1/** -*- linux-c -*- ***********************************************************
2 * Linux PPP over Ethernet (PPPoX/PPPoE) Sockets
3 *
4 * PPPoX --- Generic PPP encapsulation socket family
5 * PPPoE --- PPP over Ethernet (RFC 2516)
6 *
7 *
8 * Version:	0.7.0
9 *
10 * 070228 :	Fix to allow multiple sessions with same remote MAC and same
11 *		session id by including the local device ifindex in the
12 *		tuple identifying a session. This also ensures packets can't
13 *		be injected into a session from interfaces other than the one
14 *		specified by userspace. Florian Zumbiehl <florz@florz.de>
15 *		(Oh, BTW, this one is YYMMDD, in case you were wondering ...)
16 * 220102 :	Fix module use count on failure in pppoe_create, pppox_sk -acme
17 * 030700 :	Fixed connect logic to allow for disconnect.
18 * 270700 :	Fixed potential SMP problems; we must protect against
19 *		simultaneous invocation of ppp_input
20 *		and ppp_unregister_channel.
21 * 040800 :	Respect reference count mechanisms on net-devices.
22 * 200800 :	fix kfree(skb) in pppoe_rcv (acme)
23 *		Module reference count is decremented in the right spot now,
24 *		guards against sock_put not actually freeing the sk
25 *		in pppoe_release.
26 * 051000 :	Initialization cleanup.
27 * 111100 :	Fix recvmsg.
28 * 050101 :	Fix PADT procesing.
29 * 140501 :	Use pppoe_rcv_core to handle all backlog. (Alexey)
30 * 170701 :	Do not lock_sock with rwlock held. (DaveM)
31 *		Ignore discovery frames if user has socket
32 *		locked. (DaveM)
33 *		Ignore return value of dev_queue_xmit in __pppoe_xmit
34 *		or else we may kfree an SKB twice. (DaveM)
35 * 190701 :	When doing copies of skb's in __pppoe_xmit, always delete
36 *		the original skb that was passed in on success, never on
37 *		failure.  Delete the copy of the skb on failure to avoid
38 *		a memory leak.
39 * 081001 :	Misc. cleanup (licence string, non-blocking, prevent
40 *		reference of device on close).
41 * 121301 :	New ppp channels interface; cannot unregister a channel
42 *		from interrupts.  Thus, we mark the socket as a ZOMBIE
43 *		and do the unregistration later.
44 * 081002 :	seq_file support for proc stuff -acme
45 * 111602 :	Merge all 2.4 fixes into 2.5/2.6 tree.  Label 2.5/2.6
46 *		as version 0.7.  Spacing cleanup.
47 * Author:	Michal Ostrowski <mostrows@speakeasy.net>
48 * Contributors:
49 * 		Arnaldo Carvalho de Melo <acme@conectiva.com.br>
50 *		David S. Miller (davem@redhat.com)
51 *
52 * License:
53 *		This program is free software; you can redistribute it and/or
54 *		modify it under the terms of the GNU General Public License
55 *		as published by the Free Software Foundation; either version
56 *		2 of the License, or (at your option) any later version.
57 *
58 */
59
60#include <linux/string.h>
61#include <linux/module.h>
62#include <linux/kernel.h>
63#include <linux/slab.h>
64#include <linux/errno.h>
65#include <linux/netdevice.h>
66#include <linux/net.h>
67#include <linux/inetdevice.h>
68#include <linux/etherdevice.h>
69#include <linux/skbuff.h>
70#include <linux/init.h>
71#include <linux/if_ether.h>
72#include <linux/if_pppox.h>
73#include <linux/ppp_channel.h>
74#include <linux/ppp_defs.h>
75#include <linux/ppp-ioctl.h>
76#include <linux/notifier.h>
77#include <linux/file.h>
78#include <linux/proc_fs.h>
79#include <linux/seq_file.h>
80
81#include <linux/nsproxy.h>
82#include <net/net_namespace.h>
83#include <net/netns/generic.h>
84#include <net/sock.h>
85
86#include <asm/uaccess.h>
87
88#define PPPOE_HASH_BITS 4
89#define PPPOE_HASH_SIZE (1 << PPPOE_HASH_BITS)
90#define PPPOE_HASH_MASK	(PPPOE_HASH_SIZE - 1)
91
92static int __pppoe_xmit(struct sock *sk, struct sk_buff *skb);
93
94static const struct proto_ops pppoe_ops;
95static const struct ppp_channel_ops pppoe_chan_ops;
96
97/* per-net private data for this module */
98static int pppoe_net_id __read_mostly;
99struct pppoe_net {
100	/*
101	 * we could use _single_ hash table for all
102	 * nets by injecting net id into the hash but
103	 * it would increase hash chains and add
104	 * a few additional math comparations messy
105	 * as well, moreover in case of SMP less locking
106	 * controversy here
107	 */
108	struct pppox_sock *hash_table[PPPOE_HASH_SIZE];
109	rwlock_t hash_lock;
110};
111
112/*
113 * PPPoE could be in the following stages:
114 * 1) Discovery stage (to obtain remote MAC and Session ID)
115 * 2) Session stage (MAC and SID are known)
116 *
117 * Ethernet frames have a special tag for this but
118 * we use simpler approach based on session id
119 */
120static inline bool stage_session(__be16 sid)
121{
122	return sid != 0;
123}
124
125static inline struct pppoe_net *pppoe_pernet(struct net *net)
126{
127	BUG_ON(!net);
128
129	return net_generic(net, pppoe_net_id);
130}
131
132static inline int cmp_2_addr(struct pppoe_addr *a, struct pppoe_addr *b)
133{
134	return a->sid == b->sid && ether_addr_equal(a->remote, b->remote);
135}
136
137static inline int cmp_addr(struct pppoe_addr *a, __be16 sid, char *addr)
138{
139	return a->sid == sid && ether_addr_equal(a->remote, addr);
140}
141
142#if 8 % PPPOE_HASH_BITS
143#error 8 must be a multiple of PPPOE_HASH_BITS
144#endif
145
146static int hash_item(__be16 sid, unsigned char *addr)
147{
148	unsigned char hash = 0;
149	unsigned int i;
150
151	for (i = 0; i < ETH_ALEN; i++)
152		hash ^= addr[i];
153	for (i = 0; i < sizeof(sid_t) * 8; i += 8)
154		hash ^= (__force __u32)sid >> i;
155	for (i = 8; (i >>= 1) >= PPPOE_HASH_BITS;)
156		hash ^= hash >> i;
157
158	return hash & PPPOE_HASH_MASK;
159}
160
161/**********************************************************************
162 *
163 *  Set/get/delete/rehash items  (internal versions)
164 *
165 **********************************************************************/
166static struct pppox_sock *__get_item(struct pppoe_net *pn, __be16 sid,
167				unsigned char *addr, int ifindex)
168{
169	int hash = hash_item(sid, addr);
170	struct pppox_sock *ret;
171
172	ret = pn->hash_table[hash];
173	while (ret) {
174		if (cmp_addr(&ret->pppoe_pa, sid, addr) &&
175		    ret->pppoe_ifindex == ifindex)
176			return ret;
177
178		ret = ret->next;
179	}
180
181	return NULL;
182}
183
184static int __set_item(struct pppoe_net *pn, struct pppox_sock *po)
185{
186	int hash = hash_item(po->pppoe_pa.sid, po->pppoe_pa.remote);
187	struct pppox_sock *ret;
188
189	ret = pn->hash_table[hash];
190	while (ret) {
191		if (cmp_2_addr(&ret->pppoe_pa, &po->pppoe_pa) &&
192		    ret->pppoe_ifindex == po->pppoe_ifindex)
193			return -EALREADY;
194
195		ret = ret->next;
196	}
197
198	po->next = pn->hash_table[hash];
199	pn->hash_table[hash] = po;
200
201	return 0;
202}
203
204static void __delete_item(struct pppoe_net *pn, __be16 sid,
205					char *addr, int ifindex)
206{
207	int hash = hash_item(sid, addr);
208	struct pppox_sock *ret, **src;
209
210	ret = pn->hash_table[hash];
211	src = &pn->hash_table[hash];
212
213	while (ret) {
214		if (cmp_addr(&ret->pppoe_pa, sid, addr) &&
215		    ret->pppoe_ifindex == ifindex) {
216			*src = ret->next;
217			break;
218		}
219
220		src = &ret->next;
221		ret = ret->next;
222	}
223}
224
225/**********************************************************************
226 *
227 *  Set/get/delete/rehash items
228 *
229 **********************************************************************/
230static inline struct pppox_sock *get_item(struct pppoe_net *pn, __be16 sid,
231					unsigned char *addr, int ifindex)
232{
233	struct pppox_sock *po;
234
235	read_lock_bh(&pn->hash_lock);
236	po = __get_item(pn, sid, addr, ifindex);
237	if (po)
238		sock_hold(sk_pppox(po));
239	read_unlock_bh(&pn->hash_lock);
240
241	return po;
242}
243
244static inline struct pppox_sock *get_item_by_addr(struct net *net,
245						struct sockaddr_pppox *sp)
246{
247	struct net_device *dev;
248	struct pppoe_net *pn;
249	struct pppox_sock *pppox_sock = NULL;
250
251	int ifindex;
252
253	rcu_read_lock();
254	dev = dev_get_by_name_rcu(net, sp->sa_addr.pppoe.dev);
255	if (dev) {
256		ifindex = dev->ifindex;
257		pn = pppoe_pernet(net);
258		pppox_sock = get_item(pn, sp->sa_addr.pppoe.sid,
259				sp->sa_addr.pppoe.remote, ifindex);
260	}
261	rcu_read_unlock();
262	return pppox_sock;
263}
264
265static inline void delete_item(struct pppoe_net *pn, __be16 sid,
266					char *addr, int ifindex)
267{
268	write_lock_bh(&pn->hash_lock);
269	__delete_item(pn, sid, addr, ifindex);
270	write_unlock_bh(&pn->hash_lock);
271}
272
273/***************************************************************************
274 *
275 *  Handler for device events.
276 *  Certain device events require that sockets be unconnected.
277 *
278 **************************************************************************/
279
280static void pppoe_flush_dev(struct net_device *dev)
281{
282	struct pppoe_net *pn;
283	int i;
284
285	pn = pppoe_pernet(dev_net(dev));
286	write_lock_bh(&pn->hash_lock);
287	for (i = 0; i < PPPOE_HASH_SIZE; i++) {
288		struct pppox_sock *po = pn->hash_table[i];
289		struct sock *sk;
290
291		while (po) {
292			while (po && po->pppoe_dev != dev) {
293				po = po->next;
294			}
295
296			if (!po)
297				break;
298
299			sk = sk_pppox(po);
300
301			/* We always grab the socket lock, followed by the
302			 * hash_lock, in that order.  Since we should hold the
303			 * sock lock while doing any unbinding, we need to
304			 * release the lock we're holding.  Hold a reference to
305			 * the sock so it doesn't disappear as we're jumping
306			 * between locks.
307			 */
308
309			sock_hold(sk);
310			write_unlock_bh(&pn->hash_lock);
311			lock_sock(sk);
312
313			if (po->pppoe_dev == dev &&
314			    sk->sk_state & (PPPOX_CONNECTED | PPPOX_BOUND | PPPOX_ZOMBIE)) {
315				pppox_unbind_sock(sk);
316				sk->sk_state_change(sk);
317				po->pppoe_dev = NULL;
318				dev_put(dev);
319			}
320
321			release_sock(sk);
322			sock_put(sk);
323
324			/* Restart the process from the start of the current
325			 * hash chain. We dropped locks so the world may have
326			 * change from underneath us.
327			 */
328
329			BUG_ON(pppoe_pernet(dev_net(dev)) == NULL);
330			write_lock_bh(&pn->hash_lock);
331			po = pn->hash_table[i];
332		}
333	}
334	write_unlock_bh(&pn->hash_lock);
335}
336
337static int pppoe_device_event(struct notifier_block *this,
338			      unsigned long event, void *ptr)
339{
340	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
341
342	/* Only look at sockets that are using this specific device. */
343	switch (event) {
344	case NETDEV_CHANGEADDR:
345	case NETDEV_CHANGEMTU:
346		/* A change in mtu or address is a bad thing, requiring
347		 * LCP re-negotiation.
348		 */
349
350	case NETDEV_GOING_DOWN:
351	case NETDEV_DOWN:
352		/* Find every socket on this device and kill it. */
353		pppoe_flush_dev(dev);
354		break;
355
356	default:
357		break;
358	}
359
360	return NOTIFY_DONE;
361}
362
363static struct notifier_block pppoe_notifier = {
364	.notifier_call = pppoe_device_event,
365};
366
367/************************************************************************
368 *
369 * Do the real work of receiving a PPPoE Session frame.
370 *
371 ***********************************************************************/
372static int pppoe_rcv_core(struct sock *sk, struct sk_buff *skb)
373{
374	struct pppox_sock *po = pppox_sk(sk);
375	struct pppox_sock *relay_po;
376
377	/* Backlog receive. Semantics of backlog rcv preclude any code from
378	 * executing in lock_sock()/release_sock() bounds; meaning sk->sk_state
379	 * can't change.
380	 */
381
382	if (skb->pkt_type == PACKET_OTHERHOST)
383		goto abort_kfree;
384
385	if (sk->sk_state & PPPOX_BOUND) {
386		ppp_input(&po->chan, skb);
387	} else if (sk->sk_state & PPPOX_RELAY) {
388		relay_po = get_item_by_addr(sock_net(sk),
389					    &po->pppoe_relay);
390		if (relay_po == NULL)
391			goto abort_kfree;
392
393		if ((sk_pppox(relay_po)->sk_state & PPPOX_CONNECTED) == 0)
394			goto abort_put;
395
396		if (!__pppoe_xmit(sk_pppox(relay_po), skb))
397			goto abort_put;
398
399		sock_put(sk_pppox(relay_po));
400	} else {
401		if (sock_queue_rcv_skb(sk, skb))
402			goto abort_kfree;
403	}
404
405	return NET_RX_SUCCESS;
406
407abort_put:
408	sock_put(sk_pppox(relay_po));
409
410abort_kfree:
411	kfree_skb(skb);
412	return NET_RX_DROP;
413}
414
415/************************************************************************
416 *
417 * Receive wrapper called in BH context.
418 *
419 ***********************************************************************/
420static int pppoe_rcv(struct sk_buff *skb, struct net_device *dev,
421		     struct packet_type *pt, struct net_device *orig_dev)
422{
423	struct pppoe_hdr *ph;
424	struct pppox_sock *po;
425	struct pppoe_net *pn;
426	int len;
427
428	skb = skb_share_check(skb, GFP_ATOMIC);
429	if (!skb)
430		goto out;
431
432	if (!pskb_may_pull(skb, sizeof(struct pppoe_hdr)))
433		goto drop;
434
435	ph = pppoe_hdr(skb);
436	len = ntohs(ph->length);
437
438	skb_pull_rcsum(skb, sizeof(*ph));
439	if (skb->len < len)
440		goto drop;
441
442	if (pskb_trim_rcsum(skb, len))
443		goto drop;
444
445	pn = pppoe_pernet(dev_net(dev));
446
447	/* Note that get_item does a sock_hold(), so sk_pppox(po)
448	 * is known to be safe.
449	 */
450	po = get_item(pn, ph->sid, eth_hdr(skb)->h_source, dev->ifindex);
451	if (!po)
452		goto drop;
453
454	return sk_receive_skb(sk_pppox(po), skb, 0);
455
456drop:
457	kfree_skb(skb);
458out:
459	return NET_RX_DROP;
460}
461
462static void pppoe_unbind_sock_work(struct work_struct *work)
463{
464	struct pppox_sock *po = container_of(work, struct pppox_sock,
465					     proto.pppoe.padt_work);
466	struct sock *sk = sk_pppox(po);
467
468	lock_sock(sk);
469	if (po->pppoe_dev) {
470		dev_put(po->pppoe_dev);
471		po->pppoe_dev = NULL;
472	}
473	pppox_unbind_sock(sk);
474	release_sock(sk);
475	sock_put(sk);
476}
477
478/************************************************************************
479 *
480 * Receive a PPPoE Discovery frame.
481 * This is solely for detection of PADT frames
482 *
483 ***********************************************************************/
484static int pppoe_disc_rcv(struct sk_buff *skb, struct net_device *dev,
485			  struct packet_type *pt, struct net_device *orig_dev)
486
487{
488	struct pppoe_hdr *ph;
489	struct pppox_sock *po;
490	struct pppoe_net *pn;
491
492	skb = skb_share_check(skb, GFP_ATOMIC);
493	if (!skb)
494		goto out;
495
496	if (!pskb_may_pull(skb, sizeof(struct pppoe_hdr)))
497		goto abort;
498
499	ph = pppoe_hdr(skb);
500	if (ph->code != PADT_CODE)
501		goto abort;
502
503	pn = pppoe_pernet(dev_net(dev));
504	po = get_item(pn, ph->sid, eth_hdr(skb)->h_source, dev->ifindex);
505	if (po) {
506		struct sock *sk = sk_pppox(po);
507
508		bh_lock_sock(sk);
509
510		/* If the user has locked the socket, just ignore
511		 * the packet.  With the way two rcv protocols hook into
512		 * one socket family type, we cannot (easily) distinguish
513		 * what kind of SKB it is during backlog rcv.
514		 */
515		if (sock_owned_by_user(sk) == 0) {
516			/* We're no longer connect at the PPPOE layer,
517			 * and must wait for ppp channel to disconnect us.
518			 */
519			sk->sk_state = PPPOX_ZOMBIE;
520		}
521
522		bh_unlock_sock(sk);
523		if (!schedule_work(&po->proto.pppoe.padt_work))
524			sock_put(sk);
525	}
526
527abort:
528	kfree_skb(skb);
529out:
530	return NET_RX_SUCCESS; /* Lies... :-) */
531}
532
533static struct packet_type pppoes_ptype __read_mostly = {
534	.type	= cpu_to_be16(ETH_P_PPP_SES),
535	.func	= pppoe_rcv,
536};
537
538static struct packet_type pppoed_ptype __read_mostly = {
539	.type	= cpu_to_be16(ETH_P_PPP_DISC),
540	.func	= pppoe_disc_rcv,
541};
542
543static struct proto pppoe_sk_proto __read_mostly = {
544	.name	  = "PPPOE",
545	.owner	  = THIS_MODULE,
546	.obj_size = sizeof(struct pppox_sock),
547};
548
549/***********************************************************************
550 *
551 * Initialize a new struct sock.
552 *
553 **********************************************************************/
554static int pppoe_create(struct net *net, struct socket *sock)
555{
556	struct sock *sk;
557
558	sk = sk_alloc(net, PF_PPPOX, GFP_KERNEL, &pppoe_sk_proto);
559	if (!sk)
560		return -ENOMEM;
561
562	sock_init_data(sock, sk);
563
564	sock->state	= SS_UNCONNECTED;
565	sock->ops	= &pppoe_ops;
566
567	sk->sk_backlog_rcv	= pppoe_rcv_core;
568	sk->sk_state		= PPPOX_NONE;
569	sk->sk_type		= SOCK_STREAM;
570	sk->sk_family		= PF_PPPOX;
571	sk->sk_protocol		= PX_PROTO_OE;
572
573	INIT_WORK(&pppox_sk(sk)->proto.pppoe.padt_work,
574		  pppoe_unbind_sock_work);
575
576	return 0;
577}
578
579static int pppoe_release(struct socket *sock)
580{
581	struct sock *sk = sock->sk;
582	struct pppox_sock *po;
583	struct pppoe_net *pn;
584	struct net *net = NULL;
585
586	if (!sk)
587		return 0;
588
589	lock_sock(sk);
590	if (sock_flag(sk, SOCK_DEAD)) {
591		release_sock(sk);
592		return -EBADF;
593	}
594
595	po = pppox_sk(sk);
596
597	if (po->pppoe_dev) {
598		dev_put(po->pppoe_dev);
599		po->pppoe_dev = NULL;
600	}
601
602	pppox_unbind_sock(sk);
603
604	/* Signal the death of the socket. */
605	sk->sk_state = PPPOX_DEAD;
606
607	net = sock_net(sk);
608	pn = pppoe_pernet(net);
609
610	/*
611	 * protect "po" from concurrent updates
612	 * on pppoe_flush_dev
613	 */
614	delete_item(pn, po->pppoe_pa.sid, po->pppoe_pa.remote,
615		    po->pppoe_ifindex);
616
617	sock_orphan(sk);
618	sock->sk = NULL;
619
620	skb_queue_purge(&sk->sk_receive_queue);
621	release_sock(sk);
622	sock_put(sk);
623
624	return 0;
625}
626
627static int pppoe_connect(struct socket *sock, struct sockaddr *uservaddr,
628		  int sockaddr_len, int flags)
629{
630	struct sock *sk = sock->sk;
631	struct sockaddr_pppox *sp = (struct sockaddr_pppox *)uservaddr;
632	struct pppox_sock *po = pppox_sk(sk);
633	struct net_device *dev = NULL;
634	struct pppoe_net *pn;
635	struct net *net = NULL;
636	int error;
637
638	lock_sock(sk);
639
640	error = -EINVAL;
641	if (sp->sa_protocol != PX_PROTO_OE)
642		goto end;
643
644	/* Check for already bound sockets */
645	error = -EBUSY;
646	if ((sk->sk_state & PPPOX_CONNECTED) &&
647	     stage_session(sp->sa_addr.pppoe.sid))
648		goto end;
649
650	/* Check for already disconnected sockets, on attempts to disconnect */
651	error = -EALREADY;
652	if ((sk->sk_state & PPPOX_DEAD) &&
653	     !stage_session(sp->sa_addr.pppoe.sid))
654		goto end;
655
656	error = 0;
657
658	/* Delete the old binding */
659	if (stage_session(po->pppoe_pa.sid)) {
660		pppox_unbind_sock(sk);
661		pn = pppoe_pernet(sock_net(sk));
662		delete_item(pn, po->pppoe_pa.sid,
663			    po->pppoe_pa.remote, po->pppoe_ifindex);
664		if (po->pppoe_dev) {
665			dev_put(po->pppoe_dev);
666			po->pppoe_dev = NULL;
667		}
668
669		po->pppoe_ifindex = 0;
670		memset(&po->pppoe_pa, 0, sizeof(po->pppoe_pa));
671		memset(&po->pppoe_relay, 0, sizeof(po->pppoe_relay));
672		memset(&po->chan, 0, sizeof(po->chan));
673		po->next = NULL;
674		po->num = 0;
675
676		sk->sk_state = PPPOX_NONE;
677	}
678
679	/* Re-bind in session stage only */
680	if (stage_session(sp->sa_addr.pppoe.sid)) {
681		error = -ENODEV;
682		net = sock_net(sk);
683		dev = dev_get_by_name(net, sp->sa_addr.pppoe.dev);
684		if (!dev)
685			goto err_put;
686
687		po->pppoe_dev = dev;
688		po->pppoe_ifindex = dev->ifindex;
689		pn = pppoe_pernet(net);
690		if (!(dev->flags & IFF_UP)) {
691			goto err_put;
692		}
693
694		memcpy(&po->pppoe_pa,
695		       &sp->sa_addr.pppoe,
696		       sizeof(struct pppoe_addr));
697
698		write_lock_bh(&pn->hash_lock);
699		error = __set_item(pn, po);
700		write_unlock_bh(&pn->hash_lock);
701		if (error < 0)
702			goto err_put;
703
704		po->chan.hdrlen = (sizeof(struct pppoe_hdr) +
705				   dev->hard_header_len);
706
707		po->chan.mtu = dev->mtu - sizeof(struct pppoe_hdr) - 2;
708		po->chan.private = sk;
709		po->chan.ops = &pppoe_chan_ops;
710
711		error = ppp_register_net_channel(dev_net(dev), &po->chan);
712		if (error) {
713			delete_item(pn, po->pppoe_pa.sid,
714				    po->pppoe_pa.remote, po->pppoe_ifindex);
715			goto err_put;
716		}
717
718		sk->sk_state = PPPOX_CONNECTED;
719	}
720
721	po->num = sp->sa_addr.pppoe.sid;
722
723end:
724	release_sock(sk);
725	return error;
726err_put:
727	if (po->pppoe_dev) {
728		dev_put(po->pppoe_dev);
729		po->pppoe_dev = NULL;
730	}
731	goto end;
732}
733
734static int pppoe_getname(struct socket *sock, struct sockaddr *uaddr,
735		  int *usockaddr_len, int peer)
736{
737	int len = sizeof(struct sockaddr_pppox);
738	struct sockaddr_pppox sp;
739
740	sp.sa_family	= AF_PPPOX;
741	sp.sa_protocol	= PX_PROTO_OE;
742	memcpy(&sp.sa_addr.pppoe, &pppox_sk(sock->sk)->pppoe_pa,
743	       sizeof(struct pppoe_addr));
744
745	memcpy(uaddr, &sp, len);
746
747	*usockaddr_len = len;
748
749	return 0;
750}
751
752static int pppoe_ioctl(struct socket *sock, unsigned int cmd,
753		unsigned long arg)
754{
755	struct sock *sk = sock->sk;
756	struct pppox_sock *po = pppox_sk(sk);
757	int val;
758	int err;
759
760	switch (cmd) {
761	case PPPIOCGMRU:
762		err = -ENXIO;
763		if (!(sk->sk_state & PPPOX_CONNECTED))
764			break;
765
766		err = -EFAULT;
767		if (put_user(po->pppoe_dev->mtu -
768			     sizeof(struct pppoe_hdr) -
769			     PPP_HDRLEN,
770			     (int __user *)arg))
771			break;
772		err = 0;
773		break;
774
775	case PPPIOCSMRU:
776		err = -ENXIO;
777		if (!(sk->sk_state & PPPOX_CONNECTED))
778			break;
779
780		err = -EFAULT;
781		if (get_user(val, (int __user *)arg))
782			break;
783
784		if (val < (po->pppoe_dev->mtu
785			   - sizeof(struct pppoe_hdr)
786			   - PPP_HDRLEN))
787			err = 0;
788		else
789			err = -EINVAL;
790		break;
791
792	case PPPIOCSFLAGS:
793		err = -EFAULT;
794		if (get_user(val, (int __user *)arg))
795			break;
796		err = 0;
797		break;
798
799	case PPPOEIOCSFWD:
800	{
801		struct pppox_sock *relay_po;
802
803		err = -EBUSY;
804		if (sk->sk_state & (PPPOX_BOUND | PPPOX_ZOMBIE | PPPOX_DEAD))
805			break;
806
807		err = -ENOTCONN;
808		if (!(sk->sk_state & PPPOX_CONNECTED))
809			break;
810
811		/* PPPoE address from the user specifies an outbound
812		   PPPoE address which frames are forwarded to */
813		err = -EFAULT;
814		if (copy_from_user(&po->pppoe_relay,
815				   (void __user *)arg,
816				   sizeof(struct sockaddr_pppox)))
817			break;
818
819		err = -EINVAL;
820		if (po->pppoe_relay.sa_family != AF_PPPOX ||
821		    po->pppoe_relay.sa_protocol != PX_PROTO_OE)
822			break;
823
824		/* Check that the socket referenced by the address
825		   actually exists. */
826		relay_po = get_item_by_addr(sock_net(sk), &po->pppoe_relay);
827		if (!relay_po)
828			break;
829
830		sock_put(sk_pppox(relay_po));
831		sk->sk_state |= PPPOX_RELAY;
832		err = 0;
833		break;
834	}
835
836	case PPPOEIOCDFWD:
837		err = -EALREADY;
838		if (!(sk->sk_state & PPPOX_RELAY))
839			break;
840
841		sk->sk_state &= ~PPPOX_RELAY;
842		err = 0;
843		break;
844
845	default:
846		err = -ENOTTY;
847	}
848
849	return err;
850}
851
852static int pppoe_sendmsg(struct socket *sock, struct msghdr *m,
853			 size_t total_len)
854{
855	struct sk_buff *skb;
856	struct sock *sk = sock->sk;
857	struct pppox_sock *po = pppox_sk(sk);
858	int error;
859	struct pppoe_hdr hdr;
860	struct pppoe_hdr *ph;
861	struct net_device *dev;
862	char *start;
863
864	lock_sock(sk);
865	if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED)) {
866		error = -ENOTCONN;
867		goto end;
868	}
869
870	hdr.ver = 1;
871	hdr.type = 1;
872	hdr.code = 0;
873	hdr.sid = po->num;
874
875	dev = po->pppoe_dev;
876
877	error = -EMSGSIZE;
878	if (total_len > (dev->mtu + dev->hard_header_len))
879		goto end;
880
881
882	skb = sock_wmalloc(sk, total_len + dev->hard_header_len + 32,
883			   0, GFP_KERNEL);
884	if (!skb) {
885		error = -ENOMEM;
886		goto end;
887	}
888
889	/* Reserve space for headers. */
890	skb_reserve(skb, dev->hard_header_len);
891	skb_reset_network_header(skb);
892
893	skb->dev = dev;
894
895	skb->priority = sk->sk_priority;
896	skb->protocol = cpu_to_be16(ETH_P_PPP_SES);
897
898	ph = (struct pppoe_hdr *)skb_put(skb, total_len + sizeof(struct pppoe_hdr));
899	start = (char *)&ph->tag[0];
900
901	error = memcpy_from_msg(start, m, total_len);
902	if (error < 0) {
903		kfree_skb(skb);
904		goto end;
905	}
906
907	error = total_len;
908	dev_hard_header(skb, dev, ETH_P_PPP_SES,
909			po->pppoe_pa.remote, NULL, total_len);
910
911	memcpy(ph, &hdr, sizeof(struct pppoe_hdr));
912
913	ph->length = htons(total_len);
914
915	dev_queue_xmit(skb);
916
917end:
918	release_sock(sk);
919	return error;
920}
921
922/************************************************************************
923 *
924 * xmit function for internal use.
925 *
926 ***********************************************************************/
927static int __pppoe_xmit(struct sock *sk, struct sk_buff *skb)
928{
929	struct pppox_sock *po = pppox_sk(sk);
930	struct net_device *dev = po->pppoe_dev;
931	struct pppoe_hdr *ph;
932	int data_len = skb->len;
933
934	/* The higher-level PPP code (ppp_unregister_channel()) ensures the PPP
935	 * xmit operations conclude prior to an unregistration call.  Thus
936	 * sk->sk_state cannot change, so we don't need to do lock_sock().
937	 * But, we also can't do a lock_sock since that introduces a potential
938	 * deadlock as we'd reverse the lock ordering used when calling
939	 * ppp_unregister_channel().
940	 */
941
942	if (sock_flag(sk, SOCK_DEAD) || !(sk->sk_state & PPPOX_CONNECTED))
943		goto abort;
944
945	if (!dev)
946		goto abort;
947
948	/* Copy the data if there is no space for the header or if it's
949	 * read-only.
950	 */
951	if (skb_cow_head(skb, sizeof(*ph) + dev->hard_header_len))
952		goto abort;
953
954	__skb_push(skb, sizeof(*ph));
955	skb_reset_network_header(skb);
956
957	ph = pppoe_hdr(skb);
958	ph->ver	= 1;
959	ph->type = 1;
960	ph->code = 0;
961	ph->sid	= po->num;
962	ph->length = htons(data_len);
963
964	skb->protocol = cpu_to_be16(ETH_P_PPP_SES);
965	skb->dev = dev;
966
967	dev_hard_header(skb, dev, ETH_P_PPP_SES,
968			po->pppoe_pa.remote, NULL, data_len);
969
970	dev_queue_xmit(skb);
971	return 1;
972
973abort:
974	kfree_skb(skb);
975	return 1;
976}
977
978/************************************************************************
979 *
980 * xmit function called by generic PPP driver
981 * sends PPP frame over PPPoE socket
982 *
983 ***********************************************************************/
984static int pppoe_xmit(struct ppp_channel *chan, struct sk_buff *skb)
985{
986	struct sock *sk = (struct sock *)chan->private;
987	return __pppoe_xmit(sk, skb);
988}
989
990static const struct ppp_channel_ops pppoe_chan_ops = {
991	.start_xmit = pppoe_xmit,
992};
993
994static int pppoe_recvmsg(struct socket *sock, struct msghdr *m,
995			 size_t total_len, int flags)
996{
997	struct sock *sk = sock->sk;
998	struct sk_buff *skb;
999	int error = 0;
1000
1001	if (sk->sk_state & PPPOX_BOUND) {
1002		error = -EIO;
1003		goto end;
1004	}
1005
1006	skb = skb_recv_datagram(sk, flags & ~MSG_DONTWAIT,
1007				flags & MSG_DONTWAIT, &error);
1008	if (error < 0)
1009		goto end;
1010
1011	if (skb) {
1012		total_len = min_t(size_t, total_len, skb->len);
1013		error = skb_copy_datagram_msg(skb, 0, m, total_len);
1014		if (error == 0) {
1015			consume_skb(skb);
1016			return total_len;
1017		}
1018	}
1019
1020	kfree_skb(skb);
1021end:
1022	return error;
1023}
1024
1025#ifdef CONFIG_PROC_FS
1026static int pppoe_seq_show(struct seq_file *seq, void *v)
1027{
1028	struct pppox_sock *po;
1029	char *dev_name;
1030
1031	if (v == SEQ_START_TOKEN) {
1032		seq_puts(seq, "Id       Address              Device\n");
1033		goto out;
1034	}
1035
1036	po = v;
1037	dev_name = po->pppoe_pa.dev;
1038
1039	seq_printf(seq, "%08X %pM %8s\n",
1040		po->pppoe_pa.sid, po->pppoe_pa.remote, dev_name);
1041out:
1042	return 0;
1043}
1044
1045static inline struct pppox_sock *pppoe_get_idx(struct pppoe_net *pn, loff_t pos)
1046{
1047	struct pppox_sock *po;
1048	int i;
1049
1050	for (i = 0; i < PPPOE_HASH_SIZE; i++) {
1051		po = pn->hash_table[i];
1052		while (po) {
1053			if (!pos--)
1054				goto out;
1055			po = po->next;
1056		}
1057	}
1058
1059out:
1060	return po;
1061}
1062
1063static void *pppoe_seq_start(struct seq_file *seq, loff_t *pos)
1064	__acquires(pn->hash_lock)
1065{
1066	struct pppoe_net *pn = pppoe_pernet(seq_file_net(seq));
1067	loff_t l = *pos;
1068
1069	read_lock_bh(&pn->hash_lock);
1070	return l ? pppoe_get_idx(pn, --l) : SEQ_START_TOKEN;
1071}
1072
1073static void *pppoe_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1074{
1075	struct pppoe_net *pn = pppoe_pernet(seq_file_net(seq));
1076	struct pppox_sock *po;
1077
1078	++*pos;
1079	if (v == SEQ_START_TOKEN) {
1080		po = pppoe_get_idx(pn, 0);
1081		goto out;
1082	}
1083	po = v;
1084	if (po->next)
1085		po = po->next;
1086	else {
1087		int hash = hash_item(po->pppoe_pa.sid, po->pppoe_pa.remote);
1088
1089		po = NULL;
1090		while (++hash < PPPOE_HASH_SIZE) {
1091			po = pn->hash_table[hash];
1092			if (po)
1093				break;
1094		}
1095	}
1096
1097out:
1098	return po;
1099}
1100
1101static void pppoe_seq_stop(struct seq_file *seq, void *v)
1102	__releases(pn->hash_lock)
1103{
1104	struct pppoe_net *pn = pppoe_pernet(seq_file_net(seq));
1105	read_unlock_bh(&pn->hash_lock);
1106}
1107
1108static const struct seq_operations pppoe_seq_ops = {
1109	.start		= pppoe_seq_start,
1110	.next		= pppoe_seq_next,
1111	.stop		= pppoe_seq_stop,
1112	.show		= pppoe_seq_show,
1113};
1114
1115static int pppoe_seq_open(struct inode *inode, struct file *file)
1116{
1117	return seq_open_net(inode, file, &pppoe_seq_ops,
1118			sizeof(struct seq_net_private));
1119}
1120
1121static const struct file_operations pppoe_seq_fops = {
1122	.owner		= THIS_MODULE,
1123	.open		= pppoe_seq_open,
1124	.read		= seq_read,
1125	.llseek		= seq_lseek,
1126	.release	= seq_release_net,
1127};
1128
1129#endif /* CONFIG_PROC_FS */
1130
1131static const struct proto_ops pppoe_ops = {
1132	.family		= AF_PPPOX,
1133	.owner		= THIS_MODULE,
1134	.release	= pppoe_release,
1135	.bind		= sock_no_bind,
1136	.connect	= pppoe_connect,
1137	.socketpair	= sock_no_socketpair,
1138	.accept		= sock_no_accept,
1139	.getname	= pppoe_getname,
1140	.poll		= datagram_poll,
1141	.listen		= sock_no_listen,
1142	.shutdown	= sock_no_shutdown,
1143	.setsockopt	= sock_no_setsockopt,
1144	.getsockopt	= sock_no_getsockopt,
1145	.sendmsg	= pppoe_sendmsg,
1146	.recvmsg	= pppoe_recvmsg,
1147	.mmap		= sock_no_mmap,
1148	.ioctl		= pppox_ioctl,
1149};
1150
1151static const struct pppox_proto pppoe_proto = {
1152	.create	= pppoe_create,
1153	.ioctl	= pppoe_ioctl,
1154	.owner	= THIS_MODULE,
1155};
1156
1157static __net_init int pppoe_init_net(struct net *net)
1158{
1159	struct pppoe_net *pn = pppoe_pernet(net);
1160	struct proc_dir_entry *pde;
1161
1162	rwlock_init(&pn->hash_lock);
1163
1164	pde = proc_create("pppoe", S_IRUGO, net->proc_net, &pppoe_seq_fops);
1165#ifdef CONFIG_PROC_FS
1166	if (!pde)
1167		return -ENOMEM;
1168#endif
1169
1170	return 0;
1171}
1172
1173static __net_exit void pppoe_exit_net(struct net *net)
1174{
1175	remove_proc_entry("pppoe", net->proc_net);
1176}
1177
1178static struct pernet_operations pppoe_net_ops = {
1179	.init = pppoe_init_net,
1180	.exit = pppoe_exit_net,
1181	.id   = &pppoe_net_id,
1182	.size = sizeof(struct pppoe_net),
1183};
1184
1185static int __init pppoe_init(void)
1186{
1187	int err;
1188
1189	err = register_pernet_device(&pppoe_net_ops);
1190	if (err)
1191		goto out;
1192
1193	err = proto_register(&pppoe_sk_proto, 0);
1194	if (err)
1195		goto out_unregister_net_ops;
1196
1197	err = register_pppox_proto(PX_PROTO_OE, &pppoe_proto);
1198	if (err)
1199		goto out_unregister_pppoe_proto;
1200
1201	dev_add_pack(&pppoes_ptype);
1202	dev_add_pack(&pppoed_ptype);
1203	register_netdevice_notifier(&pppoe_notifier);
1204
1205	return 0;
1206
1207out_unregister_pppoe_proto:
1208	proto_unregister(&pppoe_sk_proto);
1209out_unregister_net_ops:
1210	unregister_pernet_device(&pppoe_net_ops);
1211out:
1212	return err;
1213}
1214
1215static void __exit pppoe_exit(void)
1216{
1217	unregister_netdevice_notifier(&pppoe_notifier);
1218	dev_remove_pack(&pppoed_ptype);
1219	dev_remove_pack(&pppoes_ptype);
1220	unregister_pppox_proto(PX_PROTO_OE);
1221	proto_unregister(&pppoe_sk_proto);
1222	unregister_pernet_device(&pppoe_net_ops);
1223}
1224
1225module_init(pppoe_init);
1226module_exit(pppoe_exit);
1227
1228MODULE_AUTHOR("Michal Ostrowski <mostrows@speakeasy.net>");
1229MODULE_DESCRIPTION("PPP over Ethernet driver");
1230MODULE_LICENSE("GPL");
1231MODULE_ALIAS_NETPROTO(PF_PPPOX);
1232