1/*
2 * net/tipc/server.c: TIPC server infrastructure
3 *
4 * Copyright (c) 2012-2013, Wind River Systems
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions are met:
9 *
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the names of the copyright holders nor the names of its
16 *    contributors may be used to endorse or promote products derived from
17 *    this software without specific prior written permission.
18 *
19 * Alternatively, this software may be distributed under the terms of the
20 * GNU General Public License ("GPL") version 2 as published by the Free
21 * Software Foundation.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
24 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
27 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33 * POSSIBILITY OF SUCH DAMAGE.
34 */
35
36#include "server.h"
37#include "core.h"
38#include "socket.h"
39#include <net/sock.h>
40#include <linux/module.h>
41
42/* Number of messages to send before rescheduling */
43#define MAX_SEND_MSG_COUNT	25
44#define MAX_RECV_MSG_COUNT	25
45#define CF_CONNECTED		1
46#define CF_SERVER		2
47
48#define sock2con(x) ((struct tipc_conn *)(x)->sk_user_data)
49
50/**
51 * struct tipc_conn - TIPC connection structure
52 * @kref: reference counter to connection object
53 * @conid: connection identifier
54 * @sock: socket handler associated with connection
55 * @flags: indicates connection state
56 * @server: pointer to connected server
57 * @rwork: receive work item
58 * @usr_data: user-specified field
59 * @rx_action: what to do when connection socket is active
60 * @outqueue: pointer to first outbound message in queue
61 * @outqueue_lock: control access to the outqueue
62 * @outqueue: list of connection objects for its server
63 * @swork: send work item
64 */
65struct tipc_conn {
66	struct kref kref;
67	int conid;
68	struct socket *sock;
69	unsigned long flags;
70	struct tipc_server *server;
71	struct work_struct rwork;
72	int (*rx_action) (struct tipc_conn *con);
73	void *usr_data;
74	struct list_head outqueue;
75	spinlock_t outqueue_lock;
76	struct work_struct swork;
77};
78
79/* An entry waiting to be sent */
80struct outqueue_entry {
81	struct list_head list;
82	struct kvec iov;
83	struct sockaddr_tipc dest;
84};
85
86static void tipc_recv_work(struct work_struct *work);
87static void tipc_send_work(struct work_struct *work);
88static void tipc_clean_outqueues(struct tipc_conn *con);
89
90static void tipc_conn_kref_release(struct kref *kref)
91{
92	struct tipc_conn *con = container_of(kref, struct tipc_conn, kref);
93	struct sockaddr_tipc *saddr = con->server->saddr;
94	struct socket *sock = con->sock;
95	struct sock *sk;
96
97	if (sock) {
98		sk = sock->sk;
99		if (test_bit(CF_SERVER, &con->flags)) {
100			__module_get(sock->ops->owner);
101			__module_get(sk->sk_prot_creator->owner);
102		}
103		saddr->scope = -TIPC_NODE_SCOPE;
104		kernel_bind(sock, (struct sockaddr *)saddr, sizeof(*saddr));
105		sock_release(sock);
106		con->sock = NULL;
107	}
108
109	tipc_clean_outqueues(con);
110	kfree(con);
111}
112
113static void conn_put(struct tipc_conn *con)
114{
115	kref_put(&con->kref, tipc_conn_kref_release);
116}
117
118static void conn_get(struct tipc_conn *con)
119{
120	kref_get(&con->kref);
121}
122
123static struct tipc_conn *tipc_conn_lookup(struct tipc_server *s, int conid)
124{
125	struct tipc_conn *con;
126
127	spin_lock_bh(&s->idr_lock);
128	con = idr_find(&s->conn_idr, conid);
129	if (con)
130		conn_get(con);
131	spin_unlock_bh(&s->idr_lock);
132	return con;
133}
134
135static void sock_data_ready(struct sock *sk)
136{
137	struct tipc_conn *con;
138
139	read_lock(&sk->sk_callback_lock);
140	con = sock2con(sk);
141	if (con && test_bit(CF_CONNECTED, &con->flags)) {
142		conn_get(con);
143		if (!queue_work(con->server->rcv_wq, &con->rwork))
144			conn_put(con);
145	}
146	read_unlock(&sk->sk_callback_lock);
147}
148
149static void sock_write_space(struct sock *sk)
150{
151	struct tipc_conn *con;
152
153	read_lock(&sk->sk_callback_lock);
154	con = sock2con(sk);
155	if (con && test_bit(CF_CONNECTED, &con->flags)) {
156		conn_get(con);
157		if (!queue_work(con->server->send_wq, &con->swork))
158			conn_put(con);
159	}
160	read_unlock(&sk->sk_callback_lock);
161}
162
163static void tipc_register_callbacks(struct socket *sock, struct tipc_conn *con)
164{
165	struct sock *sk = sock->sk;
166
167	write_lock_bh(&sk->sk_callback_lock);
168
169	sk->sk_data_ready = sock_data_ready;
170	sk->sk_write_space = sock_write_space;
171	sk->sk_user_data = con;
172
173	con->sock = sock;
174
175	write_unlock_bh(&sk->sk_callback_lock);
176}
177
178static void tipc_unregister_callbacks(struct tipc_conn *con)
179{
180	struct sock *sk = con->sock->sk;
181
182	write_lock_bh(&sk->sk_callback_lock);
183	sk->sk_user_data = NULL;
184	write_unlock_bh(&sk->sk_callback_lock);
185}
186
187static void tipc_close_conn(struct tipc_conn *con)
188{
189	struct tipc_server *s = con->server;
190
191	if (test_and_clear_bit(CF_CONNECTED, &con->flags)) {
192		if (con->conid)
193			s->tipc_conn_shutdown(con->conid, con->usr_data);
194
195		spin_lock_bh(&s->idr_lock);
196		idr_remove(&s->conn_idr, con->conid);
197		s->idr_in_use--;
198		spin_unlock_bh(&s->idr_lock);
199
200		tipc_unregister_callbacks(con);
201
202		/* We shouldn't flush pending works as we may be in the
203		 * thread. In fact the races with pending rx/tx work structs
204		 * are harmless for us here as we have already deleted this
205		 * connection from server connection list and set
206		 * sk->sk_user_data to 0 before releasing connection object.
207		 */
208		kernel_sock_shutdown(con->sock, SHUT_RDWR);
209
210		conn_put(con);
211	}
212}
213
214static struct tipc_conn *tipc_alloc_conn(struct tipc_server *s)
215{
216	struct tipc_conn *con;
217	int ret;
218
219	con = kzalloc(sizeof(struct tipc_conn), GFP_ATOMIC);
220	if (!con)
221		return ERR_PTR(-ENOMEM);
222
223	kref_init(&con->kref);
224	INIT_LIST_HEAD(&con->outqueue);
225	spin_lock_init(&con->outqueue_lock);
226	INIT_WORK(&con->swork, tipc_send_work);
227	INIT_WORK(&con->rwork, tipc_recv_work);
228
229	spin_lock_bh(&s->idr_lock);
230	ret = idr_alloc(&s->conn_idr, con, 0, 0, GFP_ATOMIC);
231	if (ret < 0) {
232		kfree(con);
233		spin_unlock_bh(&s->idr_lock);
234		return ERR_PTR(-ENOMEM);
235	}
236	con->conid = ret;
237	s->idr_in_use++;
238	spin_unlock_bh(&s->idr_lock);
239
240	set_bit(CF_CONNECTED, &con->flags);
241	con->server = s;
242
243	return con;
244}
245
246static int tipc_receive_from_sock(struct tipc_conn *con)
247{
248	struct msghdr msg = {};
249	struct tipc_server *s = con->server;
250	struct sockaddr_tipc addr;
251	struct kvec iov;
252	void *buf;
253	int ret;
254
255	buf = kmem_cache_alloc(s->rcvbuf_cache, GFP_ATOMIC);
256	if (!buf) {
257		ret = -ENOMEM;
258		goto out_close;
259	}
260
261	iov.iov_base = buf;
262	iov.iov_len = s->max_rcvbuf_size;
263	msg.msg_name = &addr;
264	ret = kernel_recvmsg(con->sock, &msg, &iov, 1, iov.iov_len,
265			     MSG_DONTWAIT);
266	if (ret <= 0) {
267		kmem_cache_free(s->rcvbuf_cache, buf);
268		goto out_close;
269	}
270
271	s->tipc_conn_recvmsg(sock_net(con->sock->sk), con->conid, &addr,
272			     con->usr_data, buf, ret);
273
274	kmem_cache_free(s->rcvbuf_cache, buf);
275
276	return 0;
277
278out_close:
279	if (ret != -EWOULDBLOCK)
280		tipc_close_conn(con);
281	else if (ret == 0)
282		/* Don't return success if we really got EOF */
283		ret = -EAGAIN;
284
285	return ret;
286}
287
288static int tipc_accept_from_sock(struct tipc_conn *con)
289{
290	struct tipc_server *s = con->server;
291	struct socket *sock = con->sock;
292	struct socket *newsock;
293	struct tipc_conn *newcon;
294	int ret;
295
296	ret = kernel_accept(sock, &newsock, O_NONBLOCK);
297	if (ret < 0)
298		return ret;
299
300	newcon = tipc_alloc_conn(con->server);
301	if (IS_ERR(newcon)) {
302		ret = PTR_ERR(newcon);
303		sock_release(newsock);
304		return ret;
305	}
306
307	newcon->rx_action = tipc_receive_from_sock;
308	tipc_register_callbacks(newsock, newcon);
309
310	/* Notify that new connection is incoming */
311	newcon->usr_data = s->tipc_conn_new(newcon->conid);
312
313	/* Wake up receive process in case of 'SYN+' message */
314	newsock->sk->sk_data_ready(newsock->sk);
315	return ret;
316}
317
318static struct socket *tipc_create_listen_sock(struct tipc_conn *con)
319{
320	struct tipc_server *s = con->server;
321	struct socket *sock = NULL;
322	int ret;
323
324	ret = __sock_create(s->net, AF_TIPC, SOCK_SEQPACKET, 0, &sock, 1);
325	if (ret < 0)
326		return NULL;
327	ret = kernel_setsockopt(sock, SOL_TIPC, TIPC_IMPORTANCE,
328				(char *)&s->imp, sizeof(s->imp));
329	if (ret < 0)
330		goto create_err;
331	ret = kernel_bind(sock, (struct sockaddr *)s->saddr, sizeof(*s->saddr));
332	if (ret < 0)
333		goto create_err;
334
335	switch (s->type) {
336	case SOCK_STREAM:
337	case SOCK_SEQPACKET:
338		con->rx_action = tipc_accept_from_sock;
339
340		ret = kernel_listen(sock, 0);
341		if (ret < 0)
342			goto create_err;
343		break;
344	case SOCK_DGRAM:
345	case SOCK_RDM:
346		con->rx_action = tipc_receive_from_sock;
347		break;
348	default:
349		pr_err("Unknown socket type %d\n", s->type);
350		goto create_err;
351	}
352
353	/* As server's listening socket owner and creator is the same module,
354	 * we have to decrease TIPC module reference count to guarantee that
355	 * it remains zero after the server socket is created, otherwise,
356	 * executing "rmmod" command is unable to make TIPC module deleted
357	 * after TIPC module is inserted successfully.
358	 *
359	 * However, the reference count is ever increased twice in
360	 * sock_create_kern(): one is to increase the reference count of owner
361	 * of TIPC socket's proto_ops struct; another is to increment the
362	 * reference count of owner of TIPC proto struct. Therefore, we must
363	 * decrement the module reference count twice to ensure that it keeps
364	 * zero after server's listening socket is created. Of course, we
365	 * must bump the module reference count twice as well before the socket
366	 * is closed.
367	 */
368	module_put(sock->ops->owner);
369	module_put(sock->sk->sk_prot_creator->owner);
370	set_bit(CF_SERVER, &con->flags);
371
372	return sock;
373
374create_err:
375	kernel_sock_shutdown(sock, SHUT_RDWR);
376	sock_release(sock);
377	return NULL;
378}
379
380static int tipc_open_listening_sock(struct tipc_server *s)
381{
382	struct socket *sock;
383	struct tipc_conn *con;
384
385	con = tipc_alloc_conn(s);
386	if (IS_ERR(con))
387		return PTR_ERR(con);
388
389	sock = tipc_create_listen_sock(con);
390	if (!sock) {
391		idr_remove(&s->conn_idr, con->conid);
392		s->idr_in_use--;
393		kfree(con);
394		return -EINVAL;
395	}
396
397	tipc_register_callbacks(sock, con);
398	return 0;
399}
400
401static struct outqueue_entry *tipc_alloc_entry(void *data, int len)
402{
403	struct outqueue_entry *entry;
404	void *buf;
405
406	entry = kmalloc(sizeof(struct outqueue_entry), GFP_ATOMIC);
407	if (!entry)
408		return NULL;
409
410	buf = kmalloc(len, GFP_ATOMIC);
411	if (!buf) {
412		kfree(entry);
413		return NULL;
414	}
415
416	memcpy(buf, data, len);
417	entry->iov.iov_base = buf;
418	entry->iov.iov_len = len;
419
420	return entry;
421}
422
423static void tipc_free_entry(struct outqueue_entry *e)
424{
425	kfree(e->iov.iov_base);
426	kfree(e);
427}
428
429static void tipc_clean_outqueues(struct tipc_conn *con)
430{
431	struct outqueue_entry *e, *safe;
432
433	spin_lock_bh(&con->outqueue_lock);
434	list_for_each_entry_safe(e, safe, &con->outqueue, list) {
435		list_del(&e->list);
436		tipc_free_entry(e);
437	}
438	spin_unlock_bh(&con->outqueue_lock);
439}
440
441int tipc_conn_sendmsg(struct tipc_server *s, int conid,
442		      struct sockaddr_tipc *addr, void *data, size_t len)
443{
444	struct outqueue_entry *e;
445	struct tipc_conn *con;
446
447	con = tipc_conn_lookup(s, conid);
448	if (!con)
449		return -EINVAL;
450
451	e = tipc_alloc_entry(data, len);
452	if (!e) {
453		conn_put(con);
454		return -ENOMEM;
455	}
456
457	if (addr)
458		memcpy(&e->dest, addr, sizeof(struct sockaddr_tipc));
459
460	spin_lock_bh(&con->outqueue_lock);
461	list_add_tail(&e->list, &con->outqueue);
462	spin_unlock_bh(&con->outqueue_lock);
463
464	if (test_bit(CF_CONNECTED, &con->flags)) {
465		if (!queue_work(s->send_wq, &con->swork))
466			conn_put(con);
467	} else {
468		conn_put(con);
469	}
470	return 0;
471}
472
473void tipc_conn_terminate(struct tipc_server *s, int conid)
474{
475	struct tipc_conn *con;
476
477	con = tipc_conn_lookup(s, conid);
478	if (con) {
479		tipc_close_conn(con);
480		conn_put(con);
481	}
482}
483
484static void tipc_send_to_sock(struct tipc_conn *con)
485{
486	int count = 0;
487	struct tipc_server *s = con->server;
488	struct outqueue_entry *e;
489	struct msghdr msg;
490	int ret;
491
492	spin_lock_bh(&con->outqueue_lock);
493	while (1) {
494		e = list_entry(con->outqueue.next, struct outqueue_entry,
495			       list);
496		if ((struct list_head *) e == &con->outqueue)
497			break;
498		spin_unlock_bh(&con->outqueue_lock);
499
500		memset(&msg, 0, sizeof(msg));
501		msg.msg_flags = MSG_DONTWAIT;
502
503		if (s->type == SOCK_DGRAM || s->type == SOCK_RDM) {
504			msg.msg_name = &e->dest;
505			msg.msg_namelen = sizeof(struct sockaddr_tipc);
506		}
507		ret = kernel_sendmsg(con->sock, &msg, &e->iov, 1,
508				     e->iov.iov_len);
509		if (ret == -EWOULDBLOCK || ret == 0) {
510			cond_resched();
511			goto out;
512		} else if (ret < 0) {
513			goto send_err;
514		}
515
516		/* Don't starve users filling buffers */
517		if (++count >= MAX_SEND_MSG_COUNT) {
518			cond_resched();
519			count = 0;
520		}
521
522		spin_lock_bh(&con->outqueue_lock);
523		list_del(&e->list);
524		tipc_free_entry(e);
525	}
526	spin_unlock_bh(&con->outqueue_lock);
527out:
528	return;
529
530send_err:
531	tipc_close_conn(con);
532}
533
534static void tipc_recv_work(struct work_struct *work)
535{
536	struct tipc_conn *con = container_of(work, struct tipc_conn, rwork);
537	int count = 0;
538
539	while (test_bit(CF_CONNECTED, &con->flags)) {
540		if (con->rx_action(con))
541			break;
542
543		/* Don't flood Rx machine */
544		if (++count >= MAX_RECV_MSG_COUNT) {
545			cond_resched();
546			count = 0;
547		}
548	}
549	conn_put(con);
550}
551
552static void tipc_send_work(struct work_struct *work)
553{
554	struct tipc_conn *con = container_of(work, struct tipc_conn, swork);
555
556	if (test_bit(CF_CONNECTED, &con->flags))
557		tipc_send_to_sock(con);
558
559	conn_put(con);
560}
561
562static void tipc_work_stop(struct tipc_server *s)
563{
564	destroy_workqueue(s->rcv_wq);
565	destroy_workqueue(s->send_wq);
566}
567
568static int tipc_work_start(struct tipc_server *s)
569{
570	s->rcv_wq = alloc_workqueue("tipc_rcv", WQ_UNBOUND, 1);
571	if (!s->rcv_wq) {
572		pr_err("can't start tipc receive workqueue\n");
573		return -ENOMEM;
574	}
575
576	s->send_wq = alloc_workqueue("tipc_send", WQ_UNBOUND, 1);
577	if (!s->send_wq) {
578		pr_err("can't start tipc send workqueue\n");
579		destroy_workqueue(s->rcv_wq);
580		return -ENOMEM;
581	}
582
583	return 0;
584}
585
586int tipc_server_start(struct tipc_server *s)
587{
588	int ret;
589
590	spin_lock_init(&s->idr_lock);
591	idr_init(&s->conn_idr);
592	s->idr_in_use = 0;
593
594	s->rcvbuf_cache = kmem_cache_create(s->name, s->max_rcvbuf_size,
595					    0, SLAB_HWCACHE_ALIGN, NULL);
596	if (!s->rcvbuf_cache)
597		return -ENOMEM;
598
599	ret = tipc_work_start(s);
600	if (ret < 0) {
601		kmem_cache_destroy(s->rcvbuf_cache);
602		return ret;
603	}
604	ret = tipc_open_listening_sock(s);
605	if (ret < 0) {
606		tipc_work_stop(s);
607		kmem_cache_destroy(s->rcvbuf_cache);
608		return ret;
609	}
610	return ret;
611}
612
613void tipc_server_stop(struct tipc_server *s)
614{
615	struct tipc_conn *con;
616	int total = 0;
617	int id;
618
619	spin_lock_bh(&s->idr_lock);
620	for (id = 0; total < s->idr_in_use; id++) {
621		con = idr_find(&s->conn_idr, id);
622		if (con) {
623			total++;
624			spin_unlock_bh(&s->idr_lock);
625			tipc_close_conn(con);
626			spin_lock_bh(&s->idr_lock);
627		}
628	}
629	spin_unlock_bh(&s->idr_lock);
630
631	tipc_work_stop(s);
632	kmem_cache_destroy(s->rcvbuf_cache);
633	idr_destroy(&s->conn_idr);
634}
635