ipv4: add a sock pointer to ip_queue_xmit()
[sfrench/cifs-2.6.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_tunnel_free(struct l2tp_tunnel *tunnel);
116
117 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
118 {
119         return sk->sk_user_data;
120 }
121
122 static inline struct l2tp_net *l2tp_pernet(struct net *net)
123 {
124         BUG_ON(!net);
125
126         return net_generic(net, l2tp_net_id);
127 }
128
129 /* Tunnel reference counts. Incremented per session that is added to
130  * the tunnel.
131  */
132 static inline void l2tp_tunnel_inc_refcount_1(struct l2tp_tunnel *tunnel)
133 {
134         atomic_inc(&tunnel->ref_count);
135 }
136
137 static inline void l2tp_tunnel_dec_refcount_1(struct l2tp_tunnel *tunnel)
138 {
139         if (atomic_dec_and_test(&tunnel->ref_count))
140                 l2tp_tunnel_free(tunnel);
141 }
142 #ifdef L2TP_REFCNT_DEBUG
143 #define l2tp_tunnel_inc_refcount(_t)                                    \
144 do {                                                                    \
145         pr_debug("l2tp_tunnel_inc_refcount: %s:%d %s: cnt=%d\n",        \
146                  __func__, __LINE__, (_t)->name,                        \
147                  atomic_read(&_t->ref_count));                          \
148         l2tp_tunnel_inc_refcount_1(_t);                                 \
149 } while (0)
150 #define l2tp_tunnel_dec_refcount(_t)
151 do {                                                                    \
152         pr_debug("l2tp_tunnel_dec_refcount: %s:%d %s: cnt=%d\n",        \
153                  __func__, __LINE__, (_t)->name,                        \
154                  atomic_read(&_t->ref_count));                          \
155         l2tp_tunnel_dec_refcount_1(_t);                                 \
156 } while (0)
157 #else
158 #define l2tp_tunnel_inc_refcount(t) l2tp_tunnel_inc_refcount_1(t)
159 #define l2tp_tunnel_dec_refcount(t) l2tp_tunnel_dec_refcount_1(t)
160 #endif
161
162 /* Session hash global list for L2TPv3.
163  * The session_id SHOULD be random according to RFC3931, but several
164  * L2TP implementations use incrementing session_ids.  So we do a real
165  * hash on the session_id, rather than a simple bitmask.
166  */
167 static inline struct hlist_head *
168 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
169 {
170         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
171
172 }
173
174 /* Lookup the tunnel socket, possibly involving the fs code if the socket is
175  * owned by userspace.  A struct sock returned from this function must be
176  * released using l2tp_tunnel_sock_put once you're done with it.
177  */
178 static struct sock *l2tp_tunnel_sock_lookup(struct l2tp_tunnel *tunnel)
179 {
180         int err = 0;
181         struct socket *sock = NULL;
182         struct sock *sk = NULL;
183
184         if (!tunnel)
185                 goto out;
186
187         if (tunnel->fd >= 0) {
188                 /* Socket is owned by userspace, who might be in the process
189                  * of closing it.  Look the socket up using the fd to ensure
190                  * consistency.
191                  */
192                 sock = sockfd_lookup(tunnel->fd, &err);
193                 if (sock)
194                         sk = sock->sk;
195         } else {
196                 /* Socket is owned by kernelspace */
197                 sk = tunnel->sock;
198                 sock_hold(sk);
199         }
200
201 out:
202         return sk;
203 }
204
205 /* Drop a reference to a tunnel socket obtained via. l2tp_tunnel_sock_put */
206 static void l2tp_tunnel_sock_put(struct sock *sk)
207 {
208         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
209         if (tunnel) {
210                 if (tunnel->fd >= 0) {
211                         /* Socket is owned by userspace */
212                         sockfd_put(sk->sk_socket);
213                 }
214                 sock_put(sk);
215         }
216         sock_put(sk);
217 }
218
219 /* Lookup a session by id in the global session list
220  */
221 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
222 {
223         struct l2tp_net *pn = l2tp_pernet(net);
224         struct hlist_head *session_list =
225                 l2tp_session_id_hash_2(pn, session_id);
226         struct l2tp_session *session;
227
228         rcu_read_lock_bh();
229         hlist_for_each_entry_rcu(session, session_list, global_hlist) {
230                 if (session->session_id == session_id) {
231                         rcu_read_unlock_bh();
232                         return session;
233                 }
234         }
235         rcu_read_unlock_bh();
236
237         return NULL;
238 }
239
240 /* Session hash list.
241  * The session_id SHOULD be random according to RFC2661, but several
242  * L2TP implementations (Cisco and Microsoft) use incrementing
243  * session_ids.  So we do a real hash on the session_id, rather than a
244  * simple bitmask.
245  */
246 static inline struct hlist_head *
247 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
248 {
249         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
250 }
251
252 /* Lookup a session by id
253  */
254 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
255 {
256         struct hlist_head *session_list;
257         struct l2tp_session *session;
258
259         /* In L2TPv3, session_ids are unique over all tunnels and we
260          * sometimes need to look them up before we know the
261          * tunnel.
262          */
263         if (tunnel == NULL)
264                 return l2tp_session_find_2(net, session_id);
265
266         session_list = l2tp_session_id_hash(tunnel, session_id);
267         read_lock_bh(&tunnel->hlist_lock);
268         hlist_for_each_entry(session, session_list, hlist) {
269                 if (session->session_id == session_id) {
270                         read_unlock_bh(&tunnel->hlist_lock);
271                         return session;
272                 }
273         }
274         read_unlock_bh(&tunnel->hlist_lock);
275
276         return NULL;
277 }
278 EXPORT_SYMBOL_GPL(l2tp_session_find);
279
280 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
281 {
282         int hash;
283         struct l2tp_session *session;
284         int count = 0;
285
286         read_lock_bh(&tunnel->hlist_lock);
287         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
288                 hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
289                         if (++count > nth) {
290                                 read_unlock_bh(&tunnel->hlist_lock);
291                                 return session;
292                         }
293                 }
294         }
295
296         read_unlock_bh(&tunnel->hlist_lock);
297
298         return NULL;
299 }
300 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
301
302 /* Lookup a session by interface name.
303  * This is very inefficient but is only used by management interfaces.
304  */
305 struct l2tp_session *l2tp_session_find_by_ifname(struct net *net, char *ifname)
306 {
307         struct l2tp_net *pn = l2tp_pernet(net);
308         int hash;
309         struct l2tp_session *session;
310
311         rcu_read_lock_bh();
312         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
313                 hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
314                         if (!strcmp(session->ifname, ifname)) {
315                                 rcu_read_unlock_bh();
316                                 return session;
317                         }
318                 }
319         }
320
321         rcu_read_unlock_bh();
322
323         return NULL;
324 }
325 EXPORT_SYMBOL_GPL(l2tp_session_find_by_ifname);
326
327 /* Lookup a tunnel by id
328  */
329 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
330 {
331         struct l2tp_tunnel *tunnel;
332         struct l2tp_net *pn = l2tp_pernet(net);
333
334         rcu_read_lock_bh();
335         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
336                 if (tunnel->tunnel_id == tunnel_id) {
337                         rcu_read_unlock_bh();
338                         return tunnel;
339                 }
340         }
341         rcu_read_unlock_bh();
342
343         return NULL;
344 }
345 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
346
347 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
348 {
349         struct l2tp_net *pn = l2tp_pernet(net);
350         struct l2tp_tunnel *tunnel;
351         int count = 0;
352
353         rcu_read_lock_bh();
354         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
355                 if (++count > nth) {
356                         rcu_read_unlock_bh();
357                         return tunnel;
358                 }
359         }
360
361         rcu_read_unlock_bh();
362
363         return NULL;
364 }
365 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
366
367 /*****************************************************************************
368  * Receive data handling
369  *****************************************************************************/
370
371 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
372  * number.
373  */
374 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
375 {
376         struct sk_buff *skbp;
377         struct sk_buff *tmp;
378         u32 ns = L2TP_SKB_CB(skb)->ns;
379
380         spin_lock_bh(&session->reorder_q.lock);
381         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
382                 if (L2TP_SKB_CB(skbp)->ns > ns) {
383                         __skb_queue_before(&session->reorder_q, skbp, skb);
384                         l2tp_dbg(session, L2TP_MSG_SEQ,
385                                  "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
386                                  session->name, ns, L2TP_SKB_CB(skbp)->ns,
387                                  skb_queue_len(&session->reorder_q));
388                         atomic_long_inc(&session->stats.rx_oos_packets);
389                         goto out;
390                 }
391         }
392
393         __skb_queue_tail(&session->reorder_q, skb);
394
395 out:
396         spin_unlock_bh(&session->reorder_q.lock);
397 }
398
399 /* Dequeue a single skb.
400  */
401 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
402 {
403         struct l2tp_tunnel *tunnel = session->tunnel;
404         int length = L2TP_SKB_CB(skb)->length;
405
406         /* We're about to requeue the skb, so return resources
407          * to its current owner (a socket receive buffer).
408          */
409         skb_orphan(skb);
410
411         atomic_long_inc(&tunnel->stats.rx_packets);
412         atomic_long_add(length, &tunnel->stats.rx_bytes);
413         atomic_long_inc(&session->stats.rx_packets);
414         atomic_long_add(length, &session->stats.rx_bytes);
415
416         if (L2TP_SKB_CB(skb)->has_seq) {
417                 /* Bump our Nr */
418                 session->nr++;
419                 session->nr &= session->nr_max;
420
421                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
422                          session->name, session->nr);
423         }
424
425         /* call private receive handler */
426         if (session->recv_skb != NULL)
427                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
428         else
429                 kfree_skb(skb);
430
431         if (session->deref)
432                 (*session->deref)(session);
433 }
434
435 /* Dequeue skbs from the session's reorder_q, subject to packet order.
436  * Skbs that have been in the queue for too long are simply discarded.
437  */
438 static void l2tp_recv_dequeue(struct l2tp_session *session)
439 {
440         struct sk_buff *skb;
441         struct sk_buff *tmp;
442
443         /* If the pkt at the head of the queue has the nr that we
444          * expect to send up next, dequeue it and any other
445          * in-sequence packets behind it.
446          */
447 start:
448         spin_lock_bh(&session->reorder_q.lock);
449         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
450                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
451                         atomic_long_inc(&session->stats.rx_seq_discards);
452                         atomic_long_inc(&session->stats.rx_errors);
453                         l2tp_dbg(session, L2TP_MSG_SEQ,
454                                  "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
455                                  session->name, L2TP_SKB_CB(skb)->ns,
456                                  L2TP_SKB_CB(skb)->length, session->nr,
457                                  skb_queue_len(&session->reorder_q));
458                         session->reorder_skip = 1;
459                         __skb_unlink(skb, &session->reorder_q);
460                         kfree_skb(skb);
461                         if (session->deref)
462                                 (*session->deref)(session);
463                         continue;
464                 }
465
466                 if (L2TP_SKB_CB(skb)->has_seq) {
467                         if (session->reorder_skip) {
468                                 l2tp_dbg(session, L2TP_MSG_SEQ,
469                                          "%s: advancing nr to next pkt: %u -> %u",
470                                          session->name, session->nr,
471                                          L2TP_SKB_CB(skb)->ns);
472                                 session->reorder_skip = 0;
473                                 session->nr = L2TP_SKB_CB(skb)->ns;
474                         }
475                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
476                                 l2tp_dbg(session, L2TP_MSG_SEQ,
477                                          "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
478                                          session->name, L2TP_SKB_CB(skb)->ns,
479                                          L2TP_SKB_CB(skb)->length, session->nr,
480                                          skb_queue_len(&session->reorder_q));
481                                 goto out;
482                         }
483                 }
484                 __skb_unlink(skb, &session->reorder_q);
485
486                 /* Process the skb. We release the queue lock while we
487                  * do so to let other contexts process the queue.
488                  */
489                 spin_unlock_bh(&session->reorder_q.lock);
490                 l2tp_recv_dequeue_skb(session, skb);
491                 goto start;
492         }
493
494 out:
495         spin_unlock_bh(&session->reorder_q.lock);
496 }
497
498 static inline int l2tp_verify_udp_checksum(struct sock *sk,
499                                            struct sk_buff *skb)
500 {
501         struct udphdr *uh = udp_hdr(skb);
502         u16 ulen = ntohs(uh->len);
503         __wsum psum;
504
505         if (sk->sk_no_check || skb_csum_unnecessary(skb))
506                 return 0;
507
508 #if IS_ENABLED(CONFIG_IPV6)
509         if (sk->sk_family == PF_INET6 && !l2tp_tunnel(sk)->v4mapped) {
510                 if (!uh->check) {
511                         LIMIT_NETDEBUG(KERN_INFO "L2TP: IPv6: checksum is 0\n");
512                         return 1;
513                 }
514                 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
515                     !csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
516                                      &ipv6_hdr(skb)->daddr, ulen,
517                                      IPPROTO_UDP, skb->csum)) {
518                         skb->ip_summed = CHECKSUM_UNNECESSARY;
519                         return 0;
520                 }
521                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
522                                                          &ipv6_hdr(skb)->daddr,
523                                                          skb->len, IPPROTO_UDP,
524                                                          0));
525         } else
526 #endif
527         {
528                 struct inet_sock *inet;
529                 if (!uh->check)
530                         return 0;
531                 inet = inet_sk(sk);
532                 psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr,
533                                           ulen, IPPROTO_UDP, 0);
534
535                 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
536                     !csum_fold(csum_add(psum, skb->csum)))
537                         return 0;
538                 skb->csum = psum;
539         }
540
541         return __skb_checksum_complete(skb);
542 }
543
544 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
545 {
546         u32 nws;
547
548         if (nr >= session->nr)
549                 nws = nr - session->nr;
550         else
551                 nws = (session->nr_max + 1) - (session->nr - nr);
552
553         return nws < session->nr_window_size;
554 }
555
556 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
557  * acceptable, else non-zero.
558  */
559 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
560 {
561         if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
562                 /* Packet sequence number is outside allowed window.
563                  * Discard it.
564                  */
565                 l2tp_dbg(session, L2TP_MSG_SEQ,
566                          "%s: pkt %u len %d discarded, outside window, nr=%u\n",
567                          session->name, L2TP_SKB_CB(skb)->ns,
568                          L2TP_SKB_CB(skb)->length, session->nr);
569                 goto discard;
570         }
571
572         if (session->reorder_timeout != 0) {
573                 /* Packet reordering enabled. Add skb to session's
574                  * reorder queue, in order of ns.
575                  */
576                 l2tp_recv_queue_skb(session, skb);
577                 goto out;
578         }
579
580         /* Packet reordering disabled. Discard out-of-sequence packets, while
581          * tracking the number if in-sequence packets after the first OOS packet
582          * is seen. After nr_oos_count_max in-sequence packets, reset the
583          * sequence number to re-enable packet reception.
584          */
585         if (L2TP_SKB_CB(skb)->ns == session->nr) {
586                 skb_queue_tail(&session->reorder_q, skb);
587         } else {
588                 u32 nr_oos = L2TP_SKB_CB(skb)->ns;
589                 u32 nr_next = (session->nr_oos + 1) & session->nr_max;
590
591                 if (nr_oos == nr_next)
592                         session->nr_oos_count++;
593                 else
594                         session->nr_oos_count = 0;
595
596                 session->nr_oos = nr_oos;
597                 if (session->nr_oos_count > session->nr_oos_count_max) {
598                         session->reorder_skip = 1;
599                         l2tp_dbg(session, L2TP_MSG_SEQ,
600                                  "%s: %d oos packets received. Resetting sequence numbers\n",
601                                  session->name, session->nr_oos_count);
602                 }
603                 if (!session->reorder_skip) {
604                         atomic_long_inc(&session->stats.rx_seq_discards);
605                         l2tp_dbg(session, L2TP_MSG_SEQ,
606                                  "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
607                                  session->name, L2TP_SKB_CB(skb)->ns,
608                                  L2TP_SKB_CB(skb)->length, session->nr,
609                                  skb_queue_len(&session->reorder_q));
610                         goto discard;
611                 }
612                 skb_queue_tail(&session->reorder_q, skb);
613         }
614
615 out:
616         return 0;
617
618 discard:
619         return 1;
620 }
621
622 /* Do receive processing of L2TP data frames. We handle both L2TPv2
623  * and L2TPv3 data frames here.
624  *
625  * L2TPv2 Data Message Header
626  *
627  *  0                   1                   2                   3
628  *  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
629  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
630  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
631  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
632  * |           Tunnel ID           |           Session ID          |
633  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
634  * |             Ns (opt)          |             Nr (opt)          |
635  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
636  * |      Offset Size (opt)        |    Offset pad... (opt)
637  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
638  *
639  * Data frames are marked by T=0. All other fields are the same as
640  * those in L2TP control frames.
641  *
642  * L2TPv3 Data Message Header
643  *
644  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
645  * |                      L2TP Session Header                      |
646  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
647  * |                      L2-Specific Sublayer                     |
648  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
649  * |                        Tunnel Payload                      ...
650  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
651  *
652  * L2TPv3 Session Header Over IP
653  *
654  *  0                   1                   2                   3
655  *  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
656  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
657  * |                           Session ID                          |
658  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
659  * |               Cookie (optional, maximum 64 bits)...
660  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
661  *                                                                 |
662  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
663  *
664  * L2TPv3 L2-Specific Sublayer Format
665  *
666  *  0                   1                   2                   3
667  *  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
668  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
669  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
670  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
671  *
672  * Cookie value, sublayer format and offset (pad) are negotiated with
673  * the peer when the session is set up. Unlike L2TPv2, we do not need
674  * to parse the packet header to determine if optional fields are
675  * present.
676  *
677  * Caller must already have parsed the frame and determined that it is
678  * a data (not control) frame before coming here. Fields up to the
679  * session-id have already been parsed and ptr points to the data
680  * after the session-id.
681  */
682 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
683                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
684                       int length, int (*payload_hook)(struct sk_buff *skb))
685 {
686         struct l2tp_tunnel *tunnel = session->tunnel;
687         int offset;
688         u32 ns, nr;
689
690         /* The ref count is increased since we now hold a pointer to
691          * the session. Take care to decrement the refcnt when exiting
692          * this function from now on...
693          */
694         l2tp_session_inc_refcount(session);
695         if (session->ref)
696                 (*session->ref)(session);
697
698         /* Parse and check optional cookie */
699         if (session->peer_cookie_len > 0) {
700                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
701                         l2tp_info(tunnel, L2TP_MSG_DATA,
702                                   "%s: cookie mismatch (%u/%u). Discarding.\n",
703                                   tunnel->name, tunnel->tunnel_id,
704                                   session->session_id);
705                         atomic_long_inc(&session->stats.rx_cookie_discards);
706                         goto discard;
707                 }
708                 ptr += session->peer_cookie_len;
709         }
710
711         /* Handle the optional sequence numbers. Sequence numbers are
712          * in different places for L2TPv2 and L2TPv3.
713          *
714          * If we are the LAC, enable/disable sequence numbers under
715          * the control of the LNS.  If no sequence numbers present but
716          * we were expecting them, discard frame.
717          */
718         ns = nr = 0;
719         L2TP_SKB_CB(skb)->has_seq = 0;
720         if (tunnel->version == L2TP_HDR_VER_2) {
721                 if (hdrflags & L2TP_HDRFLAG_S) {
722                         ns = ntohs(*(__be16 *) ptr);
723                         ptr += 2;
724                         nr = ntohs(*(__be16 *) ptr);
725                         ptr += 2;
726
727                         /* Store L2TP info in the skb */
728                         L2TP_SKB_CB(skb)->ns = ns;
729                         L2TP_SKB_CB(skb)->has_seq = 1;
730
731                         l2tp_dbg(session, L2TP_MSG_SEQ,
732                                  "%s: recv data ns=%u, nr=%u, session nr=%u\n",
733                                  session->name, ns, nr, session->nr);
734                 }
735         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
736                 u32 l2h = ntohl(*(__be32 *) ptr);
737
738                 if (l2h & 0x40000000) {
739                         ns = l2h & 0x00ffffff;
740
741                         /* Store L2TP info in the skb */
742                         L2TP_SKB_CB(skb)->ns = ns;
743                         L2TP_SKB_CB(skb)->has_seq = 1;
744
745                         l2tp_dbg(session, L2TP_MSG_SEQ,
746                                  "%s: recv data ns=%u, session nr=%u\n",
747                                  session->name, ns, session->nr);
748                 }
749         }
750
751         /* Advance past L2-specific header, if present */
752         ptr += session->l2specific_len;
753
754         if (L2TP_SKB_CB(skb)->has_seq) {
755                 /* Received a packet with sequence numbers. If we're the LNS,
756                  * check if we sre sending sequence numbers and if not,
757                  * configure it so.
758                  */
759                 if ((!session->lns_mode) && (!session->send_seq)) {
760                         l2tp_info(session, L2TP_MSG_SEQ,
761                                   "%s: requested to enable seq numbers by LNS\n",
762                                   session->name);
763                         session->send_seq = -1;
764                         l2tp_session_set_header_len(session, tunnel->version);
765                 }
766         } else {
767                 /* No sequence numbers.
768                  * If user has configured mandatory sequence numbers, discard.
769                  */
770                 if (session->recv_seq) {
771                         l2tp_warn(session, L2TP_MSG_SEQ,
772                                   "%s: recv data has no seq numbers when required. Discarding.\n",
773                                   session->name);
774                         atomic_long_inc(&session->stats.rx_seq_discards);
775                         goto discard;
776                 }
777
778                 /* If we're the LAC and we're sending sequence numbers, the
779                  * LNS has requested that we no longer send sequence numbers.
780                  * If we're the LNS and we're sending sequence numbers, the
781                  * LAC is broken. Discard the frame.
782                  */
783                 if ((!session->lns_mode) && (session->send_seq)) {
784                         l2tp_info(session, L2TP_MSG_SEQ,
785                                   "%s: requested to disable seq numbers by LNS\n",
786                                   session->name);
787                         session->send_seq = 0;
788                         l2tp_session_set_header_len(session, tunnel->version);
789                 } else if (session->send_seq) {
790                         l2tp_warn(session, L2TP_MSG_SEQ,
791                                   "%s: recv data has no seq numbers when required. Discarding.\n",
792                                   session->name);
793                         atomic_long_inc(&session->stats.rx_seq_discards);
794                         goto discard;
795                 }
796         }
797
798         /* Session data offset is handled differently for L2TPv2 and
799          * L2TPv3. For L2TPv2, there is an optional 16-bit value in
800          * the header. For L2TPv3, the offset is negotiated using AVPs
801          * in the session setup control protocol.
802          */
803         if (tunnel->version == L2TP_HDR_VER_2) {
804                 /* If offset bit set, skip it. */
805                 if (hdrflags & L2TP_HDRFLAG_O) {
806                         offset = ntohs(*(__be16 *)ptr);
807                         ptr += 2 + offset;
808                 }
809         } else
810                 ptr += session->offset;
811
812         offset = ptr - optr;
813         if (!pskb_may_pull(skb, offset))
814                 goto discard;
815
816         __skb_pull(skb, offset);
817
818         /* If caller wants to process the payload before we queue the
819          * packet, do so now.
820          */
821         if (payload_hook)
822                 if ((*payload_hook)(skb))
823                         goto discard;
824
825         /* Prepare skb for adding to the session's reorder_q.  Hold
826          * packets for max reorder_timeout or 1 second if not
827          * reordering.
828          */
829         L2TP_SKB_CB(skb)->length = length;
830         L2TP_SKB_CB(skb)->expires = jiffies +
831                 (session->reorder_timeout ? session->reorder_timeout : HZ);
832
833         /* Add packet to the session's receive queue. Reordering is done here, if
834          * enabled. Saved L2TP protocol info is stored in skb->sb[].
835          */
836         if (L2TP_SKB_CB(skb)->has_seq) {
837                 if (l2tp_recv_data_seq(session, skb))
838                         goto discard;
839         } else {
840                 /* No sequence numbers. Add the skb to the tail of the
841                  * reorder queue. This ensures that it will be
842                  * delivered after all previous sequenced skbs.
843                  */
844                 skb_queue_tail(&session->reorder_q, skb);
845         }
846
847         /* Try to dequeue as many skbs from reorder_q as we can. */
848         l2tp_recv_dequeue(session);
849
850         l2tp_session_dec_refcount(session);
851
852         return;
853
854 discard:
855         atomic_long_inc(&session->stats.rx_errors);
856         kfree_skb(skb);
857
858         if (session->deref)
859                 (*session->deref)(session);
860
861         l2tp_session_dec_refcount(session);
862 }
863 EXPORT_SYMBOL(l2tp_recv_common);
864
865 /* Drop skbs from the session's reorder_q
866  */
867 int l2tp_session_queue_purge(struct l2tp_session *session)
868 {
869         struct sk_buff *skb = NULL;
870         BUG_ON(!session);
871         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
872         while ((skb = skb_dequeue(&session->reorder_q))) {
873                 atomic_long_inc(&session->stats.rx_errors);
874                 kfree_skb(skb);
875                 if (session->deref)
876                         (*session->deref)(session);
877         }
878         return 0;
879 }
880 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
881
882 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
883  * here. The skb is not on a list when we get here.
884  * Returns 0 if the packet was a data packet and was successfully passed on.
885  * Returns 1 if the packet was not a good data packet and could not be
886  * forwarded.  All such packets are passed up to userspace to deal with.
887  */
888 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
889                               int (*payload_hook)(struct sk_buff *skb))
890 {
891         struct l2tp_session *session = NULL;
892         unsigned char *ptr, *optr;
893         u16 hdrflags;
894         u32 tunnel_id, session_id;
895         u16 version;
896         int length;
897
898         if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
899                 goto discard_bad_csum;
900
901         /* UDP always verifies the packet length. */
902         __skb_pull(skb, sizeof(struct udphdr));
903
904         /* Short packet? */
905         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
906                 l2tp_info(tunnel, L2TP_MSG_DATA,
907                           "%s: recv short packet (len=%d)\n",
908                           tunnel->name, skb->len);
909                 goto error;
910         }
911
912         /* Trace packet contents, if enabled */
913         if (tunnel->debug & L2TP_MSG_DATA) {
914                 length = min(32u, skb->len);
915                 if (!pskb_may_pull(skb, length))
916                         goto error;
917
918                 pr_debug("%s: recv\n", tunnel->name);
919                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
920         }
921
922         /* Point to L2TP header */
923         optr = ptr = skb->data;
924
925         /* Get L2TP header flags */
926         hdrflags = ntohs(*(__be16 *) ptr);
927
928         /* Check protocol version */
929         version = hdrflags & L2TP_HDR_VER_MASK;
930         if (version != tunnel->version) {
931                 l2tp_info(tunnel, L2TP_MSG_DATA,
932                           "%s: recv protocol version mismatch: got %d expected %d\n",
933                           tunnel->name, version, tunnel->version);
934                 goto error;
935         }
936
937         /* Get length of L2TP packet */
938         length = skb->len;
939
940         /* If type is control packet, it is handled by userspace. */
941         if (hdrflags & L2TP_HDRFLAG_T) {
942                 l2tp_dbg(tunnel, L2TP_MSG_DATA,
943                          "%s: recv control packet, len=%d\n",
944                          tunnel->name, length);
945                 goto error;
946         }
947
948         /* Skip flags */
949         ptr += 2;
950
951         if (tunnel->version == L2TP_HDR_VER_2) {
952                 /* If length is present, skip it */
953                 if (hdrflags & L2TP_HDRFLAG_L)
954                         ptr += 2;
955
956                 /* Extract tunnel and session ID */
957                 tunnel_id = ntohs(*(__be16 *) ptr);
958                 ptr += 2;
959                 session_id = ntohs(*(__be16 *) ptr);
960                 ptr += 2;
961         } else {
962                 ptr += 2;       /* skip reserved bits */
963                 tunnel_id = tunnel->tunnel_id;
964                 session_id = ntohl(*(__be32 *) ptr);
965                 ptr += 4;
966         }
967
968         /* Find the session context */
969         session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
970         if (!session || !session->recv_skb) {
971                 /* Not found? Pass to userspace to deal with */
972                 l2tp_info(tunnel, L2TP_MSG_DATA,
973                           "%s: no session found (%u/%u). Passing up.\n",
974                           tunnel->name, tunnel_id, session_id);
975                 goto error;
976         }
977
978         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
979
980         return 0;
981
982 discard_bad_csum:
983         LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
984         UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
985         atomic_long_inc(&tunnel->stats.rx_errors);
986         kfree_skb(skb);
987
988         return 0;
989
990 error:
991         /* Put UDP header back */
992         __skb_push(skb, sizeof(struct udphdr));
993
994         return 1;
995 }
996
997 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
998  * Return codes:
999  * 0 : success.
1000  * <0: error
1001  * >0: skb should be passed up to userspace as UDP.
1002  */
1003 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1004 {
1005         struct l2tp_tunnel *tunnel;
1006
1007         tunnel = l2tp_sock_to_tunnel(sk);
1008         if (tunnel == NULL)
1009                 goto pass_up;
1010
1011         l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
1012                  tunnel->name, skb->len);
1013
1014         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
1015                 goto pass_up_put;
1016
1017         sock_put(sk);
1018         return 0;
1019
1020 pass_up_put:
1021         sock_put(sk);
1022 pass_up:
1023         return 1;
1024 }
1025 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
1026
1027 /************************************************************************
1028  * Transmit handling
1029  ***********************************************************************/
1030
1031 /* Build an L2TP header for the session into the buffer provided.
1032  */
1033 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
1034 {
1035         struct l2tp_tunnel *tunnel = session->tunnel;
1036         __be16 *bufp = buf;
1037         __be16 *optr = buf;
1038         u16 flags = L2TP_HDR_VER_2;
1039         u32 tunnel_id = tunnel->peer_tunnel_id;
1040         u32 session_id = session->peer_session_id;
1041
1042         if (session->send_seq)
1043                 flags |= L2TP_HDRFLAG_S;
1044
1045         /* Setup L2TP header. */
1046         *bufp++ = htons(flags);
1047         *bufp++ = htons(tunnel_id);
1048         *bufp++ = htons(session_id);
1049         if (session->send_seq) {
1050                 *bufp++ = htons(session->ns);
1051                 *bufp++ = 0;
1052                 session->ns++;
1053                 session->ns &= 0xffff;
1054                 l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
1055                          session->name, session->ns);
1056         }
1057
1058         return bufp - optr;
1059 }
1060
1061 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
1062 {
1063         struct l2tp_tunnel *tunnel = session->tunnel;
1064         char *bufp = buf;
1065         char *optr = bufp;
1066
1067         /* Setup L2TP header. The header differs slightly for UDP and
1068          * IP encapsulations. For UDP, there is 4 bytes of flags.
1069          */
1070         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1071                 u16 flags = L2TP_HDR_VER_3;
1072                 *((__be16 *) bufp) = htons(flags);
1073                 bufp += 2;
1074                 *((__be16 *) bufp) = 0;
1075                 bufp += 2;
1076         }
1077
1078         *((__be32 *) bufp) = htonl(session->peer_session_id);
1079         bufp += 4;
1080         if (session->cookie_len) {
1081                 memcpy(bufp, &session->cookie[0], session->cookie_len);
1082                 bufp += session->cookie_len;
1083         }
1084         if (session->l2specific_len) {
1085                 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
1086                         u32 l2h = 0;
1087                         if (session->send_seq) {
1088                                 l2h = 0x40000000 | session->ns;
1089                                 session->ns++;
1090                                 session->ns &= 0xffffff;
1091                                 l2tp_dbg(session, L2TP_MSG_SEQ,
1092                                          "%s: updated ns to %u\n",
1093                                          session->name, session->ns);
1094                         }
1095
1096                         *((__be32 *) bufp) = htonl(l2h);
1097                 }
1098                 bufp += session->l2specific_len;
1099         }
1100         if (session->offset)
1101                 bufp += session->offset;
1102
1103         return bufp - optr;
1104 }
1105
1106 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1107                           struct flowi *fl, size_t data_len)
1108 {
1109         struct l2tp_tunnel *tunnel = session->tunnel;
1110         unsigned int len = skb->len;
1111         int error;
1112
1113         /* Debug */
1114         if (session->send_seq)
1115                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes, ns=%u\n",
1116                          session->name, data_len, session->ns - 1);
1117         else
1118                 l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %Zd bytes\n",
1119                          session->name, data_len);
1120
1121         if (session->debug & L2TP_MSG_DATA) {
1122                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1123                 unsigned char *datap = skb->data + uhlen;
1124
1125                 pr_debug("%s: xmit\n", session->name);
1126                 print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1127                                      datap, min_t(size_t, 32, len - uhlen));
1128         }
1129
1130         /* Queue the packet to IP for output */
1131         skb->local_df = 1;
1132 #if IS_ENABLED(CONFIG_IPV6)
1133         if (tunnel->sock->sk_family == PF_INET6 && !tunnel->v4mapped)
1134                 error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1135         else
1136 #endif
1137                 error = ip_queue_xmit(tunnel->sock, skb, fl);
1138
1139         /* Update stats */
1140         if (error >= 0) {
1141                 atomic_long_inc(&tunnel->stats.tx_packets);
1142                 atomic_long_add(len, &tunnel->stats.tx_bytes);
1143                 atomic_long_inc(&session->stats.tx_packets);
1144                 atomic_long_add(len, &session->stats.tx_bytes);
1145         } else {
1146                 atomic_long_inc(&tunnel->stats.tx_errors);
1147                 atomic_long_inc(&session->stats.tx_errors);
1148         }
1149
1150         return 0;
1151 }
1152
1153 #if IS_ENABLED(CONFIG_IPV6)
1154 static void l2tp_xmit_ipv6_csum(struct sock *sk, struct sk_buff *skb,
1155                                 int udp_len)
1156 {
1157         struct ipv6_pinfo *np = inet6_sk(sk);
1158         struct udphdr *uh = udp_hdr(skb);
1159
1160         if (!skb_dst(skb) || !skb_dst(skb)->dev ||
1161             !(skb_dst(skb)->dev->features & NETIF_F_IPV6_CSUM)) {
1162                 __wsum csum = skb_checksum(skb, 0, udp_len, 0);
1163                 skb->ip_summed = CHECKSUM_UNNECESSARY;
1164                 uh->check = csum_ipv6_magic(&np->saddr, &sk->sk_v6_daddr, udp_len,
1165                                             IPPROTO_UDP, csum);
1166                 if (uh->check == 0)
1167                         uh->check = CSUM_MANGLED_0;
1168         } else {
1169                 skb->ip_summed = CHECKSUM_PARTIAL;
1170                 skb->csum_start = skb_transport_header(skb) - skb->head;
1171                 skb->csum_offset = offsetof(struct udphdr, check);
1172                 uh->check = ~csum_ipv6_magic(&np->saddr, &sk->sk_v6_daddr,
1173                                              udp_len, IPPROTO_UDP, 0);
1174         }
1175 }
1176 #endif
1177
1178 /* If caller requires the skb to have a ppp header, the header must be
1179  * inserted in the skb data before calling this function.
1180  */
1181 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1182 {
1183         int data_len = skb->len;
1184         struct l2tp_tunnel *tunnel = session->tunnel;
1185         struct sock *sk = tunnel->sock;
1186         struct flowi *fl;
1187         struct udphdr *uh;
1188         struct inet_sock *inet;
1189         __wsum csum;
1190         int headroom;
1191         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1192         int udp_len;
1193         int ret = NET_XMIT_SUCCESS;
1194
1195         /* Check that there's enough headroom in the skb to insert IP,
1196          * UDP and L2TP headers. If not enough, expand it to
1197          * make room. Adjust truesize.
1198          */
1199         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1200                 uhlen + hdr_len;
1201         if (skb_cow_head(skb, headroom)) {
1202                 kfree_skb(skb);
1203                 return NET_XMIT_DROP;
1204         }
1205
1206         /* Setup L2TP header */
1207         session->build_header(session, __skb_push(skb, hdr_len));
1208
1209         /* Reset skb netfilter state */
1210         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1211         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1212                               IPSKB_REROUTED);
1213         nf_reset(skb);
1214
1215         bh_lock_sock(sk);
1216         if (sock_owned_by_user(sk)) {
1217                 kfree_skb(skb);
1218                 ret = NET_XMIT_DROP;
1219                 goto out_unlock;
1220         }
1221
1222         /* Get routing info from the tunnel socket */
1223         skb_dst_drop(skb);
1224         skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1225
1226         inet = inet_sk(sk);
1227         fl = &inet->cork.fl;
1228         switch (tunnel->encap) {
1229         case L2TP_ENCAPTYPE_UDP:
1230                 /* Setup UDP header */
1231                 __skb_push(skb, sizeof(*uh));
1232                 skb_reset_transport_header(skb);
1233                 uh = udp_hdr(skb);
1234                 uh->source = inet->inet_sport;
1235                 uh->dest = inet->inet_dport;
1236                 udp_len = uhlen + hdr_len + data_len;
1237                 uh->len = htons(udp_len);
1238                 uh->check = 0;
1239
1240                 /* Calculate UDP checksum if configured to do so */
1241 #if IS_ENABLED(CONFIG_IPV6)
1242                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1243                         l2tp_xmit_ipv6_csum(sk, skb, udp_len);
1244                 else
1245 #endif
1246                 if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1247                         skb->ip_summed = CHECKSUM_NONE;
1248                 else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1249                          (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1250                         skb->ip_summed = CHECKSUM_COMPLETE;
1251                         csum = skb_checksum(skb, 0, udp_len, 0);
1252                         uh->check = csum_tcpudp_magic(inet->inet_saddr,
1253                                                       inet->inet_daddr,
1254                                                       udp_len, IPPROTO_UDP, csum);
1255                         if (uh->check == 0)
1256                                 uh->check = CSUM_MANGLED_0;
1257                 } else {
1258                         skb->ip_summed = CHECKSUM_PARTIAL;
1259                         skb->csum_start = skb_transport_header(skb) - skb->head;
1260                         skb->csum_offset = offsetof(struct udphdr, check);
1261                         uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1262                                                        inet->inet_daddr,
1263                                                        udp_len, IPPROTO_UDP, 0);
1264                 }
1265                 break;
1266
1267         case L2TP_ENCAPTYPE_IP:
1268                 break;
1269         }
1270
1271         l2tp_xmit_core(session, skb, fl, data_len);
1272 out_unlock:
1273         bh_unlock_sock(sk);
1274
1275         return ret;
1276 }
1277 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1278
1279 /*****************************************************************************
1280  * Tinnel and session create/destroy.
1281  *****************************************************************************/
1282
1283 /* Tunnel socket destruct hook.
1284  * The tunnel context is deleted only when all session sockets have been
1285  * closed.
1286  */
1287 static void l2tp_tunnel_destruct(struct sock *sk)
1288 {
1289         struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1290         struct l2tp_net *pn;
1291
1292         if (tunnel == NULL)
1293                 goto end;
1294
1295         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1296
1297
1298         /* Disable udp encapsulation */
1299         switch (tunnel->encap) {
1300         case L2TP_ENCAPTYPE_UDP:
1301                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1302                 (udp_sk(sk))->encap_type = 0;
1303                 (udp_sk(sk))->encap_rcv = NULL;
1304                 (udp_sk(sk))->encap_destroy = NULL;
1305                 break;
1306         case L2TP_ENCAPTYPE_IP:
1307                 break;
1308         }
1309
1310         /* Remove hooks into tunnel socket */
1311         sk->sk_destruct = tunnel->old_sk_destruct;
1312         sk->sk_user_data = NULL;
1313         tunnel->sock = NULL;
1314
1315         /* Remove the tunnel struct from the tunnel list */
1316         pn = l2tp_pernet(tunnel->l2tp_net);
1317         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1318         list_del_rcu(&tunnel->list);
1319         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1320         atomic_dec(&l2tp_tunnel_count);
1321
1322         l2tp_tunnel_closeall(tunnel);
1323         l2tp_tunnel_dec_refcount(tunnel);
1324
1325         /* Call the original destructor */
1326         if (sk->sk_destruct)
1327                 (*sk->sk_destruct)(sk);
1328 end:
1329         return;
1330 }
1331
1332 /* When the tunnel is closed, all the attached sessions need to go too.
1333  */
1334 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1335 {
1336         int hash;
1337         struct hlist_node *walk;
1338         struct hlist_node *tmp;
1339         struct l2tp_session *session;
1340
1341         BUG_ON(tunnel == NULL);
1342
1343         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1344                   tunnel->name);
1345
1346         write_lock_bh(&tunnel->hlist_lock);
1347         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1348 again:
1349                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1350                         session = hlist_entry(walk, struct l2tp_session, hlist);
1351
1352                         l2tp_info(session, L2TP_MSG_CONTROL,
1353                                   "%s: closing session\n", session->name);
1354
1355                         hlist_del_init(&session->hlist);
1356
1357                         if (session->ref != NULL)
1358                                 (*session->ref)(session);
1359
1360                         write_unlock_bh(&tunnel->hlist_lock);
1361
1362                         __l2tp_session_unhash(session);
1363                         l2tp_session_queue_purge(session);
1364
1365                         if (session->session_close != NULL)
1366                                 (*session->session_close)(session);
1367
1368                         if (session->deref != NULL)
1369                                 (*session->deref)(session);
1370
1371                         l2tp_session_dec_refcount(session);
1372
1373                         write_lock_bh(&tunnel->hlist_lock);
1374
1375                         /* Now restart from the beginning of this hash
1376                          * chain.  We always remove a session from the
1377                          * list so we are guaranteed to make forward
1378                          * progress.
1379                          */
1380                         goto again;
1381                 }
1382         }
1383         write_unlock_bh(&tunnel->hlist_lock);
1384 }
1385 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1386
1387 /* Tunnel socket destroy hook for UDP encapsulation */
1388 static void l2tp_udp_encap_destroy(struct sock *sk)
1389 {
1390         struct l2tp_tunnel *tunnel = l2tp_sock_to_tunnel(sk);
1391         if (tunnel) {
1392                 l2tp_tunnel_closeall(tunnel);
1393                 sock_put(sk);
1394         }
1395 }
1396
1397 /* Really kill the tunnel.
1398  * Come here only when all sessions have been cleared from the tunnel.
1399  */
1400 static void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1401 {
1402         BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1403         BUG_ON(tunnel->sock != NULL);
1404         l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: free...\n", tunnel->name);
1405         kfree_rcu(tunnel, rcu);
1406 }
1407
1408 /* Workqueue tunnel deletion function */
1409 static void l2tp_tunnel_del_work(struct work_struct *work)
1410 {
1411         struct l2tp_tunnel *tunnel = NULL;
1412         struct socket *sock = NULL;
1413         struct sock *sk = NULL;
1414
1415         tunnel = container_of(work, struct l2tp_tunnel, del_work);
1416         sk = l2tp_tunnel_sock_lookup(tunnel);
1417         if (!sk)
1418                 return;
1419
1420         sock = sk->sk_socket;
1421
1422         /* If the tunnel socket was created by userspace, then go through the
1423          * inet layer to shut the socket down, and let userspace close it.
1424          * Otherwise, if we created the socket directly within the kernel, use
1425          * the sk API to release it here.
1426          * In either case the tunnel resources are freed in the socket
1427          * destructor when the tunnel socket goes away.
1428          */
1429         if (tunnel->fd >= 0) {
1430                 if (sock)
1431                         inet_shutdown(sock, 2);
1432         } else {
1433                 if (sock)
1434                         kernel_sock_shutdown(sock, SHUT_RDWR);
1435                 sk_release_kernel(sk);
1436         }
1437
1438         l2tp_tunnel_sock_put(sk);
1439 }
1440
1441 /* Create a socket for the tunnel, if one isn't set up by
1442  * userspace. This is used for static tunnels where there is no
1443  * managing L2TP daemon.
1444  *
1445  * Since we don't want these sockets to keep a namespace alive by
1446  * themselves, we drop the socket's namespace refcount after creation.
1447  * These sockets are freed when the namespace exits using the pernet
1448  * exit hook.
1449  */
1450 static int l2tp_tunnel_sock_create(struct net *net,
1451                                 u32 tunnel_id,
1452                                 u32 peer_tunnel_id,
1453                                 struct l2tp_tunnel_cfg *cfg,
1454                                 struct socket **sockp)
1455 {
1456         int err = -EINVAL;
1457         struct socket *sock = NULL;
1458         struct sockaddr_in udp_addr = {0};
1459         struct sockaddr_l2tpip ip_addr = {0};
1460 #if IS_ENABLED(CONFIG_IPV6)
1461         struct sockaddr_in6 udp6_addr = {0};
1462         struct sockaddr_l2tpip6 ip6_addr = {0};
1463 #endif
1464
1465         switch (cfg->encap) {
1466         case L2TP_ENCAPTYPE_UDP:
1467 #if IS_ENABLED(CONFIG_IPV6)
1468                 if (cfg->local_ip6 && cfg->peer_ip6) {
1469                         err = sock_create_kern(AF_INET6, SOCK_DGRAM, 0, &sock);
1470                         if (err < 0)
1471                                 goto out;
1472
1473                         sk_change_net(sock->sk, net);
1474
1475                         udp6_addr.sin6_family = AF_INET6;
1476                         memcpy(&udp6_addr.sin6_addr, cfg->local_ip6,
1477                                sizeof(udp6_addr.sin6_addr));
1478                         udp6_addr.sin6_port = htons(cfg->local_udp_port);
1479                         err = kernel_bind(sock, (struct sockaddr *) &udp6_addr,
1480                                           sizeof(udp6_addr));
1481                         if (err < 0)
1482                                 goto out;
1483
1484                         udp6_addr.sin6_family = AF_INET6;
1485                         memcpy(&udp6_addr.sin6_addr, cfg->peer_ip6,
1486                                sizeof(udp6_addr.sin6_addr));
1487                         udp6_addr.sin6_port = htons(cfg->peer_udp_port);
1488                         err = kernel_connect(sock,
1489                                              (struct sockaddr *) &udp6_addr,
1490                                              sizeof(udp6_addr), 0);
1491                         if (err < 0)
1492                                 goto out;
1493                 } else
1494 #endif
1495                 {
1496                         err = sock_create_kern(AF_INET, SOCK_DGRAM, 0, &sock);
1497                         if (err < 0)
1498                                 goto out;
1499
1500                         sk_change_net(sock->sk, net);
1501
1502                         udp_addr.sin_family = AF_INET;
1503                         udp_addr.sin_addr = cfg->local_ip;
1504                         udp_addr.sin_port = htons(cfg->local_udp_port);
1505                         err = kernel_bind(sock, (struct sockaddr *) &udp_addr,
1506                                           sizeof(udp_addr));
1507                         if (err < 0)
1508                                 goto out;
1509
1510                         udp_addr.sin_family = AF_INET;
1511                         udp_addr.sin_addr = cfg->peer_ip;
1512                         udp_addr.sin_port = htons(cfg->peer_udp_port);
1513                         err = kernel_connect(sock,
1514                                              (struct sockaddr *) &udp_addr,
1515                                              sizeof(udp_addr), 0);
1516                         if (err < 0)
1517                                 goto out;
1518                 }
1519
1520                 if (!cfg->use_udp_checksums)
1521                         sock->sk->sk_no_check = UDP_CSUM_NOXMIT;
1522
1523                 break;
1524
1525         case L2TP_ENCAPTYPE_IP:
1526 #if IS_ENABLED(CONFIG_IPV6)
1527                 if (cfg->local_ip6 && cfg->peer_ip6) {
1528                         err = sock_create_kern(AF_INET6, SOCK_DGRAM,
1529                                           IPPROTO_L2TP, &sock);
1530                         if (err < 0)
1531                                 goto out;
1532
1533                         sk_change_net(sock->sk, net);
1534
1535                         ip6_addr.l2tp_family = AF_INET6;
1536                         memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1537                                sizeof(ip6_addr.l2tp_addr));
1538                         ip6_addr.l2tp_conn_id = tunnel_id;
1539                         err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1540                                           sizeof(ip6_addr));
1541                         if (err < 0)
1542                                 goto out;
1543
1544                         ip6_addr.l2tp_family = AF_INET6;
1545                         memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1546                                sizeof(ip6_addr.l2tp_addr));
1547                         ip6_addr.l2tp_conn_id = peer_tunnel_id;
1548                         err = kernel_connect(sock,
1549                                              (struct sockaddr *) &ip6_addr,
1550                                              sizeof(ip6_addr), 0);
1551                         if (err < 0)
1552                                 goto out;
1553                 } else
1554 #endif
1555                 {
1556                         err = sock_create_kern(AF_INET, SOCK_DGRAM,
1557                                           IPPROTO_L2TP, &sock);
1558                         if (err < 0)
1559                                 goto out;
1560
1561                         sk_change_net(sock->sk, net);
1562
1563                         ip_addr.l2tp_family = AF_INET;
1564                         ip_addr.l2tp_addr = cfg->local_ip;
1565                         ip_addr.l2tp_conn_id = tunnel_id;
1566                         err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1567                                           sizeof(ip_addr));
1568                         if (err < 0)
1569                                 goto out;
1570
1571                         ip_addr.l2tp_family = AF_INET;
1572                         ip_addr.l2tp_addr = cfg->peer_ip;
1573                         ip_addr.l2tp_conn_id = peer_tunnel_id;
1574                         err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1575                                              sizeof(ip_addr), 0);
1576                         if (err < 0)
1577                                 goto out;
1578                 }
1579                 break;
1580
1581         default:
1582                 goto out;
1583         }
1584
1585 out:
1586         *sockp = sock;
1587         if ((err < 0) && sock) {
1588                 kernel_sock_shutdown(sock, SHUT_RDWR);
1589                 sk_release_kernel(sock->sk);
1590                 *sockp = NULL;
1591         }
1592
1593         return err;
1594 }
1595
1596 static struct lock_class_key l2tp_socket_class;
1597
1598 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)
1599 {
1600         struct l2tp_tunnel *tunnel = NULL;
1601         int err;
1602         struct socket *sock = NULL;
1603         struct sock *sk = NULL;
1604         struct l2tp_net *pn;
1605         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1606
1607         /* Get the tunnel socket from the fd, which was opened by
1608          * the userspace L2TP daemon. If not specified, create a
1609          * kernel socket.
1610          */
1611         if (fd < 0) {
1612                 err = l2tp_tunnel_sock_create(net, tunnel_id, peer_tunnel_id,
1613                                 cfg, &sock);
1614                 if (err < 0)
1615                         goto err;
1616         } else {
1617                 sock = sockfd_lookup(fd, &err);
1618                 if (!sock) {
1619                         pr_err("tunl %u: sockfd_lookup(fd=%d) returned %d\n",
1620                                tunnel_id, fd, err);
1621                         err = -EBADF;
1622                         goto err;
1623                 }
1624
1625                 /* Reject namespace mismatches */
1626                 if (!net_eq(sock_net(sock->sk), net)) {
1627                         pr_err("tunl %u: netns mismatch\n", tunnel_id);
1628                         err = -EINVAL;
1629                         goto err;
1630                 }
1631         }
1632
1633         sk = sock->sk;
1634
1635         if (cfg != NULL)
1636                 encap = cfg->encap;
1637
1638         /* Quick sanity checks */
1639         switch (encap) {
1640         case L2TP_ENCAPTYPE_UDP:
1641                 err = -EPROTONOSUPPORT;
1642                 if (sk->sk_protocol != IPPROTO_UDP) {
1643                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1644                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1645                         goto err;
1646                 }
1647                 break;
1648         case L2TP_ENCAPTYPE_IP:
1649                 err = -EPROTONOSUPPORT;
1650                 if (sk->sk_protocol != IPPROTO_L2TP) {
1651                         pr_err("tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1652                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1653                         goto err;
1654                 }
1655                 break;
1656         }
1657
1658         /* Check if this socket has already been prepped */
1659         tunnel = l2tp_tunnel(sk);
1660         if (tunnel != NULL) {
1661                 /* This socket has already been prepped */
1662                 err = -EBUSY;
1663                 goto err;
1664         }
1665
1666         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1667         if (tunnel == NULL) {
1668                 err = -ENOMEM;
1669                 goto err;
1670         }
1671
1672         tunnel->version = version;
1673         tunnel->tunnel_id = tunnel_id;
1674         tunnel->peer_tunnel_id = peer_tunnel_id;
1675         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1676
1677         tunnel->magic = L2TP_TUNNEL_MAGIC;
1678         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1679         rwlock_init(&tunnel->hlist_lock);
1680
1681         /* The net we belong to */
1682         tunnel->l2tp_net = net;
1683         pn = l2tp_pernet(net);
1684
1685         if (cfg != NULL)
1686                 tunnel->debug = cfg->debug;
1687
1688 #if IS_ENABLED(CONFIG_IPV6)
1689         if (sk->sk_family == PF_INET6) {
1690                 struct ipv6_pinfo *np = inet6_sk(sk);
1691
1692                 if (ipv6_addr_v4mapped(&np->saddr) &&
1693                     ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
1694                         struct inet_sock *inet = inet_sk(sk);
1695
1696                         tunnel->v4mapped = true;
1697                         inet->inet_saddr = np->saddr.s6_addr32[3];
1698                         inet->inet_rcv_saddr = sk->sk_v6_rcv_saddr.s6_addr32[3];
1699                         inet->inet_daddr = sk->sk_v6_daddr.s6_addr32[3];
1700                 } else {
1701                         tunnel->v4mapped = false;
1702                 }
1703         }
1704 #endif
1705
1706         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1707         tunnel->encap = encap;
1708         if (encap == L2TP_ENCAPTYPE_UDP) {
1709                 /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1710                 udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1711                 udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1712                 udp_sk(sk)->encap_destroy = l2tp_udp_encap_destroy;
1713 #if IS_ENABLED(CONFIG_IPV6)
1714                 if (sk->sk_family == PF_INET6 && !tunnel->v4mapped)
1715                         udpv6_encap_enable();
1716                 else
1717 #endif
1718                 udp_encap_enable();
1719         }
1720
1721         sk->sk_user_data = tunnel;
1722
1723         /* Hook on the tunnel socket destructor so that we can cleanup
1724          * if the tunnel socket goes away.
1725          */
1726         tunnel->old_sk_destruct = sk->sk_destruct;
1727         sk->sk_destruct = &l2tp_tunnel_destruct;
1728         tunnel->sock = sk;
1729         tunnel->fd = fd;
1730         lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class, "l2tp_sock");
1731
1732         sk->sk_allocation = GFP_ATOMIC;
1733
1734         /* Init delete workqueue struct */
1735         INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1736
1737         /* Add tunnel to our list */
1738         INIT_LIST_HEAD(&tunnel->list);
1739         atomic_inc(&l2tp_tunnel_count);
1740
1741         /* Bump the reference count. The tunnel context is deleted
1742          * only when this drops to zero. Must be done before list insertion
1743          */
1744         l2tp_tunnel_inc_refcount(tunnel);
1745         spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1746         list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1747         spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1748
1749         err = 0;
1750 err:
1751         if (tunnelp)
1752                 *tunnelp = tunnel;
1753
1754         /* If tunnel's socket was created by the kernel, it doesn't
1755          *  have a file.
1756          */
1757         if (sock && sock->file)
1758                 sockfd_put(sock);
1759
1760         return err;
1761 }
1762 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1763
1764 /* This function is used by the netlink TUNNEL_DELETE command.
1765  */
1766 int l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1767 {
1768         l2tp_tunnel_closeall(tunnel);
1769         return (false == queue_work(l2tp_wq, &tunnel->del_work));
1770 }
1771 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1772
1773 /* Really kill the session.
1774  */
1775 void l2tp_session_free(struct l2tp_session *session)
1776 {
1777         struct l2tp_tunnel *tunnel = session->tunnel;
1778
1779         BUG_ON(atomic_read(&session->ref_count) != 0);
1780
1781         if (tunnel) {
1782                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1783                 if (session->session_id != 0)
1784                         atomic_dec(&l2tp_session_count);
1785                 sock_put(tunnel->sock);
1786                 session->tunnel = NULL;
1787                 l2tp_tunnel_dec_refcount(tunnel);
1788         }
1789
1790         kfree(session);
1791 }
1792 EXPORT_SYMBOL_GPL(l2tp_session_free);
1793
1794 /* Remove an l2tp session from l2tp_core's hash lists.
1795  * Provides a tidyup interface for pseudowire code which can't just route all
1796  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1797  * callback.
1798  */
1799 void __l2tp_session_unhash(struct l2tp_session *session)
1800 {
1801         struct l2tp_tunnel *tunnel = session->tunnel;
1802
1803         /* Remove the session from core hashes */
1804         if (tunnel) {
1805                 /* Remove from the per-tunnel hash */
1806                 write_lock_bh(&tunnel->hlist_lock);
1807                 hlist_del_init(&session->hlist);
1808                 write_unlock_bh(&tunnel->hlist_lock);
1809
1810                 /* For L2TPv3 we have a per-net hash: remove from there, too */
1811                 if (tunnel->version != L2TP_HDR_VER_2) {
1812                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1813                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1814                         hlist_del_init_rcu(&session->global_hlist);
1815                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1816                         synchronize_rcu();
1817                 }
1818         }
1819 }
1820 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1821
1822 /* This function is used by the netlink SESSION_DELETE command and by
1823    pseudowire modules.
1824  */
1825 int l2tp_session_delete(struct l2tp_session *session)
1826 {
1827         if (session->ref)
1828                 (*session->ref)(session);
1829         __l2tp_session_unhash(session);
1830         l2tp_session_queue_purge(session);
1831         if (session->session_close != NULL)
1832                 (*session->session_close)(session);
1833         if (session->deref)
1834                 (*session->deref)(session);
1835         l2tp_session_dec_refcount(session);
1836         return 0;
1837 }
1838 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1839
1840 /* We come here whenever a session's send_seq, cookie_len or
1841  * l2specific_len parameters are set.
1842  */
1843 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1844 {
1845         if (version == L2TP_HDR_VER_2) {
1846                 session->hdr_len = 6;
1847                 if (session->send_seq)
1848                         session->hdr_len += 4;
1849         } else {
1850                 session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1851                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1852                         session->hdr_len += 4;
1853         }
1854
1855 }
1856 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1857
1858 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)
1859 {
1860         struct l2tp_session *session;
1861
1862         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1863         if (session != NULL) {
1864                 session->magic = L2TP_SESSION_MAGIC;
1865                 session->tunnel = tunnel;
1866
1867                 session->session_id = session_id;
1868                 session->peer_session_id = peer_session_id;
1869                 session->nr = 0;
1870                 if (tunnel->version == L2TP_HDR_VER_2)
1871                         session->nr_max = 0xffff;
1872                 else
1873                         session->nr_max = 0xffffff;
1874                 session->nr_window_size = session->nr_max / 2;
1875                 session->nr_oos_count_max = 4;
1876
1877                 /* Use NR of first received packet */
1878                 session->reorder_skip = 1;
1879
1880                 sprintf(&session->name[0], "sess %u/%u",
1881                         tunnel->tunnel_id, session->session_id);
1882
1883                 skb_queue_head_init(&session->reorder_q);
1884
1885                 INIT_HLIST_NODE(&session->hlist);
1886                 INIT_HLIST_NODE(&session->global_hlist);
1887
1888                 /* Inherit debug options from tunnel */
1889                 session->debug = tunnel->debug;
1890
1891                 if (cfg) {
1892                         session->pwtype = cfg->pw_type;
1893                         session->debug = cfg->debug;
1894                         session->mtu = cfg->mtu;
1895                         session->mru = cfg->mru;
1896                         session->send_seq = cfg->send_seq;
1897                         session->recv_seq = cfg->recv_seq;
1898                         session->lns_mode = cfg->lns_mode;
1899                         session->reorder_timeout = cfg->reorder_timeout;
1900                         session->offset = cfg->offset;
1901                         session->l2specific_type = cfg->l2specific_type;
1902                         session->l2specific_len = cfg->l2specific_len;
1903                         session->cookie_len = cfg->cookie_len;
1904                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1905                         session->peer_cookie_len = cfg->peer_cookie_len;
1906                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1907                 }
1908
1909                 if (tunnel->version == L2TP_HDR_VER_2)
1910                         session->build_header = l2tp_build_l2tpv2_header;
1911                 else
1912                         session->build_header = l2tp_build_l2tpv3_header;
1913
1914                 l2tp_session_set_header_len(session, tunnel->version);
1915
1916                 /* Bump the reference count. The session context is deleted
1917                  * only when this drops to zero.
1918                  */
1919                 l2tp_session_inc_refcount(session);
1920                 l2tp_tunnel_inc_refcount(tunnel);
1921
1922                 /* Ensure tunnel socket isn't deleted */
1923                 sock_hold(tunnel->sock);
1924
1925                 /* Add session to the tunnel's hash list */
1926                 write_lock_bh(&tunnel->hlist_lock);
1927                 hlist_add_head(&session->hlist,
1928                                l2tp_session_id_hash(tunnel, session_id));
1929                 write_unlock_bh(&tunnel->hlist_lock);
1930
1931                 /* And to the global session list if L2TPv3 */
1932                 if (tunnel->version != L2TP_HDR_VER_2) {
1933                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1934
1935                         spin_lock_bh(&pn->l2tp_session_hlist_lock);
1936                         hlist_add_head_rcu(&session->global_hlist,
1937                                            l2tp_session_id_hash_2(pn, session_id));
1938                         spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1939                 }
1940
1941                 /* Ignore management session in session count value */
1942                 if (session->session_id != 0)
1943                         atomic_inc(&l2tp_session_count);
1944         }
1945
1946         return session;
1947 }
1948 EXPORT_SYMBOL_GPL(l2tp_session_create);
1949
1950 /*****************************************************************************
1951  * Init and cleanup
1952  *****************************************************************************/
1953
1954 static __net_init int l2tp_init_net(struct net *net)
1955 {
1956         struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1957         int hash;
1958
1959         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1960         spin_lock_init(&pn->l2tp_tunnel_list_lock);
1961
1962         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1963                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1964
1965         spin_lock_init(&pn->l2tp_session_hlist_lock);
1966
1967         return 0;
1968 }
1969
1970 static __net_exit void l2tp_exit_net(struct net *net)
1971 {
1972         struct l2tp_net *pn = l2tp_pernet(net);
1973         struct l2tp_tunnel *tunnel = NULL;
1974
1975         rcu_read_lock_bh();
1976         list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1977                 (void)l2tp_tunnel_delete(tunnel);
1978         }
1979         rcu_read_unlock_bh();
1980 }
1981
1982 static struct pernet_operations l2tp_net_ops = {
1983         .init = l2tp_init_net,
1984         .exit = l2tp_exit_net,
1985         .id   = &l2tp_net_id,
1986         .size = sizeof(struct l2tp_net),
1987 };
1988
1989 static int __init l2tp_init(void)
1990 {
1991         int rc = 0;
1992
1993         rc = register_pernet_device(&l2tp_net_ops);
1994         if (rc)
1995                 goto out;
1996
1997         l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1998         if (!l2tp_wq) {
1999                 pr_err("alloc_workqueue failed\n");
2000                 rc = -ENOMEM;
2001                 goto out;
2002         }
2003
2004         pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
2005
2006 out:
2007         return rc;
2008 }
2009
2010 static void __exit l2tp_exit(void)
2011 {
2012         unregister_pernet_device(&l2tp_net_ops);
2013         if (l2tp_wq) {
2014                 destroy_workqueue(l2tp_wq);
2015                 l2tp_wq = NULL;
2016         }
2017 }
2018
2019 module_init(l2tp_init);
2020 module_exit(l2tp_exit);
2021
2022 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
2023 MODULE_DESCRIPTION("L2TP core");
2024 MODULE_LICENSE("GPL");
2025 MODULE_VERSION(L2TP_DRV_VERSION);
2026