         l2tp_session_dec_refcount(session);
850
851         return;
852
853 discard:
854         atomic_long_inc(&session->stats.rx_errors);
855         kfree_skb(skb);
856
857         if (session->deref)
858                 (*session->deref)(session);
859
860         l2tp_session_dec_refcount(session);
861 }
862 EXPORT_SYMBOL(l2tp_recv_common);
863
864 /* Drop skbs from the session's reorder_q
865  */
866 int l2tp_session_queue_purge(struct l2tp_session *session)
867 {
868         struct sk_buff *skb = NULL;
869         BUG_ON(!session);
870         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
871         while ((skb = skb_dequeue(&session->reorder_q))) {
872                 atomic_long_inc(&session->stats.rx_errors);
873                 kfree_skb(skb);
874                 if (session->deref)
875                         (*session->deref)(session);
876         }
877         return 0;
878 }
879 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
880
881 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
882  * here. The skb is not on a list when we get here.
883  * Returns 0 if the packet was a data packet and was successfully passed on.
884  * Returns 1 if the packet was not a good data packet and could not be
885  * forwarded.  All such packets are passed up to userspace to deal with.
886  */
887 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
888                               int (*payload_hook)(struct sk_buff *skb))
889 {
890         struct l2tp_session *session = NULL;
891         unsigned char *ptr, *optr;
892         u16 hdrflags;
893         u32 tunnel_id, session_id;
894         u16 version;
895         int length;
896
897         if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
898                 goto discard_bad_csum;
899
900         /* UDP always verifies the packet length. */
901         __skb_pull(skb, sizeof(struct udphdr));
902
903         /* Short packet? */
904         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
905                 l2tp_info(tunnel, L2TP_MSG_DATA,
906                           "%s: recv short packet (len=%d)\n",
907                           tunnel->name, skb->len);
908                 goto error;
909         }
910
911         /* Trace packet contents, if enabled */
912         if (tunnel->debug & L2TP_MSG_DATA) {
913                 length = min(32u, skb->len);
914                 if (!pskb_may_pull(skb, length))
915                         goto error;
916
917                 pr_debug("%s: recv\n", tunnel->name);
918                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
919         }
920
921         /* Point to L2TP header */
922         optr = ptr = skb->data;
923
924         /* Get L2TP header flags */
925         hdrflags = ntohs(*(__be16 *) ptr);
926
927         /* Check protocol version */
928         version = hdrflags & L2TP_HDR_VER_MASK;
929         if (version != tunnel->version) {
930                 l2tp_info(tunnel, L2TP_MSG_DATA,
931                           "%s: recv protocol version mismatch: got %d expected %d\n",
932                           tunnel->name, version, tunnel->version);
933                 goto error;
934         }
935
936         /* Get length of L2TP packet */
937         length = skb->len;
938
939         /* If type is control packet, it is handled by userspace. */
940         if (hdrflags & L2TP_HDRFLAG_T) {
941                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
942                          "%s: recv control packet, len=%d\n",
943                          tunnel->name, length);
944                 goto error;
945         }
946
947         /* Skip flags */
948         ptr += 2;
949
950         if (tunnel->version == L2TP_HDR_VER_2) {
951                 /* If length is present, skip it */
952                 if (hdrflags & L2TP_HDRFLAG_L)
953                         ptr += 2;
954
955                 /* Extract tunnel and session ID */
956                 tunnel_id = ntohs(*(__be16 *) ptr);
957                 ptr += 2;
958                 session_id = ntohs(*(__be16 *) ptr);
959                 ptr += 2;
960         } else {
961                 ptr += 2;       /* skip reserved bits */
962                 tunnel_id = tunnel->tunnel_id;
963                 session_id = ntohl(*(__be32 *) ptr);
964                 ptr += 4;
965         }
966
967         /* Find the session context */
968         session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
969         if (!session || !session->recv_skb) {
970                 /* Not found? Pass to userspace to deal with */
971                 l2tp_info(tunnel, L2TP_MSG_DATA,
972                           "%s: no session found (%u/%u). Passing up.\n",
973                           tunnel->name, tunnel_id, session_id);
974                 goto error;
975         }
976
977         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
978
979         return 0;
980
981 discard_bad_csum:
982         LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
983         UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
984         atomic_long_inc(&tunnel->stats.rx_errors);
985         kfree_skb(skb);
986
987         return 0;
988
989 error:
990         /* Put UDP header back */
991         __skb_push(skb, sizeof(struct udphdr));
992
993         return 1;
994 }
995
996 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
997  * Return codes:
998  * 0 : success.
999  * <0: error
1000  * >0: skb should be passed up to userspace as UDP.
1001  */
1002 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1003 {
1004         struct l2tp_tunnel *tunnel;
1005
1006         tunnel = l2tp_sock_to_tunnel(sk);
1007         if (tunnel == NULL)
1008                 goto pass_up;
1009
1010         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
1011                  tunnel->name, skb->len);
1012
1013         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
1014                 goto pass_up_put;
1015
1016         sock_put(sk);
1017         return 0;
1018
1019 pass_up_put:
1020         sock_put(sk);
1021 pass_up:
1022         return 1;
1023 }
1024 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
1025
1026 /************************************************************************
1027  * Transmit handling
1028  ***********************************************************************/
1029
1030 /* Build an L2TP header for the session into the buffer provided.
1031  */
1032 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
1033 {
1034         struct l2tp_tunnel *tunnel = session->tunnel;
1035         __be16 *bufp = buf;
1036         __be16 *optr = buf;
1037         u16 flags = L2TP_HDR_VER_2;
1038         u32 tunnel_id = tunnel->peer_tunnel_id;
1039         u32 session_id = session->peer_session_id;
1040
1041         if (session->send_seq)
1042                 flags |= L2TP_HDRFLAG_S;
1043
1044         /* Setup L2TP header. */
1045         *bufp++ = htons(flags);
1046         *bufp++ = htons(tunnel_id);
1047         *bufp++ = htons(session_id);
1048         if (session->send_seq) {
1049                 *bufp++ = htons(session->ns);
1050                 *bufp++ = 0;
1051                 session->ns++;
1052                 session->ns &= 0xffff;
1053                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
1054                          session->name, session->ns);
1055         }
1056
1057         return bufp - optr;
1058 }
1059
1060 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1061 {
1062         struct l2tp_tunnel *tunnel = session->tunnel;
1063         char *bufp = buf;
1064         char *optr = bufp;
1065
1066         /* Setup L2TP header. The header differs slightly for UDP and
1067          * IP encapsulations. For UDP, there is 4 bytes of flags.
1068          */
1069         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1070                 u16 flags = L2TP_HDR_VER_3;
1071                 *((__be16 *) bufp) = htons(flags);
1072                 bufp += 2;
1073                 *((__be16 *) bufp) = 0;
1074                 bufp += 2;
1075         }
1076
1077         *((__be32 *) bufp) = htonl(session->peer_session_id);
1078         bufp += 4;
1079         if (session->cookie_len) {
1080                 memcpy(bufp, &session->cookie[0], session->cookie_len);
1081                 bufp += session->cookie_len;
1082         }
1083         if (session->l2specific_len) {
1084                 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1085                         u32 l2h = 0;
1086                         if (session->send_seq) {
1087                                 l2h = 0x40000000 | session->ns;
1088                                 session->ns++;
1089                                 session->ns &= 0xffffff;
1090                                 l2tp_dbg(session, L2TP_MSG_SEQ,
1091                                          "%s: updated ns to %u\n",
1092                                          session->name, session->ns);
1093                         }
1094
1095                         *((__be32 *) bufp) = htonl(l2h);
1096                 }
1097                 bufp += session->l2specific_len;
1098         }
1099         if (session->offset)
1100                 bufp += session->offset;
1101
1102         return bufp - optr;
1103 }
1104
1105 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1106                           struct flowi *fl, size_t data_len)
1107 {
1108         struct l2tp_tunnel *tunnel = session->tunnel;
1109         unsigned int len = skb->len;
1110         int error;
1111
1112         /* Debug */
1113         if (session->send_seq)
1114                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes, ns=%u\n",
1115                          session->name, data_len, session->ns - 1);
1116         else
1117                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes\n",
1118                          session->name, data_len);
1119
1120         if (session->debug & L2TP_MSG_DATA) {
1121                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1122                 unsigned char *datap = skb->data + uhlen;
1123
1124                 pr_debug("%s: xmit\n", session->name);
1125                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1126                                      datap, min_t(size_t, 32, len - uhlen));
1127         }
1128
1129         /* Queue the packet to IP for output */
1130         skb->local_df = 1;
1131 #if IS_ENABLED(CONFIG_IPV6)
1132         if (skb->sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1133                 error = inet6_csk_xmit(skb, NULL);
1134         else
1135 #endif
1136                 error = ip_queue_xmit(skb, fl);
1137
1138         /* Update stats */
1139         if (error >= 0) {
1140                 atomic_long_inc(&tunnel->stats.tx_packets);
1141                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1142                 atomic_long_inc(&session->stats.tx_packets);
1143                 atomic_long_add(len, &session->stats.tx_bytes);
1144         } else {
1145                 atomic_long_inc(&tunnel->stats.tx_errors);
1146                 atomic_long_inc(&session->stats.tx_errors);
1147         }
1148
1149         return 0;
1150 }
1151
1152 /* Automatically called when the skb is freed.
1153  */
1154 static void l2tp_sock_wfree(struct sk_buff *skb)
1155 {
1156         sock_put(skb->sk);
1157 }
1158
1159 /* For data skbs that we transmit, we associate with the tunnel socket
1160  * but don't do accounting.
1161  */
1162 static inline void l2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1163 {
1164         sock_hold(sk);
1165         skb->sk = sk;
1166         skb->destructor = l2tp_sock_wfree;
1167 }
1168
1169 #if IS_ENABLED(CONFIG_IPV6)
1170 static void l2tp_xmit_ipv6_csum(struct sock *sk, struct sk_buff *skb,
1171                                 int udp_len)
1172 {
1173         struct ipv6_pinfo *np = inet6_sk(sk);
1174         struct udphdr *uh = udp_hdr(skb);
1175
1176         if (!skb_dst(skb) || !skb_dst(skb)->dev ||
1177             !(skb_dst(skb)->dev->features & NETIF_F_IPV6_CSUM)) {
1178                 __wsum csum = skb_checksum(skb, 0, udp_len, 0);
1179                 skb->ip_summed = CHECKSUM_UNNECESSARY;
1180                 uh->check = csum_ipv6_magic(&np->saddr, &np->daddr, udp_len,
1181                                             IPPROTO_UDP, csum);
1182                 if (uh->check == 0)
1183                         uh->check = CSUM_MANGLED_0;
1184         } else {
1185                 skb->ip_summed = CHECKSUM_PARTIAL;
1186                 skb->csum_start = skb_transport_header(skb) - skb->head;
1187                 skb->csum_offset = offsetof(struct udphdr, check);
1188                 uh->check = ~csum_ipv6_magic(&np->saddr, &np->daddr,
1189                                              udp_len, IPPROTO_UDP, 0);
1190         }
1191 }
1192 #endif
1193
1194 /* If caller requires the skb to have a ppp header, the header must be
1195  * inserted in the skb data before calling this function.
1196  */
1197 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1198 {
1199         int data_len = skb->len;
1200         struct l2tp_tunnel *tunnel = session->tunnel;
1201         struct sock *sk = tunnel->sock;
1202         struct flowi *fl;
1203         struct udphdr *uh;
1204         struct inet_sock *inet;
1205         __wsum csum;
1206         int headroom;
1207         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1208         int udp_len;
1209         int ret = NET_XMIT_SUCCESS;
1210
1211         /* Check that there's enough headroom in the skb to insert IP,
1212          * UDP and L2TP headers. If not enough, expand it to
1213          * make room. Adjust truesize.
1214          */
1215         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1216                 uhlen + hdr_len;
1217         if (skb_cow_head(skb, headroom)) {
1218                 kfree_skb(skb);
1219                 return NET_XMIT_DROP;
1220         }
1221
1222         skb_orphan(skb);
1223         /* Setup L2TP header */
1224         session->build_header(session, __skb_push(skb, hdr_len));
1225
1226         /* Reset skb netfilter state */
1227         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1228         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1229                               IPSKB_REROUTED);
1230         nf_reset(skb);
1231
1232         bh_lock_sock(sk);
1233         if (sock_owned_by_user(sk)) {
1234                 kfree_skb(skb);
1235                 ret = NET_XMIT_DROP;
1236                 goto out_unlock;
1237         }
1238
1239         /* Get routing info from the tunnel socket */
1240         skb_dst_drop(skb);
1241         skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1242
1243         inet = inet_sk(sk);
1244         fl = &inet->cork.fl;
1245         switch (tunnel->encap) {
1246         case L2TP_ENCAPTYPE_UDP:
1247                 /* Setup UDP header */
1248                 __skb_push(skb, sizeof(*uh));
1249                 skb_reset_transport_header(skb);
1250                 uh = udp_hdr(skb);
1251                 uh->source = inet->inet_sport;
1252                 uh->dest = inet->inet_dport;
1253                 udp_len = uhlen + hdr_len + data_len;
1254                 uh->len = htons(udp_len);
1255                 uh->check = 0;
1256
1257                 /* Calculate UDP checksum if configured to do so */
1258 #if IS_ENABLED(CONFIG_IPV6)
1259                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1260                         l2tp_xmit_ipv6_csum(sk, skb, udp_len);
1261                 else
1262 #endif
1263                 if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1264                         skb->ip_summed = CHECKSUM_NONE;
1265                 else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1266                          (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1267                         skb->ip_summed = CHECKSUM_COMPLETE;
1268                         csum = skb_checksum(skb, 0, udp_len, 0);
1269                         uh->check = csum_tcpudp_magic(inet->inet_saddr,
1270                                                       inet->inet_daddr,
1271                                                       udp_len, IPPROTO_UDP, csum);
1272                         if (uh->check == 0)
1273                                 uh->check = CSUM_MANGLED_0;
1274                 } else {
1275                         skb->ip_summed = CHECKSUM_PARTIAL;
1276                         skb->csum_start = skb_transport_header(skb) - skb->head;
1277                         skb->csum_offset = offsetof(struct udphdr, check);
1278                         uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1279                                                        inet->inet_daddr,
1280                                                        udp_len, IPPROTO_UDP, 0);
1281                 }
1282                 break;
1283
1284         case L2TP_ENCAPTYPE_IP:
1285                 break;
1286         }
1287
1288         l2tp_skb_set_owner_w(skb, sk);
1289
1290         l2tp_xmit_core(session, skb, fl, data_len);
1291 out_unlock:
1292         bh_unlock_sock(sk);
1293
1294         return ret;
1295 }
1296 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1297
1298 /*****************************************************************************
1299  * Tinnel and session create/destroy.
1300  *****************************************************************************/
1301
1302 /* Tunnel socket destruct hook.
1303  * The tunnel context is deleted only when all session sockets have been
1304  * closed.
1305  */
1306 static void l2tp_tunnel_destruct(struct sock *sk)
1307 {
1308         struct l2tp_tunnel *tunnel;
1309         struct l2tp_net *pn;
1310
1311         tunnel = sk->sk_user_data;
1312         if (tunnel == NULL)
1313                 goto end;
1314
1315         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1316
1317
1318         /* Disable udp encapsulation */
1319         switch (tunnel->encap) {
1320         case L2TP_ENCAPTYPE_UDP:
1321                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1322                 (udp_sk(sk))->encap_type = 0;
1323                 (udp_sk(sk))->encap_rcv = NULL;
1324                 (udp_sk(sk))->encap_destroy = NULL;
1325                 break;
1326         case L2TP_ENCAPTYPE_IP:
1327                 break;
1328         }
1329
1330         /* Remove hooks into tunnel socket */
1331         sk->sk_destruct = tunnel->old_sk_destruct;
1332         sk->sk_user_data = NULL;
1333         tunnel->sock = NULL;
1334
1335         /* Remove the tunnel struct from the tunnel list */
1336         pn = l2tp_pernet(tunnel->l2tp_net);
1337         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1338         list_del_rcu(&tunnel->list);
1339         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1340         atomic_dec(&l2tp_tunnel_count);
1341
1342         l2tp_tunnel_closeall(tunnel);
1343         l2tp_tunnel_dec_refcount(tunnel);
1344
1345         /* Call the original destructor */
1346         if (sk->sk_destruct)
1347                 (*sk->sk_destruct)(sk);
1348 end:
1349         return;
1350 }
1351
1352 /* When the tunnel is closed, all the attached sessions need to go too.
1353  */
1354 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1355 {
1356         int hash;
1357         struct hlist_node *walk;
1358         struct hlist_node *tmp;
1359         struct l2tp_session *session;
1360
1361         BUG_ON(tunnel == NULL);
1362
1363         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1364                   tunnel->name);
1365
1366         write_lock_bh(&tunnel->hlist_lock);
1367         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1368 again:
1369                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1370                         session = hlist_entry(walk, struct l2tp_session, hlist);
1371
1372                         l2tp_info(session, L2TP_MSG_CONTROL,
1373                                   "%s: closing session\n", session->name);
1374
1375                         hlist_del_init(&session->hlist);
1376
1377                         if (session->ref != NULL)
1378                                 (*session->ref)(session);
1379
1380                         write_unlock_bh(&tunnel->hlist_lock);
1381
1382                         __l2tp_session_unhash(session);
1383                         l2tp_session_queue_purge(session);
1384
1385                         if (session->session_close != NULL)
1386                                 (*session->session_close)(session);
1387
1388                         if (session->deref != NULL)
1389                                 (*session->deref)(session);
1390
1391                         l2tp_session_dec_refcount(session);
1392
1393                         write_lock_bh(&tunnel->hlist_lock);
1394
1395                         /* Now restart from the beginning of this hash
1396                          * chain.  We always remove a session from the
1397                          * list so we are guaranteed to make forward
1398                          * progress.
1399                          */
1400                         goto again;
1401                 }
1402         }
1403         write_unlock_bh(&tunnel->hlist_lock);
1404 }
1405 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1406
1407 /* Tunnel socket destroy hook for UDP encapsulation */
1408 static void l2tp_udp_encap_destroy(struct sock *sk)
1409 {
1410         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
1411         if (tunnel) {
1412                 l2tp_tunnel_closeall(tunnel);
1413                 sock_put(sk);
1414         }
1415 }
1416
1417 /* Really kill the tunnel.
1418  * Come here only when all sessions have been cleared from the tunnel.
1419  */
1420 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1421 {
1422         BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1423         BUG_ON(tunnel->sock != NULL);
1424         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: free...\n", tunnel->name);
1425         kfree_rcu(tunnel, rcu);
1426 }
1427
1428 /* Workqueue tunnel deletion function */
1429 static void l2tp_tunnel_del_work(struct work_struct *work)
1430 {
1431         struct l2tp_tunnel *tunnel = NULL;
1432         struct socket *sock = NULL;
1433         struct sock *sk = NULL;
1434
1435         tunnel = container_of(work, struct l2tp_tunnel, del_work);
1436         sk = l2tp_tunnel_sock_lookup(tunnel);
1437         if (!sk)
1438                 return;
1439
1440         sock = sk->sk_socket;
1441
1442         /* If the tunnel socket was created by userspace, then go through the
1443          * inet layer to shut the socket down, and let userspace close it.
1444          * Otherwise, if we created the socket directly within the kernel, use
1445          * the sk API to release it here.
1446          * In either case the tunnel resources are freed in the socket
1447          * destructor when the tunnel socket goes away.
1448          */
1449         if (tunnel->fd >= 0) {
1450                 if (sock)
1451                         inet_shutdown(sock, 2);
1452         } else {
1453                 if (sock)
1454                         kernel_sock_shutdown(sock, SHUT_RDWR);
1455                 sk_release_kernel(sk);
1456         }
1457
1458         l2tp_tunnel_sock_put(sk);
1459 }
1460
1461 /* Create a socket for the tunnel, if one isn't set up by
1462  * userspace. This is used for static tunnels where there is no
1463  * managing L2TP daemon.
1464  *
1465  * Since we don't want these sockets to keep a namespace alive by
1466  * themselves, we drop the socket's namespace refcount after creation.
1467  * These sockets are freed when the namespace exits using the pernet
1468  * exit hook.
1469  */
1470 static int l2tp_tunnel_sock_create(struct net *net,
1471                                 u32 tunnel_id,
1472                                 u32 peer_tunnel_id,
1473                                 struct l2tp_tunnel_cfg *cfg,
1474                                 struct socket **sockp)
1475 {
1476         int err = -EINVAL;
1477         struct socket *sock = NULL;
1478         struct sockaddr_in udp_addr = {0};
1479         struct sockaddr_l2tpip ip_addr = {0};
1480 #if IS_ENABLED(CONFIG_IPV6)
1481         struct sockaddr_in6 udp6_addr = {0};
1482         struct sockaddr_l2tpip6 ip6_addr = {0};
1483 #endif
1484
1485         switch (cfg->encap) {
1486         case L2TP_ENCAPTYPE_UDP:
1487 #if IS_ENABLED(CONFIG_IPV6)
1488                 if (cfg->local_ip6 && cfg->peer_ip6) {
1489                         err = sock_create_kern(AF_INET6, SOCK_DGRAM, 0, &sock);
1490                         if (err < 0)
1491                                 goto out;
1492
1493                         sk_change_net(sock->sk, net);
1494
1495                         udp6_addr.sin6_family = AF_INET6;
1496                         memcpy(&udp6_addr.sin6_addr, cfg->local_ip6,
1497                                sizeof(udp6_addr.sin6_addr));
1498                         udp6_addr.sin6_port = htons(cfg->local_udp_port);
1499                         err = kernel_bind(sock, (struct sockaddr *) &udp6_addr,
1500                                           sizeof(udp6_addr));
1501                         if (err < 0)
1502                                 goto out;
1503
1504                         udp6_addr.sin6_family = AF_INET6;
1505                         memcpy(&udp6_addr.sin6_addr, cfg->peer_ip6,
1506                                sizeof(udp6_addr.sin6_addr));
1507                         udp6_addr.sin6_port = htons(cfg->peer_udp_port);
1508                         err = kernel_connect(sock,
1509                                              (struct sockaddr *) &udp6_addr,
1510                                              sizeof(udp6_addr), 0);
1511                         if (err < 0)
1512                                 goto out;
1513                 } else
1514 #endif
1515                 {
1516                         err = sock_create_kern(AF_INET, SOCK_DGRAM, 0, &sock);
1517                         if (err < 0)
1518                                 goto out;
1519
1520                         sk_change_net(sock->sk, net);
1521
1522                         udp_addr.sin_family = AF_INET;
1523                         udp_addr.sin_addr = cfg->local_ip;
1524                         udp_addr.sin_port = htons(cfg->local_udp_port);
1525                         err = kernel_bind(sock, (struct sockaddr *) &udp_addr,
1526                                           sizeof(udp_addr));
1527                         if (err < 0)
1528                                 goto out;
1529
1530                         udp_addr.sin_family = AF_INET;
1531                         udp_addr.sin_addr = cfg->peer_ip;
1532                         udp_addr.sin_port = htons(cfg->peer_udp_port);
1533                         err = kernel_connect(sock,
1534                                              (struct sockaddr *) &udp_addr,
1535                                              sizeof(udp_addr), 0);
1536                         if (err < 0)
1537                                 goto out;
1538                 }
1539
1540                 if (!cfg->use_udp_checksums)
1541                         sock->sk->sk_no_check = UDP_CSUM_NOXMIT;
1542
1543                 break;
1544
1545         case L2TP_ENCAPTYPE_IP:
1546 #if IS_ENABLED(CONFIG_IPV6)
1547                 if (cfg->local_ip6 && cfg->peer_ip6) {
1548                         err = sock_create_kern(AF_INET6, SOCK_DGRAM,
1549                                           IPPROTO_L2TP, &sock);
1550                         if (err < 0)
1551                                 goto out;
1552
1553                         sk_change_net(sock->sk, net);
1554
1555                         ip6_addr.l2tp_family = AF_INET6;
1556                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1557                                sizeof(ip6_addr.l2tp_addr));
1558                         ip6_addr.l2tp_conn_id = tunnel_id;
1559                         err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1560                                           sizeof(ip6_addr));
1561                         if (err < 0)
1562                                 goto out;
1563
1564                         ip6_addr.l2tp_family = AF_INET6;
1565                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1566                                sizeof(ip6_addr.l2tp_addr));
1567                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1568                         err = kernel_connect(sock,
1569                                              (struct sockaddr *) &ip6_addr,
1570                                              sizeof(ip6_addr), 0);
1571                         if (err < 0)
1572                                 goto out;
1573                 } else
1574 #endif
1575                 {
1576                         err = sock_create_kern(AF_INET, SOCK_DGRAM,
1577                                           IPPROTO_L2TP, &sock);
1578                         if (err < 0)
1579                                 goto out;
1580
1581                         sk_change_net(sock->sk, net);
1582
1583                         ip_addr.l2tp_family = AF_INET;
1584                         ip_addr.l2tp_addr = cfg->local_ip;
1585                         ip_addr.l2tp_conn_id = tunnel_id;
1586                         err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1587                                           sizeof(ip_addr));
1588                         if (err < 0)
1589                                 goto out;
1590
1591                         ip_addr.l2tp_family = AF_INET;
1592                         ip_addr.l2tp_addr = cfg->peer_ip;
1593                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1594                         err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1595                                              sizeof(ip_addr), 0);
1596                         if (err < 0)
1597                                 goto out;
1598                 }
1599                 break;
1600
1601         default:
1602                 goto out;
1603         }
1604
1605 out:
1606         *sockp = sock;
1607         if ((err < 0) && sock) {
1608                 kernel_sock_shutdown(sock, SHUT_RDWR);
1609                 sk_release_kernel(sock->sk);
1610                 *sockp = NULL;
1611         }
1612
1613         return err;
1614 }
1615
1616 static struct lock_class_key l2tp_socket_class;
1617
1618 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)
1619 {
1620         struct l2tp_tunnel *tunnel = NULL;
1621         int err;
1622         struct socket *sock = NULL;
1623         struct sock *sk = NULL;
1624         struct l2tp_net *pn;
1625         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1626
1627         /* Get the tunnel socket from the fd, which was opened by
1628          * the userspace L2TP daemon. If not specified, create a
1629          * kernel socket.
1630          */
1631         if (fd < 0) {
1632                 err = l2tp_tunnel_sock_create(net, tunnel_id, peer_tunnel_id,
1633                                 cfg, &sock);
1634                 if (err < 0)
1635                         goto err;
1636         } else {
1637                 sock = sockfd_lookup(fd, &err);
1638                 if (!sock) {
1639                         pr_err("tunl %u: sockfd_lookup(fd=%d) returned %d\n",
1640                                tunnel_id, fd, err);
1641                         err = -EBADF;
1642                         goto err;
1643                 }
1644
1645                 /* Reject namespace mismatches */
1646                 if (!net_eq(sock_net(sock->sk), net)) {
1647                         pr_err("tunl %u: netns mismatch\n", tunnel_id);
1648                         err = -EINVAL;
1649                         goto err;
1650                 }
1651         }
1652
1653         sk = sock->sk;
1654
1655         if (cfg != NULL)
1656                 encap = cfg->encap;
1657
1658         /* Quick sanity checks */
1659         switch (encap) {
1660         case L2TP_ENCAPTYPE_UDP:
1661                 err = -EPROTONOSUPPORT;
1662                 if (sk->sk_protocol != IPPROTO_UDP) {
1663                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1664                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1665                         goto err;
1666                 }
1667                 break;
1668         case L2TP_ENCAPTYPE_IP:
1669                 err = -EPROTONOSUPPORT;
1670                 if (sk->sk_protocol != IPPROTO_L2TP) {
1671                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1672                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1673                         goto err;
1674                 }
1675                 break;
1676         }
1677
1678         /* Check if this socket has already been prepped */
1679         tunnel = (struct l2tp_tunnel *)sk->sk_user_data;
1680         if (tunnel != NULL) {
1681                 /* This socket has already been prepped */
1682                 err = -EBUSY;
1683                 goto err;
1684         }
1685
1686         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1687         if (tunnel == NULL) {
1688                 err = -ENOMEM;
1689                 goto err;
1690         }
1691
1692         tunnel->version = version;
1693         tunnel->tunnel_id = tunnel_id;
1694         tunnel->peer_tunnel_id = peer_tunnel_id;
1695         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1696
1697         tunnel->magic = L2TP_TUNNEL_MAGIC;
1698         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1699         rwlock_init(&tunnel->hlist_lock);
1700
1701         /* The net we belong to */
1702         tunnel->l2tp_net = net;
1703         pn = l2tp_pernet(net);
1704
1705         if (cfg != NULL)
1706                 tunnel->debug = cfg->debug;
1707
1708 #if IS_ENABLED(CONFIG_IPV6)
1709         if (sk->sk_family == PF_INET6) {
1710                 struct ipv6_pinfo *np = inet6_sk(sk);
1711
1712                 if (ipv6_addr_v4mapped(&np->saddr) &&
1713                     ipv6_addr_v4mapped(&np->daddr)) {
1714                         struct inet_sock *inet = inet_sk(sk);
1715
1716                         tunnel->v4mapped = true;
1717                         inet->inet_saddr = np->saddr.s6_addr32[3];
1718                         inet->inet_rcv_saddr = np->rcv_saddr.s6_addr32[3];
1719                         inet->inet_daddr = np->daddr.s6_addr32[3];
1720                 } else {
1721                         tunnel->v4mapped = false;
1722                 }
1723         }
1724 #endif
1725
1726         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1727         tunnel->encap = encap;
1728         if (encap == L2TP_ENCAPTYPE_UDP) {
1729                 /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1730                 udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1731                 udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1732                 udp_sk(sk)->encap_destroy = l2tp_udp_encap_destroy;
1733 #if IS_ENABLED(CONFIG_IPV6)
1734                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1735                         udpv6_encap_enable();
1736                 else
1737 #endif
1738                 udp_encap_enable();
1739         }
1740
1741         sk->sk_user_data = tunnel;
1742
1743         /* Hook on the tunnel socket destructor so that we can cleanup
1744          * if the tunnel socket goes away.
1745          */
1746         tunnel->old_sk_destruct = sk->sk_destruct;
1747         sk->sk_destruct = &l2tp_tunnel_destruct;
1748         tunnel->sock = sk;
1749         tunnel->fd = fd;
1750         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class, "l2tp_sock");
1751
1752         sk->sk_allocation = GFP_ATOMIC;
1753
1754         /* Init delete workqueue struct */
1755         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1756
1757         /* Add tunnel to our list */
1758         INIT_LIST_HEAD(&tunnel->list);
1759         atomic_inc(&l2tp_tunnel_count);
1760
1761         /* Bump the reference count. The tunnel context is deleted
1762          * only when this drops to zero. Must be done before list insertion
1763          */
1764         l2tp_tunnel_inc_refcount(tunnel);
1765         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1766         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1767         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1768
1769         err = 0;
1770 err:
1771         if (tunnelp)
1772                 *tunnelp = tunnel;
1773
1774         /* If tunnel's socket was created by the kernel, it doesn't
1775          *  have a file.
1776          */
1777         if (sock && sock->file)
1778                 sockfd_put(sock);
1779
1780         return err;
1781 }
1782 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1783
1784 /* This function is used by the netlink TUNNEL_DELETE command.
1785  */
1786 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1787 {
1788         l2tp_tunnel_closeall(tunnel);
1789         return (false == queue_work(l2tp_wq, &tunnel->del_work));
1790 }
1791 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1792
1793 /* Really kill the session.
1794  */
1795 void l2tp_session_free(struct l2tp_session *session)
1796 {
1797         struct l2tp_tunnel *tunnel = session->tunnel;
1798
1799         BUG_ON(atomic_read(&session->ref_count) != 0);
1800
1801         if (tunnel) {
1802                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1803                 if (session->session_id != 0)
1804                         atomic_dec(&l2tp_session_count);
1805                 sock_put(tunnel->sock);
1806                 session->tunnel = NULL;
1807                 l2tp_tunnel_dec_refcount(tunnel);
1808         }
1809
1810         kfree(session);
1811
1812         return;
1813 }
1814 EXPORT_SYMBOL_GPL(l2tp_session_free);
1815
1816 /* Remove an l2tp session from l2tp_core's hash lists.
1817  * Provides a tidyup interface for pseudowire code which can't just route all
1818  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1819  * callback.
1820  */
1821 void __l2tp_session_unhash(struct l2tp_session *session)
1822 {
1823         struct l2tp_tunnel *tunnel = session->tunnel;
1824
1825         /* Remove the session from core hashes */
1826         if (tunnel) {
1827                 /* Remove from the per-tunnel hash */
1828                 write_lock_bh(&tunnel->hlist_lock);
1829                 hlist_del_init(&session->hlist);
1830                 write_unlock_bh(&tunnel->hlist_lock);
1831
1832                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1833                 if (tunnel->version != L2TP_HDR_VER_2) {
1834                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1835                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1836                         hlist_del_init_rcu(&session->global_hlist);
1837                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1838                         synchronize_rcu();
1839                 }
1840         }
1841 }
1842 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1843
1844 /* This function is used by the netlink SESSION_DELETE command and by
1845    pseudowire modules.
1846  */
1847 int l2tp_session_delete(struct l2tp_session *session)
1848 {
1849         if (session->ref)
1850                 (*session->ref)(session);
1851         __l2tp_session_unhash(session);
1852         l2tp_session_queue_purge(session);
1853         if (session->session_close != NULL)
1854                 (*session->session_close)(session);
1855         if (session->deref)
1856                 (*session->deref)(session);
1857         l2tp_session_dec_refcount(session);
1858         return 0;
1859 }
1860 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1861
1862 /* We come here whenever a session's send_seq, cookie_len or
1863  * l2specific_len parameters are set.
1864  */
1865 static void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1866 {
1867         if (version == L2TP_HDR_VER_2) {
1868                 session->hdr_len = 6;
1869                 if (session->send_seq)
1870                         session->hdr_len += 4;
1871         } else {
1872                 session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1873                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1874                         session->hdr_len += 4;
1875         }
1876
1877 }
1878
1879 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)
1880 {
1881         struct l2tp_session *session;
1882
1883         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1884         if (session != NULL) {
1885                 session->magic = L2TP_SESSION_MAGIC;
1886                 session->tunnel = tunnel;
1887
1888                 session->session_id = session_id;
1889                 session->peer_session_id = peer_session_id;
1890                 session->nr = 0;
1891                 if (tunnel->version == L2TP_HDR_VER_2)
1892                         session->nr_max = 0xffff;
1893                 else
1894                         session->nr_max = 0xffffff;
1895                 session->nr_window_size = session->nr_max / 2;
1896                 session->nr_oos_count_max = 4;
1897
1898                 /* Use NR of first received packet */
1899                 session->reorder_skip = 1;
1900
1901                 sprintf(&session->name[0], "sess %u/%u",
1902                         tunnel->tunnel_id, session->session_id);
1903
1904                 skb_queue_head_init(&session->reorder_q);
1905
1906                 INIT_HLIST_NODE(&session->hlist);
1907                 INIT_HLIST_NODE(&session->global_hlist);
1908
1909                 /* Inherit debug options from tunnel */
1910                 session->debug = tunnel->debug;
1911
1912                 if (cfg) {
1913                         session->pwtype = cfg->pw_type;
1914                         session->debug = cfg->debug;
1915                         session->mtu = cfg->mtu;
1916                         session->mru = cfg->mru;
1917                         session->send_seq = cfg->send_seq;
1918                         session->recv_seq = cfg->recv_seq;
1919                         session->lns_mode = cfg->lns_mode;
1920                         session->reorder_timeout = cfg->reorder_timeout;
1921                         session->offset = cfg->offset;
1922                         session->l2specific_type = cfg->l2specific_type;
1923                         session->l2specific_len = cfg->l2specific_len;
1924                         session->cookie_len = cfg->cookie_len;
1925                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1926                         session->peer_cookie_len = cfg->peer_cookie_len;
1927                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1928                 }
1929
1930                 if (tunnel->version == L2TP_HDR_VER_2)
1931                         session->build_header = l2tp_build_l2tpv2_header;
1932                 else
1933                         session->build_header = l2tp_build_l2tpv3_header;
1934
1935                 l2tp_session_set_header_len(session, tunnel->version);
1936
1937                 /* Bump the reference count. The session context is deleted
1938                  * only when this drops to zero.
1939                  */
1940                 l2tp_session_inc_refcount(session);
1941                 l2tp_tunnel_inc_refcount(tunnel);
1942
1943                 /* Ensure tunnel socket isn't deleted */
1944                 sock_hold(tunnel->sock);
1945
1946                 /* Add session to the tunnel's hash list */
1947                 write_lock_bh(&tunnel->hlist_lock);
1948                 hlist_add_head(&session->hlist,
1949                                l2tp_session_id_hash(tunnel, session_id));
1950                 write_unlock_bh(&tunnel->hlist_lock);
1951
1952                 /* And to the global session list if L2TPv3 */
1953                 if (tunnel->version != L2TP_HDR_VER_2) {
1954                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1955
1956                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1957                         hlist_add_head_rcu(&session->global_hlist,
1958                                            l2tp_session_id_hash_2(pn, session_id));
1959                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1960                 }
1961
1962                 /* Ignore management session in session count value */
1963                 if (session->session_id != 0)
1964                         atomic_inc(&l2tp_session_count);
1965         }
1966
1967         return session;
1968 }
1969 EXPORT_SYMBOL_GPL(l2tp_session_create);
1970
1971 /*****************************************************************************
1972  * Init and cleanup
1973  *****************************************************************************/
1974
1975 static __net_init int l2tp_init_net(struct net *net)
1976 {
1977         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1978         int hash;
1979
1980         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1981         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1982
1983         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1984                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1985
1986         spin_lock_init(&pn->l2tp_session_hlist_lock);
1987
1988         return 0;
1989 }
1990
1991 static __net_exit void l2tp_exit_net(struct net *net)
1992 {
1993         struct l2tp_net *pn = l2tp_pernet(net);
1994         struct l2tp_tunnel *tunnel = NULL;
1995
1996         rcu_read_lock_bh();
1997         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1998                 (void)l2tp_tunnel_delete(tunnel);
1999         }
2000         rcu_read_unlock_bh();
2001 }
2002
2003 static struct pernet_operations l2tp_net_ops = {
2004         .init = l2tp_init_net,
2005         .exit = l2tp_exit_net,
2006         .id   = &l2tp_net_id,
2007         .size = sizeof(struct l2tp_net),
2008 };
2009
2010 static int __init l2tp_init(void)
2011 {
2012         int rc = 0;
2013
2014         rc = register_pernet_device(&l2tp_net_ops);
2015         if (rc)
2016                 goto out;
2017
2018         l2tp_wq = alloc_workqueue("l2tp", WQ_NON_REENTRANT | WQ_UNBOUND, 0);
2019         if (!l2tp_wq) {
2020                 pr_err("alloc_workqueue failed\n");
2021                 rc = -ENOMEM;
2022                 goto out;
2023         }
2024
2025         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
2026
2027 out:
2028         return rc;
2029 }
2030
2031 static void __exit l2tp_exit(void)
2032 {
2033         unregister_pernet_device(&l2tp_net_ops);
2034         if (l2tp_wq) {
2035                 destroy_workqueue(l2tp_wq);
2036                 l2tp_wq = NULL;
2037         }
2038 }
2039
2040 module_init(l2tp_init);
2041 module_exit(l2tp_exit);
2042
2043 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
2044 MODULE_DESCRIPTION("L2TP core");
2045 MODULE_LICENSE("GPL");
2046 MODULE_VERSION(L2TP_DRV_VERSION);
2047