l2tp: Add L2TPv3 IP encapsulation (no UDP) support
[pandora-kernel.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 #include <linux/module.h>
22 #include <linux/string.h>
23 #include <linux/list.h>
24 #include <linux/uaccess.h>
25
26 #include <linux/kernel.h>
27 #include <linux/spinlock.h>
28 #include <linux/kthread.h>
29 #include <linux/sched.h>
30 #include <linux/slab.h>
31 #include <linux/errno.h>
32 #include <linux/jiffies.h>
33
34 #include <linux/netdevice.h>
35 #include <linux/net.h>
36 #include <linux/inetdevice.h>
37 #include <linux/skbuff.h>
38 #include <linux/init.h>
39 #include <linux/in.h>
40 #include <linux/ip.h>
41 #include <linux/udp.h>
42 #include <linux/l2tp.h>
43 #include <linux/hash.h>
44 #include <linux/sort.h>
45 #include <linux/file.h>
46 #include <linux/nsproxy.h>
47 #include <net/net_namespace.h>
48 #include <net/netns/generic.h>
49 #include <net/dst.h>
50 #include <net/ip.h>
51 #include <net/udp.h>
52 #include <net/xfrm.h>
53 #include <net/protocol.h>
54
55 #include <asm/byteorder.h>
56 #include <asm/atomic.h>
57
58 #include "l2tp_core.h"
59
60 #define L2TP_DRV_VERSION        "V2.0"
61
62 /* L2TP header constants */
63 #define L2TP_HDRFLAG_T     0x8000
64 #define L2TP_HDRFLAG_L     0x4000
65 #define L2TP_HDRFLAG_S     0x0800
66 #define L2TP_HDRFLAG_O     0x0200
67 #define L2TP_HDRFLAG_P     0x0100
68
69 #define L2TP_HDR_VER_MASK  0x000F
70 #define L2TP_HDR_VER_2     0x0002
71 #define L2TP_HDR_VER_3     0x0003
72
73 /* L2TPv3 default L2-specific sublayer */
74 #define L2TP_SLFLAG_S      0x40000000
75 #define L2TP_SL_SEQ_MASK   0x00ffffff
76
77 #define L2TP_HDR_SIZE_SEQ               10
78 #define L2TP_HDR_SIZE_NOSEQ             6
79
80 /* Default trace flags */
81 #define L2TP_DEFAULT_DEBUG_FLAGS        0
82
83 #define PRINTK(_mask, _type, _lvl, _fmt, args...)                       \
84         do {                                                            \
85                 if ((_mask) & (_type))                                  \
86                         printk(_lvl "L2TP: " _fmt, ##args);             \
87         } while (0)
88
89 /* Private data stored for received packets in the skb.
90  */
91 struct l2tp_skb_cb {
92         u32                     ns;
93         u16                     has_seq;
94         u16                     length;
95         unsigned long           expires;
96 };
97
98 #define L2TP_SKB_CB(skb)        ((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
99
100 static atomic_t l2tp_tunnel_count;
101 static atomic_t l2tp_session_count;
102
103 /* per-net private data for this module */
104 static unsigned int l2tp_net_id;
105 struct l2tp_net {
106         struct list_head l2tp_tunnel_list;
107         rwlock_t l2tp_tunnel_list_lock;
108         struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
109         rwlock_t l2tp_session_hlist_lock;
110 };
111
112 static inline struct l2tp_net *l2tp_pernet(struct net *net)
113 {
114         BUG_ON(!net);
115
116         return net_generic(net, l2tp_net_id);
117 }
118
119 /* Session hash global list for L2TPv3.
120  * The session_id SHOULD be random according to RFC3931, but several
121  * L2TP implementations use incrementing session_ids.  So we do a real
122  * hash on the session_id, rather than a simple bitmask.
123  */
124 static inline struct hlist_head *
125 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
126 {
127         return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
128
129 }
130
131 /* Lookup a session by id in the global session list
132  */
133 static struct l2tp_session *l2tp_session_find_2(struct net *net, u32 session_id)
134 {
135         struct l2tp_net *pn = l2tp_pernet(net);
136         struct hlist_head *session_list =
137                 l2tp_session_id_hash_2(pn, session_id);
138         struct l2tp_session *session;
139         struct hlist_node *walk;
140
141         read_lock_bh(&pn->l2tp_session_hlist_lock);
142         hlist_for_each_entry(session, walk, session_list, global_hlist) {
143                 if (session->session_id == session_id) {
144                         read_unlock_bh(&pn->l2tp_session_hlist_lock);
145                         return session;
146                 }
147         }
148         read_unlock_bh(&pn->l2tp_session_hlist_lock);
149
150         return NULL;
151 }
152
153 /* Session hash list.
154  * The session_id SHOULD be random according to RFC2661, but several
155  * L2TP implementations (Cisco and Microsoft) use incrementing
156  * session_ids.  So we do a real hash on the session_id, rather than a
157  * simple bitmask.
158  */
159 static inline struct hlist_head *
160 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
161 {
162         return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
163 }
164
165 /* Lookup a session by id
166  */
167 struct l2tp_session *l2tp_session_find(struct net *net, struct l2tp_tunnel *tunnel, u32 session_id)
168 {
169         struct hlist_head *session_list;
170         struct l2tp_session *session;
171         struct hlist_node *walk;
172
173         /* In L2TPv3, session_ids are unique over all tunnels and we
174          * sometimes need to look them up before we know the
175          * tunnel.
176          */
177         if (tunnel == NULL)
178                 return l2tp_session_find_2(net, session_id);
179
180         session_list = l2tp_session_id_hash(tunnel, session_id);
181         read_lock_bh(&tunnel->hlist_lock);
182         hlist_for_each_entry(session, walk, session_list, hlist) {
183                 if (session->session_id == session_id) {
184                         read_unlock_bh(&tunnel->hlist_lock);
185                         return session;
186                 }
187         }
188         read_unlock_bh(&tunnel->hlist_lock);
189
190         return NULL;
191 }
192 EXPORT_SYMBOL_GPL(l2tp_session_find);
193
194 struct l2tp_session *l2tp_session_find_nth(struct l2tp_tunnel *tunnel, int nth)
195 {
196         int hash;
197         struct hlist_node *walk;
198         struct l2tp_session *session;
199         int count = 0;
200
201         read_lock_bh(&tunnel->hlist_lock);
202         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
203                 hlist_for_each_entry(session, walk, &tunnel->session_hlist[hash], hlist) {
204                         if (++count > nth) {
205                                 read_unlock_bh(&tunnel->hlist_lock);
206                                 return session;
207                         }
208                 }
209         }
210
211         read_unlock_bh(&tunnel->hlist_lock);
212
213         return NULL;
214 }
215 EXPORT_SYMBOL_GPL(l2tp_session_find_nth);
216
217 /* Lookup a tunnel by id
218  */
219 struct l2tp_tunnel *l2tp_tunnel_find(struct net *net, u32 tunnel_id)
220 {
221         struct l2tp_tunnel *tunnel;
222         struct l2tp_net *pn = l2tp_pernet(net);
223
224         read_lock_bh(&pn->l2tp_tunnel_list_lock);
225         list_for_each_entry(tunnel, &pn->l2tp_tunnel_list, list) {
226                 if (tunnel->tunnel_id == tunnel_id) {
227                         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
228                         return tunnel;
229                 }
230         }
231         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
232
233         return NULL;
234 }
235 EXPORT_SYMBOL_GPL(l2tp_tunnel_find);
236
237 struct l2tp_tunnel *l2tp_tunnel_find_nth(struct net *net, int nth)
238 {
239         struct l2tp_net *pn = l2tp_pernet(net);
240         struct l2tp_tunnel *tunnel;
241         int count = 0;
242
243         read_lock_bh(&pn->l2tp_tunnel_list_lock);
244         list_for_each_entry(tunnel, &pn->l2tp_tunnel_list, list) {
245                 if (++count > nth) {
246                         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
247                         return tunnel;
248                 }
249         }
250
251         read_unlock_bh(&pn->l2tp_tunnel_list_lock);
252
253         return NULL;
254 }
255 EXPORT_SYMBOL_GPL(l2tp_tunnel_find_nth);
256
257 /*****************************************************************************
258  * Receive data handling
259  *****************************************************************************/
260
261 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
262  * number.
263  */
264 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
265 {
266         struct sk_buff *skbp;
267         struct sk_buff *tmp;
268         u32 ns = L2TP_SKB_CB(skb)->ns;
269
270         spin_lock_bh(&session->reorder_q.lock);
271         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
272                 if (L2TP_SKB_CB(skbp)->ns > ns) {
273                         __skb_queue_before(&session->reorder_q, skbp, skb);
274                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
275                                "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
276                                session->name, ns, L2TP_SKB_CB(skbp)->ns,
277                                skb_queue_len(&session->reorder_q));
278                         session->stats.rx_oos_packets++;
279                         goto out;
280                 }
281         }
282
283         __skb_queue_tail(&session->reorder_q, skb);
284
285 out:
286         spin_unlock_bh(&session->reorder_q.lock);
287 }
288
289 /* Dequeue a single skb.
290  */
291 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
292 {
293         struct l2tp_tunnel *tunnel = session->tunnel;
294         int length = L2TP_SKB_CB(skb)->length;
295
296         /* We're about to requeue the skb, so return resources
297          * to its current owner (a socket receive buffer).
298          */
299         skb_orphan(skb);
300
301         tunnel->stats.rx_packets++;
302         tunnel->stats.rx_bytes += length;
303         session->stats.rx_packets++;
304         session->stats.rx_bytes += length;
305
306         if (L2TP_SKB_CB(skb)->has_seq) {
307                 /* Bump our Nr */
308                 session->nr++;
309                 if (tunnel->version == L2TP_HDR_VER_2)
310                         session->nr &= 0xffff;
311                 else
312                         session->nr &= 0xffffff;
313
314                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
315                        "%s: updated nr to %hu\n", session->name, session->nr);
316         }
317
318         /* call private receive handler */
319         if (session->recv_skb != NULL)
320                 (*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
321         else
322                 kfree_skb(skb);
323
324         if (session->deref)
325                 (*session->deref)(session);
326 }
327
328 /* Dequeue skbs from the session's reorder_q, subject to packet order.
329  * Skbs that have been in the queue for too long are simply discarded.
330  */
331 static void l2tp_recv_dequeue(struct l2tp_session *session)
332 {
333         struct sk_buff *skb;
334         struct sk_buff *tmp;
335
336         /* If the pkt at the head of the queue has the nr that we
337          * expect to send up next, dequeue it and any other
338          * in-sequence packets behind it.
339          */
340         spin_lock_bh(&session->reorder_q.lock);
341         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
342                 if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
343                         session->stats.rx_seq_discards++;
344                         session->stats.rx_errors++;
345                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
346                                "%s: oos pkt %u len %d discarded (too old), "
347                                "waiting for %u, reorder_q_len=%d\n",
348                                session->name, L2TP_SKB_CB(skb)->ns,
349                                L2TP_SKB_CB(skb)->length, session->nr,
350                                skb_queue_len(&session->reorder_q));
351                         __skb_unlink(skb, &session->reorder_q);
352                         kfree_skb(skb);
353                         if (session->deref)
354                                 (*session->deref)(session);
355                         continue;
356                 }
357
358                 if (L2TP_SKB_CB(skb)->has_seq) {
359                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
360                                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
361                                        "%s: holding oos pkt %u len %d, "
362                                        "waiting for %u, reorder_q_len=%d\n",
363                                        session->name, L2TP_SKB_CB(skb)->ns,
364                                        L2TP_SKB_CB(skb)->length, session->nr,
365                                        skb_queue_len(&session->reorder_q));
366                                 goto out;
367                         }
368                 }
369                 __skb_unlink(skb, &session->reorder_q);
370
371                 /* Process the skb. We release the queue lock while we
372                  * do so to let other contexts process the queue.
373                  */
374                 spin_unlock_bh(&session->reorder_q.lock);
375                 l2tp_recv_dequeue_skb(session, skb);
376                 spin_lock_bh(&session->reorder_q.lock);
377         }
378
379 out:
380         spin_unlock_bh(&session->reorder_q.lock);
381 }
382
383 static inline int l2tp_verify_udp_checksum(struct sock *sk,
384                                            struct sk_buff *skb)
385 {
386         struct udphdr *uh = udp_hdr(skb);
387         u16 ulen = ntohs(uh->len);
388         struct inet_sock *inet;
389         __wsum psum;
390
391         if (sk->sk_no_check || skb_csum_unnecessary(skb) || !uh->check)
392                 return 0;
393
394         inet = inet_sk(sk);
395         psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr, ulen,
396                                   IPPROTO_UDP, 0);
397
398         if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
399             !csum_fold(csum_add(psum, skb->csum)))
400                 return 0;
401
402         skb->csum = psum;
403
404         return __skb_checksum_complete(skb);
405 }
406
407 /* Do receive processing of L2TP data frames. We handle both L2TPv2
408  * and L2TPv3 data frames here.
409  *
410  * L2TPv2 Data Message Header
411  *
412  *  0                   1                   2                   3
413  *  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
414  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
415  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
416  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
417  * |           Tunnel ID           |           Session ID          |
418  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
419  * |             Ns (opt)          |             Nr (opt)          |
420  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
421  * |      Offset Size (opt)        |    Offset pad... (opt)
422  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
423  *
424  * Data frames are marked by T=0. All other fields are the same as
425  * those in L2TP control frames.
426  *
427  * L2TPv3 Data Message Header
428  *
429  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
430  * |                      L2TP Session Header                      |
431  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
432  * |                      L2-Specific Sublayer                     |
433  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
434  * |                        Tunnel Payload                      ...
435  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
436  *
437  * L2TPv3 Session Header Over IP
438  *
439  *  0                   1                   2                   3
440  *  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
441  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
442  * |                           Session ID                          |
443  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
444  * |               Cookie (optional, maximum 64 bits)...
445  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
446  *                                                                 |
447  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
448  *
449  * L2TPv3 L2-Specific Sublayer Format
450  *
451  *  0                   1                   2                   3
452  *  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
453  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
454  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
455  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
456  *
457  * Cookie value, sublayer format and offset (pad) are negotiated with
458  * the peer when the session is set up. Unlike L2TPv2, we do not need
459  * to parse the packet header to determine if optional fields are
460  * present.
461  *
462  * Caller must already have parsed the frame and determined that it is
463  * a data (not control) frame before coming here. Fields up to the
464  * session-id have already been parsed and ptr points to the data
465  * after the session-id.
466  */
467 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
468                       unsigned char *ptr, unsigned char *optr, u16 hdrflags,
469                       int length, int (*payload_hook)(struct sk_buff *skb))
470 {
471         struct l2tp_tunnel *tunnel = session->tunnel;
472         int offset;
473         u32 ns, nr;
474
475         /* The ref count is increased since we now hold a pointer to
476          * the session. Take care to decrement the refcnt when exiting
477          * this function from now on...
478          */
479         l2tp_session_inc_refcount(session);
480         if (session->ref)
481                 (*session->ref)(session);
482
483         /* Parse and check optional cookie */
484         if (session->peer_cookie_len > 0) {
485                 if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
486                         PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
487                                "%s: cookie mismatch (%u/%u). Discarding.\n",
488                                tunnel->name, tunnel->tunnel_id, session->session_id);
489                         session->stats.rx_cookie_discards++;
490                         goto discard;
491                 }
492                 ptr += session->peer_cookie_len;
493         }
494
495         /* Handle the optional sequence numbers. Sequence numbers are
496          * in different places for L2TPv2 and L2TPv3.
497          *
498          * If we are the LAC, enable/disable sequence numbers under
499          * the control of the LNS.  If no sequence numbers present but
500          * we were expecting them, discard frame.
501          */
502         ns = nr = 0;
503         L2TP_SKB_CB(skb)->has_seq = 0;
504         if (tunnel->version == L2TP_HDR_VER_2) {
505                 if (hdrflags & L2TP_HDRFLAG_S) {
506                         ns = ntohs(*(__be16 *) ptr);
507                         ptr += 2;
508                         nr = ntohs(*(__be16 *) ptr);
509                         ptr += 2;
510
511                         /* Store L2TP info in the skb */
512                         L2TP_SKB_CB(skb)->ns = ns;
513                         L2TP_SKB_CB(skb)->has_seq = 1;
514
515                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
516                                "%s: recv data ns=%u, nr=%u, session nr=%u\n",
517                                session->name, ns, nr, session->nr);
518                 }
519         } else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
520                 u32 l2h = ntohl(*(__be32 *) ptr);
521
522                 if (l2h & 0x40000000) {
523                         ns = l2h & 0x00ffffff;
524
525                         /* Store L2TP info in the skb */
526                         L2TP_SKB_CB(skb)->ns = ns;
527                         L2TP_SKB_CB(skb)->has_seq = 1;
528
529                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
530                                "%s: recv data ns=%u, session nr=%u\n",
531                                session->name, ns, session->nr);
532                 }
533         }
534
535         /* Advance past L2-specific header, if present */
536         ptr += session->l2specific_len;
537
538         if (L2TP_SKB_CB(skb)->has_seq) {
539                 /* Received a packet with sequence numbers. If we're the LNS,
540                  * check if we sre sending sequence numbers and if not,
541                  * configure it so.
542                  */
543                 if ((!session->lns_mode) && (!session->send_seq)) {
544                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_INFO,
545                                "%s: requested to enable seq numbers by LNS\n",
546                                session->name);
547                         session->send_seq = -1;
548                         l2tp_session_set_header_len(session, tunnel->version);
549                 }
550         } else {
551                 /* No sequence numbers.
552                  * If user has configured mandatory sequence numbers, discard.
553                  */
554                 if (session->recv_seq) {
555                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_WARNING,
556                                "%s: recv data has no seq numbers when required. "
557                                "Discarding\n", session->name);
558                         session->stats.rx_seq_discards++;
559                         goto discard;
560                 }
561
562                 /* If we're the LAC and we're sending sequence numbers, the
563                  * LNS has requested that we no longer send sequence numbers.
564                  * If we're the LNS and we're sending sequence numbers, the
565                  * LAC is broken. Discard the frame.
566                  */
567                 if ((!session->lns_mode) && (session->send_seq)) {
568                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_INFO,
569                                "%s: requested to disable seq numbers by LNS\n",
570                                session->name);
571                         session->send_seq = 0;
572                         l2tp_session_set_header_len(session, tunnel->version);
573                 } else if (session->send_seq) {
574                         PRINTK(session->debug, L2TP_MSG_SEQ, KERN_WARNING,
575                                "%s: recv data has no seq numbers when required. "
576                                "Discarding\n", session->name);
577                         session->stats.rx_seq_discards++;
578                         goto discard;
579                 }
580         }
581
582         /* Session data offset is handled differently for L2TPv2 and
583          * L2TPv3. For L2TPv2, there is an optional 16-bit value in
584          * the header. For L2TPv3, the offset is negotiated using AVPs
585          * in the session setup control protocol.
586          */
587         if (tunnel->version == L2TP_HDR_VER_2) {
588                 /* If offset bit set, skip it. */
589                 if (hdrflags & L2TP_HDRFLAG_O) {
590                         offset = ntohs(*(__be16 *)ptr);
591                         ptr += 2 + offset;
592                 }
593         } else
594                 ptr += session->offset;
595
596         offset = ptr - optr;
597         if (!pskb_may_pull(skb, offset))
598                 goto discard;
599
600         __skb_pull(skb, offset);
601
602         /* If caller wants to process the payload before we queue the
603          * packet, do so now.
604          */
605         if (payload_hook)
606                 if ((*payload_hook)(skb))
607                         goto discard;
608
609         /* Prepare skb for adding to the session's reorder_q.  Hold
610          * packets for max reorder_timeout or 1 second if not
611          * reordering.
612          */
613         L2TP_SKB_CB(skb)->length = length;
614         L2TP_SKB_CB(skb)->expires = jiffies +
615                 (session->reorder_timeout ? session->reorder_timeout : HZ);
616
617         /* Add packet to the session's receive queue. Reordering is done here, if
618          * enabled. Saved L2TP protocol info is stored in skb->sb[].
619          */
620         if (L2TP_SKB_CB(skb)->has_seq) {
621                 if (session->reorder_timeout != 0) {
622                         /* Packet reordering enabled. Add skb to session's
623                          * reorder queue, in order of ns.
624                          */
625                         l2tp_recv_queue_skb(session, skb);
626                 } else {
627                         /* Packet reordering disabled. Discard out-of-sequence
628                          * packets
629                          */
630                         if (L2TP_SKB_CB(skb)->ns != session->nr) {
631                                 session->stats.rx_seq_discards++;
632                                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
633                                        "%s: oos pkt %u len %d discarded, "
634                                        "waiting for %u, reorder_q_len=%d\n",
635                                        session->name, L2TP_SKB_CB(skb)->ns,
636                                        L2TP_SKB_CB(skb)->length, session->nr,
637                                        skb_queue_len(&session->reorder_q));
638                                 goto discard;
639                         }
640                         skb_queue_tail(&session->reorder_q, skb);
641                 }
642         } else {
643                 /* No sequence numbers. Add the skb to the tail of the
644                  * reorder queue. This ensures that it will be
645                  * delivered after all previous sequenced skbs.
646                  */
647                 skb_queue_tail(&session->reorder_q, skb);
648         }
649
650         /* Try to dequeue as many skbs from reorder_q as we can. */
651         l2tp_recv_dequeue(session);
652
653         l2tp_session_dec_refcount(session);
654
655         return;
656
657 discard:
658         session->stats.rx_errors++;
659         kfree_skb(skb);
660
661         if (session->deref)
662                 (*session->deref)(session);
663
664         l2tp_session_dec_refcount(session);
665 }
666 EXPORT_SYMBOL(l2tp_recv_common);
667
668 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
669  * here. The skb is not on a list when we get here.
670  * Returns 0 if the packet was a data packet and was successfully passed on.
671  * Returns 1 if the packet was not a good data packet and could not be
672  * forwarded.  All such packets are passed up to userspace to deal with.
673  */
674 int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
675                        int (*payload_hook)(struct sk_buff *skb))
676 {
677         struct l2tp_session *session = NULL;
678         unsigned char *ptr, *optr;
679         u16 hdrflags;
680         u32 tunnel_id, session_id;
681         int offset;
682         u16 version;
683         int length;
684
685         if (tunnel->sock && l2tp_verify_udp_checksum(tunnel->sock, skb))
686                 goto discard_bad_csum;
687
688         /* UDP always verifies the packet length. */
689         __skb_pull(skb, sizeof(struct udphdr));
690
691         /* Short packet? */
692         if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
693                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
694                        "%s: recv short packet (len=%d)\n", tunnel->name, skb->len);
695                 goto error;
696         }
697
698         /* Point to L2TP header */
699         optr = ptr = skb->data;
700
701         /* Trace packet contents, if enabled */
702         if (tunnel->debug & L2TP_MSG_DATA) {
703                 length = min(32u, skb->len);
704                 if (!pskb_may_pull(skb, length))
705                         goto error;
706
707                 printk(KERN_DEBUG "%s: recv: ", tunnel->name);
708
709                 offset = 0;
710                 do {
711                         printk(" %02X", ptr[offset]);
712                 } while (++offset < length);
713
714                 printk("\n");
715         }
716
717         /* Get L2TP header flags */
718         hdrflags = ntohs(*(__be16 *) ptr);
719
720         /* Check protocol version */
721         version = hdrflags & L2TP_HDR_VER_MASK;
722         if (version != tunnel->version) {
723                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
724                        "%s: recv protocol version mismatch: got %d expected %d\n",
725                        tunnel->name, version, tunnel->version);
726                 goto error;
727         }
728
729         /* Get length of L2TP packet */
730         length = skb->len;
731
732         /* If type is control packet, it is handled by userspace. */
733         if (hdrflags & L2TP_HDRFLAG_T) {
734                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_DEBUG,
735                        "%s: recv control packet, len=%d\n", tunnel->name, length);
736                 goto error;
737         }
738
739         /* Skip flags */
740         ptr += 2;
741
742         if (tunnel->version == L2TP_HDR_VER_2) {
743                 /* If length is present, skip it */
744                 if (hdrflags & L2TP_HDRFLAG_L)
745                         ptr += 2;
746
747                 /* Extract tunnel and session ID */
748                 tunnel_id = ntohs(*(__be16 *) ptr);
749                 ptr += 2;
750                 session_id = ntohs(*(__be16 *) ptr);
751                 ptr += 2;
752         } else {
753                 ptr += 2;       /* skip reserved bits */
754                 tunnel_id = tunnel->tunnel_id;
755                 session_id = ntohl(*(__be32 *) ptr);
756                 ptr += 4;
757         }
758
759         /* Find the session context */
760         session = l2tp_session_find(tunnel->l2tp_net, tunnel, session_id);
761         if (!session) {
762                 /* Not found? Pass to userspace to deal with */
763                 PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_INFO,
764                        "%s: no session found (%u/%u). Passing up.\n",
765                        tunnel->name, tunnel_id, session_id);
766                 goto error;
767         }
768
769         l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
770
771         return 0;
772
773 discard_bad_csum:
774         LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
775         UDP_INC_STATS_USER(tunnel->l2tp_net, UDP_MIB_INERRORS, 0);
776         tunnel->stats.rx_errors++;
777         kfree_skb(skb);
778
779         return 0;
780
781 error:
782         /* Put UDP header back */
783         __skb_push(skb, sizeof(struct udphdr));
784
785         return 1;
786 }
787 EXPORT_SYMBOL_GPL(l2tp_udp_recv_core);
788
789 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
790  * Return codes:
791  * 0 : success.
792  * <0: error
793  * >0: skb should be passed up to userspace as UDP.
794  */
795 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
796 {
797         struct l2tp_tunnel *tunnel;
798
799         tunnel = l2tp_sock_to_tunnel(sk);
800         if (tunnel == NULL)
801                 goto pass_up;
802
803         PRINTK(tunnel->debug, L2TP_MSG_DATA, KERN_DEBUG,
804                "%s: received %d bytes\n", tunnel->name, skb->len);
805
806         if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
807                 goto pass_up_put;
808
809         sock_put(sk);
810         return 0;
811
812 pass_up_put:
813         sock_put(sk);
814 pass_up:
815         return 1;
816 }
817 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
818
819 /************************************************************************
820  * Transmit handling
821  ***********************************************************************/
822
823 /* Build an L2TP header for the session into the buffer provided.
824  */
825 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
826 {
827         struct l2tp_tunnel *tunnel = session->tunnel;
828         __be16 *bufp = buf;
829         __be16 *optr = buf;
830         u16 flags = L2TP_HDR_VER_2;
831         u32 tunnel_id = tunnel->peer_tunnel_id;
832         u32 session_id = session->peer_session_id;
833
834         if (session->send_seq)
835                 flags |= L2TP_HDRFLAG_S;
836
837         /* Setup L2TP header. */
838         *bufp++ = htons(flags);
839         *bufp++ = htons(tunnel_id);
840         *bufp++ = htons(session_id);
841         if (session->send_seq) {
842                 *bufp++ = htons(session->ns);
843                 *bufp++ = 0;
844                 session->ns++;
845                 session->ns &= 0xffff;
846                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
847                        "%s: updated ns to %u\n", session->name, session->ns);
848         }
849
850         return bufp - optr;
851 }
852
853 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
854 {
855         struct l2tp_tunnel *tunnel = session->tunnel;
856         char *bufp = buf;
857         char *optr = bufp;
858
859         /* Setup L2TP header. The header differs slightly for UDP and
860          * IP encapsulations. For UDP, there is 4 bytes of flags.
861          */
862         if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
863                 u16 flags = L2TP_HDR_VER_3;
864                 *((__be16 *) bufp) = htons(flags);
865                 bufp += 2;
866                 *((__be16 *) bufp) = 0;
867                 bufp += 2;
868         }
869
870         *((__be32 *) bufp) = htonl(session->peer_session_id);
871         bufp += 4;
872         if (session->cookie_len) {
873                 memcpy(bufp, &session->cookie[0], session->cookie_len);
874                 bufp += session->cookie_len;
875         }
876         if (session->l2specific_len) {
877                 if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
878                         u32 l2h = 0;
879                         if (session->send_seq) {
880                                 l2h = 0x40000000 | session->ns;
881                                 session->ns++;
882                                 session->ns &= 0xffffff;
883                                 PRINTK(session->debug, L2TP_MSG_SEQ, KERN_DEBUG,
884                                        "%s: updated ns to %u\n", session->name, session->ns);
885                         }
886
887                         *((__be32 *) bufp) = htonl(l2h);
888                 }
889                 bufp += session->l2specific_len;
890         }
891         if (session->offset)
892                 bufp += session->offset;
893
894         return bufp - optr;
895 }
896
897 int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, size_t data_len)
898 {
899         struct l2tp_tunnel *tunnel = session->tunnel;
900         unsigned int len = skb->len;
901         int error;
902
903         /* Debug */
904         if (session->send_seq)
905                 PRINTK(session->debug, L2TP_MSG_DATA, KERN_DEBUG,
906                        "%s: send %Zd bytes, ns=%u\n", session->name,
907                        data_len, session->ns - 1);
908         else
909                 PRINTK(session->debug, L2TP_MSG_DATA, KERN_DEBUG,
910                        "%s: send %Zd bytes\n", session->name, data_len);
911
912         if (session->debug & L2TP_MSG_DATA) {
913                 int i;
914                 int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
915                 unsigned char *datap = skb->data + uhlen;
916
917                 printk(KERN_DEBUG "%s: xmit:", session->name);
918                 for (i = 0; i < (len - uhlen); i++) {
919                         printk(" %02X", *datap++);
920                         if (i == 31) {
921                                 printk(" ...");
922                                 break;
923                         }
924                 }
925                 printk("\n");
926         }
927
928         /* Queue the packet to IP for output */
929         error = ip_queue_xmit(skb, 1);
930
931         /* Update stats */
932         if (error >= 0) {
933                 tunnel->stats.tx_packets++;
934                 tunnel->stats.tx_bytes += len;
935                 session->stats.tx_packets++;
936                 session->stats.tx_bytes += len;
937         } else {
938                 tunnel->stats.tx_errors++;
939                 session->stats.tx_errors++;
940         }
941
942         return 0;
943 }
944 EXPORT_SYMBOL_GPL(l2tp_xmit_core);
945
946 /* Automatically called when the skb is freed.
947  */
948 static void l2tp_sock_wfree(struct sk_buff *skb)
949 {
950         sock_put(skb->sk);
951 }
952
953 /* For data skbs that we transmit, we associate with the tunnel socket
954  * but don't do accounting.
955  */
956 static inline void l2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
957 {
958         sock_hold(sk);
959         skb->sk = sk;
960         skb->destructor = l2tp_sock_wfree;
961 }
962
963 /* If caller requires the skb to have a ppp header, the header must be
964  * inserted in the skb data before calling this function.
965  */
966 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
967 {
968         int data_len = skb->len;
969         struct l2tp_tunnel *tunnel = session->tunnel;
970         struct sock *sk = tunnel->sock;
971         struct udphdr *uh;
972         struct inet_sock *inet;
973         __wsum csum;
974         int old_headroom;
975         int new_headroom;
976         int headroom;
977         int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
978         int udp_len;
979
980         /* Check that there's enough headroom in the skb to insert IP,
981          * UDP and L2TP headers. If not enough, expand it to
982          * make room. Adjust truesize.
983          */
984         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
985                 uhlen + hdr_len;
986         old_headroom = skb_headroom(skb);
987         if (skb_cow_head(skb, headroom))
988                 goto abort;
989
990         new_headroom = skb_headroom(skb);
991         skb_orphan(skb);
992         skb->truesize += new_headroom - old_headroom;
993
994         /* Setup L2TP header */
995         session->build_header(session, __skb_push(skb, hdr_len));
996
997         /* Reset skb netfilter state */
998         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
999         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1000                               IPSKB_REROUTED);
1001         nf_reset(skb);
1002
1003         /* Get routing info from the tunnel socket */
1004         skb_dst_drop(skb);
1005         skb_dst_set(skb, dst_clone(__sk_dst_get(sk)));
1006
1007         switch (tunnel->encap) {
1008         case L2TP_ENCAPTYPE_UDP:
1009                 /* Setup UDP header */
1010                 inet = inet_sk(sk);
1011                 __skb_push(skb, sizeof(*uh));
1012                 skb_reset_transport_header(skb);
1013                 uh = udp_hdr(skb);
1014                 uh->source = inet->inet_sport;
1015                 uh->dest = inet->inet_dport;
1016                 udp_len = uhlen + hdr_len + data_len;
1017                 uh->len = htons(udp_len);
1018                 uh->check = 0;
1019
1020                 /* Calculate UDP checksum if configured to do so */
1021                 if (sk->sk_no_check == UDP_CSUM_NOXMIT)
1022                         skb->ip_summed = CHECKSUM_NONE;
1023                 else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1024                          (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1025                         skb->ip_summed = CHECKSUM_COMPLETE;
1026                         csum = skb_checksum(skb, 0, udp_len, 0);
1027                         uh->check = csum_tcpudp_magic(inet->inet_saddr,
1028                                                       inet->inet_daddr,
1029                                                       udp_len, IPPROTO_UDP, csum);
1030                         if (uh->check == 0)
1031                                 uh->check = CSUM_MANGLED_0;
1032                 } else {
1033                         skb->ip_summed = CHECKSUM_PARTIAL;
1034                         skb->csum_start = skb_transport_header(skb) - skb->head;
1035                         skb->csum_offset = offsetof(struct udphdr, check);
1036                         uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1037                                                        inet->inet_daddr,
1038                                                        udp_len, IPPROTO_UDP, 0);
1039                 }
1040                 break;
1041
1042         case L2TP_ENCAPTYPE_IP:
1043                 break;
1044         }
1045
1046         l2tp_skb_set_owner_w(skb, sk);
1047
1048         l2tp_xmit_core(session, skb, data_len);
1049
1050 abort:
1051         return 0;
1052 }
1053 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1054
1055 /*****************************************************************************
1056  * Tinnel and session create/destroy.
1057  *****************************************************************************/
1058
1059 /* Tunnel socket destruct hook.
1060  * The tunnel context is deleted only when all session sockets have been
1061  * closed.
1062  */
1063 void l2tp_tunnel_destruct(struct sock *sk)
1064 {
1065         struct l2tp_tunnel *tunnel;
1066
1067         tunnel = sk->sk_user_data;
1068         if (tunnel == NULL)
1069                 goto end;
1070
1071         PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1072                "%s: closing...\n", tunnel->name);
1073
1074         /* Close all sessions */
1075         l2tp_tunnel_closeall(tunnel);
1076
1077         switch (tunnel->encap) {
1078         case L2TP_ENCAPTYPE_UDP:
1079                 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1080                 (udp_sk(sk))->encap_type = 0;
1081                 (udp_sk(sk))->encap_rcv = NULL;
1082                 break;
1083         case L2TP_ENCAPTYPE_IP:
1084                 break;
1085         }
1086
1087         /* Remove hooks into tunnel socket */
1088         tunnel->sock = NULL;
1089         sk->sk_destruct = tunnel->old_sk_destruct;
1090         sk->sk_user_data = NULL;
1091
1092         /* Call the original destructor */
1093         if (sk->sk_destruct)
1094                 (*sk->sk_destruct)(sk);
1095
1096         /* We're finished with the socket */
1097         l2tp_tunnel_dec_refcount(tunnel);
1098
1099 end:
1100         return;
1101 }
1102 EXPORT_SYMBOL(l2tp_tunnel_destruct);
1103
1104 /* When the tunnel is closed, all the attached sessions need to go too.
1105  */
1106 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1107 {
1108         int hash;
1109         struct hlist_node *walk;
1110         struct hlist_node *tmp;
1111         struct l2tp_session *session;
1112
1113         BUG_ON(tunnel == NULL);
1114
1115         PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1116                "%s: closing all sessions...\n", tunnel->name);
1117
1118         write_lock_bh(&tunnel->hlist_lock);
1119         for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1120 again:
1121                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1122                         session = hlist_entry(walk, struct l2tp_session, hlist);
1123
1124                         PRINTK(session->debug, L2TP_MSG_CONTROL, KERN_INFO,
1125                                "%s: closing session\n", session->name);
1126
1127                         hlist_del_init(&session->hlist);
1128
1129                         /* Since we should hold the sock lock while
1130                          * doing any unbinding, we need to release the
1131                          * lock we're holding before taking that lock.
1132                          * Hold a reference to the sock so it doesn't
1133                          * disappear as we're jumping between locks.
1134                          */
1135                         if (session->ref != NULL)
1136                                 (*session->ref)(session);
1137
1138                         write_unlock_bh(&tunnel->hlist_lock);
1139
1140                         if (tunnel->version != L2TP_HDR_VER_2) {
1141                                 struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1142
1143                                 write_lock_bh(&pn->l2tp_session_hlist_lock);
1144                                 hlist_del_init(&session->global_hlist);
1145                                 write_unlock_bh(&pn->l2tp_session_hlist_lock);
1146                         }
1147
1148                         if (session->session_close != NULL)
1149                                 (*session->session_close)(session);
1150
1151                         if (session->deref != NULL)
1152                                 (*session->deref)(session);
1153
1154                         write_lock_bh(&tunnel->hlist_lock);
1155
1156                         /* Now restart from the beginning of this hash
1157                          * chain.  We always remove a session from the
1158                          * list so we are guaranteed to make forward
1159                          * progress.
1160                          */
1161                         goto again;
1162                 }
1163         }
1164         write_unlock_bh(&tunnel->hlist_lock);
1165 }
1166 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1167
1168 /* Really kill the tunnel.
1169  * Come here only when all sessions have been cleared from the tunnel.
1170  */
1171 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
1172 {
1173         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1174
1175         BUG_ON(atomic_read(&tunnel->ref_count) != 0);
1176         BUG_ON(tunnel->sock != NULL);
1177
1178         PRINTK(tunnel->debug, L2TP_MSG_CONTROL, KERN_INFO,
1179                "%s: free...\n", tunnel->name);
1180
1181         /* Remove from tunnel list */
1182         write_lock_bh(&pn->l2tp_tunnel_list_lock);
1183         list_del_init(&tunnel->list);
1184         write_unlock_bh(&pn->l2tp_tunnel_list_lock);
1185
1186         atomic_dec(&l2tp_tunnel_count);
1187         kfree(tunnel);
1188 }
1189 EXPORT_SYMBOL_GPL(l2tp_tunnel_free);
1190
1191 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)
1192 {
1193         struct l2tp_tunnel *tunnel = NULL;
1194         int err;
1195         struct socket *sock = NULL;
1196         struct sock *sk = NULL;
1197         struct l2tp_net *pn;
1198         enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1199
1200         /* Get the tunnel socket from the fd, which was opened by
1201          * the userspace L2TP daemon.
1202          */
1203         err = -EBADF;
1204         sock = sockfd_lookup(fd, &err);
1205         if (!sock) {
1206                 printk(KERN_ERR "tunl %hu: sockfd_lookup(fd=%d) returned %d\n",
1207                        tunnel_id, fd, err);
1208                 goto err;
1209         }
1210
1211         sk = sock->sk;
1212
1213         if (cfg != NULL)
1214                 encap = cfg->encap;
1215
1216         /* Quick sanity checks */
1217         switch (encap) {
1218         case L2TP_ENCAPTYPE_UDP:
1219                 err = -EPROTONOSUPPORT;
1220                 if (sk->sk_protocol != IPPROTO_UDP) {
1221                         printk(KERN_ERR "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1222                                tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1223                         goto err;
1224                 }
1225                 break;
1226         case L2TP_ENCAPTYPE_IP:
1227                 err = -EPROTONOSUPPORT;
1228                 if (sk->sk_protocol != IPPROTO_L2TP) {
1229                         printk(KERN_ERR "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1230                                tunnel_id, fd, sk->sk_protocol, IPPROTO_L2TP);
1231                         goto err;
1232                 }
1233                 break;
1234         }
1235
1236         /* Check if this socket has already been prepped */
1237         tunnel = (struct l2tp_tunnel *)sk->sk_user_data;
1238         if (tunnel != NULL) {
1239                 /* This socket has already been prepped */
1240                 err = -EBUSY;
1241                 goto err;
1242         }
1243
1244         tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1245         if (tunnel == NULL) {
1246                 err = -ENOMEM;
1247                 goto err;
1248         }
1249
1250         tunnel->version = version;
1251         tunnel->tunnel_id = tunnel_id;
1252         tunnel->peer_tunnel_id = peer_tunnel_id;
1253         tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1254
1255         tunnel->magic = L2TP_TUNNEL_MAGIC;
1256         sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1257         rwlock_init(&tunnel->hlist_lock);
1258
1259         /* The net we belong to */
1260         tunnel->l2tp_net = net;
1261         pn = l2tp_pernet(net);
1262
1263         if (cfg != NULL)
1264                 tunnel->debug = cfg->debug;
1265
1266         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1267         tunnel->encap = encap;
1268         if (encap == L2TP_ENCAPTYPE_UDP) {
1269                 /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1270                 udp_sk(sk)->encap_type = UDP_ENCAP_L2TPINUDP;
1271                 udp_sk(sk)->encap_rcv = l2tp_udp_encap_recv;
1272         }
1273
1274         sk->sk_user_data = tunnel;
1275
1276         /* Hook on the tunnel socket destructor so that we can cleanup
1277          * if the tunnel socket goes away.
1278          */
1279         tunnel->old_sk_destruct = sk->sk_destruct;
1280         sk->sk_destruct = &l2tp_tunnel_destruct;
1281         tunnel->sock = sk;
1282         sk->sk_allocation = GFP_ATOMIC;
1283
1284         /* Add tunnel to our list */
1285         INIT_LIST_HEAD(&tunnel->list);
1286         write_lock_bh(&pn->l2tp_tunnel_list_lock);
1287         list_add(&tunnel->list, &pn->l2tp_tunnel_list);
1288         write_unlock_bh(&pn->l2tp_tunnel_list_lock);
1289         atomic_inc(&l2tp_tunnel_count);
1290
1291         /* Bump the reference count. The tunnel context is deleted
1292          * only when this drops to zero.
1293          */
1294         l2tp_tunnel_inc_refcount(tunnel);
1295
1296         err = 0;
1297 err:
1298         if (tunnelp)
1299                 *tunnelp = tunnel;
1300
1301         if (sock)
1302                 sockfd_put(sock);
1303
1304         return err;
1305 }
1306 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1307
1308 /* Really kill the session.
1309  */
1310 void l2tp_session_free(struct l2tp_session *session)
1311 {
1312         struct l2tp_tunnel *tunnel;
1313
1314         BUG_ON(atomic_read(&session->ref_count) != 0);
1315
1316         tunnel = session->tunnel;
1317         if (tunnel != NULL) {
1318                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1319
1320                 /* Delete the session from the hash */
1321                 write_lock_bh(&tunnel->hlist_lock);
1322                 hlist_del_init(&session->hlist);
1323                 write_unlock_bh(&tunnel->hlist_lock);
1324
1325                 /* Unlink from the global hash if not L2TPv2 */
1326                 if (tunnel->version != L2TP_HDR_VER_2) {
1327                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1328
1329                         write_lock_bh(&pn->l2tp_session_hlist_lock);
1330                         hlist_del_init(&session->global_hlist);
1331                         write_unlock_bh(&pn->l2tp_session_hlist_lock);
1332                 }
1333
1334                 if (session->session_id != 0)
1335                         atomic_dec(&l2tp_session_count);
1336
1337                 sock_put(tunnel->sock);
1338
1339                 /* This will delete the tunnel context if this
1340                  * is the last session on the tunnel.
1341                  */
1342                 session->tunnel = NULL;
1343                 l2tp_tunnel_dec_refcount(tunnel);
1344         }
1345
1346         kfree(session);
1347
1348         return;
1349 }
1350 EXPORT_SYMBOL_GPL(l2tp_session_free);
1351
1352 /* We come here whenever a session's send_seq, cookie_len or
1353  * l2specific_len parameters are set.
1354  */
1355 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1356 {
1357         if (version == L2TP_HDR_VER_2) {
1358                 session->hdr_len = 6;
1359                 if (session->send_seq)
1360                         session->hdr_len += 4;
1361         } else {
1362                 session->hdr_len = 4 + session->cookie_len + session->l2specific_len + session->offset;
1363                 if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1364                         session->hdr_len += 4;
1365         }
1366
1367 }
1368 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1369
1370 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)
1371 {
1372         struct l2tp_session *session;
1373
1374         session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1375         if (session != NULL) {
1376                 session->magic = L2TP_SESSION_MAGIC;
1377                 session->tunnel = tunnel;
1378
1379                 session->session_id = session_id;
1380                 session->peer_session_id = peer_session_id;
1381                 session->nr = 1;
1382
1383                 sprintf(&session->name[0], "sess %u/%u",
1384                         tunnel->tunnel_id, session->session_id);
1385
1386                 skb_queue_head_init(&session->reorder_q);
1387
1388                 INIT_HLIST_NODE(&session->hlist);
1389                 INIT_HLIST_NODE(&session->global_hlist);
1390
1391                 /* Inherit debug options from tunnel */
1392                 session->debug = tunnel->debug;
1393
1394                 if (cfg) {
1395                         session->pwtype = cfg->pw_type;
1396                         session->debug = cfg->debug;
1397                         session->mtu = cfg->mtu;
1398                         session->mru = cfg->mru;
1399                         session->send_seq = cfg->send_seq;
1400                         session->recv_seq = cfg->recv_seq;
1401                         session->lns_mode = cfg->lns_mode;
1402                         session->reorder_timeout = cfg->reorder_timeout;
1403                         session->offset = cfg->offset;
1404                         session->l2specific_type = cfg->l2specific_type;
1405                         session->l2specific_len = cfg->l2specific_len;
1406                         session->cookie_len = cfg->cookie_len;
1407                         memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1408                         session->peer_cookie_len = cfg->peer_cookie_len;
1409                         memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1410                 }
1411
1412                 if (tunnel->version == L2TP_HDR_VER_2)
1413                         session->build_header = l2tp_build_l2tpv2_header;
1414                 else
1415                         session->build_header = l2tp_build_l2tpv3_header;
1416
1417                 l2tp_session_set_header_len(session, tunnel->version);
1418
1419                 /* Bump the reference count. The session context is deleted
1420                  * only when this drops to zero.
1421                  */
1422                 l2tp_session_inc_refcount(session);
1423                 l2tp_tunnel_inc_refcount(tunnel);
1424
1425                 /* Ensure tunnel socket isn't deleted */
1426                 sock_hold(tunnel->sock);
1427
1428                 /* Add session to the tunnel's hash list */
1429                 write_lock_bh(&tunnel->hlist_lock);
1430                 hlist_add_head(&session->hlist,
1431                                l2tp_session_id_hash(tunnel, session_id));
1432                 write_unlock_bh(&tunnel->hlist_lock);
1433
1434                 /* And to the global session list if L2TPv3 */
1435                 if (tunnel->version != L2TP_HDR_VER_2) {
1436                         struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1437
1438                         write_lock_bh(&pn->l2tp_session_hlist_lock);
1439                         hlist_add_head(&session->global_hlist,
1440                                        l2tp_session_id_hash_2(pn, session_id));
1441                         write_unlock_bh(&pn->l2tp_session_hlist_lock);
1442                 }
1443
1444                 /* Ignore management session in session count value */
1445                 if (session->session_id != 0)
1446                         atomic_inc(&l2tp_session_count);
1447         }
1448
1449         return session;
1450 }
1451 EXPORT_SYMBOL_GPL(l2tp_session_create);
1452
1453 /*****************************************************************************
1454  * Init and cleanup
1455  *****************************************************************************/
1456
1457 static __net_init int l2tp_init_net(struct net *net)
1458 {
1459         struct l2tp_net *pn;
1460         int err;
1461         int hash;
1462
1463         pn = kzalloc(sizeof(*pn), GFP_KERNEL);
1464         if (!pn)
1465                 return -ENOMEM;
1466
1467         INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1468         rwlock_init(&pn->l2tp_tunnel_list_lock);
1469
1470         for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1471                 INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1472
1473         rwlock_init(&pn->l2tp_session_hlist_lock);
1474
1475         err = net_assign_generic(net, l2tp_net_id, pn);
1476         if (err)
1477                 goto out;
1478
1479         return 0;
1480
1481 out:
1482         kfree(pn);
1483         return err;
1484 }
1485
1486 static __net_exit void l2tp_exit_net(struct net *net)
1487 {
1488         struct l2tp_net *pn;
1489
1490         pn = net_generic(net, l2tp_net_id);
1491         /*
1492          * if someone has cached our net then
1493          * further net_generic call will return NULL
1494          */
1495         net_assign_generic(net, l2tp_net_id, NULL);
1496         kfree(pn);
1497 }
1498
1499 static struct pernet_operations l2tp_net_ops = {
1500         .init = l2tp_init_net,
1501         .exit = l2tp_exit_net,
1502         .id   = &l2tp_net_id,
1503         .size = sizeof(struct l2tp_net),
1504 };
1505
1506 static int __init l2tp_init(void)
1507 {
1508         int rc = 0;
1509
1510         rc = register_pernet_device(&l2tp_net_ops);
1511         if (rc)
1512                 goto out;
1513
1514         printk(KERN_INFO "L2TP core driver, %s\n", L2TP_DRV_VERSION);
1515
1516 out:
1517         return rc;
1518 }
1519
1520 static void __exit l2tp_exit(void)
1521 {
1522         unregister_pernet_device(&l2tp_net_ops);
1523 }
1524
1525 module_init(l2tp_init);
1526 module_exit(l2tp_exit);
1527
1528 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1529 MODULE_DESCRIPTION("L2TP core");
1530 MODULE_LICENSE("GPL");
1531 MODULE_VERSION(L2TP_DRV_VERSION);
1532