ondemand: Solve a big performance issue by counting IOWAIT time as busy
[linux/fpc-iii.git] / drivers / net / pppol2tp.c
blob449a9825200d5775d8bf221a36657ec87588f8fc
1 /*****************************************************************************
2 * Linux PPP over L2TP (PPPoX/PPPoL2TP) Sockets
4 * PPPoX --- Generic PPP encapsulation socket family
5 * PPPoL2TP --- PPP over L2TP (RFC 2661)
7 * Version: 1.0.0
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)
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.
24 /* This driver handles only L2TP data frames; control frames are handled by a
25 * userspace application.
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.
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.
36 * Here's example userspace code to create a socket for sending/receiving data
37 * over an L2TP session:-
39 * struct sockaddr_pppol2tp sax;
40 * int fd;
41 * int session_fd;
43 * fd = socket(AF_PPPOX, SOCK_DGRAM, PX_PROTO_OL2TP);
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;
56 * session_fd = connect(fd, (struct sockaddr *)&sax, sizeof(sax));
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.
63 #include <linux/module.h>
64 #include <linux/string.h>
65 #include <linux/list.h>
66 #include <asm/uaccess.h>
68 #include <linux/kernel.h>
69 #include <linux/spinlock.h>
70 #include <linux/kthread.h>
71 #include <linux/sched.h>
72 #include <linux/slab.h>
73 #include <linux/errno.h>
74 #include <linux/jiffies.h>
76 #include <linux/netdevice.h>
77 #include <linux/net.h>
78 #include <linux/inetdevice.h>
79 #include <linux/skbuff.h>
80 #include <linux/init.h>
81 #include <linux/ip.h>
82 #include <linux/udp.h>
83 #include <linux/if_pppox.h>
84 #include <linux/if_pppol2tp.h>
85 #include <net/sock.h>
86 #include <linux/ppp_channel.h>
87 #include <linux/ppp_defs.h>
88 #include <linux/if_ppp.h>
89 #include <linux/file.h>
90 #include <linux/hash.h>
91 #include <linux/sort.h>
92 #include <linux/proc_fs.h>
93 #include <linux/nsproxy.h>
94 #include <net/net_namespace.h>
95 #include <net/netns/generic.h>
96 #include <net/dst.h>
97 #include <net/ip.h>
98 #include <net/udp.h>
99 #include <net/xfrm.h>
101 #include <asm/byteorder.h>
102 #include <asm/atomic.h>
105 #define PPPOL2TP_DRV_VERSION "V1.0"
107 /* L2TP header constants */
108 #define L2TP_HDRFLAG_T 0x8000
109 #define L2TP_HDRFLAG_L 0x4000
110 #define L2TP_HDRFLAG_S 0x0800
111 #define L2TP_HDRFLAG_O 0x0200
112 #define L2TP_HDRFLAG_P 0x0100
114 #define L2TP_HDR_VER_MASK 0x000F
115 #define L2TP_HDR_VER 0x0002
117 /* Space for UDP, L2TP and PPP headers */
118 #define PPPOL2TP_HEADER_OVERHEAD 40
120 /* Just some random numbers */
121 #define L2TP_TUNNEL_MAGIC 0x42114DDA
122 #define L2TP_SESSION_MAGIC 0x0C04EB7D
124 #define PPPOL2TP_HASH_BITS 4
125 #define PPPOL2TP_HASH_SIZE (1 << PPPOL2TP_HASH_BITS)
127 /* Default trace flags */
128 #define PPPOL2TP_DEFAULT_DEBUG_FLAGS 0
130 #define PRINTK(_mask, _type, _lvl, _fmt, args...) \
131 do { \
132 if ((_mask) & (_type)) \
133 printk(_lvl "PPPOL2TP: " _fmt, ##args); \
134 } while(0)
136 /* Number of bytes to build transmit L2TP headers.
137 * Unfortunately the size is different depending on whether sequence numbers
138 * are enabled.
140 #define PPPOL2TP_L2TP_HDR_SIZE_SEQ 10
141 #define PPPOL2TP_L2TP_HDR_SIZE_NOSEQ 6
143 struct pppol2tp_tunnel;
145 /* Describes a session. It is the sk_user_data field in the PPPoL2TP
146 * socket. Contains information to determine incoming packets and transmit
147 * outgoing ones.
149 struct pppol2tp_session
151 int magic; /* should be
152 * L2TP_SESSION_MAGIC */
153 int owner; /* pid that opened the socket */
155 struct sock *sock; /* Pointer to the session
156 * PPPoX socket */
157 struct sock *tunnel_sock; /* Pointer to the tunnel UDP
158 * socket */
160 struct pppol2tp_addr tunnel_addr; /* Description of tunnel */
162 struct pppol2tp_tunnel *tunnel; /* back pointer to tunnel
163 * context */
165 char name[20]; /* "sess xxxxx/yyyyy", where
166 * x=tunnel_id, y=session_id */
167 int mtu;
168 int mru;
169 int flags; /* accessed by PPPIOCGFLAGS.
170 * Unused. */
171 unsigned recv_seq:1; /* expect receive packets with
172 * sequence numbers? */
173 unsigned send_seq:1; /* send packets with sequence
174 * numbers? */
175 unsigned lns_mode:1; /* behave as LNS? LAC enables
176 * sequence numbers under
177 * control of LNS. */
178 int debug; /* bitmask of debug message
179 * categories */
180 int reorder_timeout; /* configured reorder timeout
181 * (in jiffies) */
182 u16 nr; /* session NR state (receive) */
183 u16 ns; /* session NR state (send) */
184 struct sk_buff_head reorder_q; /* receive reorder queue */
185 struct pppol2tp_ioc_stats stats;
186 struct hlist_node hlist; /* Hash list node */
189 /* The sk_user_data field of the tunnel's UDP socket. It contains info to track
190 * all the associated sessions so incoming packets can be sorted out
192 struct pppol2tp_tunnel
194 int magic; /* Should be L2TP_TUNNEL_MAGIC */
195 rwlock_t hlist_lock; /* protect session_hlist */
196 struct hlist_head session_hlist[PPPOL2TP_HASH_SIZE];
197 /* hashed list of sessions,
198 * hashed by id */
199 int debug; /* bitmask of debug message
200 * categories */
201 char name[12]; /* "tunl xxxxx" */
202 struct pppol2tp_ioc_stats stats;
204 void (*old_sk_destruct)(struct sock *);
206 struct sock *sock; /* Parent socket */
207 struct list_head list; /* Keep a list of all open
208 * prepared sockets */
209 struct net *pppol2tp_net; /* the net we belong to */
211 atomic_t ref_count;
214 /* Private data stored for received packets in the skb.
216 struct pppol2tp_skb_cb {
217 u16 ns;
218 u16 nr;
219 u16 has_seq;
220 u16 length;
221 unsigned long expires;
224 #define PPPOL2TP_SKB_CB(skb) ((struct pppol2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
226 static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb);
227 static void pppol2tp_tunnel_free(struct pppol2tp_tunnel *tunnel);
229 static atomic_t pppol2tp_tunnel_count;
230 static atomic_t pppol2tp_session_count;
231 static struct ppp_channel_ops pppol2tp_chan_ops = { pppol2tp_xmit , NULL };
232 static const struct proto_ops pppol2tp_ops;
234 /* per-net private data for this module */
235 static int pppol2tp_net_id __read_mostly;
236 struct pppol2tp_net {
237 struct list_head pppol2tp_tunnel_list;
238 rwlock_t pppol2tp_tunnel_list_lock;
241 static inline struct pppol2tp_net *pppol2tp_pernet(struct net *net)
243 BUG_ON(!net);
245 return net_generic(net, pppol2tp_net_id);
248 /* Helpers to obtain tunnel/session contexts from sockets.
250 static inline struct pppol2tp_session *pppol2tp_sock_to_session(struct sock *sk)
252 struct pppol2tp_session *session;
254 if (sk == NULL)
255 return NULL;
257 sock_hold(sk);
258 session = (struct pppol2tp_session *)(sk->sk_user_data);
259 if (session == NULL) {
260 sock_put(sk);
261 goto out;
264 BUG_ON(session->magic != L2TP_SESSION_MAGIC);
265 out:
266 return session;
269 static inline struct pppol2tp_tunnel *pppol2tp_sock_to_tunnel(struct sock *sk)
271 struct pppol2tp_tunnel *tunnel;
273 if (sk == NULL)
274 return NULL;
276 sock_hold(sk);
277 tunnel = (struct pppol2tp_tunnel *)(sk->sk_user_data);
278 if (tunnel == NULL) {
279 sock_put(sk);
280 goto out;
283 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
284 out:
285 return tunnel;
288 /* Tunnel reference counts. Incremented per session that is added to
289 * the tunnel.
291 static inline void pppol2tp_tunnel_inc_refcount(struct pppol2tp_tunnel *tunnel)
293 atomic_inc(&tunnel->ref_count);
296 static inline void pppol2tp_tunnel_dec_refcount(struct pppol2tp_tunnel *tunnel)
298 if (atomic_dec_and_test(&tunnel->ref_count))
299 pppol2tp_tunnel_free(tunnel);
302 /* Session hash list.
303 * The session_id SHOULD be random according to RFC2661, but several
304 * L2TP implementations (Cisco and Microsoft) use incrementing
305 * session_ids. So we do a real hash on the session_id, rather than a
306 * simple bitmask.
308 static inline struct hlist_head *
309 pppol2tp_session_id_hash(struct pppol2tp_tunnel *tunnel, u16 session_id)
311 unsigned long hash_val = (unsigned long) session_id;
312 return &tunnel->session_hlist[hash_long(hash_val, PPPOL2TP_HASH_BITS)];
315 /* Lookup a session by id
317 static struct pppol2tp_session *
318 pppol2tp_session_find(struct pppol2tp_tunnel *tunnel, u16 session_id)
320 struct hlist_head *session_list =
321 pppol2tp_session_id_hash(tunnel, session_id);
322 struct pppol2tp_session *session;
323 struct hlist_node *walk;
325 read_lock_bh(&tunnel->hlist_lock);
326 hlist_for_each_entry(session, walk, session_list, hlist) {
327 if (session->tunnel_addr.s_session == session_id) {
328 read_unlock_bh(&tunnel->hlist_lock);
329 return session;
332 read_unlock_bh(&tunnel->hlist_lock);
334 return NULL;
337 /* Lookup a tunnel by id
339 static struct pppol2tp_tunnel *pppol2tp_tunnel_find(struct net *net, u16 tunnel_id)
341 struct pppol2tp_tunnel *tunnel;
342 struct pppol2tp_net *pn = pppol2tp_pernet(net);
344 read_lock_bh(&pn->pppol2tp_tunnel_list_lock);
345 list_for_each_entry(tunnel, &pn->pppol2tp_tunnel_list, list) {
346 if (tunnel->stats.tunnel_id == tunnel_id) {
347 read_unlock_bh(&pn->pppol2tp_tunnel_list_lock);
348 return tunnel;
351 read_unlock_bh(&pn->pppol2tp_tunnel_list_lock);
353 return NULL;
356 /*****************************************************************************
357 * Receive data handling
358 *****************************************************************************/
360 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
361 * number.
363 static void pppol2tp_recv_queue_skb(struct pppol2tp_session *session, struct sk_buff *skb)
365 struct sk_buff *skbp;
366 struct sk_buff *tmp;
367 u16 ns = PPPOL2TP_SKB_CB(skb)->ns;
369 spin_lock_bh(&session->reorder_q.lock);
370 skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
371 if (PPPOL2TP_SKB_CB(skbp)->ns > ns) {
372 __skb_queue_before(&session->reorder_q, skbp, skb);
373 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
374 "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
375 session->name, ns, PPPOL2TP_SKB_CB(skbp)->ns,
376 skb_queue_len(&session->reorder_q));
377 session->stats.rx_oos_packets++;
378 goto out;
382 __skb_queue_tail(&session->reorder_q, skb);
384 out:
385 spin_unlock_bh(&session->reorder_q.lock);
388 /* Dequeue a single skb.
390 static void pppol2tp_recv_dequeue_skb(struct pppol2tp_session *session, struct sk_buff *skb)
392 struct pppol2tp_tunnel *tunnel = session->tunnel;
393 int length = PPPOL2TP_SKB_CB(skb)->length;
394 struct sock *session_sock = NULL;
396 /* We're about to requeue the skb, so return resources
397 * to its current owner (a socket receive buffer).
399 skb_orphan(skb);
401 tunnel->stats.rx_packets++;
402 tunnel->stats.rx_bytes += length;
403 session->stats.rx_packets++;
404 session->stats.rx_bytes += length;
406 if (PPPOL2TP_SKB_CB(skb)->has_seq) {
407 /* Bump our Nr */
408 session->nr++;
409 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
410 "%s: updated nr to %hu\n", session->name, session->nr);
413 /* If the socket is bound, send it in to PPP's input queue. Otherwise
414 * queue it on the session socket.
416 session_sock = session->sock;
417 if (session_sock->sk_state & PPPOX_BOUND) {
418 struct pppox_sock *po;
419 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
420 "%s: recv %d byte data frame, passing to ppp\n",
421 session->name, length);
423 /* We need to forget all info related to the L2TP packet
424 * gathered in the skb as we are going to reuse the same
425 * skb for the inner packet.
426 * Namely we need to:
427 * - reset xfrm (IPSec) information as it applies to
428 * the outer L2TP packet and not to the inner one
429 * - release the dst to force a route lookup on the inner
430 * IP packet since skb->dst currently points to the dst
431 * of the UDP tunnel
432 * - reset netfilter information as it doesn't apply
433 * to the inner packet either
435 secpath_reset(skb);
436 skb_dst_drop(skb);
437 nf_reset(skb);
439 po = pppox_sk(session_sock);
440 ppp_input(&po->chan, skb);
441 } else {
442 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
443 "%s: socket not bound\n", session->name);
445 /* Not bound. Nothing we can do, so discard. */
446 session->stats.rx_errors++;
447 kfree_skb(skb);
450 sock_put(session->sock);
453 /* Dequeue skbs from the session's reorder_q, subject to packet order.
454 * Skbs that have been in the queue for too long are simply discarded.
456 static void pppol2tp_recv_dequeue(struct pppol2tp_session *session)
458 struct sk_buff *skb;
459 struct sk_buff *tmp;
461 /* If the pkt at the head of the queue has the nr that we
462 * expect to send up next, dequeue it and any other
463 * in-sequence packets behind it.
465 spin_lock_bh(&session->reorder_q.lock);
466 skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
467 if (time_after(jiffies, PPPOL2TP_SKB_CB(skb)->expires)) {
468 session->stats.rx_seq_discards++;
469 session->stats.rx_errors++;
470 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
471 "%s: oos pkt %hu len %d discarded (too old), "
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 __skb_unlink(skb, &session->reorder_q);
477 kfree_skb(skb);
478 sock_put(session->sock);
479 continue;
482 if (PPPOL2TP_SKB_CB(skb)->has_seq) {
483 if (PPPOL2TP_SKB_CB(skb)->ns != session->nr) {
484 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
485 "%s: holding oos pkt %hu len %d, "
486 "waiting for %hu, reorder_q_len=%d\n",
487 session->name, PPPOL2TP_SKB_CB(skb)->ns,
488 PPPOL2TP_SKB_CB(skb)->length, session->nr,
489 skb_queue_len(&session->reorder_q));
490 goto out;
493 __skb_unlink(skb, &session->reorder_q);
495 /* Process the skb. We release the queue lock while we
496 * do so to let other contexts process the queue.
498 spin_unlock_bh(&session->reorder_q.lock);
499 pppol2tp_recv_dequeue_skb(session, skb);
500 spin_lock_bh(&session->reorder_q.lock);
503 out:
504 spin_unlock_bh(&session->reorder_q.lock);
507 static inline int pppol2tp_verify_udp_checksum(struct sock *sk,
508 struct sk_buff *skb)
510 struct udphdr *uh = udp_hdr(skb);
511 u16 ulen = ntohs(uh->len);
512 struct inet_sock *inet;
513 __wsum psum;
515 if (sk->sk_no_check || skb_csum_unnecessary(skb) || !uh->check)
516 return 0;
518 inet = inet_sk(sk);
519 psum = csum_tcpudp_nofold(inet->inet_saddr, inet->inet_daddr, ulen,
520 IPPROTO_UDP, 0);
522 if ((skb->ip_summed == CHECKSUM_COMPLETE) &&
523 !csum_fold(csum_add(psum, skb->csum)))
524 return 0;
526 skb->csum = psum;
528 return __skb_checksum_complete(skb);
531 /* Internal receive frame. Do the real work of receiving an L2TP data frame
532 * here. The skb is not on a list when we get here.
533 * Returns 0 if the packet was a data packet and was successfully passed on.
534 * Returns 1 if the packet was not a good data packet and could not be
535 * forwarded. All such packets are passed up to userspace to deal with.
537 static int pppol2tp_recv_core(struct sock *sock, struct sk_buff *skb)
539 struct pppol2tp_session *session = NULL;
540 struct pppol2tp_tunnel *tunnel;
541 unsigned char *ptr, *optr;
542 u16 hdrflags;
543 u16 tunnel_id, session_id;
544 int length;
545 int offset;
547 tunnel = pppol2tp_sock_to_tunnel(sock);
548 if (tunnel == NULL)
549 goto no_tunnel;
551 if (tunnel->sock && pppol2tp_verify_udp_checksum(tunnel->sock, skb))
552 goto discard_bad_csum;
554 /* UDP always verifies the packet length. */
555 __skb_pull(skb, sizeof(struct udphdr));
557 /* Short packet? */
558 if (!pskb_may_pull(skb, 12)) {
559 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
560 "%s: recv short packet (len=%d)\n", tunnel->name, skb->len);
561 goto error;
564 /* Point to L2TP header */
565 optr = ptr = skb->data;
567 /* Get L2TP header flags */
568 hdrflags = ntohs(*(__be16*)ptr);
570 /* Trace packet contents, if enabled */
571 if (tunnel->debug & PPPOL2TP_MSG_DATA) {
572 length = min(16u, skb->len);
573 if (!pskb_may_pull(skb, length))
574 goto error;
576 printk(KERN_DEBUG "%s: recv: ", tunnel->name);
578 offset = 0;
579 do {
580 printk(" %02X", ptr[offset]);
581 } while (++offset < length);
583 printk("\n");
586 /* Get length of L2TP packet */
587 length = skb->len;
589 /* If type is control packet, it is handled by userspace. */
590 if (hdrflags & L2TP_HDRFLAG_T) {
591 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
592 "%s: recv control packet, len=%d\n", tunnel->name, length);
593 goto error;
596 /* Skip flags */
597 ptr += 2;
599 /* If length is present, skip it */
600 if (hdrflags & L2TP_HDRFLAG_L)
601 ptr += 2;
603 /* Extract tunnel and session ID */
604 tunnel_id = ntohs(*(__be16 *) ptr);
605 ptr += 2;
606 session_id = ntohs(*(__be16 *) ptr);
607 ptr += 2;
609 /* Find the session context */
610 session = pppol2tp_session_find(tunnel, session_id);
611 if (!session) {
612 /* Not found? Pass to userspace to deal with */
613 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_INFO,
614 "%s: no socket found (%hu/%hu). Passing up.\n",
615 tunnel->name, tunnel_id, session_id);
616 goto error;
618 sock_hold(session->sock);
620 /* The ref count on the socket was increased by the above call since
621 * we now hold a pointer to the session. Take care to do sock_put()
622 * when exiting this function from now on...
625 /* Handle the optional sequence numbers. If we are the LAC,
626 * enable/disable sequence numbers under the control of the LNS. If
627 * no sequence numbers present but we were expecting them, discard
628 * frame.
630 if (hdrflags & L2TP_HDRFLAG_S) {
631 u16 ns, nr;
632 ns = ntohs(*(__be16 *) ptr);
633 ptr += 2;
634 nr = ntohs(*(__be16 *) ptr);
635 ptr += 2;
637 /* Received a packet with sequence numbers. If we're the LNS,
638 * check if we sre sending sequence numbers and if not,
639 * configure it so.
641 if ((!session->lns_mode) && (!session->send_seq)) {
642 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_INFO,
643 "%s: requested to enable seq numbers by LNS\n",
644 session->name);
645 session->send_seq = -1;
648 /* Store L2TP info in the skb */
649 PPPOL2TP_SKB_CB(skb)->ns = ns;
650 PPPOL2TP_SKB_CB(skb)->nr = nr;
651 PPPOL2TP_SKB_CB(skb)->has_seq = 1;
653 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
654 "%s: recv data ns=%hu, nr=%hu, session nr=%hu\n",
655 session->name, ns, nr, session->nr);
656 } else {
657 /* No sequence numbers.
658 * If user has configured mandatory sequence numbers, discard.
660 if (session->recv_seq) {
661 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_WARNING,
662 "%s: recv data has no seq numbers when required. "
663 "Discarding\n", session->name);
664 session->stats.rx_seq_discards++;
665 goto discard;
668 /* If we're the LAC and we're sending sequence numbers, the
669 * LNS has requested that we no longer send sequence numbers.
670 * If we're the LNS and we're sending sequence numbers, the
671 * LAC is broken. Discard the frame.
673 if ((!session->lns_mode) && (session->send_seq)) {
674 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_INFO,
675 "%s: requested to disable seq numbers by LNS\n",
676 session->name);
677 session->send_seq = 0;
678 } else if (session->send_seq) {
679 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_WARNING,
680 "%s: recv data has no seq numbers when required. "
681 "Discarding\n", session->name);
682 session->stats.rx_seq_discards++;
683 goto discard;
686 /* Store L2TP info in the skb */
687 PPPOL2TP_SKB_CB(skb)->has_seq = 0;
690 /* If offset bit set, skip it. */
691 if (hdrflags & L2TP_HDRFLAG_O) {
692 offset = ntohs(*(__be16 *)ptr);
693 ptr += 2 + offset;
696 offset = ptr - optr;
697 if (!pskb_may_pull(skb, offset))
698 goto discard;
700 __skb_pull(skb, offset);
702 /* Skip PPP header, if present. In testing, Microsoft L2TP clients
703 * don't send the PPP header (PPP header compression enabled), but
704 * other clients can include the header. So we cope with both cases
705 * here. The PPP header is always FF03 when using L2TP.
707 * Note that skb->data[] isn't dereferenced from a u16 ptr here since
708 * the field may be unaligned.
710 if (!pskb_may_pull(skb, 2))
711 goto discard;
713 if ((skb->data[0] == 0xff) && (skb->data[1] == 0x03))
714 skb_pull(skb, 2);
716 /* Prepare skb for adding to the session's reorder_q. Hold
717 * packets for max reorder_timeout or 1 second if not
718 * reordering.
720 PPPOL2TP_SKB_CB(skb)->length = length;
721 PPPOL2TP_SKB_CB(skb)->expires = jiffies +
722 (session->reorder_timeout ? session->reorder_timeout : HZ);
724 /* Add packet to the session's receive queue. Reordering is done here, if
725 * enabled. Saved L2TP protocol info is stored in skb->sb[].
727 if (PPPOL2TP_SKB_CB(skb)->has_seq) {
728 if (session->reorder_timeout != 0) {
729 /* Packet reordering enabled. Add skb to session's
730 * reorder queue, in order of ns.
732 pppol2tp_recv_queue_skb(session, skb);
733 } else {
734 /* Packet reordering disabled. Discard out-of-sequence
735 * packets
737 if (PPPOL2TP_SKB_CB(skb)->ns != session->nr) {
738 session->stats.rx_seq_discards++;
739 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
740 "%s: oos pkt %hu len %d discarded, "
741 "waiting for %hu, reorder_q_len=%d\n",
742 session->name, PPPOL2TP_SKB_CB(skb)->ns,
743 PPPOL2TP_SKB_CB(skb)->length, session->nr,
744 skb_queue_len(&session->reorder_q));
745 goto discard;
747 skb_queue_tail(&session->reorder_q, skb);
749 } else {
750 /* No sequence numbers. Add the skb to the tail of the
751 * reorder queue. This ensures that it will be
752 * delivered after all previous sequenced skbs.
754 skb_queue_tail(&session->reorder_q, skb);
757 /* Try to dequeue as many skbs from reorder_q as we can. */
758 pppol2tp_recv_dequeue(session);
759 sock_put(sock);
761 return 0;
763 discard:
764 session->stats.rx_errors++;
765 kfree_skb(skb);
766 sock_put(session->sock);
767 sock_put(sock);
769 return 0;
771 discard_bad_csum:
772 LIMIT_NETDEBUG("%s: UDP: bad checksum\n", tunnel->name);
773 UDP_INC_STATS_USER(&init_net, UDP_MIB_INERRORS, 0);
774 tunnel->stats.rx_errors++;
775 kfree_skb(skb);
776 sock_put(sock);
778 return 0;
780 error:
781 /* Put UDP header back */
782 __skb_push(skb, sizeof(struct udphdr));
783 sock_put(sock);
785 no_tunnel:
786 return 1;
789 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
790 * Return codes:
791 * 0 : success.
792 * <0: error
793 * >0: skb should be passed up to userspace as UDP.
795 static int pppol2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
797 struct pppol2tp_tunnel *tunnel;
799 tunnel = pppol2tp_sock_to_tunnel(sk);
800 if (tunnel == NULL)
801 goto pass_up;
803 PRINTK(tunnel->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
804 "%s: received %d bytes\n", tunnel->name, skb->len);
806 if (pppol2tp_recv_core(sk, skb))
807 goto pass_up_put;
809 sock_put(sk);
810 return 0;
812 pass_up_put:
813 sock_put(sk);
814 pass_up:
815 return 1;
818 /* Receive message. This is the recvmsg for the PPPoL2TP socket.
820 static int pppol2tp_recvmsg(struct kiocb *iocb, struct socket *sock,
821 struct msghdr *msg, size_t len,
822 int flags)
824 int err;
825 struct sk_buff *skb;
826 struct sock *sk = sock->sk;
828 err = -EIO;
829 if (sk->sk_state & PPPOX_BOUND)
830 goto end;
832 msg->msg_namelen = 0;
834 err = 0;
835 skb = skb_recv_datagram(sk, flags & ~MSG_DONTWAIT,
836 flags & MSG_DONTWAIT, &err);
837 if (!skb)
838 goto end;
840 if (len > skb->len)
841 len = skb->len;
842 else if (len < skb->len)
843 msg->msg_flags |= MSG_TRUNC;
845 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, len);
846 if (likely(err == 0))
847 err = len;
849 kfree_skb(skb);
850 end:
851 return err;
854 /************************************************************************
855 * Transmit handling
856 ***********************************************************************/
858 /* Tell how big L2TP headers are for a particular session. This
859 * depends on whether sequence numbers are being used.
861 static inline int pppol2tp_l2tp_header_len(struct pppol2tp_session *session)
863 if (session->send_seq)
864 return PPPOL2TP_L2TP_HDR_SIZE_SEQ;
866 return PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
869 /* Build an L2TP header for the session into the buffer provided.
871 static void pppol2tp_build_l2tp_header(struct pppol2tp_session *session,
872 void *buf)
874 __be16 *bufp = buf;
875 u16 flags = L2TP_HDR_VER;
877 if (session->send_seq)
878 flags |= L2TP_HDRFLAG_S;
880 /* Setup L2TP header.
881 * FIXME: Can this ever be unaligned? Is direct dereferencing of
882 * 16-bit header fields safe here for all architectures?
884 *bufp++ = htons(flags);
885 *bufp++ = htons(session->tunnel_addr.d_tunnel);
886 *bufp++ = htons(session->tunnel_addr.d_session);
887 if (session->send_seq) {
888 *bufp++ = htons(session->ns);
889 *bufp++ = 0;
890 session->ns++;
891 PRINTK(session->debug, PPPOL2TP_MSG_SEQ, KERN_DEBUG,
892 "%s: updated ns to %hu\n", session->name, session->ns);
896 /* This is the sendmsg for the PPPoL2TP pppol2tp_session socket. We come here
897 * when a user application does a sendmsg() on the session socket. L2TP and
898 * PPP headers must be inserted into the user's data.
900 static int pppol2tp_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
901 size_t total_len)
903 static const unsigned char ppph[2] = { 0xff, 0x03 };
904 struct sock *sk = sock->sk;
905 struct inet_sock *inet;
906 __wsum csum;
907 struct sk_buff *skb;
908 int error;
909 int hdr_len;
910 struct pppol2tp_session *session;
911 struct pppol2tp_tunnel *tunnel;
912 struct udphdr *uh;
913 unsigned int len;
914 struct sock *sk_tun;
915 u16 udp_len;
917 error = -ENOTCONN;
918 if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
919 goto error;
921 /* Get session and tunnel contexts */
922 error = -EBADF;
923 session = pppol2tp_sock_to_session(sk);
924 if (session == NULL)
925 goto error;
927 sk_tun = session->tunnel_sock;
928 tunnel = pppol2tp_sock_to_tunnel(sk_tun);
929 if (tunnel == NULL)
930 goto error_put_sess;
932 /* What header length is configured for this session? */
933 hdr_len = pppol2tp_l2tp_header_len(session);
935 /* Allocate a socket buffer */
936 error = -ENOMEM;
937 skb = sock_wmalloc(sk, NET_SKB_PAD + sizeof(struct iphdr) +
938 sizeof(struct udphdr) + hdr_len +
939 sizeof(ppph) + total_len,
940 0, GFP_KERNEL);
941 if (!skb)
942 goto error_put_sess_tun;
944 /* Reserve space for headers. */
945 skb_reserve(skb, NET_SKB_PAD);
946 skb_reset_network_header(skb);
947 skb_reserve(skb, sizeof(struct iphdr));
948 skb_reset_transport_header(skb);
950 /* Build UDP header */
951 inet = inet_sk(sk_tun);
952 udp_len = hdr_len + sizeof(ppph) + total_len;
953 uh = (struct udphdr *) skb->data;
954 uh->source = inet->inet_sport;
955 uh->dest = inet->inet_dport;
956 uh->len = htons(udp_len);
957 uh->check = 0;
958 skb_put(skb, sizeof(struct udphdr));
960 /* Build L2TP header */
961 pppol2tp_build_l2tp_header(session, skb->data);
962 skb_put(skb, hdr_len);
964 /* Add PPP header */
965 skb->data[0] = ppph[0];
966 skb->data[1] = ppph[1];
967 skb_put(skb, 2);
969 /* Copy user data into skb */
970 error = memcpy_fromiovec(skb->data, m->msg_iov, total_len);
971 if (error < 0) {
972 kfree_skb(skb);
973 goto error_put_sess_tun;
975 skb_put(skb, total_len);
977 /* Calculate UDP checksum if configured to do so */
978 if (sk_tun->sk_no_check == UDP_CSUM_NOXMIT)
979 skb->ip_summed = CHECKSUM_NONE;
980 else if (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM)) {
981 skb->ip_summed = CHECKSUM_COMPLETE;
982 csum = skb_checksum(skb, 0, udp_len, 0);
983 uh->check = csum_tcpudp_magic(inet->inet_saddr,
984 inet->inet_daddr,
985 udp_len, IPPROTO_UDP, csum);
986 if (uh->check == 0)
987 uh->check = CSUM_MANGLED_0;
988 } else {
989 skb->ip_summed = CHECKSUM_PARTIAL;
990 skb->csum_start = skb_transport_header(skb) - skb->head;
991 skb->csum_offset = offsetof(struct udphdr, check);
992 uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
993 inet->inet_daddr,
994 udp_len, IPPROTO_UDP, 0);
997 /* Debug */
998 if (session->send_seq)
999 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1000 "%s: send %Zd bytes, ns=%hu\n", session->name,
1001 total_len, session->ns - 1);
1002 else
1003 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1004 "%s: send %Zd bytes\n", session->name, total_len);
1006 if (session->debug & PPPOL2TP_MSG_DATA) {
1007 int i;
1008 unsigned char *datap = skb->data;
1010 printk(KERN_DEBUG "%s: xmit:", session->name);
1011 for (i = 0; i < total_len; i++) {
1012 printk(" %02X", *datap++);
1013 if (i == 15) {
1014 printk(" ...");
1015 break;
1018 printk("\n");
1021 /* Queue the packet to IP for output */
1022 len = skb->len;
1023 error = ip_queue_xmit(skb, 1);
1025 /* Update stats */
1026 if (error >= 0) {
1027 tunnel->stats.tx_packets++;
1028 tunnel->stats.tx_bytes += len;
1029 session->stats.tx_packets++;
1030 session->stats.tx_bytes += len;
1031 } else {
1032 tunnel->stats.tx_errors++;
1033 session->stats.tx_errors++;
1036 return error;
1038 error_put_sess_tun:
1039 sock_put(session->tunnel_sock);
1040 error_put_sess:
1041 sock_put(sk);
1042 error:
1043 return error;
1046 /* Automatically called when the skb is freed.
1048 static void pppol2tp_sock_wfree(struct sk_buff *skb)
1050 sock_put(skb->sk);
1053 /* For data skbs that we transmit, we associate with the tunnel socket
1054 * but don't do accounting.
1056 static inline void pppol2tp_skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1058 sock_hold(sk);
1059 skb->sk = sk;
1060 skb->destructor = pppol2tp_sock_wfree;
1063 /* Transmit function called by generic PPP driver. Sends PPP frame
1064 * over PPPoL2TP socket.
1066 * This is almost the same as pppol2tp_sendmsg(), but rather than
1067 * being called with a msghdr from userspace, it is called with a skb
1068 * from the kernel.
1070 * The supplied skb from ppp doesn't have enough headroom for the
1071 * insertion of L2TP, UDP and IP headers so we need to allocate more
1072 * headroom in the skb. This will create a cloned skb. But we must be
1073 * careful in the error case because the caller will expect to free
1074 * the skb it supplied, not our cloned skb. So we take care to always
1075 * leave the original skb unfreed if we return an error.
1077 static int pppol2tp_xmit(struct ppp_channel *chan, struct sk_buff *skb)
1079 static const u8 ppph[2] = { 0xff, 0x03 };
1080 struct sock *sk = (struct sock *) chan->private;
1081 struct sock *sk_tun;
1082 int hdr_len;
1083 u16 udp_len;
1084 struct pppol2tp_session *session;
1085 struct pppol2tp_tunnel *tunnel;
1086 int rc;
1087 int headroom;
1088 int data_len = skb->len;
1089 struct inet_sock *inet;
1090 __wsum csum;
1091 struct udphdr *uh;
1092 unsigned int len;
1093 int old_headroom;
1094 int new_headroom;
1096 if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
1097 goto abort;
1099 /* Get session and tunnel contexts from the socket */
1100 session = pppol2tp_sock_to_session(sk);
1101 if (session == NULL)
1102 goto abort;
1104 sk_tun = session->tunnel_sock;
1105 if (sk_tun == NULL)
1106 goto abort_put_sess;
1107 tunnel = pppol2tp_sock_to_tunnel(sk_tun);
1108 if (tunnel == NULL)
1109 goto abort_put_sess;
1111 /* What header length is configured for this session? */
1112 hdr_len = pppol2tp_l2tp_header_len(session);
1114 /* Check that there's enough headroom in the skb to insert IP,
1115 * UDP and L2TP and PPP headers. If not enough, expand it to
1116 * make room. Adjust truesize.
1118 headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1119 sizeof(struct udphdr) + hdr_len + sizeof(ppph);
1120 old_headroom = skb_headroom(skb);
1121 if (skb_cow_head(skb, headroom))
1122 goto abort_put_sess_tun;
1124 new_headroom = skb_headroom(skb);
1125 skb_orphan(skb);
1126 skb->truesize += new_headroom - old_headroom;
1128 /* Setup PPP header */
1129 __skb_push(skb, sizeof(ppph));
1130 skb->data[0] = ppph[0];
1131 skb->data[1] = ppph[1];
1133 /* Setup L2TP header */
1134 pppol2tp_build_l2tp_header(session, __skb_push(skb, hdr_len));
1136 udp_len = sizeof(struct udphdr) + hdr_len + sizeof(ppph) + data_len;
1138 /* Setup UDP header */
1139 inet = inet_sk(sk_tun);
1140 __skb_push(skb, sizeof(*uh));
1141 skb_reset_transport_header(skb);
1142 uh = udp_hdr(skb);
1143 uh->source = inet->inet_sport;
1144 uh->dest = inet->inet_dport;
1145 uh->len = htons(udp_len);
1146 uh->check = 0;
1148 /* Debug */
1149 if (session->send_seq)
1150 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1151 "%s: send %d bytes, ns=%hu\n", session->name,
1152 data_len, session->ns - 1);
1153 else
1154 PRINTK(session->debug, PPPOL2TP_MSG_DATA, KERN_DEBUG,
1155 "%s: send %d bytes\n", session->name, data_len);
1157 if (session->debug & PPPOL2TP_MSG_DATA) {
1158 int i;
1159 unsigned char *datap = skb->data;
1161 printk(KERN_DEBUG "%s: xmit:", session->name);
1162 for (i = 0; i < data_len; i++) {
1163 printk(" %02X", *datap++);
1164 if (i == 31) {
1165 printk(" ...");
1166 break;
1169 printk("\n");
1172 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1173 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1174 IPSKB_REROUTED);
1175 nf_reset(skb);
1177 /* Get routing info from the tunnel socket */
1178 skb_dst_drop(skb);
1179 skb_dst_set(skb, dst_clone(__sk_dst_get(sk_tun)));
1180 pppol2tp_skb_set_owner_w(skb, sk_tun);
1182 /* Calculate UDP checksum if configured to do so */
1183 if (sk_tun->sk_no_check == UDP_CSUM_NOXMIT)
1184 skb->ip_summed = CHECKSUM_NONE;
1185 else if ((skb_dst(skb) && skb_dst(skb)->dev) &&
1186 (!(skb_dst(skb)->dev->features & NETIF_F_V4_CSUM))) {
1187 skb->ip_summed = CHECKSUM_COMPLETE;
1188 csum = skb_checksum(skb, 0, udp_len, 0);
1189 uh->check = csum_tcpudp_magic(inet->inet_saddr,
1190 inet->inet_daddr,
1191 udp_len, IPPROTO_UDP, csum);
1192 if (uh->check == 0)
1193 uh->check = CSUM_MANGLED_0;
1194 } else {
1195 skb->ip_summed = CHECKSUM_PARTIAL;
1196 skb->csum_start = skb_transport_header(skb) - skb->head;
1197 skb->csum_offset = offsetof(struct udphdr, check);
1198 uh->check = ~csum_tcpudp_magic(inet->inet_saddr,
1199 inet->inet_daddr,
1200 udp_len, IPPROTO_UDP, 0);
1203 /* Queue the packet to IP for output */
1204 len = skb->len;
1205 rc = ip_queue_xmit(skb, 1);
1207 /* Update stats */
1208 if (rc >= 0) {
1209 tunnel->stats.tx_packets++;
1210 tunnel->stats.tx_bytes += len;
1211 session->stats.tx_packets++;
1212 session->stats.tx_bytes += len;
1213 } else {
1214 tunnel->stats.tx_errors++;
1215 session->stats.tx_errors++;
1218 sock_put(sk_tun);
1219 sock_put(sk);
1220 return 1;
1222 abort_put_sess_tun:
1223 sock_put(sk_tun);
1224 abort_put_sess:
1225 sock_put(sk);
1226 abort:
1227 /* Free the original skb */
1228 kfree_skb(skb);
1229 return 1;
1232 /*****************************************************************************
1233 * Session (and tunnel control) socket create/destroy.
1234 *****************************************************************************/
1236 /* When the tunnel UDP socket is closed, all the attached sockets need to go
1237 * too.
1239 static void pppol2tp_tunnel_closeall(struct pppol2tp_tunnel *tunnel)
1241 int hash;
1242 struct hlist_node *walk;
1243 struct hlist_node *tmp;
1244 struct pppol2tp_session *session;
1245 struct sock *sk;
1247 BUG_ON(tunnel == NULL);
1249 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1250 "%s: closing all sessions...\n", tunnel->name);
1252 write_lock_bh(&tunnel->hlist_lock);
1253 for (hash = 0; hash < PPPOL2TP_HASH_SIZE; hash++) {
1254 again:
1255 hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1256 struct sk_buff *skb;
1258 session = hlist_entry(walk, struct pppol2tp_session, hlist);
1260 sk = session->sock;
1262 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1263 "%s: closing session\n", session->name);
1265 hlist_del_init(&session->hlist);
1267 /* Since we should hold the sock lock while
1268 * doing any unbinding, we need to release the
1269 * lock we're holding before taking that lock.
1270 * Hold a reference to the sock so it doesn't
1271 * disappear as we're jumping between locks.
1273 sock_hold(sk);
1274 write_unlock_bh(&tunnel->hlist_lock);
1275 lock_sock(sk);
1277 if (sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND)) {
1278 pppox_unbind_sock(sk);
1279 sk->sk_state = PPPOX_DEAD;
1280 sk->sk_state_change(sk);
1283 /* Purge any queued data */
1284 skb_queue_purge(&sk->sk_receive_queue);
1285 skb_queue_purge(&sk->sk_write_queue);
1286 while ((skb = skb_dequeue(&session->reorder_q))) {
1287 kfree_skb(skb);
1288 sock_put(sk);
1291 release_sock(sk);
1292 sock_put(sk);
1294 /* Now restart from the beginning of this hash
1295 * chain. We always remove a session from the
1296 * list so we are guaranteed to make forward
1297 * progress.
1299 write_lock_bh(&tunnel->hlist_lock);
1300 goto again;
1303 write_unlock_bh(&tunnel->hlist_lock);
1306 /* Really kill the tunnel.
1307 * Come here only when all sessions have been cleared from the tunnel.
1309 static void pppol2tp_tunnel_free(struct pppol2tp_tunnel *tunnel)
1311 struct pppol2tp_net *pn = pppol2tp_pernet(tunnel->pppol2tp_net);
1313 /* Remove from socket list */
1314 write_lock_bh(&pn->pppol2tp_tunnel_list_lock);
1315 list_del_init(&tunnel->list);
1316 write_unlock_bh(&pn->pppol2tp_tunnel_list_lock);
1318 atomic_dec(&pppol2tp_tunnel_count);
1319 kfree(tunnel);
1322 /* Tunnel UDP socket destruct hook.
1323 * The tunnel context is deleted only when all session sockets have been
1324 * closed.
1326 static void pppol2tp_tunnel_destruct(struct sock *sk)
1328 struct pppol2tp_tunnel *tunnel;
1330 tunnel = sk->sk_user_data;
1331 if (tunnel == NULL)
1332 goto end;
1334 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1335 "%s: closing...\n", tunnel->name);
1337 /* Close all sessions */
1338 pppol2tp_tunnel_closeall(tunnel);
1340 /* No longer an encapsulation socket. See net/ipv4/udp.c */
1341 (udp_sk(sk))->encap_type = 0;
1342 (udp_sk(sk))->encap_rcv = NULL;
1344 /* Remove hooks into tunnel socket */
1345 tunnel->sock = NULL;
1346 sk->sk_destruct = tunnel->old_sk_destruct;
1347 sk->sk_user_data = NULL;
1349 /* Call original (UDP) socket descructor */
1350 if (sk->sk_destruct != NULL)
1351 (*sk->sk_destruct)(sk);
1353 pppol2tp_tunnel_dec_refcount(tunnel);
1355 end:
1356 return;
1359 /* Really kill the session socket. (Called from sock_put() if
1360 * refcnt == 0.)
1362 static void pppol2tp_session_destruct(struct sock *sk)
1364 struct pppol2tp_session *session = NULL;
1366 if (sk->sk_user_data != NULL) {
1367 struct pppol2tp_tunnel *tunnel;
1369 session = sk->sk_user_data;
1370 if (session == NULL)
1371 goto out;
1373 BUG_ON(session->magic != L2TP_SESSION_MAGIC);
1375 /* Don't use pppol2tp_sock_to_tunnel() here to
1376 * get the tunnel context because the tunnel
1377 * socket might have already been closed (its
1378 * sk->sk_user_data will be NULL) so use the
1379 * session's private tunnel ptr instead.
1381 tunnel = session->tunnel;
1382 if (tunnel != NULL) {
1383 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1385 /* If session_id is zero, this is a null
1386 * session context, which was created for a
1387 * socket that is being used only to manage
1388 * tunnels.
1390 if (session->tunnel_addr.s_session != 0) {
1391 /* Delete the session socket from the
1392 * hash
1394 write_lock_bh(&tunnel->hlist_lock);
1395 hlist_del_init(&session->hlist);
1396 write_unlock_bh(&tunnel->hlist_lock);
1398 atomic_dec(&pppol2tp_session_count);
1401 /* This will delete the tunnel context if this
1402 * is the last session on the tunnel.
1404 session->tunnel = NULL;
1405 session->tunnel_sock = NULL;
1406 pppol2tp_tunnel_dec_refcount(tunnel);
1410 kfree(session);
1411 out:
1412 return;
1415 /* Called when the PPPoX socket (session) is closed.
1417 static int pppol2tp_release(struct socket *sock)
1419 struct sock *sk = sock->sk;
1420 struct pppol2tp_session *session;
1421 int error;
1423 if (!sk)
1424 return 0;
1426 error = -EBADF;
1427 lock_sock(sk);
1428 if (sock_flag(sk, SOCK_DEAD) != 0)
1429 goto error;
1431 pppox_unbind_sock(sk);
1433 /* Signal the death of the socket. */
1434 sk->sk_state = PPPOX_DEAD;
1435 sock_orphan(sk);
1436 sock->sk = NULL;
1438 session = pppol2tp_sock_to_session(sk);
1440 /* Purge any queued data */
1441 skb_queue_purge(&sk->sk_receive_queue);
1442 skb_queue_purge(&sk->sk_write_queue);
1443 if (session != NULL) {
1444 struct sk_buff *skb;
1445 while ((skb = skb_dequeue(&session->reorder_q))) {
1446 kfree_skb(skb);
1447 sock_put(sk);
1449 sock_put(sk);
1452 release_sock(sk);
1454 /* This will delete the session context via
1455 * pppol2tp_session_destruct() if the socket's refcnt drops to
1456 * zero.
1458 sock_put(sk);
1460 return 0;
1462 error:
1463 release_sock(sk);
1464 return error;
1467 /* Internal function to prepare a tunnel (UDP) socket to have PPPoX
1468 * sockets attached to it.
1470 static struct sock *pppol2tp_prepare_tunnel_socket(struct net *net,
1471 int fd, u16 tunnel_id, int *error)
1473 int err;
1474 struct socket *sock = NULL;
1475 struct sock *sk;
1476 struct pppol2tp_tunnel *tunnel;
1477 struct pppol2tp_net *pn;
1478 struct sock *ret = NULL;
1480 /* Get the tunnel UDP socket from the fd, which was opened by
1481 * the userspace L2TP daemon.
1483 err = -EBADF;
1484 sock = sockfd_lookup(fd, &err);
1485 if (!sock) {
1486 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1487 "tunl %hu: sockfd_lookup(fd=%d) returned %d\n",
1488 tunnel_id, fd, err);
1489 goto err;
1492 sk = sock->sk;
1494 /* Quick sanity checks */
1495 err = -EPROTONOSUPPORT;
1496 if (sk->sk_protocol != IPPROTO_UDP) {
1497 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1498 "tunl %hu: fd %d wrong protocol, got %d, expected %d\n",
1499 tunnel_id, fd, sk->sk_protocol, IPPROTO_UDP);
1500 goto err;
1502 err = -EAFNOSUPPORT;
1503 if (sock->ops->family != AF_INET) {
1504 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_ERR,
1505 "tunl %hu: fd %d wrong family, got %d, expected %d\n",
1506 tunnel_id, fd, sock->ops->family, AF_INET);
1507 goto err;
1510 err = -ENOTCONN;
1512 /* Check if this socket has already been prepped */
1513 tunnel = (struct pppol2tp_tunnel *)sk->sk_user_data;
1514 if (tunnel != NULL) {
1515 /* User-data field already set */
1516 err = -EBUSY;
1517 BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1519 /* This socket has already been prepped */
1520 ret = tunnel->sock;
1521 goto out;
1524 /* This socket is available and needs prepping. Create a new tunnel
1525 * context and init it.
1527 sk->sk_user_data = tunnel = kzalloc(sizeof(struct pppol2tp_tunnel), GFP_KERNEL);
1528 if (sk->sk_user_data == NULL) {
1529 err = -ENOMEM;
1530 goto err;
1533 tunnel->magic = L2TP_TUNNEL_MAGIC;
1534 sprintf(&tunnel->name[0], "tunl %hu", tunnel_id);
1536 tunnel->stats.tunnel_id = tunnel_id;
1537 tunnel->debug = PPPOL2TP_DEFAULT_DEBUG_FLAGS;
1539 /* Hook on the tunnel socket destructor so that we can cleanup
1540 * if the tunnel socket goes away.
1542 tunnel->old_sk_destruct = sk->sk_destruct;
1543 sk->sk_destruct = pppol2tp_tunnel_destruct;
1545 tunnel->sock = sk;
1546 sk->sk_allocation = GFP_ATOMIC;
1548 /* Misc init */
1549 rwlock_init(&tunnel->hlist_lock);
1551 /* The net we belong to */
1552 tunnel->pppol2tp_net = net;
1553 pn = pppol2tp_pernet(net);
1555 /* Add tunnel to our list */
1556 INIT_LIST_HEAD(&tunnel->list);
1557 write_lock_bh(&pn->pppol2tp_tunnel_list_lock);
1558 list_add(&tunnel->list, &pn->pppol2tp_tunnel_list);
1559 write_unlock_bh(&pn->pppol2tp_tunnel_list_lock);
1560 atomic_inc(&pppol2tp_tunnel_count);
1562 /* Bump the reference count. The tunnel context is deleted
1563 * only when this drops to zero.
1565 pppol2tp_tunnel_inc_refcount(tunnel);
1567 /* Mark socket as an encapsulation socket. See net/ipv4/udp.c */
1568 (udp_sk(sk))->encap_type = UDP_ENCAP_L2TPINUDP;
1569 (udp_sk(sk))->encap_rcv = pppol2tp_udp_encap_recv;
1571 ret = tunnel->sock;
1573 *error = 0;
1574 out:
1575 if (sock)
1576 sockfd_put(sock);
1578 return ret;
1580 err:
1581 *error = err;
1582 goto out;
1585 static struct proto pppol2tp_sk_proto = {
1586 .name = "PPPOL2TP",
1587 .owner = THIS_MODULE,
1588 .obj_size = sizeof(struct pppox_sock),
1591 /* socket() handler. Initialize a new struct sock.
1593 static int pppol2tp_create(struct net *net, struct socket *sock)
1595 int error = -ENOMEM;
1596 struct sock *sk;
1598 sk = sk_alloc(net, PF_PPPOX, GFP_KERNEL, &pppol2tp_sk_proto);
1599 if (!sk)
1600 goto out;
1602 sock_init_data(sock, sk);
1604 sock->state = SS_UNCONNECTED;
1605 sock->ops = &pppol2tp_ops;
1607 sk->sk_backlog_rcv = pppol2tp_recv_core;
1608 sk->sk_protocol = PX_PROTO_OL2TP;
1609 sk->sk_family = PF_PPPOX;
1610 sk->sk_state = PPPOX_NONE;
1611 sk->sk_type = SOCK_STREAM;
1612 sk->sk_destruct = pppol2tp_session_destruct;
1614 error = 0;
1616 out:
1617 return error;
1620 /* connect() handler. Attach a PPPoX socket to a tunnel UDP socket
1622 static int pppol2tp_connect(struct socket *sock, struct sockaddr *uservaddr,
1623 int sockaddr_len, int flags)
1625 struct sock *sk = sock->sk;
1626 struct sockaddr_pppol2tp *sp = (struct sockaddr_pppol2tp *) uservaddr;
1627 struct pppox_sock *po = pppox_sk(sk);
1628 struct sock *tunnel_sock = NULL;
1629 struct pppol2tp_session *session = NULL;
1630 struct pppol2tp_tunnel *tunnel;
1631 struct dst_entry *dst;
1632 int error = 0;
1634 lock_sock(sk);
1636 error = -EINVAL;
1637 if (sp->sa_protocol != PX_PROTO_OL2TP)
1638 goto end;
1640 /* Check for already bound sockets */
1641 error = -EBUSY;
1642 if (sk->sk_state & PPPOX_CONNECTED)
1643 goto end;
1645 /* We don't supporting rebinding anyway */
1646 error = -EALREADY;
1647 if (sk->sk_user_data)
1648 goto end; /* socket is already attached */
1650 /* Don't bind if s_tunnel is 0 */
1651 error = -EINVAL;
1652 if (sp->pppol2tp.s_tunnel == 0)
1653 goto end;
1655 /* Special case: prepare tunnel socket if s_session and
1656 * d_session is 0. Otherwise look up tunnel using supplied
1657 * tunnel id.
1659 if ((sp->pppol2tp.s_session == 0) && (sp->pppol2tp.d_session == 0)) {
1660 tunnel_sock = pppol2tp_prepare_tunnel_socket(sock_net(sk),
1661 sp->pppol2tp.fd,
1662 sp->pppol2tp.s_tunnel,
1663 &error);
1664 if (tunnel_sock == NULL)
1665 goto end;
1667 sock_hold(tunnel_sock);
1668 tunnel = tunnel_sock->sk_user_data;
1669 } else {
1670 tunnel = pppol2tp_tunnel_find(sock_net(sk), sp->pppol2tp.s_tunnel);
1672 /* Error if we can't find the tunnel */
1673 error = -ENOENT;
1674 if (tunnel == NULL)
1675 goto end;
1677 tunnel_sock = tunnel->sock;
1680 /* Check that this session doesn't already exist */
1681 error = -EEXIST;
1682 session = pppol2tp_session_find(tunnel, sp->pppol2tp.s_session);
1683 if (session != NULL)
1684 goto end;
1686 /* Allocate and initialize a new session context. */
1687 session = kzalloc(sizeof(struct pppol2tp_session), GFP_KERNEL);
1688 if (session == NULL) {
1689 error = -ENOMEM;
1690 goto end;
1693 skb_queue_head_init(&session->reorder_q);
1695 session->magic = L2TP_SESSION_MAGIC;
1696 session->owner = current->pid;
1697 session->sock = sk;
1698 session->tunnel = tunnel;
1699 session->tunnel_sock = tunnel_sock;
1700 session->tunnel_addr = sp->pppol2tp;
1701 sprintf(&session->name[0], "sess %hu/%hu",
1702 session->tunnel_addr.s_tunnel,
1703 session->tunnel_addr.s_session);
1705 session->stats.tunnel_id = session->tunnel_addr.s_tunnel;
1706 session->stats.session_id = session->tunnel_addr.s_session;
1708 INIT_HLIST_NODE(&session->hlist);
1710 /* Inherit debug options from tunnel */
1711 session->debug = tunnel->debug;
1713 /* Default MTU must allow space for UDP/L2TP/PPP
1714 * headers.
1716 session->mtu = session->mru = 1500 - PPPOL2TP_HEADER_OVERHEAD;
1718 /* If PMTU discovery was enabled, use the MTU that was discovered */
1719 dst = sk_dst_get(sk);
1720 if (dst != NULL) {
1721 u32 pmtu = dst_mtu(__sk_dst_get(sk));
1722 if (pmtu != 0)
1723 session->mtu = session->mru = pmtu -
1724 PPPOL2TP_HEADER_OVERHEAD;
1725 dst_release(dst);
1728 /* Special case: if source & dest session_id == 0x0000, this socket is
1729 * being created to manage the tunnel. Don't add the session to the
1730 * session hash list, just set up the internal context for use by
1731 * ioctl() and sockopt() handlers.
1733 if ((session->tunnel_addr.s_session == 0) &&
1734 (session->tunnel_addr.d_session == 0)) {
1735 error = 0;
1736 sk->sk_user_data = session;
1737 goto out_no_ppp;
1740 /* Get tunnel context from the tunnel socket */
1741 tunnel = pppol2tp_sock_to_tunnel(tunnel_sock);
1742 if (tunnel == NULL) {
1743 error = -EBADF;
1744 goto end;
1747 /* Right now, because we don't have a way to push the incoming skb's
1748 * straight through the UDP layer, the only header we need to worry
1749 * about is the L2TP header. This size is different depending on
1750 * whether sequence numbers are enabled for the data channel.
1752 po->chan.hdrlen = PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
1754 po->chan.private = sk;
1755 po->chan.ops = &pppol2tp_chan_ops;
1756 po->chan.mtu = session->mtu;
1758 error = ppp_register_net_channel(sock_net(sk), &po->chan);
1759 if (error)
1760 goto end_put_tun;
1762 /* This is how we get the session context from the socket. */
1763 sk->sk_user_data = session;
1765 /* Add session to the tunnel's hash list */
1766 write_lock_bh(&tunnel->hlist_lock);
1767 hlist_add_head(&session->hlist,
1768 pppol2tp_session_id_hash(tunnel,
1769 session->tunnel_addr.s_session));
1770 write_unlock_bh(&tunnel->hlist_lock);
1772 atomic_inc(&pppol2tp_session_count);
1774 out_no_ppp:
1775 pppol2tp_tunnel_inc_refcount(tunnel);
1776 sk->sk_state = PPPOX_CONNECTED;
1777 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1778 "%s: created\n", session->name);
1780 end_put_tun:
1781 sock_put(tunnel_sock);
1782 end:
1783 release_sock(sk);
1785 if (error != 0) {
1786 if (session)
1787 PRINTK(session->debug,
1788 PPPOL2TP_MSG_CONTROL, KERN_WARNING,
1789 "%s: connect failed: %d\n",
1790 session->name, error);
1791 else
1792 PRINTK(-1, PPPOL2TP_MSG_CONTROL, KERN_WARNING,
1793 "connect failed: %d\n", error);
1796 return error;
1799 /* getname() support.
1801 static int pppol2tp_getname(struct socket *sock, struct sockaddr *uaddr,
1802 int *usockaddr_len, int peer)
1804 int len = sizeof(struct sockaddr_pppol2tp);
1805 struct sockaddr_pppol2tp sp;
1806 int error = 0;
1807 struct pppol2tp_session *session;
1809 error = -ENOTCONN;
1810 if (sock->sk->sk_state != PPPOX_CONNECTED)
1811 goto end;
1813 session = pppol2tp_sock_to_session(sock->sk);
1814 if (session == NULL) {
1815 error = -EBADF;
1816 goto end;
1819 sp.sa_family = AF_PPPOX;
1820 sp.sa_protocol = PX_PROTO_OL2TP;
1821 memcpy(&sp.pppol2tp, &session->tunnel_addr,
1822 sizeof(struct pppol2tp_addr));
1824 memcpy(uaddr, &sp, len);
1826 *usockaddr_len = len;
1828 error = 0;
1829 sock_put(sock->sk);
1831 end:
1832 return error;
1835 /****************************************************************************
1836 * ioctl() handlers.
1838 * The PPPoX socket is created for L2TP sessions: tunnels have their own UDP
1839 * sockets. However, in order to control kernel tunnel features, we allow
1840 * userspace to create a special "tunnel" PPPoX socket which is used for
1841 * control only. Tunnel PPPoX sockets have session_id == 0 and simply allow
1842 * the user application to issue L2TP setsockopt(), getsockopt() and ioctl()
1843 * calls.
1844 ****************************************************************************/
1846 /* Session ioctl helper.
1848 static int pppol2tp_session_ioctl(struct pppol2tp_session *session,
1849 unsigned int cmd, unsigned long arg)
1851 struct ifreq ifr;
1852 int err = 0;
1853 struct sock *sk = session->sock;
1854 int val = (int) arg;
1856 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_DEBUG,
1857 "%s: pppol2tp_session_ioctl(cmd=%#x, arg=%#lx)\n",
1858 session->name, cmd, arg);
1860 sock_hold(sk);
1862 switch (cmd) {
1863 case SIOCGIFMTU:
1864 err = -ENXIO;
1865 if (!(sk->sk_state & PPPOX_CONNECTED))
1866 break;
1868 err = -EFAULT;
1869 if (copy_from_user(&ifr, (void __user *) arg, sizeof(struct ifreq)))
1870 break;
1871 ifr.ifr_mtu = session->mtu;
1872 if (copy_to_user((void __user *) arg, &ifr, sizeof(struct ifreq)))
1873 break;
1875 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1876 "%s: get mtu=%d\n", session->name, session->mtu);
1877 err = 0;
1878 break;
1880 case SIOCSIFMTU:
1881 err = -ENXIO;
1882 if (!(sk->sk_state & PPPOX_CONNECTED))
1883 break;
1885 err = -EFAULT;
1886 if (copy_from_user(&ifr, (void __user *) arg, sizeof(struct ifreq)))
1887 break;
1889 session->mtu = ifr.ifr_mtu;
1891 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1892 "%s: set mtu=%d\n", session->name, session->mtu);
1893 err = 0;
1894 break;
1896 case PPPIOCGMRU:
1897 err = -ENXIO;
1898 if (!(sk->sk_state & PPPOX_CONNECTED))
1899 break;
1901 err = -EFAULT;
1902 if (put_user(session->mru, (int __user *) arg))
1903 break;
1905 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1906 "%s: get mru=%d\n", session->name, session->mru);
1907 err = 0;
1908 break;
1910 case PPPIOCSMRU:
1911 err = -ENXIO;
1912 if (!(sk->sk_state & PPPOX_CONNECTED))
1913 break;
1915 err = -EFAULT;
1916 if (get_user(val,(int __user *) arg))
1917 break;
1919 session->mru = val;
1920 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1921 "%s: set mru=%d\n", session->name, session->mru);
1922 err = 0;
1923 break;
1925 case PPPIOCGFLAGS:
1926 err = -EFAULT;
1927 if (put_user(session->flags, (int __user *) arg))
1928 break;
1930 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1931 "%s: get flags=%d\n", session->name, session->flags);
1932 err = 0;
1933 break;
1935 case PPPIOCSFLAGS:
1936 err = -EFAULT;
1937 if (get_user(val, (int __user *) arg))
1938 break;
1939 session->flags = val;
1940 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1941 "%s: set flags=%d\n", session->name, session->flags);
1942 err = 0;
1943 break;
1945 case PPPIOCGL2TPSTATS:
1946 err = -ENXIO;
1947 if (!(sk->sk_state & PPPOX_CONNECTED))
1948 break;
1950 if (copy_to_user((void __user *) arg, &session->stats,
1951 sizeof(session->stats)))
1952 break;
1953 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
1954 "%s: get L2TP stats\n", session->name);
1955 err = 0;
1956 break;
1958 default:
1959 err = -ENOSYS;
1960 break;
1963 sock_put(sk);
1965 return err;
1968 /* Tunnel ioctl helper.
1970 * Note the special handling for PPPIOCGL2TPSTATS below. If the ioctl data
1971 * specifies a session_id, the session ioctl handler is called. This allows an
1972 * application to retrieve session stats via a tunnel socket.
1974 static int pppol2tp_tunnel_ioctl(struct pppol2tp_tunnel *tunnel,
1975 unsigned int cmd, unsigned long arg)
1977 int err = 0;
1978 struct sock *sk = tunnel->sock;
1979 struct pppol2tp_ioc_stats stats_req;
1981 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_DEBUG,
1982 "%s: pppol2tp_tunnel_ioctl(cmd=%#x, arg=%#lx)\n", tunnel->name,
1983 cmd, arg);
1985 sock_hold(sk);
1987 switch (cmd) {
1988 case PPPIOCGL2TPSTATS:
1989 err = -ENXIO;
1990 if (!(sk->sk_state & PPPOX_CONNECTED))
1991 break;
1993 if (copy_from_user(&stats_req, (void __user *) arg,
1994 sizeof(stats_req))) {
1995 err = -EFAULT;
1996 break;
1998 if (stats_req.session_id != 0) {
1999 /* resend to session ioctl handler */
2000 struct pppol2tp_session *session =
2001 pppol2tp_session_find(tunnel, stats_req.session_id);
2002 if (session != NULL)
2003 err = pppol2tp_session_ioctl(session, cmd, arg);
2004 else
2005 err = -EBADR;
2006 break;
2008 #ifdef CONFIG_XFRM
2009 tunnel->stats.using_ipsec = (sk->sk_policy[0] || sk->sk_policy[1]) ? 1 : 0;
2010 #endif
2011 if (copy_to_user((void __user *) arg, &tunnel->stats,
2012 sizeof(tunnel->stats))) {
2013 err = -EFAULT;
2014 break;
2016 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2017 "%s: get L2TP stats\n", tunnel->name);
2018 err = 0;
2019 break;
2021 default:
2022 err = -ENOSYS;
2023 break;
2026 sock_put(sk);
2028 return err;
2031 /* Main ioctl() handler.
2032 * Dispatch to tunnel or session helpers depending on the socket.
2034 static int pppol2tp_ioctl(struct socket *sock, unsigned int cmd,
2035 unsigned long arg)
2037 struct sock *sk = sock->sk;
2038 struct pppol2tp_session *session;
2039 struct pppol2tp_tunnel *tunnel;
2040 int err;
2042 if (!sk)
2043 return 0;
2045 err = -EBADF;
2046 if (sock_flag(sk, SOCK_DEAD) != 0)
2047 goto end;
2049 err = -ENOTCONN;
2050 if ((sk->sk_user_data == NULL) ||
2051 (!(sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND))))
2052 goto end;
2054 /* Get session context from the socket */
2055 err = -EBADF;
2056 session = pppol2tp_sock_to_session(sk);
2057 if (session == NULL)
2058 goto end;
2060 /* Special case: if session's session_id is zero, treat ioctl as a
2061 * tunnel ioctl
2063 if ((session->tunnel_addr.s_session == 0) &&
2064 (session->tunnel_addr.d_session == 0)) {
2065 err = -EBADF;
2066 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2067 if (tunnel == NULL)
2068 goto end_put_sess;
2070 err = pppol2tp_tunnel_ioctl(tunnel, cmd, arg);
2071 sock_put(session->tunnel_sock);
2072 goto end_put_sess;
2075 err = pppol2tp_session_ioctl(session, cmd, arg);
2077 end_put_sess:
2078 sock_put(sk);
2079 end:
2080 return err;
2083 /*****************************************************************************
2084 * setsockopt() / getsockopt() support.
2086 * The PPPoX socket is created for L2TP sessions: tunnels have their own UDP
2087 * sockets. In order to control kernel tunnel features, we allow userspace to
2088 * create a special "tunnel" PPPoX socket which is used for control only.
2089 * Tunnel PPPoX sockets have session_id == 0 and simply allow the user
2090 * application to issue L2TP setsockopt(), getsockopt() and ioctl() calls.
2091 *****************************************************************************/
2093 /* Tunnel setsockopt() helper.
2095 static int pppol2tp_tunnel_setsockopt(struct sock *sk,
2096 struct pppol2tp_tunnel *tunnel,
2097 int optname, int val)
2099 int err = 0;
2101 switch (optname) {
2102 case PPPOL2TP_SO_DEBUG:
2103 tunnel->debug = val;
2104 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2105 "%s: set debug=%x\n", tunnel->name, tunnel->debug);
2106 break;
2108 default:
2109 err = -ENOPROTOOPT;
2110 break;
2113 return err;
2116 /* Session setsockopt helper.
2118 static int pppol2tp_session_setsockopt(struct sock *sk,
2119 struct pppol2tp_session *session,
2120 int optname, int val)
2122 int err = 0;
2124 switch (optname) {
2125 case PPPOL2TP_SO_RECVSEQ:
2126 if ((val != 0) && (val != 1)) {
2127 err = -EINVAL;
2128 break;
2130 session->recv_seq = val ? -1 : 0;
2131 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2132 "%s: set recv_seq=%d\n", session->name,
2133 session->recv_seq);
2134 break;
2136 case PPPOL2TP_SO_SENDSEQ:
2137 if ((val != 0) && (val != 1)) {
2138 err = -EINVAL;
2139 break;
2141 session->send_seq = val ? -1 : 0;
2143 struct sock *ssk = session->sock;
2144 struct pppox_sock *po = pppox_sk(ssk);
2145 po->chan.hdrlen = val ? PPPOL2TP_L2TP_HDR_SIZE_SEQ :
2146 PPPOL2TP_L2TP_HDR_SIZE_NOSEQ;
2148 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2149 "%s: set send_seq=%d\n", session->name, session->send_seq);
2150 break;
2152 case PPPOL2TP_SO_LNSMODE:
2153 if ((val != 0) && (val != 1)) {
2154 err = -EINVAL;
2155 break;
2157 session->lns_mode = val ? -1 : 0;
2158 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2159 "%s: set lns_mode=%d\n", session->name,
2160 session->lns_mode);
2161 break;
2163 case PPPOL2TP_SO_DEBUG:
2164 session->debug = val;
2165 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2166 "%s: set debug=%x\n", session->name, session->debug);
2167 break;
2169 case PPPOL2TP_SO_REORDERTO:
2170 session->reorder_timeout = msecs_to_jiffies(val);
2171 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2172 "%s: set reorder_timeout=%d\n", session->name,
2173 session->reorder_timeout);
2174 break;
2176 default:
2177 err = -ENOPROTOOPT;
2178 break;
2181 return err;
2184 /* Main setsockopt() entry point.
2185 * Does API checks, then calls either the tunnel or session setsockopt
2186 * handler, according to whether the PPPoL2TP socket is a for a regular
2187 * session or the special tunnel type.
2189 static int pppol2tp_setsockopt(struct socket *sock, int level, int optname,
2190 char __user *optval, unsigned int optlen)
2192 struct sock *sk = sock->sk;
2193 struct pppol2tp_session *session = sk->sk_user_data;
2194 struct pppol2tp_tunnel *tunnel;
2195 int val;
2196 int err;
2198 if (level != SOL_PPPOL2TP)
2199 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
2201 if (optlen < sizeof(int))
2202 return -EINVAL;
2204 if (get_user(val, (int __user *)optval))
2205 return -EFAULT;
2207 err = -ENOTCONN;
2208 if (sk->sk_user_data == NULL)
2209 goto end;
2211 /* Get session context from the socket */
2212 err = -EBADF;
2213 session = pppol2tp_sock_to_session(sk);
2214 if (session == NULL)
2215 goto end;
2217 /* Special case: if session_id == 0x0000, treat as operation on tunnel
2219 if ((session->tunnel_addr.s_session == 0) &&
2220 (session->tunnel_addr.d_session == 0)) {
2221 err = -EBADF;
2222 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2223 if (tunnel == NULL)
2224 goto end_put_sess;
2226 err = pppol2tp_tunnel_setsockopt(sk, tunnel, optname, val);
2227 sock_put(session->tunnel_sock);
2228 } else
2229 err = pppol2tp_session_setsockopt(sk, session, optname, val);
2231 err = 0;
2233 end_put_sess:
2234 sock_put(sk);
2235 end:
2236 return err;
2239 /* Tunnel getsockopt helper. Called with sock locked.
2241 static int pppol2tp_tunnel_getsockopt(struct sock *sk,
2242 struct pppol2tp_tunnel *tunnel,
2243 int optname, int *val)
2245 int err = 0;
2247 switch (optname) {
2248 case PPPOL2TP_SO_DEBUG:
2249 *val = tunnel->debug;
2250 PRINTK(tunnel->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2251 "%s: get debug=%x\n", tunnel->name, tunnel->debug);
2252 break;
2254 default:
2255 err = -ENOPROTOOPT;
2256 break;
2259 return err;
2262 /* Session getsockopt helper. Called with sock locked.
2264 static int pppol2tp_session_getsockopt(struct sock *sk,
2265 struct pppol2tp_session *session,
2266 int optname, int *val)
2268 int err = 0;
2270 switch (optname) {
2271 case PPPOL2TP_SO_RECVSEQ:
2272 *val = session->recv_seq;
2273 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2274 "%s: get recv_seq=%d\n", session->name, *val);
2275 break;
2277 case PPPOL2TP_SO_SENDSEQ:
2278 *val = session->send_seq;
2279 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2280 "%s: get send_seq=%d\n", session->name, *val);
2281 break;
2283 case PPPOL2TP_SO_LNSMODE:
2284 *val = session->lns_mode;
2285 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2286 "%s: get lns_mode=%d\n", session->name, *val);
2287 break;
2289 case PPPOL2TP_SO_DEBUG:
2290 *val = session->debug;
2291 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2292 "%s: get debug=%d\n", session->name, *val);
2293 break;
2295 case PPPOL2TP_SO_REORDERTO:
2296 *val = (int) jiffies_to_msecs(session->reorder_timeout);
2297 PRINTK(session->debug, PPPOL2TP_MSG_CONTROL, KERN_INFO,
2298 "%s: get reorder_timeout=%d\n", session->name, *val);
2299 break;
2301 default:
2302 err = -ENOPROTOOPT;
2305 return err;
2308 /* Main getsockopt() entry point.
2309 * Does API checks, then calls either the tunnel or session getsockopt
2310 * handler, according to whether the PPPoX socket is a for a regular session
2311 * or the special tunnel type.
2313 static int pppol2tp_getsockopt(struct socket *sock, int level,
2314 int optname, char __user *optval, int __user *optlen)
2316 struct sock *sk = sock->sk;
2317 struct pppol2tp_session *session = sk->sk_user_data;
2318 struct pppol2tp_tunnel *tunnel;
2319 int val, len;
2320 int err;
2322 if (level != SOL_PPPOL2TP)
2323 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
2325 if (get_user(len, (int __user *) optlen))
2326 return -EFAULT;
2328 len = min_t(unsigned int, len, sizeof(int));
2330 if (len < 0)
2331 return -EINVAL;
2333 err = -ENOTCONN;
2334 if (sk->sk_user_data == NULL)
2335 goto end;
2337 /* Get the session context */
2338 err = -EBADF;
2339 session = pppol2tp_sock_to_session(sk);
2340 if (session == NULL)
2341 goto end;
2343 /* Special case: if session_id == 0x0000, treat as operation on tunnel */
2344 if ((session->tunnel_addr.s_session == 0) &&
2345 (session->tunnel_addr.d_session == 0)) {
2346 err = -EBADF;
2347 tunnel = pppol2tp_sock_to_tunnel(session->tunnel_sock);
2348 if (tunnel == NULL)
2349 goto end_put_sess;
2351 err = pppol2tp_tunnel_getsockopt(sk, tunnel, optname, &val);
2352 sock_put(session->tunnel_sock);
2353 } else
2354 err = pppol2tp_session_getsockopt(sk, session, optname, &val);
2356 err = -EFAULT;
2357 if (put_user(len, (int __user *) optlen))
2358 goto end_put_sess;
2360 if (copy_to_user((void __user *) optval, &val, len))
2361 goto end_put_sess;
2363 err = 0;
2365 end_put_sess:
2366 sock_put(sk);
2367 end:
2368 return err;
2371 /*****************************************************************************
2372 * /proc filesystem for debug
2373 *****************************************************************************/
2375 #ifdef CONFIG_PROC_FS
2377 #include <linux/seq_file.h>
2379 struct pppol2tp_seq_data {
2380 struct seq_net_private p;
2381 struct pppol2tp_tunnel *tunnel; /* current tunnel */
2382 struct pppol2tp_session *session; /* NULL means get first session in tunnel */
2385 static struct pppol2tp_session *next_session(struct pppol2tp_tunnel *tunnel, struct pppol2tp_session *curr)
2387 struct pppol2tp_session *session = NULL;
2388 struct hlist_node *walk;
2389 int found = 0;
2390 int next = 0;
2391 int i;
2393 read_lock_bh(&tunnel->hlist_lock);
2394 for (i = 0; i < PPPOL2TP_HASH_SIZE; i++) {
2395 hlist_for_each_entry(session, walk, &tunnel->session_hlist[i], hlist) {
2396 if (curr == NULL) {
2397 found = 1;
2398 goto out;
2400 if (session == curr) {
2401 next = 1;
2402 continue;
2404 if (next) {
2405 found = 1;
2406 goto out;
2410 out:
2411 read_unlock_bh(&tunnel->hlist_lock);
2412 if (!found)
2413 session = NULL;
2415 return session;
2418 static struct pppol2tp_tunnel *next_tunnel(struct pppol2tp_net *pn,
2419 struct pppol2tp_tunnel *curr)
2421 struct pppol2tp_tunnel *tunnel = NULL;
2423 read_lock_bh(&pn->pppol2tp_tunnel_list_lock);
2424 if (list_is_last(&curr->list, &pn->pppol2tp_tunnel_list)) {
2425 goto out;
2427 tunnel = list_entry(curr->list.next, struct pppol2tp_tunnel, list);
2428 out:
2429 read_unlock_bh(&pn->pppol2tp_tunnel_list_lock);
2431 return tunnel;
2434 static void *pppol2tp_seq_start(struct seq_file *m, loff_t *offs)
2436 struct pppol2tp_seq_data *pd = SEQ_START_TOKEN;
2437 struct pppol2tp_net *pn;
2438 loff_t pos = *offs;
2440 if (!pos)
2441 goto out;
2443 BUG_ON(m->private == NULL);
2444 pd = m->private;
2445 pn = pppol2tp_pernet(seq_file_net(m));
2447 if (pd->tunnel == NULL) {
2448 if (!list_empty(&pn->pppol2tp_tunnel_list))
2449 pd->tunnel = list_entry(pn->pppol2tp_tunnel_list.next, struct pppol2tp_tunnel, list);
2450 } else {
2451 pd->session = next_session(pd->tunnel, pd->session);
2452 if (pd->session == NULL) {
2453 pd->tunnel = next_tunnel(pn, pd->tunnel);
2457 /* NULL tunnel and session indicates end of list */
2458 if ((pd->tunnel == NULL) && (pd->session == NULL))
2459 pd = NULL;
2461 out:
2462 return pd;
2465 static void *pppol2tp_seq_next(struct seq_file *m, void *v, loff_t *pos)
2467 (*pos)++;
2468 return NULL;
2471 static void pppol2tp_seq_stop(struct seq_file *p, void *v)
2473 /* nothing to do */
2476 static void pppol2tp_seq_tunnel_show(struct seq_file *m, void *v)
2478 struct pppol2tp_tunnel *tunnel = v;
2480 seq_printf(m, "\nTUNNEL '%s', %c %d\n",
2481 tunnel->name,
2482 (tunnel == tunnel->sock->sk_user_data) ? 'Y':'N',
2483 atomic_read(&tunnel->ref_count) - 1);
2484 seq_printf(m, " %08x %llu/%llu/%llu %llu/%llu/%llu\n",
2485 tunnel->debug,
2486 (unsigned long long)tunnel->stats.tx_packets,
2487 (unsigned long long)tunnel->stats.tx_bytes,
2488 (unsigned long long)tunnel->stats.tx_errors,
2489 (unsigned long long)tunnel->stats.rx_packets,
2490 (unsigned long long)tunnel->stats.rx_bytes,
2491 (unsigned long long)tunnel->stats.rx_errors);
2494 static void pppol2tp_seq_session_show(struct seq_file *m, void *v)
2496 struct pppol2tp_session *session = v;
2498 seq_printf(m, " SESSION '%s' %08X/%d %04X/%04X -> "
2499 "%04X/%04X %d %c\n",
2500 session->name,
2501 ntohl(session->tunnel_addr.addr.sin_addr.s_addr),
2502 ntohs(session->tunnel_addr.addr.sin_port),
2503 session->tunnel_addr.s_tunnel,
2504 session->tunnel_addr.s_session,
2505 session->tunnel_addr.d_tunnel,
2506 session->tunnel_addr.d_session,
2507 session->sock->sk_state,
2508 (session == session->sock->sk_user_data) ?
2509 'Y' : 'N');
2510 seq_printf(m, " %d/%d/%c/%c/%s %08x %u\n",
2511 session->mtu, session->mru,
2512 session->recv_seq ? 'R' : '-',
2513 session->send_seq ? 'S' : '-',
2514 session->lns_mode ? "LNS" : "LAC",
2515 session->debug,
2516 jiffies_to_msecs(session->reorder_timeout));
2517 seq_printf(m, " %hu/%hu %llu/%llu/%llu %llu/%llu/%llu\n",
2518 session->nr, session->ns,
2519 (unsigned long long)session->stats.tx_packets,
2520 (unsigned long long)session->stats.tx_bytes,
2521 (unsigned long long)session->stats.tx_errors,
2522 (unsigned long long)session->stats.rx_packets,
2523 (unsigned long long)session->stats.rx_bytes,
2524 (unsigned long long)session->stats.rx_errors);
2527 static int pppol2tp_seq_show(struct seq_file *m, void *v)
2529 struct pppol2tp_seq_data *pd = v;
2531 /* display header on line 1 */
2532 if (v == SEQ_START_TOKEN) {
2533 seq_puts(m, "PPPoL2TP driver info, " PPPOL2TP_DRV_VERSION "\n");
2534 seq_puts(m, "TUNNEL name, user-data-ok session-count\n");
2535 seq_puts(m, " debug tx-pkts/bytes/errs rx-pkts/bytes/errs\n");
2536 seq_puts(m, " SESSION name, addr/port src-tid/sid "
2537 "dest-tid/sid state user-data-ok\n");
2538 seq_puts(m, " mtu/mru/rcvseq/sendseq/lns debug reorderto\n");
2539 seq_puts(m, " nr/ns tx-pkts/bytes/errs rx-pkts/bytes/errs\n");
2540 goto out;
2543 /* Show the tunnel or session context.
2545 if (pd->session == NULL)
2546 pppol2tp_seq_tunnel_show(m, pd->tunnel);
2547 else
2548 pppol2tp_seq_session_show(m, pd->session);
2550 out:
2551 return 0;
2554 static const struct seq_operations pppol2tp_seq_ops = {
2555 .start = pppol2tp_seq_start,
2556 .next = pppol2tp_seq_next,
2557 .stop = pppol2tp_seq_stop,
2558 .show = pppol2tp_seq_show,
2561 /* Called when our /proc file is opened. We allocate data for use when
2562 * iterating our tunnel / session contexts and store it in the private
2563 * data of the seq_file.
2565 static int pppol2tp_proc_open(struct inode *inode, struct file *file)
2567 return seq_open_net(inode, file, &pppol2tp_seq_ops,
2568 sizeof(struct pppol2tp_seq_data));
2571 static const struct file_operations pppol2tp_proc_fops = {
2572 .owner = THIS_MODULE,
2573 .open = pppol2tp_proc_open,
2574 .read = seq_read,
2575 .llseek = seq_lseek,
2576 .release = seq_release_net,
2579 #endif /* CONFIG_PROC_FS */
2581 /*****************************************************************************
2582 * Init and cleanup
2583 *****************************************************************************/
2585 static const struct proto_ops pppol2tp_ops = {
2586 .family = AF_PPPOX,
2587 .owner = THIS_MODULE,
2588 .release = pppol2tp_release,
2589 .bind = sock_no_bind,
2590 .connect = pppol2tp_connect,
2591 .socketpair = sock_no_socketpair,
2592 .accept = sock_no_accept,
2593 .getname = pppol2tp_getname,
2594 .poll = datagram_poll,
2595 .listen = sock_no_listen,
2596 .shutdown = sock_no_shutdown,
2597 .setsockopt = pppol2tp_setsockopt,
2598 .getsockopt = pppol2tp_getsockopt,
2599 .sendmsg = pppol2tp_sendmsg,
2600 .recvmsg = pppol2tp_recvmsg,
2601 .mmap = sock_no_mmap,
2602 .ioctl = pppox_ioctl,
2605 static struct pppox_proto pppol2tp_proto = {
2606 .create = pppol2tp_create,
2607 .ioctl = pppol2tp_ioctl
2610 static __net_init int pppol2tp_init_net(struct net *net)
2612 struct pppol2tp_net *pn = pppol2tp_pernet(net);
2613 struct proc_dir_entry *pde;
2615 INIT_LIST_HEAD(&pn->pppol2tp_tunnel_list);
2616 rwlock_init(&pn->pppol2tp_tunnel_list_lock);
2618 pde = proc_net_fops_create(net, "pppol2tp", S_IRUGO, &pppol2tp_proc_fops);
2619 #ifdef CONFIG_PROC_FS
2620 if (!pde)
2621 return -ENOMEM;
2622 #endif
2624 return 0;
2627 static __net_exit void pppol2tp_exit_net(struct net *net)
2629 proc_net_remove(net, "pppol2tp");
2632 static struct pernet_operations pppol2tp_net_ops = {
2633 .init = pppol2tp_init_net,
2634 .exit = pppol2tp_exit_net,
2635 .id = &pppol2tp_net_id,
2636 .size = sizeof(struct pppol2tp_net),
2639 static int __init pppol2tp_init(void)
2641 int err;
2643 err = proto_register(&pppol2tp_sk_proto, 0);
2644 if (err)
2645 goto out;
2646 err = register_pppox_proto(PX_PROTO_OL2TP, &pppol2tp_proto);
2647 if (err)
2648 goto out_unregister_pppol2tp_proto;
2650 err = register_pernet_device(&pppol2tp_net_ops);
2651 if (err)
2652 goto out_unregister_pppox_proto;
2654 printk(KERN_INFO "PPPoL2TP kernel driver, %s\n",
2655 PPPOL2TP_DRV_VERSION);
2657 out:
2658 return err;
2659 out_unregister_pppox_proto:
2660 unregister_pppox_proto(PX_PROTO_OL2TP);
2661 out_unregister_pppol2tp_proto:
2662 proto_unregister(&pppol2tp_sk_proto);
2663 goto out;
2666 static void __exit pppol2tp_exit(void)
2668 unregister_pppox_proto(PX_PROTO_OL2TP);
2669 unregister_pernet_device(&pppol2tp_net_ops);
2670 proto_unregister(&pppol2tp_sk_proto);
2673 module_init(pppol2tp_init);
2674 module_exit(pppol2tp_exit);
2676 MODULE_AUTHOR("Martijn van Oosterhout <kleptog@svana.org>, "
2677 "James Chapman <jchapman@katalix.com>");
2678 MODULE_DESCRIPTION("PPP over L2TP over UDP");
2679 MODULE_LICENSE("GPL");
2680 MODULE_VERSION(PPPOL2TP_DRV_VERSION);