USB: don't use reset-resume if drivers don't support it
[pandora-kernel.git] / drivers / net / pppol2tp.c
1 /*****************************************************************************
2  * Linux PPP over L2TP (PPPoX/PPPoL2TP) Sockets
3  *
4  * PPPoX    --- Generic PPP encapsulation socket family
5  * PPPoL2TP --- PPP over L2TP (RFC 2661)
6  *
7  * Version:     1.0.0
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  * License:
17  *              This program is free software; you can redistribute it and/or
18  *              modify it under the terms of the GNU General Public License
19  *              as published by the Free Software Foundation; either version
20  *              2 of the License, or (at your option) any later version.
21  *
22  */
23
24 /* This driver handles only L2TP data frames; control frames are handled by a
25  * userspace application.
26  *
27  * To send data in an L2TP session, userspace opens a PPPoL2TP socket and
28  * attaches it to a bound UDP socket with local tunnel_id / session_id and
29  * peer tunnel_id / session_id set. Data can then be sent or received using
30  * regular socket sendmsg() / recvmsg() calls. Kernel parameters of the socket
31  * can be read or modified using ioctl() or [gs]etsockopt() calls.
32  *
33  * When a PPPoL2TP socket is connected with local and peer session_id values
34  * zero, the socket is treated as a special tunnel management socket.
35  *
36  * Here's example userspace code to create a socket for sending/receiving data
37  * over an L2TP session:-
38  *
39  *      struct sockaddr_pppol2tp sax;
40  *      int fd;
41  *      int session_fd;
42  *
43  *      fd = socket(AF_PPPOX, SOCK_DGRAM, PX_PROTO_OL2TP);
44  *
45  *      sax.sa_family = AF_PPPOX;
46  *      sax.sa_protocol = PX_PROTO_OL2TP;
47  *      sax.pppol2tp.fd = tunnel_fd;    // bound UDP socket
48  *      sax.pppol2tp.addr.sin_addr.s_addr = addr->sin_addr.s_addr;
49  *      sax.pppol2tp.addr.sin_port = addr->sin_port;
50  *      sax.pppol2tp.addr.sin_family = AF_INET;
51  *      sax.pppol2tp.s_tunnel  = tunnel_id;
52  *      sax.pppol2tp.s_session = session_id;
53  *      sax.pppol2tp.d_tunnel  = peer_tunnel_id;
54  *      sax.pppol2tp.d_session = peer_session_id;
55  *
56  *      session_fd = connect(fd, (struct sockaddr *)&sax, sizeof(sax));
57  *
58  * A pppd plugin that allows PPP traffic to be carried over L2TP using
59  * this driver is available from the OpenL2TP project at
60  * http://openl2tp.sourceforge.net.
61  */
62
63 #include <linux/module.h>
64 #include <linux/version.h>
65 #include <linux/string.h>
66 #include <linux/list.h>
67 #include <asm/uaccess.h>
68
69 #include <linux/kernel.h>
70 #include <linux/spinlock.h>
71 #include <linux/kthread.h>
72 #include <linux/sched.h>
73 #include <linux/slab.h>
74 #include <linux/errno.h>
75 #include <linux/jiffies.h>
76
77 #include <linux/netdevice.h>
78 #include <linux/net.h>
79 #include <linux/inetdevice.h>
80 #include <linux/skbuff.h>
81 #include <linux/init.h>
82 #include <linux/ip.h>
83 #include <linux/udp.h>
84 #include <linux/if_pppox.h>
85 #include <linux/if_pppol2tp.h>
86 #include <net/sock.h>
87 #include <linux/ppp_channel.h>
88 #include <linux/ppp_defs.h>
89 #include <linux/if_ppp.h>
90 #include <linux/file.h>
91 #include <linux/hash.h>
92 #include <linux/sort.h>
93 #include <linux/proc_fs.h>
94 #include <net/net_namespace.h>
95 #include <net/dst.h>
96 #include <net/ip.h>
97 #include <net/udp.h>
98 #include <net/xfrm.h>
99
100 #include <asm/byteorder.h>
101 #include <asm/atomic.h>
102
103
104 #define PPPOL2TP_DRV_VERSION    "V1.0"
105
106 /* L2TP header constants */
107 #define L2TP_HDRFLAG_T     0x8000
108 #define L2TP_HDRFLAG_L     0x4000
109 #define L2TP_HDRFLAG_S     0x0800
110 #define L2TP_HDRFLAG_O     0x0200
111 #define L2TP_HDRFLAG_P     0x0100
112
113 #define L2TP_HDR_VER_MASK  0x000F
114 #define L2TP_HDR_VER       0x0002
115
116 /* Space for UDP, L2TP and PPP headers */
117 #define PPPOL2TP_HEADER_OVERHEAD        40
118
119 /* Just some random numbers */
120 #define L2TP_TUNNEL_MAGIC       0x42114DDA
121 #define L2TP_SESSION_MAGIC      0x0C04EB7D
122
123 #define PPPOL2TP_HASH_BITS      4
124 #define PPPOL2TP_HASH_SIZE      (1 << PPPOL2TP_HASH_BITS)
125
126 /* Default trace flags */
127 #define PPPOL2TP_DEFAULT_DEBUG_FLAGS    0
128
129 #define PRINTK(_mask, _type, _lvl, _fmt, args...)                       \
130         do {                                                            \
131                 if ((_mask) & (_type))                                  \
132                         printk(_lvl "PPPOL2TP: " _fmt, ##args);         \
133         } while(0)
134
135 /* Number of bytes to build transmit L2TP headers.
136  * Unfortunately the size is different depending on whether sequence numbers
137  * are enabled.
138  */
139 #define PPPOL2TP_L2TP_HDR_SIZE_SEQ              10
140 #define PPPOL2TP_L2TP_HDR_SIZE_NOSEQ            6
141
142 struct pppol2tp_tunnel;
143
144 /* Describes a session. It is the sk_user_data field in the PPPoL2TP
145  * socket. Contains information to determine incoming packets and transmit
146  * outgoing ones.
147  */
148 struct pppol2tp_session
149 {
150         int                     magic;          /* should be
151                                                  * L2TP_SESSION_MAGIC */
152         int                     owner;          /* pid that opened the socket */
153
154         struct sock             *sock;          /* Pointer to the session
155                                                  * PPPoX socket */
156         struct sock             *tunnel_sock;   /* Pointer to the tunnel UDP
157                                                  * socket */
158
159         struct pppol2tp_addr    tunnel_addr;    /* Description of tunnel */
160
161         struct pppol2tp_tunnel  *tunnel;        /* back pointer to tunnel
162                                                  * context */
163
164         char                    name[20];       /* "sess xxxxx/yyyyy", where
165                                                  * x=tunnel_id, y=session_id */
166         int                     mtu;
167         int                     mru;
168         int                     flags;          /* accessed by PPPIOCGFLAGS.
169                                                  * Unused. */
170         unsigned                recv_seq:1;     /* expect receive packets with
171                                                  * sequence numbers? */
172         unsigned                send_seq:1;     /* send packets with sequence
173                                                  * numbers? */
174         unsigned                lns_mode:1;     /* behave as LNS? LAC enables
175                                                  * sequence numbers under
176                                                  * control of LNS. */
177         int                     debug;          /* bitmask of debug message
178                                                  * categories */
179         int                     reorder_timeout; /* configured reorder timeout
180                                                   * (in jiffies) */
181         u16                     nr;             /* session NR state (receive) */
182         u16                     ns;             /* session NR state (send) */
183         struct sk_buff_head     reorder_q;      /* receive reorder queue */
184         struct pppol2tp_ioc_stats stats;
185         struct hlist_node       hlist;          /* Hash list node */
186 };
187
188 /* The sk_user_data field of the tunnel's UDP socket. It contains info to track
189  * all the associated sessions so incoming packets can be sorted out
190  */
191 struct pppol2tp_tunnel
192 {
193         int                     magic;          /* Should be L2TP_TUNNEL_MAGIC */
194         rwlock_t                hlist_lock;     /* protect session_hlist */
195         struct hlist_head       session_hlist[PPPOL2TP_HASH_SIZE];
196                                                 /* hashed list of sessions,
197                                                  * hashed by id */
198         int                     debug;          /* bitmask of debug message
199                                                  * categories */
200         char                    name[12];       /* "tunl xxxxx" */
201         struct pppol2tp_ioc_stats stats;
202
203         void (*old_sk_destruct)(struct sock *);
204
205         struct sock             *sock;          /* Parent socket */
206         struct list_head        list;           /* Keep a list of all open
207                                                  * prepared sockets */
208
209         atomic_t                ref_count;
210 };
211
212 /* Private data stored for received packets in the skb.
213  */
214 struct pppol2tp_skb_cb {
215         u16                     ns;
216         u16                     nr;
217         u16                     has_seq;
218         u16                     length;
219         unsigned long           expires;
220 };
221
222 #define PPPOL2TP_SKB_CB(skb)    ((struct pppol2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
223
224 static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb);
225 static void pppol2tp_tunnel_free(struct pppol2tp_tunnel *tunnel);
226
227 static atomic_t pppol2tp_tunnel_count;
228 static atomic_t pppol2tp_session_count;
229 static struct ppp_channel_ops pppol2tp_chan_ops = { pppol2tp_xmit , NULL };
230 static struct proto_ops pppol2tp_ops;
231 static LIST_HEAD(pppol2tp_tunnel_list);
232 static DEFINE_RWLOCK(pppol2tp_tunnel_list_lock);
233
234 /* Helpers to obtain tunnel/session contexts from sockets.
235  */
236 static inline struct pppol2tp_session *pppol2tp_sock_to_session(struct sock *sk)
237 {
238         struct pppol2tp_session *session;
239
240         if (sk == NULL)
241                 return NULL;
242
243         sock_hold(sk);
244         session = (struct pppol2tp_session *)(sk->sk_user_data);
245         if (session == NULL) {
246                 sock_put(sk);
247                 goto out;
248         }
249
250         BUG_ON(session->magic != L2TP_SESSION_MAGIC);
251 out:
252         return session;
253 }
254
255 static inline struct pppol2tp_tunnel *pppol2tp_sock_to_tunnel(struct sock *sk)
256 {
257         struct pppol2tp_tunnel *tunnel;
258
259         if (sk == NULL)
260                 return NULL;
261
262         sock_hold(sk);
263         tunnel = (struct pppol2tp_tunnel *)(sk->sk_user_data);
264         if (tunnel == NULL) {
265                 sock_put(sk);
266                 goto out;
267         }
268
269         BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
270 out:
271         return tunnel;
272 }
273
274 /* Tunnel reference counts. Incremented per session that is added to
275  * the tunnel.
276  */
277 static inline void pppol2tp_tunnel_inc_refcount(struct pppol2tp_tunnel *tunnel)
278 {
279         atomic_inc(&tunnel->ref_count);
280 }
281
282 static inline void pppol2tp_tunnel_dec_refcount(struct pppol2tp_tunnel *tunnel)
283 {
284         if (atomic_dec_and_test(&tunnel->ref_count))
285                 pppol2tp_tunnel_free(tunnel);
286 }
287
288 /* Session hash list.
289  * The session_id SHOULD be random according to RFC2661, but several
290  * L2TP implementations (Cisco and Microsoft) use incrementing
291  * session_ids.  So we do a real hash on the session_id, rather than a
292  * simple bitmask.
293  */
294 static inline struct hlist_head *
295 pppol2tp_session_id_hash(struct pppol2tp_tunnel *tunnel, u16 session_id)
296 {
297         unsigned long hash_val = (unsigned long) session_id;
298         return &tunnel->session_hlist[hash_long(hash_val, PPPOL2TP_HASH_BITS)];
299 }
300
301 /* Lookup a session by id
302  */
303 static struct pppol2tp_session *
304 pppol2tp_session_find(struct pppol2tp_tunnel *tunnel, u16 session_id)
305 {
306         struct hlist_head *session_list =
307                 pppol2tp_session_id_hash(tunnel, session_id);
308         struct pppol2tp_session *session;
309         struct hlist_node *walk;
310
311         read_lock_bh(&tunnel->hlist_lock);
312         hlist_for_each_entry(session, walk, session_list, hlist) {
313                 if (session->tunnel_addr.s_session == session_id) {
314                         read_unlock_bh(&tunnel->hlist_lock);
315                         return session;
316                 }
317         }
318         read_unlock_bh(&tunnel->hlist_lock);
319
320         return NULL;
321 }
322
323 /* Lookup a tunnel by id
324  */
325 static struct pppol2tp_tunnel *pppol2tp_tunnel_find(u16 tunnel_id)
326 {
327         struct pppol2tp_tunnel *tunnel = NULL;
328
329         read_lock_bh(&pppol2tp_tunnel_list_lock);
330         list_for_each_entry(tunnel, &pppol2tp_tunnel_list, list) {
331                 if (tunnel->stats.tunnel_id == tunnel_id) {
332                         read_unlock_bh(&pppol2tp_tunnel_list_lock);
333                         return tunnel;
334                 }
335         }
336         read_unlock_bh(&pppol2tp_tunnel_list_lock);
337
338         return NULL;
339 }
340
341 /*****************************************************************************
342  * Receive data handling
343  *****************************************************************************/
344
345 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
346  * number.
347  */
348 static void pppol2tp_recv_queue_skb(struct pppol2tp_session *session, struct sk_buff *skb)
349 {
350         struct sk_buff *skbp;
351         struct sk_buff *tmp;
352         u16 ns = PPPOL2TP_SKB_CB(skb)->ns;
353
354         spin_lock_bh(&session->reorder_q.lock);
355         skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
356                 if (PPPOL2TP_SKB_CB(skbp)->ns > ns) {
357                         __skb_insert(skb, skbp->prev, skbp, &session->reorder_q);
358                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
359                                "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
360                                session->name, ns, PPPOL2TP_SKB_CB(skbp)->ns,
361                                skb_queue_len(&session->reorder_q));
362                         session->stats.rx_oos_packets++;
363                         goto out;
364                 }
365         }
366
367         __skb_queue_tail(&session->reorder_q, skb);
368
369 out:
370         spin_unlock_bh(&session->reorder_q.lock);
371 }
372
373 /* Dequeue a single skb.
374  */
375 static void pppol2tp_recv_dequeue_skb(struct pppol2tp_session *session, struct sk_buff *skb)
376 {
377         struct pppol2tp_tunnel *tunnel = session->tunnel;
378         int length = PPPOL2TP_SKB_CB(skb)->length;
379         struct sock *session_sock = NULL;
380
381         /* We're about to requeue the skb, so return resources
382          * to its current owner (a socket receive buffer).
383          */
384         skb_orphan(skb);
385
386         tunnel->stats.rx_packets++;
387         tunnel->stats.rx_bytes += length;
388         session->stats.rx_packets++;
389         session->stats.rx_bytes += length;
390
391         if (PPPOL2TP_SKB_CB(skb)->has_seq) {
392                 /* Bump our Nr */
393                 session->nr++;
394                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
395                        "%s: updated nr to %hu\n", session->name, session->nr);
396         }
397
398         /* If the socket is bound, send it in to PPP's input queue. Otherwise
399          * queue it on the session socket.
400          */
401         session_sock = session->sock;
402         if (session_sock->sk_state & PPPOX_BOUND) {
403                 struct pppox_sock *po;
404                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
405                        "%s: recv %d byte data frame, passing to ppp\n",
406                        session->name, length);
407
408                 /* We need to forget all info related to the L2TP packet
409                  * gathered in the skb as we are going to reuse the same
410                  * skb for the inner packet.
411                  * Namely we need to:
412                  * - reset xfrm (IPSec) information as it applies to
413                  *   the outer L2TP packet and not to the inner one
414                  * - release the dst to force a route lookup on the inner
415                  *   IP packet since skb->dst currently points to the dst
416                  *   of the UDP tunnel
417                  * - reset netfilter information as it doesn't apply
418                  *   to the inner packet either
419                  */
420                 secpath_reset(skb);
421                 dst_release(skb->dst);
422                 skb->dst = NULL;
423                 nf_reset(skb);
424
425                 po = pppox_sk(session_sock);
426                 ppp_input(&po->chan, skb);
427         } else {
428                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
429                        "%s: socket not bound\n", session->name);
430
431                 /* Not bound. Nothing we can do, so discard. */
432                 session->stats.rx_errors++;
433                 kfree_skb(skb);
434         }
435
436         sock_put(session->sock);
437 }
438
439 /* Dequeue skbs from the session's reorder_q, subject to packet order.
440  * Skbs that have been in the queue for too long are simply discarded.
441  */
442 static void pppol2tp_recv_dequeue(struct pppol2tp_session *session)
443 {
444         struct sk_buff *skb;
445         struct sk_buff *tmp;
446
447         /* If the pkt at the head of the queue has the nr that we
448          * expect to send up next, dequeue it and any other
449          * in-sequence packets behind it.
450          */
451         spin_lock_bh(&session->reorder_q.lock);
452         skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
453                 if (time_after(jiffies, PPPOL2TP_SKB_CB(skb)->expires)) {
454                         session->stats.rx_seq_discards++;
455                         session->stats.rx_errors++;
456                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
457                                "%s: oos pkt %hu len %d discarded (too old), "
458                                "waiting for %hu, reorder_q_len=%d\n",
459                                session->name, PPPOL2TP_SKB_CB(skb)->ns,
460                                PPPOL2TP_SKB_CB(skb)->length, session->nr,
461                                skb_queue_len(&session->reorder_q));
462                         __skb_unlink(skb, &session->reorder_q);
463                         kfree_skb(skb);
464                         sock_put(session->sock);
465                         continue;
466                 }
467
468                 if (PPPOL2TP_SKB_CB(skb)->has_seq) {
469                         if (PPPOL2TP_SKB_CB(skb)->ns != session->nr) {
470                                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
471                                        "%s: holding oos pkt %hu len %d, "
472                                        "waiting for %hu, reorder_q_len=%d\n",
473                                        session->name, PPPOL2TP_SKB_CB(skb)->ns,
474                                        PPPOL2TP_SKB_CB(skb)->length, session->nr,
475                                        skb_queue_len(&session->reorder_q));
476                                 goto out;
477                         }
478                 }
479                 __skb_unlink(skb, &session->reorder_q);
480
481                 /* Process the skb. We release the queue lock while we
482                  * do so to let other contexts process the queue.
483                  */
484                 spin_unlock_bh(&session->reorder_q.lock);
485                 pppol2tp_recv_dequeue_skb(session, skb);
486                 spin_lock_bh(&session->reorder_q.lock);
487         }
488
489 out:
490         spin_unlock_bh(&session->reorder_q.lock);
491 }
492
493 /* Internal receive frame. Do the real work of receiving an L2TP data frame
494  * here. The skb is not on a list when we get here.
495  * Returns 0 if the packet was a data packet and was successfully passed on.
496  * Returns 1 if the packet was not a good data packet and could not be
497  * forwarded.  All such packets are passed up to userspace to deal with.
498  */
499 static int pppol2tp_recv_core(struct sock *sock, struct sk_buff *skb)
500 {
501         struct pppol2tp_session *session = NULL;
502         struct pppol2tp_tunnel *tunnel;
503         unsigned char *ptr, *optr;
504         u16 hdrflags;
505         u16 tunnel_id, session_id;
506         int length;
507         int offset;
508
509         tunnel = pppol2tp_sock_to_tunnel(sock);
510         if (tunnel == NULL)
511                 goto no_tunnel;
512
513         /* UDP always verifies the packet length. */
514         __skb_pull(skb, sizeof(struct udphdr));
515
516         /* Short packet? */
517         if (!pskb_may_pull(skb, 12)) {
518                 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
519                        "%s: recv short packet (len=%d)\n", tunnel->name, skb->len);
520                 goto error;
521         }
522
523         /* Point to L2TP header */
524         optr = ptr = skb->data;
525
526         /* Get L2TP header flags */
527         hdrflags = ntohs(*(__be16*)ptr);
528
529         /* Trace packet contents, if enabled */
530         if (tunnel->debug & PPPOL2TP_MSG_DATA) {
531                 length = min(16u, skb->len);
532                 if (!pskb_may_pull(skb, length))
533                         goto error;
534
535                 printk(KERN_DEBUG "%s: recv: ", tunnel->name);
536
537                 offset = 0;
538                 do {
539                         printk(" %02X", ptr[offset]);
540                 } while (++offset < length);
541
542                 printk("\n");
543         }
544
545         /* Get length of L2TP packet */
546         length = skb->len;
547
548         /* If type is control packet, it is handled by userspace. */
549         if (hdrflags & L2TP_HDRFLAG_T) {
550                 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
551                        "%s: recv control packet, len=%d\n", tunnel->name, length);
552                 goto error;
553         }
554
555         /* Skip flags */
556         ptr += 2;
557
558         /* If length is present, skip it */
559         if (hdrflags & L2TP_HDRFLAG_L)
560                 ptr += 2;
561
562         /* Extract tunnel and session ID */
563         tunnel_id = ntohs(*(__be16 *) ptr);
564         ptr += 2;
565         session_id = ntohs(*(__be16 *) ptr);
566         ptr += 2;
567
568         /* Find the session context */
569         session = pppol2tp_session_find(tunnel, session_id);
570         if (!session) {
571                 /* Not found? Pass to userspace to deal with */
572                 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
573                        "%s: no socket found (%hu/%hu). Passing up.\n",
574                        tunnel->name, tunnel_id, session_id);
575                 goto error;
576         }
577         sock_hold(session->sock);
578
579         /* The ref count on the socket was increased by the above call since
580          * we now hold a pointer to the session. Take care to do sock_put()
581          * when exiting this function from now on...
582          */
583
584         /* Handle the optional sequence numbers.  If we are the LAC,
585          * enable/disable sequence numbers under the control of the LNS.  If
586          * no sequence numbers present but we were expecting them, discard
587          * frame.
588          */
589         if (hdrflags & L2TP_HDRFLAG_S) {
590                 u16 ns, nr;
591                 ns = ntohs(*(__be16 *) ptr);
592                 ptr += 2;
593                 nr = ntohs(*(__be16 *) ptr);
594                 ptr += 2;
595
596                 /* Received a packet with sequence numbers. If we're the LNS,
597                  * check if we sre sending sequence numbers and if not,
598                  * configure it so.
599                  */
600                 if ((!session->lns_mode) && (!session->send_seq)) {
601                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_INFO,
602                                "%s: requested to enable seq numbers by LNS\n",
603                                session->name);
604                         session->send_seq = -1;
605                 }
606
607                 /* Store L2TP info in the skb */
608                 PPPOL2TP_SKB_CB(skb)->ns = ns;
609                 PPPOL2TP_SKB_CB(skb)->nr = nr;
610                 PPPOL2TP_SKB_CB(skb)->has_seq = 1;
611
612                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
613                        "%s: recv data ns=%hu, nr=%hu, session nr=%hu\n",
614                        session->name, ns, nr, session->nr);
615         } else {
616                 /* No sequence numbers.
617                  * If user has configured mandatory sequence numbers, discard.
618                  */
619                 if (session->recv_seq) {
620                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_WARNING,
621                                "%s: recv data has no seq numbers when required. "
622                                "Discarding\n", session->name);
623                         session->stats.rx_seq_discards++;
624                         goto discard;
625                 }
626
627                 /* If we're the LAC and we're sending sequence numbers, the
628                  * LNS has requested that we no longer send sequence numbers.
629                  * If we're the LNS and we're sending sequence numbers, the
630                  * LAC is broken. Discard the frame.
631                  */
632                 if ((!session->lns_mode) && (session->send_seq)) {
633                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_INFO,
634                                "%s: requested to disable seq numbers by LNS\n",
635                                session->name);
636                         session->send_seq = 0;
637                 } else if (session->send_seq) {
638                         PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_WARNING,
639                                "%s: recv data has no seq numbers when required. "
640                                "Discarding\n", session->name);
641                         session->stats.rx_seq_discards++;
642                         goto discard;
643                 }
644
645                 /* Store L2TP info in the skb */
646                 PPPOL2TP_SKB_CB(skb)->has_seq = 0;
647         }
648
649         /* If offset bit set, skip it. */
650         if (hdrflags & L2TP_HDRFLAG_O) {
651                 offset = ntohs(*(__be16 *)ptr);
652                 ptr += 2 + offset;
653         }
654
655         offset = ptr - optr;
656         if (!pskb_may_pull(skb, offset))
657                 goto discard;
658
659         __skb_pull(skb, offset);
660
661         /* Skip PPP header, if present.  In testing, Microsoft L2TP clients
662          * don't send the PPP header (PPP header compression enabled), but
663          * other clients can include the header. So we cope with both cases
664          * here. The PPP header is always FF03 when using L2TP.
665          *
666          * Note that skb->data[] isn't dereferenced from a u16 ptr here since
667          * the field may be unaligned.
668          */
669         if (!pskb_may_pull(skb, 2))
670                 goto discard;
671
672         if ((skb->data[0] == 0xff) && (skb->data[1] == 0x03))
673                 skb_pull(skb, 2);
674
675         /* Prepare skb for adding to the session's reorder_q.  Hold
676          * packets for max reorder_timeout or 1 second if not
677          * reordering.
678          */
679         PPPOL2TP_SKB_CB(skb)->length = length;
680         PPPOL2TP_SKB_CB(skb)->expires = jiffies +
681                 (session->reorder_timeout ? session->reorder_timeout : HZ);
682
683         /* Add packet to the session's receive queue. Reordering is done here, if
684          * enabled. Saved L2TP protocol info is stored in skb->sb[].
685          */
686         if (PPPOL2TP_SKB_CB(skb)->has_seq) {
687                 if (session->reorder_timeout != 0) {
688                         /* Packet reordering enabled. Add skb to session's
689                          * reorder queue, in order of ns.
690                          */
691                         pppol2tp_recv_queue_skb(session, skb);
692                 } else {
693                         /* Packet reordering disabled. Discard out-of-sequence
694                          * packets
695                          */
696                         if (PPPOL2TP_SKB_CB(skb)->ns != session->nr) {
697                                 session->stats.rx_seq_discards++;
698                                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
699                                        "%s: oos pkt %hu len %d discarded, "
700                                        "waiting for %hu, reorder_q_len=%d\n",
701                                        session->name, PPPOL2TP_SKB_CB(skb)->ns,
702                                        PPPOL2TP_SKB_CB(skb)->length, session->nr,
703                                        skb_queue_len(&session->reorder_q));
704                                 goto discard;
705                         }
706                         skb_queue_tail(&session->reorder_q, skb);
707                 }
708         } else {
709                 /* No sequence numbers. Add the skb to the tail of the
710                  * reorder queue. This ensures that it will be
711                  * delivered after all previous sequenced skbs.
712                  */
713                 skb_queue_tail(&session->reorder_q, skb);
714         }
715
716         /* Try to dequeue as many skbs from reorder_q as we can. */
717         pppol2tp_recv_dequeue(session);
718
719         return 0;
720
721 discard:
722         session->stats.rx_errors++;
723         kfree_skb(skb);
724         sock_put(session->sock);
725         sock_put(sock);
726
727         return 0;
728
729 error:
730         /* Put UDP header back */
731         __skb_push(skb, sizeof(struct udphdr));
732         sock_put(sock);
733
734 no_tunnel:
735         return 1;
736 }
737
738 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
739  * Return codes:
740  * 0 : success.
741  * <0: error
742  * >0: skb should be passed up to userspace as UDP.
743  */
744 static int pppol2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
745 {
746         struct pppol2tp_tunnel *tunnel;
747
748         tunnel = pppol2tp_sock_to_tunnel(sk);
749         if (tunnel == NULL)
750                 goto pass_up;
751
752         PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
753                "%s: received %d bytes\n", tunnel->name, skb->len);
754
755         if (pppol2tp_recv_core(sk, skb))
756                 goto pass_up_put;
757
758         sock_put(sk);
759         return 0;
760
761 pass_up_put:
762         sock_put(sk);
763 pass_up:
764         return 1;
765 }
766
767 /* Receive message. This is the recvmsg for the PPPoL2TP socket.
768  */
769 static int pppol2tp_recvmsg(struct kiocb *iocb, struct socket *sock,
770                             struct msghdr *msg, size_t len,
771                             int flags)
772 {
773         int err;
774         struct sk_buff *skb;
775         struct sock *sk = sock->sk;
776
777         err = -EIO;
778         if (sk->sk_state & PPPOX_BOUND)
779                 goto end;
780
781         msg->msg_namelen = 0;
782
783         err = 0;
784         skb = skb_recv_datagram(sk, flags & ~MSG_DONTWAIT,
785                                 flags & MSG_DONTWAIT, &err);
786         if (skb) {
787                 err = memcpy_toiovec(msg->msg_iov, (unsigned char *) skb->data,
788                                      skb->len);
789                 if (err < 0)
790                         goto do_skb_free;
791                 err = skb->len;
792         }
793 do_skb_free:
794         kfree_skb(skb);
795 end:
796         return err;
797 }
798
799 /************************************************************************
800  * Transmit handling
801  ***********************************************************************/
802
803 /* Tell how big L2TP headers are for a particular session. This
804  * depends on whether sequence numbers are being used.
805  */
806 static inline int pppol2tp_l2tp_header_len(struct pppol2tp_session *session)
807 {
808         if (session->send_seq)
809                 return PPPOL2TP_L2TP_HDR_SIZE_SEQ;
810
811         return PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
812 }
813
814 /* Build an L2TP header for the session into the buffer provided.
815  */
816 static void pppol2tp_build_l2tp_header(struct pppol2tp_session *session,
817                                        void *buf)
818 {
819         __be16 *bufp = buf;
820         u16 flags = L2TP_HDR_VER;
821
822         if (session->send_seq)
823                 flags |= L2TP_HDRFLAG_S;
824
825         /* Setup L2TP header.
826          * FIXME: Can this ever be unaligned? Is direct dereferencing of
827          * 16-bit header fields safe here for all architectures?
828          */
829         *bufp++ = htons(flags);
830         *bufp++ = htons(session->tunnel_addr.d_tunnel);
831         *bufp++ = htons(session->tunnel_addr.d_session);
832         if (session->send_seq) {
833                 *bufp++ = htons(session->ns);
834                 *bufp++ = 0;
835                 session->ns++;
836                 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
837                        "%s: updated ns to %hu\n", session->name, session->ns);
838         }
839 }
840
841 /* This is the sendmsg for the PPPoL2TP pppol2tp_session socket.  We come here
842  * when a user application does a sendmsg() on the session socket. L2TP and
843  * PPP headers must be inserted into the user's data.
844  */
845 static int pppol2tp_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
846                             size_t total_len)
847 {
848         static const unsigned char ppph[2] = { 0xff, 0x03 };
849         struct sock *sk = sock->sk;
850         struct inet_sock *inet;
851         __wsum csum = 0;
852         struct sk_buff *skb;
853         int error;
854         int hdr_len;
855         struct pppol2tp_session *session;
856         struct pppol2tp_tunnel *tunnel;
857         struct udphdr *uh;
858         unsigned int len;
859
860         error = -ENOTCONN;
861         if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
862                 goto error;
863
864         /* Get session and tunnel contexts */
865         error = -EBADF;
866         session = pppol2tp_sock_to_session(sk);
867         if (session == NULL)
868                 goto error;
869
870         tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
871         if (tunnel == NULL)
872                 goto error_put_sess;
873
874         /* What header length is configured for this session? */
875         hdr_len = pppol2tp_l2tp_header_len(session);
876
877         /* Allocate a socket buffer */
878         error = -ENOMEM;
879         skb = sock_wmalloc(sk, NET_SKB_PAD + sizeof(struct iphdr) +
880                            sizeof(struct udphdr) + hdr_len +
881                            sizeof(ppph) + total_len,
882                            0, GFP_KERNEL);
883         if (!skb)
884                 goto error_put_sess_tun;
885
886         /* Reserve space for headers. */
887         skb_reserve(skb, NET_SKB_PAD);
888         skb_reset_network_header(skb);
889         skb_reserve(skb, sizeof(struct iphdr));
890         skb_reset_transport_header(skb);
891
892         /* Build UDP header */
893         inet = inet_sk(session->tunnel_sock);
894         uh = (struct udphdr *) skb->data;
895         uh->source = inet->sport;
896         uh->dest = inet->dport;
897         uh->len = htons(hdr_len + sizeof(ppph) + total_len);
898         uh->check = 0;
899         skb_put(skb, sizeof(struct udphdr));
900
901         /* Build L2TP header */
902         pppol2tp_build_l2tp_header(session, skb->data);
903         skb_put(skb, hdr_len);
904
905         /* Add PPP header */
906         skb->data[0] = ppph[0];
907         skb->data[1] = ppph[1];
908         skb_put(skb, 2);
909
910         /* Copy user data into skb */
911         error = memcpy_fromiovec(skb->data, m->msg_iov, total_len);
912         if (error < 0) {
913                 kfree_skb(skb);
914                 goto error_put_sess_tun;
915         }
916         skb_put(skb, total_len);
917
918         /* Calculate UDP checksum if configured to do so */
919         if (session->tunnel_sock->sk_no_check != UDP_CSUM_NOXMIT)
920                 csum = udp_csum_outgoing(sk, skb);
921
922         /* Debug */
923         if (session->send_seq)
924                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
925                        "%s: send %Zd bytes, ns=%hu\n", session->name,
926                        total_len, session->ns - 1);
927         else
928                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
929                        "%s: send %Zd bytes\n", session->name, total_len);
930
931         if (session->debug & PPPOL2TP_MSG_DATA) {
932                 int i;
933                 unsigned char *datap = skb->data;
934
935                 printk(KERN_DEBUG "%s: xmit:", session->name);
936                 for (i = 0; i < total_len; i++) {
937                         printk(" %02X", *datap++);
938                         if (i == 15) {
939                                 printk(" ...");
940                                 break;
941                         }
942                 }
943                 printk("\n");
944         }
945
946         /* Queue the packet to IP for output */
947         len = skb->len;
948         error = ip_queue_xmit(skb, 1);
949
950         /* Update stats */
951         if (error >= 0) {
952                 tunnel->stats.tx_packets++;
953                 tunnel->stats.tx_bytes += len;
954                 session->stats.tx_packets++;
955                 session->stats.tx_bytes += len;
956         } else {
957                 tunnel->stats.tx_errors++;
958                 session->stats.tx_errors++;
959         }
960
961         return error;
962
963 error_put_sess_tun:
964         sock_put(session->tunnel_sock);
965 error_put_sess:
966         sock_put(sk);
967 error:
968         return error;
969 }
970
971 /* Automatically called when the skb is freed.
972  */
973 static void pppol2tp_sock_wfree(struct sk_buff *skb)
974 {
975         sock_put(skb->sk);
976 }
977
978 /* For data skbs that we transmit, we associate with the tunnel socket
979  * but don't do accounting.
980  */
981 static inline void pppol2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
982 {
983         sock_hold(sk);
984         skb->sk = sk;
985         skb->destructor = pppol2tp_sock_wfree;
986 }
987
988 /* Transmit function called by generic PPP driver.  Sends PPP frame
989  * over PPPoL2TP socket.
990  *
991  * This is almost the same as pppol2tp_sendmsg(), but rather than
992  * being called with a msghdr from userspace, it is called with a skb
993  * from the kernel.
994  *
995  * The supplied skb from ppp doesn't have enough headroom for the
996  * insertion of L2TP, UDP and IP headers so we need to allocate more
997  * headroom in the skb. This will create a cloned skb. But we must be
998  * careful in the error case because the caller will expect to free
999  * the skb it supplied, not our cloned skb. So we take care to always
1000  * leave the original skb unfreed if we return an error.
1001  */
1002 static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb)
1003 {
1004         static const u8 ppph[2] = { 0xff, 0x03 };
1005         struct sock *sk = (struct sock *) chan->private;
1006         struct sock *sk_tun;
1007         int hdr_len;
1008         struct pppol2tp_session *session;
1009         struct pppol2tp_tunnel *tunnel;
1010         int rc;
1011         int headroom;
1012         int data_len = skb->len;
1013         struct inet_sock *inet;
1014         __wsum csum = 0;
1015         struct udphdr *uh;
1016         unsigned int len;
1017         int old_headroom;
1018         int new_headroom;
1019
1020         if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
1021                 goto abort;
1022
1023         /* Get session and tunnel contexts from the socket */
1024         session = pppol2tp_sock_to_session(sk);
1025         if (session == NULL)
1026                 goto abort;
1027
1028         sk_tun = session->tunnel_sock;
1029         if (sk_tun == NULL)
1030                 goto abort_put_sess;
1031         tunnel = pppol2tp_sock_to_tunnel(sk_tun);
1032         if (tunnel == NULL)
1033                 goto abort_put_sess;
1034
1035         /* What header length is configured for this session? */
1036         hdr_len = pppol2tp_l2tp_header_len(session);
1037
1038         /* Check that there's enough headroom in the skb to insert IP,
1039          * UDP and L2TP and PPP headers. If not enough, expand it to
1040          * make room. Adjust truesize.
1041          */
1042         headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1043                 sizeof(struct udphdr) + hdr_len + sizeof(ppph);
1044         old_headroom = skb_headroom(skb);
1045         if (skb_cow_head(skb, headroom))
1046                 goto abort_put_sess_tun;
1047
1048         new_headroom = skb_headroom(skb);
1049         skb_orphan(skb);
1050         skb->truesize += new_headroom - old_headroom;
1051
1052         /* Setup PPP header */
1053         __skb_push(skb, sizeof(ppph));
1054         skb->data[0] = ppph[0];
1055         skb->data[1] = ppph[1];
1056
1057         /* Setup L2TP header */
1058         pppol2tp_build_l2tp_header(session, __skb_push(skb, hdr_len));
1059
1060         /* Setup UDP header */
1061         inet = inet_sk(sk_tun);
1062         __skb_push(skb, sizeof(*uh));
1063         skb_reset_transport_header(skb);
1064         uh = udp_hdr(skb);
1065         uh->source = inet->sport;
1066         uh->dest = inet->dport;
1067         uh->len = htons(sizeof(struct udphdr) + hdr_len + sizeof(ppph) + data_len);
1068         uh->check = 0;
1069
1070         /* *BROKEN* Calculate UDP checksum if configured to do so */
1071         if (sk_tun->sk_no_check != UDP_CSUM_NOXMIT)
1072                 csum = udp_csum_outgoing(sk_tun, skb);
1073
1074         /* Debug */
1075         if (session->send_seq)
1076                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1077                        "%s: send %d bytes, ns=%hu\n", session->name,
1078                        data_len, session->ns - 1);
1079         else
1080                 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1081                        "%s: send %d bytes\n", session->name, data_len);
1082
1083         if (session->debug & PPPOL2TP_MSG_DATA) {
1084                 int i;
1085                 unsigned char *datap = skb->data;
1086
1087                 printk(KERN_DEBUG "%s: xmit:", session->name);
1088                 for (i = 0; i < data_len; i++) {
1089                         printk(" %02X", *datap++);
1090                         if (i == 31) {
1091                                 printk(" ...");
1092                                 break;
1093                         }
1094                 }
1095                 printk("\n");
1096         }
1097
1098         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1099         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1100                               IPSKB_REROUTED);
1101         nf_reset(skb);
1102
1103         /* Get routing info from the tunnel socket */
1104         dst_release(skb->dst);
1105         skb->dst = dst_clone(__sk_dst_get(sk_tun));
1106         pppol2tp_skb_set_owner_w(skb, sk_tun);
1107
1108         /* Queue the packet to IP for output */
1109         len = skb->len;
1110         rc = ip_queue_xmit(skb, 1);
1111
1112         /* Update stats */
1113         if (rc >= 0) {
1114                 tunnel->stats.tx_packets++;
1115                 tunnel->stats.tx_bytes += len;
1116                 session->stats.tx_packets++;
1117                 session->stats.tx_bytes += len;
1118         } else {
1119                 tunnel->stats.tx_errors++;
1120                 session->stats.tx_errors++;
1121         }
1122
1123         sock_put(sk_tun);
1124         sock_put(sk);
1125         return 1;
1126
1127 abort_put_sess_tun:
1128         sock_put(sk_tun);
1129 abort_put_sess:
1130         sock_put(sk);
1131 abort:
1132         /* Free the original skb */
1133         kfree_skb(skb);
1134         return 1;
1135 }
1136
1137 /*****************************************************************************
1138  * Session (and tunnel control) socket create/destroy.
1139  *****************************************************************************/
1140
1141 /* When the tunnel UDP socket is closed, all the attached sockets need to go
1142  * too.
1143  */
1144 static void pppol2tp_tunnel_closeall(struct pppol2tp_tunnel *tunnel)
1145 {
1146         int hash;
1147         struct hlist_node *walk;
1148         struct hlist_node *tmp;
1149         struct pppol2tp_session *session;
1150         struct sock *sk;
1151
1152         if (tunnel == NULL)
1153                 BUG();
1154
1155         PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1156                "%s: closing all sessions...\n", tunnel->name);
1157
1158         write_lock_bh(&tunnel->hlist_lock);
1159         for (hash = 0; hash < PPPOL2TP_HASH_SIZE; hash++) {
1160 again:
1161                 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1162                         struct sk_buff *skb;
1163
1164                         session = hlist_entry(walk, struct pppol2tp_session, hlist);
1165
1166                         sk = session->sock;
1167
1168                         PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1169                                "%s: closing session\n", session->name);
1170
1171                         hlist_del_init(&session->hlist);
1172
1173                         /* Since we should hold the sock lock while
1174                          * doing any unbinding, we need to release the
1175                          * lock we're holding before taking that lock.
1176                          * Hold a reference to the sock so it doesn't
1177                          * disappear as we're jumping between locks.
1178                          */
1179                         sock_hold(sk);
1180                         write_unlock_bh(&tunnel->hlist_lock);
1181                         lock_sock(sk);
1182
1183                         if (sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND)) {
1184                                 pppox_unbind_sock(sk);
1185                                 sk->sk_state = PPPOX_DEAD;
1186                                 sk->sk_state_change(sk);
1187                         }
1188
1189                         /* Purge any queued data */
1190                         skb_queue_purge(&sk->sk_receive_queue);
1191                         skb_queue_purge(&sk->sk_write_queue);
1192                         while ((skb = skb_dequeue(&session->reorder_q))) {
1193                                 kfree_skb(skb);
1194                                 sock_put(sk);
1195                         }
1196
1197                         release_sock(sk);
1198                         sock_put(sk);
1199
1200                         /* Now restart from the beginning of this hash
1201                          * chain.  We always remove a session from the
1202                          * list so we are guaranteed to make forward
1203                          * progress.
1204                          */
1205                         write_lock_bh(&tunnel->hlist_lock);
1206                         goto again;
1207                 }
1208         }
1209         write_unlock_bh(&tunnel->hlist_lock);
1210 }
1211
1212 /* Really kill the tunnel.
1213  * Come here only when all sessions have been cleared from the tunnel.
1214  */
1215 static void pppol2tp_tunnel_free(struct pppol2tp_tunnel *tunnel)
1216 {
1217         /* Remove from socket list */
1218         write_lock_bh(&pppol2tp_tunnel_list_lock);
1219         list_del_init(&tunnel->list);
1220         write_unlock_bh(&pppol2tp_tunnel_list_lock);
1221
1222         atomic_dec(&pppol2tp_tunnel_count);
1223         kfree(tunnel);
1224 }
1225
1226 /* Tunnel UDP socket destruct hook.
1227  * The tunnel context is deleted only when all session sockets have been
1228  * closed.
1229  */
1230 static void pppol2tp_tunnel_destruct(struct sock *sk)
1231 {
1232         struct pppol2tp_tunnel *tunnel;
1233
1234         tunnel = sk->sk_user_data;
1235         if (tunnel == NULL)
1236                 goto end;
1237
1238         PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1239                "%s: closing...\n", tunnel->name);
1240
1241         /* Close all sessions */
1242         pppol2tp_tunnel_closeall(tunnel);
1243
1244         /* No longer an encapsulation socket. See net/ipv4/udp.c */
1245         (udp_sk(sk))->encap_type = 0;
1246         (udp_sk(sk))->encap_rcv = NULL;
1247
1248         /* Remove hooks into tunnel socket */
1249         tunnel->sock = NULL;
1250         sk->sk_destruct = tunnel->old_sk_destruct;
1251         sk->sk_user_data = NULL;
1252
1253         /* Call original (UDP) socket descructor */
1254         if (sk->sk_destruct != NULL)
1255                 (*sk->sk_destruct)(sk);
1256
1257         pppol2tp_tunnel_dec_refcount(tunnel);
1258
1259 end:
1260         return;
1261 }
1262
1263 /* Really kill the session socket. (Called from sock_put() if
1264  * refcnt == 0.)
1265  */
1266 static void pppol2tp_session_destruct(struct sock *sk)
1267 {
1268         struct pppol2tp_session *session = NULL;
1269
1270         if (sk->sk_user_data != NULL) {
1271                 struct pppol2tp_tunnel *tunnel;
1272
1273                 session = sk->sk_user_data;
1274                 if (session == NULL)
1275                         goto out;
1276
1277                 BUG_ON(session->magic != L2TP_SESSION_MAGIC);
1278
1279                 /* Don't use pppol2tp_sock_to_tunnel() here to
1280                  * get the tunnel context because the tunnel
1281                  * socket might have already been closed (its
1282                  * sk->sk_user_data will be NULL) so use the
1283                  * session's private tunnel ptr instead.
1284                  */
1285                 tunnel = session->tunnel;
1286                 if (tunnel != NULL) {
1287                         BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1288
1289                         /* If session_id is zero, this is a null
1290                          * session context, which was created for a
1291                          * socket that is being used only to manage
1292                          * tunnels.
1293                          */
1294                         if (session->tunnel_addr.s_session != 0) {
1295                                 /* Delete the session socket from the
1296                                  * hash
1297                                  */
1298                                 write_lock_bh(&tunnel->hlist_lock);
1299                                 hlist_del_init(&session->hlist);
1300                                 write_unlock_bh(&tunnel->hlist_lock);
1301
1302                                 atomic_dec(&pppol2tp_session_count);
1303                         }
1304
1305                         /* This will delete the tunnel context if this
1306                          * is the last session on the tunnel.
1307                          */
1308                         session->tunnel = NULL;
1309                         session->tunnel_sock = NULL;
1310                         pppol2tp_tunnel_dec_refcount(tunnel);
1311                 }
1312         }
1313
1314         kfree(session);
1315 out:
1316         return;
1317 }
1318
1319 /* Called when the PPPoX socket (session) is closed.
1320  */
1321 static int pppol2tp_release(struct socket *sock)
1322 {
1323         struct sock *sk = sock->sk;
1324         struct pppol2tp_session *session;
1325         int error;
1326
1327         if (!sk)
1328                 return 0;
1329
1330         error = -EBADF;
1331         lock_sock(sk);
1332         if (sock_flag(sk, SOCK_DEAD) != 0)
1333                 goto error;
1334
1335         pppox_unbind_sock(sk);
1336
1337         /* Signal the death of the socket. */
1338         sk->sk_state = PPPOX_DEAD;
1339         sock_orphan(sk);
1340         sock->sk = NULL;
1341
1342         session = pppol2tp_sock_to_session(sk);
1343
1344         /* Purge any queued data */
1345         skb_queue_purge(&sk->sk_receive_queue);
1346         skb_queue_purge(&sk->sk_write_queue);
1347         if (session != NULL) {
1348                 struct sk_buff *skb;
1349                 while ((skb = skb_dequeue(&session->reorder_q))) {
1350                         kfree_skb(skb);
1351                         sock_put(sk);
1352                 }
1353         }
1354
1355         release_sock(sk);
1356
1357         /* This will delete the session context via
1358          * pppol2tp_session_destruct() if the socket's refcnt drops to
1359          * zero.
1360          */
1361         sock_put(sk);
1362
1363         return 0;
1364
1365 error:
1366         release_sock(sk);
1367         return error;
1368 }
1369
1370 /* Internal function to prepare a tunnel (UDP) socket to have PPPoX
1371  * sockets attached to it.
1372  */
1373 static struct sock *pppol2tp_prepare_tunnel_socket(int fd, u16 tunnel_id,
1374                                                    int *error)
1375 {
1376         int err;
1377         struct socket *sock = NULL;
1378         struct sock *sk;
1379         struct pppol2tp_tunnel *tunnel;
1380         struct sock *ret = NULL;
1381
1382         /* Get the tunnel UDP socket from the fd, which was opened by
1383          * the userspace L2TP daemon.
1384          */
1385         err = -EBADF;
1386         sock = sockfd_lookup(fd, &err);
1387         if (!sock) {
1388                 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1389                        "tunl %hu: sockfd_lookup(fd=%d) returned %d\n",
1390                        tunnel_id, fd, err);
1391                 goto err;
1392         }
1393
1394         sk = sock->sk;
1395
1396         /* Quick sanity checks */
1397         err = -EPROTONOSUPPORT;
1398         if (sk->sk_protocol != IPPROTO_UDP) {
1399                 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1400                        "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1401                        tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1402                 goto err;
1403         }
1404         err = -EAFNOSUPPORT;
1405         if (sock->ops->family != AF_INET) {
1406                 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1407                        "tunl %hu: fd %d wrong family, got %d, expected %d\n",
1408                        tunnel_id, fd, sock->ops->family, AF_INET);
1409                 goto err;
1410         }
1411
1412         err = -ENOTCONN;
1413
1414         /* Check if this socket has already been prepped */
1415         tunnel = (struct pppol2tp_tunnel *)sk->sk_user_data;
1416         if (tunnel != NULL) {
1417                 /* User-data field already set */
1418                 err = -EBUSY;
1419                 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1420
1421                 /* This socket has already been prepped */
1422                 ret = tunnel->sock;
1423                 goto out;
1424         }
1425
1426         /* This socket is available and needs prepping. Create a new tunnel
1427          * context and init it.
1428          */
1429         sk->sk_user_data = tunnel = kzalloc(sizeof(struct pppol2tp_tunnel), GFP_KERNEL);
1430         if (sk->sk_user_data == NULL) {
1431                 err = -ENOMEM;
1432                 goto err;
1433         }
1434
1435         tunnel->magic = L2TP_TUNNEL_MAGIC;
1436         sprintf(&tunnel->name[0], "tunl %hu", tunnel_id);
1437
1438         tunnel->stats.tunnel_id = tunnel_id;
1439         tunnel->debug = PPPOL2TP_DEFAULT_DEBUG_FLAGS;
1440
1441         /* Hook on the tunnel socket destructor so that we can cleanup
1442          * if the tunnel socket goes away.
1443          */
1444         tunnel->old_sk_destruct = sk->sk_destruct;
1445         sk->sk_destruct = &pppol2tp_tunnel_destruct;
1446
1447         tunnel->sock = sk;
1448         sk->sk_allocation = GFP_ATOMIC;
1449
1450         /* Misc init */
1451         rwlock_init(&tunnel->hlist_lock);
1452
1453         /* Add tunnel to our list */
1454         INIT_LIST_HEAD(&tunnel->list);
1455         write_lock_bh(&pppol2tp_tunnel_list_lock);
1456         list_add(&tunnel->list, &pppol2tp_tunnel_list);
1457         write_unlock_bh(&pppol2tp_tunnel_list_lock);
1458         atomic_inc(&pppol2tp_tunnel_count);
1459
1460         /* Bump the reference count. The tunnel context is deleted
1461          * only when this drops to zero.
1462          */
1463         pppol2tp_tunnel_inc_refcount(tunnel);
1464
1465         /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1466         (udp_sk(sk))->encap_type = UDP_ENCAP_L2TPINUDP;
1467         (udp_sk(sk))->encap_rcv = pppol2tp_udp_encap_recv;
1468
1469         ret = tunnel->sock;
1470
1471         *error = 0;
1472 out:
1473         if (sock)
1474                 sockfd_put(sock);
1475
1476         return ret;
1477
1478 err:
1479         *error = err;
1480         goto out;
1481 }
1482
1483 static struct proto pppol2tp_sk_proto = {
1484         .name     = "PPPOL2TP",
1485         .owner    = THIS_MODULE,
1486         .obj_size = sizeof(struct pppox_sock),
1487 };
1488
1489 /* socket() handler. Initialize a new struct sock.
1490  */
1491 static int pppol2tp_create(struct net *net, struct socket *sock)
1492 {
1493         int error = -ENOMEM;
1494         struct sock *sk;
1495
1496         sk = sk_alloc(net, PF_PPPOX, GFP_KERNEL, &pppol2tp_sk_proto);
1497         if (!sk)
1498                 goto out;
1499
1500         sock_init_data(sock, sk);
1501
1502         sock->state  = SS_UNCONNECTED;
1503         sock->ops    = &pppol2tp_ops;
1504
1505         sk->sk_backlog_rcv = pppol2tp_recv_core;
1506         sk->sk_protocol    = PX_PROTO_OL2TP;
1507         sk->sk_family      = PF_PPPOX;
1508         sk->sk_state       = PPPOX_NONE;
1509         sk->sk_type        = SOCK_STREAM;
1510         sk->sk_destruct    = pppol2tp_session_destruct;
1511
1512         error = 0;
1513
1514 out:
1515         return error;
1516 }
1517
1518 /* connect() handler. Attach a PPPoX socket to a tunnel UDP socket
1519  */
1520 static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
1521                             int sockaddr_len, int flags)
1522 {
1523         struct sock *sk = sock->sk;
1524         struct sockaddr_pppol2tp *sp = (struct sockaddr_pppol2tp *) uservaddr;
1525         struct pppox_sock *po = pppox_sk(sk);
1526         struct sock *tunnel_sock = NULL;
1527         struct pppol2tp_session *session = NULL;
1528         struct pppol2tp_tunnel *tunnel;
1529         struct dst_entry *dst;
1530         int error = 0;
1531
1532         lock_sock(sk);
1533
1534         error = -EINVAL;
1535         if (sp->sa_protocol != PX_PROTO_OL2TP)
1536                 goto end;
1537
1538         /* Check for already bound sockets */
1539         error = -EBUSY;
1540         if (sk->sk_state & PPPOX_CONNECTED)
1541                 goto end;
1542
1543         /* We don't supporting rebinding anyway */
1544         error = -EALREADY;
1545         if (sk->sk_user_data)
1546                 goto end; /* socket is already attached */
1547
1548         /* Don't bind if s_tunnel is 0 */
1549         error = -EINVAL;
1550         if (sp->pppol2tp.s_tunnel == 0)
1551                 goto end;
1552
1553         /* Special case: prepare tunnel socket if s_session and
1554          * d_session is 0. Otherwise look up tunnel using supplied
1555          * tunnel id.
1556          */
1557         if ((sp->pppol2tp.s_session == 0) && (sp->pppol2tp.d_session == 0)) {
1558                 tunnel_sock = pppol2tp_prepare_tunnel_socket(sp->pppol2tp.fd,
1559                                                              sp->pppol2tp.s_tunnel,
1560                                                              &error);
1561                 if (tunnel_sock == NULL)
1562                         goto end;
1563
1564                 tunnel = tunnel_sock->sk_user_data;
1565         } else {
1566                 tunnel = pppol2tp_tunnel_find(sp->pppol2tp.s_tunnel);
1567
1568                 /* Error if we can't find the tunnel */
1569                 error = -ENOENT;
1570                 if (tunnel == NULL)
1571                         goto end;
1572
1573                 tunnel_sock = tunnel->sock;
1574         }
1575
1576         /* Check that this session doesn't already exist */
1577         error = -EEXIST;
1578         session = pppol2tp_session_find(tunnel, sp->pppol2tp.s_session);
1579         if (session != NULL)
1580                 goto end;
1581
1582         /* Allocate and initialize a new session context. */
1583         session = kzalloc(sizeof(struct pppol2tp_session), GFP_KERNEL);
1584         if (session == NULL) {
1585                 error = -ENOMEM;
1586                 goto end;
1587         }
1588
1589         skb_queue_head_init(&session->reorder_q);
1590
1591         session->magic       = L2TP_SESSION_MAGIC;
1592         session->owner       = current->pid;
1593         session->sock        = sk;
1594         session->tunnel      = tunnel;
1595         session->tunnel_sock = tunnel_sock;
1596         session->tunnel_addr = sp->pppol2tp;
1597         sprintf(&session->name[0], "sess %hu/%hu",
1598                 session->tunnel_addr.s_tunnel,
1599                 session->tunnel_addr.s_session);
1600
1601         session->stats.tunnel_id  = session->tunnel_addr.s_tunnel;
1602         session->stats.session_id = session->tunnel_addr.s_session;
1603
1604         INIT_HLIST_NODE(&session->hlist);
1605
1606         /* Inherit debug options from tunnel */
1607         session->debug = tunnel->debug;
1608
1609         /* Default MTU must allow space for UDP/L2TP/PPP
1610          * headers.
1611          */
1612         session->mtu = session->mru = 1500 - PPPOL2TP_HEADER_OVERHEAD;
1613
1614         /* If PMTU discovery was enabled, use the MTU that was discovered */
1615         dst = sk_dst_get(sk);
1616         if (dst != NULL) {
1617                 u32 pmtu = dst_mtu(__sk_dst_get(sk));
1618                 if (pmtu != 0)
1619                         session->mtu = session->mru = pmtu -
1620                                 PPPOL2TP_HEADER_OVERHEAD;
1621                 dst_release(dst);
1622         }
1623
1624         /* Special case: if source & dest session_id == 0x0000, this socket is
1625          * being created to manage the tunnel. Don't add the session to the
1626          * session hash list, just set up the internal context for use by
1627          * ioctl() and sockopt() handlers.
1628          */
1629         if ((session->tunnel_addr.s_session == 0) &&
1630             (session->tunnel_addr.d_session == 0)) {
1631                 error = 0;
1632                 sk->sk_user_data = session;
1633                 goto out_no_ppp;
1634         }
1635
1636         /* Get tunnel context from the tunnel socket */
1637         tunnel = pppol2tp_sock_to_tunnel(tunnel_sock);
1638         if (tunnel == NULL) {
1639                 error = -EBADF;
1640                 goto end;
1641         }
1642
1643         /* Right now, because we don't have a way to push the incoming skb's
1644          * straight through the UDP layer, the only header we need to worry
1645          * about is the L2TP header. This size is different depending on
1646          * whether sequence numbers are enabled for the data channel.
1647          */
1648         po->chan.hdrlen = PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
1649
1650         po->chan.private = sk;
1651         po->chan.ops     = &pppol2tp_chan_ops;
1652         po->chan.mtu     = session->mtu;
1653
1654         error = ppp_register_channel(&po->chan);
1655         if (error)
1656                 goto end_put_tun;
1657
1658         /* This is how we get the session context from the socket. */
1659         sk->sk_user_data = session;
1660
1661         /* Add session to the tunnel's hash list */
1662         write_lock_bh(&tunnel->hlist_lock);
1663         hlist_add_head(&session->hlist,
1664                        pppol2tp_session_id_hash(tunnel,
1665                                                 session->tunnel_addr.s_session));
1666         write_unlock_bh(&tunnel->hlist_lock);
1667
1668         atomic_inc(&pppol2tp_session_count);
1669
1670 out_no_ppp:
1671         pppol2tp_tunnel_inc_refcount(tunnel);
1672         sk->sk_state = PPPOX_CONNECTED;
1673         PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1674                "%s: created\n", session->name);
1675
1676 end_put_tun:
1677         sock_put(tunnel_sock);
1678 end:
1679         release_sock(sk);
1680
1681         if (error != 0) {
1682                 if (session)
1683                         PRINTK(session->debug,
1684                                 PPPOL2TP_MSG_CONTROL, KERN_WARNING,
1685                                 "%s: connect failed: %d\n",
1686                                 session->name, error);
1687                 else
1688                         PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_WARNING,
1689                                 "connect failed: %d\n", error);
1690         }
1691
1692         return error;
1693 }
1694
1695 /* getname() support.
1696  */
1697 static int pppol2tp_getname(struct socket *sock, struct sockaddr *uaddr,
1698                             int *usockaddr_len, int peer)
1699 {
1700         int len = sizeof(struct sockaddr_pppol2tp);
1701         struct sockaddr_pppol2tp sp;
1702         int error = 0;
1703         struct pppol2tp_session *session;
1704
1705         error = -ENOTCONN;
1706         if (sock->sk->sk_state != PPPOX_CONNECTED)
1707                 goto end;
1708
1709         session = pppol2tp_sock_to_session(sock->sk);
1710         if (session == NULL) {
1711                 error = -EBADF;
1712                 goto end;
1713         }
1714
1715         sp.sa_family    = AF_PPPOX;
1716         sp.sa_protocol  = PX_PROTO_OL2TP;
1717         memcpy(&sp.pppol2tp, &session->tunnel_addr,
1718                sizeof(struct pppol2tp_addr));
1719
1720         memcpy(uaddr, &sp, len);
1721
1722         *usockaddr_len = len;
1723
1724         error = 0;
1725         sock_put(sock->sk);
1726
1727 end:
1728         return error;
1729 }
1730
1731 /****************************************************************************
1732  * ioctl() handlers.
1733  *
1734  * The PPPoX socket is created for L2TP sessions: tunnels have their own UDP
1735  * sockets. However, in order to control kernel tunnel features, we allow
1736  * userspace to create a special "tunnel" PPPoX socket which is used for
1737  * control only.  Tunnel PPPoX sockets have session_id == 0 and simply allow
1738  * the user application to issue L2TP setsockopt(), getsockopt() and ioctl()
1739  * calls.
1740  ****************************************************************************/
1741
1742 /* Session ioctl helper.
1743  */
1744 static int pppol2tp_session_ioctl(struct pppol2tp_session *session,
1745                                   unsigned int cmd, unsigned long arg)
1746 {
1747         struct ifreq ifr;
1748         int err = 0;
1749         struct sock *sk = session->sock;
1750         int val = (int) arg;
1751
1752         PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_DEBUG,
1753                "%s: pppol2tp_session_ioctl(cmd=%#x, arg=%#lx)\n",
1754                session->name, cmd, arg);
1755
1756         sock_hold(sk);
1757
1758         switch (cmd) {
1759         case SIOCGIFMTU:
1760                 err = -ENXIO;
1761                 if (!(sk->sk_state & PPPOX_CONNECTED))
1762                         break;
1763
1764                 err = -EFAULT;
1765                 if (copy_from_user(&ifr, (void __user *) arg, sizeof(struct ifreq)))
1766                         break;
1767                 ifr.ifr_mtu = session->mtu;
1768                 if (copy_to_user((void __user *) arg, &ifr, sizeof(struct ifreq)))
1769                         break;
1770
1771                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1772                        "%s: get mtu=%d\n", session->name, session->mtu);
1773                 err = 0;
1774                 break;
1775
1776         case SIOCSIFMTU:
1777                 err = -ENXIO;
1778                 if (!(sk->sk_state & PPPOX_CONNECTED))
1779                         break;
1780
1781                 err = -EFAULT;
1782                 if (copy_from_user(&ifr, (void __user *) arg, sizeof(struct ifreq)))
1783                         break;
1784
1785                 session->mtu = ifr.ifr_mtu;
1786
1787                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1788                        "%s: set mtu=%d\n", session->name, session->mtu);
1789                 err = 0;
1790                 break;
1791
1792         case PPPIOCGMRU:
1793                 err = -ENXIO;
1794                 if (!(sk->sk_state & PPPOX_CONNECTED))
1795                         break;
1796
1797                 err = -EFAULT;
1798                 if (put_user(session->mru, (int __user *) arg))
1799                         break;
1800
1801                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1802                        "%s: get mru=%d\n", session->name, session->mru);
1803                 err = 0;
1804                 break;
1805
1806         case PPPIOCSMRU:
1807                 err = -ENXIO;
1808                 if (!(sk->sk_state & PPPOX_CONNECTED))
1809                         break;
1810
1811                 err = -EFAULT;
1812                 if (get_user(val,(int __user *) arg))
1813                         break;
1814
1815                 session->mru = val;
1816                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1817                        "%s: set mru=%d\n", session->name, session->mru);
1818                 err = 0;
1819                 break;
1820
1821         case PPPIOCGFLAGS:
1822                 err = -EFAULT;
1823                 if (put_user(session->flags, (int __user *) arg))
1824                         break;
1825
1826                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1827                        "%s: get flags=%d\n", session->name, session->flags);
1828                 err = 0;
1829                 break;
1830
1831         case PPPIOCSFLAGS:
1832                 err = -EFAULT;
1833                 if (get_user(val, (int __user *) arg))
1834                         break;
1835                 session->flags = val;
1836                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1837                        "%s: set flags=%d\n", session->name, session->flags);
1838                 err = 0;
1839                 break;
1840
1841         case PPPIOCGL2TPSTATS:
1842                 err = -ENXIO;
1843                 if (!(sk->sk_state & PPPOX_CONNECTED))
1844                         break;
1845
1846                 if (copy_to_user((void __user *) arg, &session->stats,
1847                                  sizeof(session->stats)))
1848                         break;
1849                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1850                        "%s: get L2TP stats\n", session->name);
1851                 err = 0;
1852                 break;
1853
1854         default:
1855                 err = -ENOSYS;
1856                 break;
1857         }
1858
1859         sock_put(sk);
1860
1861         return err;
1862 }
1863
1864 /* Tunnel ioctl helper.
1865  *
1866  * Note the special handling for PPPIOCGL2TPSTATS below. If the ioctl data
1867  * specifies a session_id, the session ioctl handler is called. This allows an
1868  * application to retrieve session stats via a tunnel socket.
1869  */
1870 static int pppol2tp_tunnel_ioctl(struct pppol2tp_tunnel *tunnel,
1871                                  unsigned int cmd, unsigned long arg)
1872 {
1873         int err = 0;
1874         struct sock *sk = tunnel->sock;
1875         struct pppol2tp_ioc_stats stats_req;
1876
1877         PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_DEBUG,
1878                "%s: pppol2tp_tunnel_ioctl(cmd=%#x, arg=%#lx)\n", tunnel->name,
1879                cmd, arg);
1880
1881         sock_hold(sk);
1882
1883         switch (cmd) {
1884         case PPPIOCGL2TPSTATS:
1885                 err = -ENXIO;
1886                 if (!(sk->sk_state & PPPOX_CONNECTED))
1887                         break;
1888
1889                 if (copy_from_user(&stats_req, (void __user *) arg,
1890                                    sizeof(stats_req))) {
1891                         err = -EFAULT;
1892                         break;
1893                 }
1894                 if (stats_req.session_id != 0) {
1895                         /* resend to session ioctl handler */
1896                         struct pppol2tp_session *session =
1897                                 pppol2tp_session_find(tunnel, stats_req.session_id);
1898                         if (session != NULL)
1899                                 err = pppol2tp_session_ioctl(session, cmd, arg);
1900                         else
1901                                 err = -EBADR;
1902                         break;
1903                 }
1904 #ifdef CONFIG_XFRM
1905                 tunnel->stats.using_ipsec = (sk->sk_policy[0] || sk->sk_policy[1]) ? 1 : 0;
1906 #endif
1907                 if (copy_to_user((void __user *) arg, &tunnel->stats,
1908                                  sizeof(tunnel->stats))) {
1909                         err = -EFAULT;
1910                         break;
1911                 }
1912                 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1913                        "%s: get L2TP stats\n", tunnel->name);
1914                 err = 0;
1915                 break;
1916
1917         default:
1918                 err = -ENOSYS;
1919                 break;
1920         }
1921
1922         sock_put(sk);
1923
1924         return err;
1925 }
1926
1927 /* Main ioctl() handler.
1928  * Dispatch to tunnel or session helpers depending on the socket.
1929  */
1930 static int pppol2tp_ioctl(struct socket *sock, unsigned int cmd,
1931                           unsigned long arg)
1932 {
1933         struct sock *sk = sock->sk;
1934         struct pppol2tp_session *session;
1935         struct pppol2tp_tunnel *tunnel;
1936         int err;
1937
1938         if (!sk)
1939                 return 0;
1940
1941         err = -EBADF;
1942         if (sock_flag(sk, SOCK_DEAD) != 0)
1943                 goto end;
1944
1945         err = -ENOTCONN;
1946         if ((sk->sk_user_data == NULL) ||
1947             (!(sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND))))
1948                 goto end;
1949
1950         /* Get session context from the socket */
1951         err = -EBADF;
1952         session = pppol2tp_sock_to_session(sk);
1953         if (session == NULL)
1954                 goto end;
1955
1956         /* Special case: if session's session_id is zero, treat ioctl as a
1957          * tunnel ioctl
1958          */
1959         if ((session->tunnel_addr.s_session == 0) &&
1960             (session->tunnel_addr.d_session == 0)) {
1961                 err = -EBADF;
1962                 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
1963                 if (tunnel == NULL)
1964                         goto end_put_sess;
1965
1966                 err = pppol2tp_tunnel_ioctl(tunnel, cmd, arg);
1967                 sock_put(session->tunnel_sock);
1968                 goto end_put_sess;
1969         }
1970
1971         err = pppol2tp_session_ioctl(session, cmd, arg);
1972
1973 end_put_sess:
1974         sock_put(sk);
1975 end:
1976         return err;
1977 }
1978
1979 /*****************************************************************************
1980  * setsockopt() / getsockopt() support.
1981  *
1982  * The PPPoX socket is created for L2TP sessions: tunnels have their own UDP
1983  * sockets. In order to control kernel tunnel features, we allow userspace to
1984  * create a special "tunnel" PPPoX socket which is used for control only.
1985  * Tunnel PPPoX sockets have session_id == 0 and simply allow the user
1986  * application to issue L2TP setsockopt(), getsockopt() and ioctl() calls.
1987  *****************************************************************************/
1988
1989 /* Tunnel setsockopt() helper.
1990  */
1991 static int pppol2tp_tunnel_setsockopt(struct sock *sk,
1992                                       struct pppol2tp_tunnel *tunnel,
1993                                       int optname, int val)
1994 {
1995         int err = 0;
1996
1997         switch (optname) {
1998         case PPPOL2TP_SO_DEBUG:
1999                 tunnel->debug = val;
2000                 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2001                        "%s: set debug=%x\n", tunnel->name, tunnel->debug);
2002                 break;
2003
2004         default:
2005                 err = -ENOPROTOOPT;
2006                 break;
2007         }
2008
2009         return err;
2010 }
2011
2012 /* Session setsockopt helper.
2013  */
2014 static int pppol2tp_session_setsockopt(struct sock *sk,
2015                                        struct pppol2tp_session *session,
2016                                        int optname, int val)
2017 {
2018         int err = 0;
2019
2020         switch (optname) {
2021         case PPPOL2TP_SO_RECVSEQ:
2022                 if ((val != 0) && (val != 1)) {
2023                         err = -EINVAL;
2024                         break;
2025                 }
2026                 session->recv_seq = val ? -1 : 0;
2027                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2028                        "%s: set recv_seq=%d\n", session->name,
2029                        session->recv_seq);
2030                 break;
2031
2032         case PPPOL2TP_SO_SENDSEQ:
2033                 if ((val != 0) && (val != 1)) {
2034                         err = -EINVAL;
2035                         break;
2036                 }
2037                 session->send_seq = val ? -1 : 0;
2038                 {
2039                         struct sock *ssk      = session->sock;
2040                         struct pppox_sock *po = pppox_sk(ssk);
2041                         po->chan.hdrlen = val ? PPPOL2TP_L2TP_HDR_SIZE_SEQ :
2042                                 PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
2043                 }
2044                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2045                        "%s: set send_seq=%d\n", session->name, session->send_seq);
2046                 break;
2047
2048         case PPPOL2TP_SO_LNSMODE:
2049                 if ((val != 0) && (val != 1)) {
2050                         err = -EINVAL;
2051                         break;
2052                 }
2053                 session->lns_mode = val ? -1 : 0;
2054                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2055                        "%s: set lns_mode=%d\n", session->name,
2056                        session->lns_mode);
2057                 break;
2058
2059         case PPPOL2TP_SO_DEBUG:
2060                 session->debug = val;
2061                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2062                        "%s: set debug=%x\n", session->name, session->debug);
2063                 break;
2064
2065         case PPPOL2TP_SO_REORDERTO:
2066                 session->reorder_timeout = msecs_to_jiffies(val);
2067                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2068                        "%s: set reorder_timeout=%d\n", session->name,
2069                        session->reorder_timeout);
2070                 break;
2071
2072         default:
2073                 err = -ENOPROTOOPT;
2074                 break;
2075         }
2076
2077         return err;
2078 }
2079
2080 /* Main setsockopt() entry point.
2081  * Does API checks, then calls either the tunnel or session setsockopt
2082  * handler, according to whether the PPPoL2TP socket is a for a regular
2083  * session or the special tunnel type.
2084  */
2085 static int pppol2tp_setsockopt(struct socket *sock, int level, int optname,
2086                                char __user *optval, int optlen)
2087 {
2088         struct sock *sk = sock->sk;
2089         struct pppol2tp_session *session = sk->sk_user_data;
2090         struct pppol2tp_tunnel *tunnel;
2091         int val;
2092         int err;
2093
2094         if (level != SOL_PPPOL2TP)
2095                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
2096
2097         if (optlen < sizeof(int))
2098                 return -EINVAL;
2099
2100         if (get_user(val, (int __user *)optval))
2101                 return -EFAULT;
2102
2103         err = -ENOTCONN;
2104         if (sk->sk_user_data == NULL)
2105                 goto end;
2106
2107         /* Get session context from the socket */
2108         err = -EBADF;
2109         session = pppol2tp_sock_to_session(sk);
2110         if (session == NULL)
2111                 goto end;
2112
2113         /* Special case: if session_id == 0x0000, treat as operation on tunnel
2114          */
2115         if ((session->tunnel_addr.s_session == 0) &&
2116             (session->tunnel_addr.d_session == 0)) {
2117                 err = -EBADF;
2118                 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2119                 if (tunnel == NULL)
2120                         goto end_put_sess;
2121
2122                 err = pppol2tp_tunnel_setsockopt(sk, tunnel, optname, val);
2123                 sock_put(session->tunnel_sock);
2124         } else
2125                 err = pppol2tp_session_setsockopt(sk, session, optname, val);
2126
2127         err = 0;
2128
2129 end_put_sess:
2130         sock_put(sk);
2131 end:
2132         return err;
2133 }
2134
2135 /* Tunnel getsockopt helper. Called with sock locked.
2136  */
2137 static int pppol2tp_tunnel_getsockopt(struct sock *sk,
2138                                       struct pppol2tp_tunnel *tunnel,
2139                                       int optname, int *val)
2140 {
2141         int err = 0;
2142
2143         switch (optname) {
2144         case PPPOL2TP_SO_DEBUG:
2145                 *val = tunnel->debug;
2146                 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2147                        "%s: get debug=%x\n", tunnel->name, tunnel->debug);
2148                 break;
2149
2150         default:
2151                 err = -ENOPROTOOPT;
2152                 break;
2153         }
2154
2155         return err;
2156 }
2157
2158 /* Session getsockopt helper. Called with sock locked.
2159  */
2160 static int pppol2tp_session_getsockopt(struct sock *sk,
2161                                        struct pppol2tp_session *session,
2162                                        int optname, int *val)
2163 {
2164         int err = 0;
2165
2166         switch (optname) {
2167         case PPPOL2TP_SO_RECVSEQ:
2168                 *val = session->recv_seq;
2169                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2170                        "%s: get recv_seq=%d\n", session->name, *val);
2171                 break;
2172
2173         case PPPOL2TP_SO_SENDSEQ:
2174                 *val = session->send_seq;
2175                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2176                        "%s: get send_seq=%d\n", session->name, *val);
2177                 break;
2178
2179         case PPPOL2TP_SO_LNSMODE:
2180                 *val = session->lns_mode;
2181                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2182                        "%s: get lns_mode=%d\n", session->name, *val);
2183                 break;
2184
2185         case PPPOL2TP_SO_DEBUG:
2186                 *val = session->debug;
2187                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2188                        "%s: get debug=%d\n", session->name, *val);
2189                 break;
2190
2191         case PPPOL2TP_SO_REORDERTO:
2192                 *val = (int) jiffies_to_msecs(session->reorder_timeout);
2193                 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2194                        "%s: get reorder_timeout=%d\n", session->name, *val);
2195                 break;
2196
2197         default:
2198                 err = -ENOPROTOOPT;
2199         }
2200
2201         return err;
2202 }
2203
2204 /* Main getsockopt() entry point.
2205  * Does API checks, then calls either the tunnel or session getsockopt
2206  * handler, according to whether the PPPoX socket is a for a regular session
2207  * or the special tunnel type.
2208  */
2209 static int pppol2tp_getsockopt(struct socket *sock, int level,
2210                                int optname, char __user *optval, int __user *optlen)
2211 {
2212         struct sock *sk = sock->sk;
2213         struct pppol2tp_session *session = sk->sk_user_data;
2214         struct pppol2tp_tunnel *tunnel;
2215         int val, len;
2216         int err;
2217
2218         if (level != SOL_PPPOL2TP)
2219                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
2220
2221         if (get_user(len, (int __user *) optlen))
2222                 return -EFAULT;
2223
2224         len = min_t(unsigned int, len, sizeof(int));
2225
2226         if (len < 0)
2227                 return -EINVAL;
2228
2229         err = -ENOTCONN;
2230         if (sk->sk_user_data == NULL)
2231                 goto end;
2232
2233         /* Get the session context */
2234         err = -EBADF;
2235         session = pppol2tp_sock_to_session(sk);
2236         if (session == NULL)
2237                 goto end;
2238
2239         /* Special case: if session_id == 0x0000, treat as operation on tunnel */
2240         if ((session->tunnel_addr.s_session == 0) &&
2241             (session->tunnel_addr.d_session == 0)) {
2242                 err = -EBADF;
2243                 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2244                 if (tunnel == NULL)
2245                         goto end_put_sess;
2246
2247                 err = pppol2tp_tunnel_getsockopt(sk, tunnel, optname, &val);
2248                 sock_put(session->tunnel_sock);
2249         } else
2250                 err = pppol2tp_session_getsockopt(sk, session, optname, &val);
2251
2252         err = -EFAULT;
2253         if (put_user(len, (int __user *) optlen))
2254                 goto end_put_sess;
2255
2256         if (copy_to_user((void __user *) optval, &val, len))
2257                 goto end_put_sess;
2258
2259         err = 0;
2260
2261 end_put_sess:
2262         sock_put(sk);
2263 end:
2264         return err;
2265 }
2266
2267 /*****************************************************************************
2268  * /proc filesystem for debug
2269  *****************************************************************************/
2270
2271 #ifdef CONFIG_PROC_FS
2272
2273 #include <linux/seq_file.h>
2274
2275 struct pppol2tp_seq_data {
2276         struct pppol2tp_tunnel *tunnel; /* current tunnel */
2277         struct pppol2tp_session *session; /* NULL means get first session in tunnel */
2278 };
2279
2280 static struct pppol2tp_session *next_session(struct pppol2tp_tunnel *tunnel, struct pppol2tp_session *curr)
2281 {
2282         struct pppol2tp_session *session = NULL;
2283         struct hlist_node *walk;
2284         int found = 0;
2285         int next = 0;
2286         int i;
2287
2288         read_lock_bh(&tunnel->hlist_lock);
2289         for (i = 0; i < PPPOL2TP_HASH_SIZE; i++) {
2290                 hlist_for_each_entry(session, walk, &tunnel->session_hlist[i], hlist) {
2291                         if (curr == NULL) {
2292                                 found = 1;
2293                                 goto out;
2294                         }
2295                         if (session == curr) {
2296                                 next = 1;
2297                                 continue;
2298                         }
2299                         if (next) {
2300                                 found = 1;
2301                                 goto out;
2302                         }
2303                 }
2304         }
2305 out:
2306         read_unlock_bh(&tunnel->hlist_lock);
2307         if (!found)
2308                 session = NULL;
2309
2310         return session;
2311 }
2312
2313 static struct pppol2tp_tunnel *next_tunnel(struct pppol2tp_tunnel *curr)
2314 {
2315         struct pppol2tp_tunnel *tunnel = NULL;
2316
2317         read_lock_bh(&pppol2tp_tunnel_list_lock);
2318         if (list_is_last(&curr->list, &pppol2tp_tunnel_list)) {
2319                 goto out;
2320         }
2321         tunnel = list_entry(curr->list.next, struct pppol2tp_tunnel, list);
2322 out:
2323         read_unlock_bh(&pppol2tp_tunnel_list_lock);
2324
2325         return tunnel;
2326 }
2327
2328 static void *pppol2tp_seq_start(struct seq_file *m, loff_t *offs)
2329 {
2330         struct pppol2tp_seq_data *pd = SEQ_START_TOKEN;
2331         loff_t pos = *offs;
2332
2333         if (!pos)
2334                 goto out;
2335
2336         BUG_ON(m->private == NULL);
2337         pd = m->private;
2338
2339         if (pd->tunnel == NULL) {
2340                 if (!list_empty(&pppol2tp_tunnel_list))
2341                         pd->tunnel = list_entry(pppol2tp_tunnel_list.next, struct pppol2tp_tunnel, list);
2342         } else {
2343                 pd->session = next_session(pd->tunnel, pd->session);
2344                 if (pd->session == NULL) {
2345                         pd->tunnel = next_tunnel(pd->tunnel);
2346                 }
2347         }
2348
2349         /* NULL tunnel and session indicates end of list */
2350         if ((pd->tunnel == NULL) && (pd->session == NULL))
2351                 pd = NULL;
2352
2353 out:
2354         return pd;
2355 }
2356
2357 static void *pppol2tp_seq_next(struct seq_file *m, void *v, loff_t *pos)
2358 {
2359         (*pos)++;
2360         return NULL;
2361 }
2362
2363 static void pppol2tp_seq_stop(struct seq_file *p, void *v)
2364 {
2365         /* nothing to do */
2366 }
2367
2368 static void pppol2tp_seq_tunnel_show(struct seq_file *m, void *v)
2369 {
2370         struct pppol2tp_tunnel *tunnel = v;
2371
2372         seq_printf(m, "\nTUNNEL '%s', %c %d\n",
2373                    tunnel->name,
2374                    (tunnel == tunnel->sock->sk_user_data) ? 'Y':'N',
2375                    atomic_read(&tunnel->ref_count) - 1);
2376         seq_printf(m, " %08x %llu/%llu/%llu %llu/%llu/%llu\n",
2377                    tunnel->debug,
2378                    (unsigned long long)tunnel->stats.tx_packets,
2379                    (unsigned long long)tunnel->stats.tx_bytes,
2380                    (unsigned long long)tunnel->stats.tx_errors,
2381                    (unsigned long long)tunnel->stats.rx_packets,
2382                    (unsigned long long)tunnel->stats.rx_bytes,
2383                    (unsigned long long)tunnel->stats.rx_errors);
2384 }
2385
2386 static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
2387 {
2388         struct pppol2tp_session *session = v;
2389
2390         seq_printf(m, "  SESSION '%s' %08X/%d %04X/%04X -> "
2391                    "%04X/%04X %d %c\n",
2392                    session->name,
2393                    ntohl(session->tunnel_addr.addr.sin_addr.s_addr),
2394                    ntohs(session->tunnel_addr.addr.sin_port),
2395                    session->tunnel_addr.s_tunnel,
2396                    session->tunnel_addr.s_session,
2397                    session->tunnel_addr.d_tunnel,
2398                    session->tunnel_addr.d_session,
2399                    session->sock->sk_state,
2400                    (session == session->sock->sk_user_data) ?
2401                    'Y' : 'N');
2402         seq_printf(m, "   %d/%d/%c/%c/%s %08x %u\n",
2403                    session->mtu, session->mru,
2404                    session->recv_seq ? 'R' : '-',
2405                    session->send_seq ? 'S' : '-',
2406                    session->lns_mode ? "LNS" : "LAC",
2407                    session->debug,
2408                    jiffies_to_msecs(session->reorder_timeout));
2409         seq_printf(m, "   %hu/%hu %llu/%llu/%llu %llu/%llu/%llu\n",
2410                    session->nr, session->ns,
2411                    (unsigned long long)session->stats.tx_packets,
2412                    (unsigned long long)session->stats.tx_bytes,
2413                    (unsigned long long)session->stats.tx_errors,
2414                    (unsigned long long)session->stats.rx_packets,
2415                    (unsigned long long)session->stats.rx_bytes,
2416                    (unsigned long long)session->stats.rx_errors);
2417 }
2418
2419 static int pppol2tp_seq_show(struct seq_file *m, void *v)
2420 {
2421         struct pppol2tp_seq_data *pd = v;
2422
2423         /* display header on line 1 */
2424         if (v == SEQ_START_TOKEN) {
2425                 seq_puts(m, "PPPoL2TP driver info, " PPPOL2TP_DRV_VERSION "\n");
2426                 seq_puts(m, "TUNNEL name, user-data-ok session-count\n");
2427                 seq_puts(m, " debug tx-pkts/bytes/errs rx-pkts/bytes/errs\n");
2428                 seq_puts(m, "  SESSION name, addr/port src-tid/sid "
2429                          "dest-tid/sid state user-data-ok\n");
2430                 seq_puts(m, "   mtu/mru/rcvseq/sendseq/lns debug reorderto\n");
2431                 seq_puts(m, "   nr/ns tx-pkts/bytes/errs rx-pkts/bytes/errs\n");
2432                 goto out;
2433         }
2434
2435         /* Show the tunnel or session context.
2436          */
2437         if (pd->session == NULL)
2438                 pppol2tp_seq_tunnel_show(m, pd->tunnel);
2439         else
2440                 pppol2tp_seq_session_show(m, pd->session);
2441
2442 out:
2443         return 0;
2444 }
2445
2446 static struct seq_operations pppol2tp_seq_ops = {
2447         .start          = pppol2tp_seq_start,
2448         .next           = pppol2tp_seq_next,
2449         .stop           = pppol2tp_seq_stop,
2450         .show           = pppol2tp_seq_show,
2451 };
2452
2453 /* Called when our /proc file is opened. We allocate data for use when
2454  * iterating our tunnel / session contexts and store it in the private
2455  * data of the seq_file.
2456  */
2457 static int pppol2tp_proc_open(struct inode *inode, struct file *file)
2458 {
2459         struct seq_file *m;
2460         struct pppol2tp_seq_data *pd;
2461         int ret = 0;
2462
2463         ret = seq_open(file, &pppol2tp_seq_ops);
2464         if (ret < 0)
2465                 goto out;
2466
2467         m = file->private_data;
2468
2469         /* Allocate and fill our proc_data for access later */
2470         ret = -ENOMEM;
2471         m->private = kzalloc(sizeof(struct pppol2tp_seq_data), GFP_KERNEL);
2472         if (m->private == NULL)
2473                 goto out;
2474
2475         pd = m->private;
2476         ret = 0;
2477
2478 out:
2479         return ret;
2480 }
2481
2482 /* Called when /proc file access completes.
2483  */
2484 static int pppol2tp_proc_release(struct inode *inode, struct file *file)
2485 {
2486         struct seq_file *m = (struct seq_file *)file->private_data;
2487
2488         kfree(m->private);
2489         m->private = NULL;
2490
2491         return seq_release(inode, file);
2492 }
2493
2494 static struct file_operations pppol2tp_proc_fops = {
2495         .owner          = THIS_MODULE,
2496         .open           = pppol2tp_proc_open,
2497         .read           = seq_read,
2498         .llseek         = seq_lseek,
2499         .release        = pppol2tp_proc_release,
2500 };
2501
2502 static struct proc_dir_entry *pppol2tp_proc;
2503
2504 #endif /* CONFIG_PROC_FS */
2505
2506 /*****************************************************************************
2507  * Init and cleanup
2508  *****************************************************************************/
2509
2510 static struct proto_ops pppol2tp_ops = {
2511         .family         = AF_PPPOX,
2512         .owner          = THIS_MODULE,
2513         .release        = pppol2tp_release,
2514         .bind           = sock_no_bind,
2515         .connect        = pppol2tp_connect,
2516         .socketpair     = sock_no_socketpair,
2517         .accept         = sock_no_accept,
2518         .getname        = pppol2tp_getname,
2519         .poll           = datagram_poll,
2520         .listen         = sock_no_listen,
2521         .shutdown       = sock_no_shutdown,
2522         .setsockopt     = pppol2tp_setsockopt,
2523         .getsockopt     = pppol2tp_getsockopt,
2524         .sendmsg        = pppol2tp_sendmsg,
2525         .recvmsg        = pppol2tp_recvmsg,
2526         .mmap           = sock_no_mmap,
2527         .ioctl          = pppox_ioctl,
2528 };
2529
2530 static struct pppox_proto pppol2tp_proto = {
2531         .create         = pppol2tp_create,
2532         .ioctl          = pppol2tp_ioctl
2533 };
2534
2535 static int __init pppol2tp_init(void)
2536 {
2537         int err;
2538
2539         err = proto_register(&pppol2tp_sk_proto, 0);
2540         if (err)
2541                 goto out;
2542         err = register_pppox_proto(PX_PROTO_OL2TP, &pppol2tp_proto);
2543         if (err)
2544                 goto out_unregister_pppol2tp_proto;
2545
2546 #ifdef CONFIG_PROC_FS
2547         pppol2tp_proc = proc_net_fops_create(&init_net, "pppol2tp", 0,
2548                                              &pppol2tp_proc_fops);
2549         if (!pppol2tp_proc) {
2550                 err = -ENOMEM;
2551                 goto out_unregister_pppox_proto;
2552         }
2553 #endif /* CONFIG_PROC_FS */
2554         printk(KERN_INFO "PPPoL2TP kernel driver, %s\n",
2555                PPPOL2TP_DRV_VERSION);
2556
2557 out:
2558         return err;
2559 #ifdef CONFIG_PROC_FS
2560 out_unregister_pppox_proto:
2561         unregister_pppox_proto(PX_PROTO_OL2TP);
2562 #endif
2563 out_unregister_pppol2tp_proto:
2564         proto_unregister(&pppol2tp_sk_proto);
2565         goto out;
2566 }
2567
2568 static void __exit pppol2tp_exit(void)
2569 {
2570         unregister_pppox_proto(PX_PROTO_OL2TP);
2571
2572 #ifdef CONFIG_PROC_FS
2573         remove_proc_entry("pppol2tp", init_net.proc_net);
2574 #endif
2575         proto_unregister(&pppol2tp_sk_proto);
2576 }
2577
2578 module_init(pppol2tp_init);
2579 module_exit(pppol2tp_exit);
2580
2581 MODULE_AUTHOR("Martijn van Oosterhout <kleptog@svana.org>, "
2582               "James Chapman <jchapman@katalix.com>");
2583 MODULE_DESCRIPTION("PPP over L2TP over UDP");
2584 MODULE_LICENSE("GPL");
2585 MODULE_VERSION(PPPOL2TP_DRV_VERSION);