1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2011, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include <net/sock.h>
38 
39 #include <linux/tipc.h>
40 #include <linux/tipc_config.h>
41 
42 #include "core.h"
43 #include "port.h"
44 
45 #define SS_LISTENING	-1	/* socket is listening */
46 #define SS_READY	-2	/* socket is connectionless */
47 
48 #define OVERLOAD_LIMIT_BASE	5000
49 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
50 
51 struct tipc_sock {
52 	struct sock sk;
53 	struct tipc_port *p;
54 	struct tipc_portid peer_name;
55 	long conn_timeout;
56 };
57 
58 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
59 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
60 
61 #define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
62 			(sock->state == SS_DISCONNECTING))
63 
64 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
65 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
66 static void wakeupdispatch(struct tipc_port *tport);
67 
68 static const struct proto_ops packet_ops;
69 static const struct proto_ops stream_ops;
70 static const struct proto_ops msg_ops;
71 
72 static struct proto tipc_proto;
73 
74 static int sockets_enabled;
75 
76 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
77 
78 /*
79  * Revised TIPC socket locking policy:
80  *
81  * Most socket operations take the standard socket lock when they start
82  * and hold it until they finish (or until they need to sleep).  Acquiring
83  * this lock grants the owner exclusive access to the fields of the socket
84  * data structures, with the exception of the backlog queue.  A few socket
85  * operations can be done without taking the socket lock because they only
86  * read socket information that never changes during the life of the socket.
87  *
88  * Socket operations may acquire the lock for the associated TIPC port if they
89  * need to perform an operation on the port.  If any routine needs to acquire
90  * both the socket lock and the port lock it must take the socket lock first
91  * to avoid the risk of deadlock.
92  *
93  * The dispatcher handling incoming messages cannot grab the socket lock in
94  * the standard fashion, since invoked it runs at the BH level and cannot block.
95  * Instead, it checks to see if the socket lock is currently owned by someone,
96  * and either handles the message itself or adds it to the socket's backlog
97  * queue; in the latter case the queued message is processed once the process
98  * owning the socket lock releases it.
99  *
100  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
101  * the problem of a blocked socket operation preventing any other operations
102  * from occurring.  However, applications must be careful if they have
103  * multiple threads trying to send (or receive) on the same socket, as these
104  * operations might interfere with each other.  For example, doing a connect
105  * and a receive at the same time might allow the receive to consume the
106  * ACK message meant for the connect.  While additional work could be done
107  * to try and overcome this, it doesn't seem to be worthwhile at the present.
108  *
109  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
110  * that another operation that must be performed in a non-blocking manner is
111  * not delayed for very long because the lock has already been taken.
112  *
113  * NOTE: This code assumes that certain fields of a port/socket pair are
114  * constant over its lifetime; such fields can be examined without taking
115  * the socket lock and/or port lock, and do not need to be re-read even
116  * after resuming processing after waiting.  These fields include:
117  *   - socket type
118  *   - pointer to socket sk structure (aka tipc_sock structure)
119  *   - pointer to port structure
120  *   - port reference
121  */
122 
123 /**
124  * advance_rx_queue - discard first buffer in socket receive queue
125  *
126  * Caller must hold socket lock
127  */
128 
advance_rx_queue(struct sock * sk)129 static void advance_rx_queue(struct sock *sk)
130 {
131 	buf_discard(__skb_dequeue(&sk->sk_receive_queue));
132 	atomic_dec(&tipc_queue_size);
133 }
134 
135 /**
136  * discard_rx_queue - discard all buffers in socket receive queue
137  *
138  * Caller must hold socket lock
139  */
140 
discard_rx_queue(struct sock * sk)141 static void discard_rx_queue(struct sock *sk)
142 {
143 	struct sk_buff *buf;
144 
145 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
146 		atomic_dec(&tipc_queue_size);
147 		buf_discard(buf);
148 	}
149 }
150 
151 /**
152  * reject_rx_queue - reject all buffers in socket receive queue
153  *
154  * Caller must hold socket lock
155  */
156 
reject_rx_queue(struct sock * sk)157 static void reject_rx_queue(struct sock *sk)
158 {
159 	struct sk_buff *buf;
160 
161 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
162 		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
163 		atomic_dec(&tipc_queue_size);
164 	}
165 }
166 
167 /**
168  * tipc_create - create a TIPC socket
169  * @net: network namespace (must be default network)
170  * @sock: pre-allocated socket structure
171  * @protocol: protocol indicator (must be 0)
172  * @kern: caused by kernel or by userspace?
173  *
174  * This routine creates additional data structures used by the TIPC socket,
175  * initializes them, and links them together.
176  *
177  * Returns 0 on success, errno otherwise
178  */
179 
tipc_create(struct net * net,struct socket * sock,int protocol,int kern)180 static int tipc_create(struct net *net, struct socket *sock, int protocol,
181 		       int kern)
182 {
183 	const struct proto_ops *ops;
184 	socket_state state;
185 	struct sock *sk;
186 	struct tipc_port *tp_ptr;
187 
188 	/* Validate arguments */
189 
190 	if (!net_eq(net, &init_net))
191 		return -EAFNOSUPPORT;
192 
193 	if (unlikely(protocol != 0))
194 		return -EPROTONOSUPPORT;
195 
196 	switch (sock->type) {
197 	case SOCK_STREAM:
198 		ops = &stream_ops;
199 		state = SS_UNCONNECTED;
200 		break;
201 	case SOCK_SEQPACKET:
202 		ops = &packet_ops;
203 		state = SS_UNCONNECTED;
204 		break;
205 	case SOCK_DGRAM:
206 	case SOCK_RDM:
207 		ops = &msg_ops;
208 		state = SS_READY;
209 		break;
210 	default:
211 		return -EPROTOTYPE;
212 	}
213 
214 	/* Allocate socket's protocol area */
215 
216 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
217 	if (sk == NULL)
218 		return -ENOMEM;
219 
220 	/* Allocate TIPC port for socket to use */
221 
222 	tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
223 				     TIPC_LOW_IMPORTANCE);
224 	if (unlikely(!tp_ptr)) {
225 		sk_free(sk);
226 		return -ENOMEM;
227 	}
228 
229 	/* Finish initializing socket data structures */
230 
231 	sock->ops = ops;
232 	sock->state = state;
233 
234 	sock_init_data(sock, sk);
235 	sk->sk_backlog_rcv = backlog_rcv;
236 	tipc_sk(sk)->p = tp_ptr;
237 	tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
238 
239 	spin_unlock_bh(tp_ptr->lock);
240 
241 	if (sock->state == SS_READY) {
242 		tipc_set_portunreturnable(tp_ptr->ref, 1);
243 		if (sock->type == SOCK_DGRAM)
244 			tipc_set_portunreliable(tp_ptr->ref, 1);
245 	}
246 
247 	return 0;
248 }
249 
250 /**
251  * release - destroy a TIPC socket
252  * @sock: socket to destroy
253  *
254  * This routine cleans up any messages that are still queued on the socket.
255  * For DGRAM and RDM socket types, all queued messages are rejected.
256  * For SEQPACKET and STREAM socket types, the first message is rejected
257  * and any others are discarded.  (If the first message on a STREAM socket
258  * is partially-read, it is discarded and the next one is rejected instead.)
259  *
260  * NOTE: Rejected messages are not necessarily returned to the sender!  They
261  * are returned or discarded according to the "destination droppable" setting
262  * specified for the message by the sender.
263  *
264  * Returns 0 on success, errno otherwise
265  */
266 
release(struct socket * sock)267 static int release(struct socket *sock)
268 {
269 	struct sock *sk = sock->sk;
270 	struct tipc_port *tport;
271 	struct sk_buff *buf;
272 	int res;
273 
274 	/*
275 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
276 	 * releases a pre-allocated child socket that was never used)
277 	 */
278 
279 	if (sk == NULL)
280 		return 0;
281 
282 	tport = tipc_sk_port(sk);
283 	lock_sock(sk);
284 
285 	/*
286 	 * Reject all unreceived messages, except on an active connection
287 	 * (which disconnects locally & sends a 'FIN+' to peer)
288 	 */
289 
290 	while (sock->state != SS_DISCONNECTING) {
291 		buf = __skb_dequeue(&sk->sk_receive_queue);
292 		if (buf == NULL)
293 			break;
294 		atomic_dec(&tipc_queue_size);
295 		if (TIPC_SKB_CB(buf)->handle != 0)
296 			buf_discard(buf);
297 		else {
298 			if ((sock->state == SS_CONNECTING) ||
299 			    (sock->state == SS_CONNECTED)) {
300 				sock->state = SS_DISCONNECTING;
301 				tipc_disconnect(tport->ref);
302 			}
303 			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
304 		}
305 	}
306 
307 	/*
308 	 * Delete TIPC port; this ensures no more messages are queued
309 	 * (also disconnects an active connection & sends a 'FIN-' to peer)
310 	 */
311 
312 	res = tipc_deleteport(tport->ref);
313 
314 	/* Discard any remaining (connection-based) messages in receive queue */
315 
316 	discard_rx_queue(sk);
317 
318 	/* Reject any messages that accumulated in backlog queue */
319 
320 	sock->state = SS_DISCONNECTING;
321 	release_sock(sk);
322 
323 	sock_put(sk);
324 	sock->sk = NULL;
325 
326 	return res;
327 }
328 
329 /**
330  * bind - associate or disassocate TIPC name(s) with a socket
331  * @sock: socket structure
332  * @uaddr: socket address describing name(s) and desired operation
333  * @uaddr_len: size of socket address data structure
334  *
335  * Name and name sequence binding is indicated using a positive scope value;
336  * a negative scope value unbinds the specified name.  Specifying no name
337  * (i.e. a socket address length of 0) unbinds all names from the socket.
338  *
339  * Returns 0 on success, errno otherwise
340  *
341  * NOTE: This routine doesn't need to take the socket lock since it doesn't
342  *       access any non-constant socket information.
343  */
344 
bind(struct socket * sock,struct sockaddr * uaddr,int uaddr_len)345 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
346 {
347 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
348 	u32 portref = tipc_sk_port(sock->sk)->ref;
349 
350 	if (unlikely(!uaddr_len))
351 		return tipc_withdraw(portref, 0, NULL);
352 
353 	if (uaddr_len < sizeof(struct sockaddr_tipc))
354 		return -EINVAL;
355 	if (addr->family != AF_TIPC)
356 		return -EAFNOSUPPORT;
357 
358 	if (addr->addrtype == TIPC_ADDR_NAME)
359 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
360 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
361 		return -EAFNOSUPPORT;
362 
363 	return (addr->scope > 0) ?
364 		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
365 		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
366 }
367 
368 /**
369  * get_name - get port ID of socket or peer socket
370  * @sock: socket structure
371  * @uaddr: area for returned socket address
372  * @uaddr_len: area for returned length of socket address
373  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
374  *
375  * Returns 0 on success, errno otherwise
376  *
377  * NOTE: This routine doesn't need to take the socket lock since it only
378  *       accesses socket information that is unchanging (or which changes in
379  *       a completely predictable manner).
380  */
381 
get_name(struct socket * sock,struct sockaddr * uaddr,int * uaddr_len,int peer)382 static int get_name(struct socket *sock, struct sockaddr *uaddr,
383 		    int *uaddr_len, int peer)
384 {
385 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
386 	struct tipc_sock *tsock = tipc_sk(sock->sk);
387 
388 	memset(addr, 0, sizeof(*addr));
389 	if (peer) {
390 		if ((sock->state != SS_CONNECTED) &&
391 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
392 			return -ENOTCONN;
393 		addr->addr.id.ref = tsock->peer_name.ref;
394 		addr->addr.id.node = tsock->peer_name.node;
395 	} else {
396 		addr->addr.id.ref = tsock->p->ref;
397 		addr->addr.id.node = tipc_own_addr;
398 	}
399 
400 	*uaddr_len = sizeof(*addr);
401 	addr->addrtype = TIPC_ADDR_ID;
402 	addr->family = AF_TIPC;
403 	addr->scope = 0;
404 	addr->addr.name.domain = 0;
405 
406 	return 0;
407 }
408 
409 /**
410  * poll - read and possibly block on pollmask
411  * @file: file structure associated with the socket
412  * @sock: socket for which to calculate the poll bits
413  * @wait: ???
414  *
415  * Returns pollmask value
416  *
417  * COMMENTARY:
418  * It appears that the usual socket locking mechanisms are not useful here
419  * since the pollmask info is potentially out-of-date the moment this routine
420  * exits.  TCP and other protocols seem to rely on higher level poll routines
421  * to handle any preventable race conditions, so TIPC will do the same ...
422  *
423  * TIPC sets the returned events as follows:
424  *
425  * socket state		flags set
426  * ------------		---------
427  * unconnected		no read flags
428  *			no write flags
429  *
430  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
431  *			no write flags
432  *
433  * connected		POLLIN/POLLRDNORM if data in rx queue
434  *			POLLOUT if port is not congested
435  *
436  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
437  *			no write flags
438  *
439  * listening		POLLIN if SYN in rx queue
440  *			no write flags
441  *
442  * ready		POLLIN/POLLRDNORM if data in rx queue
443  * [connectionless]	POLLOUT (since port cannot be congested)
444  *
445  * IMPORTANT: The fact that a read or write operation is indicated does NOT
446  * imply that the operation will succeed, merely that it should be performed
447  * and will not block.
448  */
449 
poll(struct file * file,struct socket * sock,poll_table * wait)450 static unsigned int poll(struct file *file, struct socket *sock,
451 			 poll_table *wait)
452 {
453 	struct sock *sk = sock->sk;
454 	u32 mask = 0;
455 
456 	poll_wait(file, sk_sleep(sk), wait);
457 
458 	switch ((int)sock->state) {
459 	case SS_READY:
460 	case SS_CONNECTED:
461 		if (!tipc_sk_port(sk)->congested)
462 			mask |= POLLOUT;
463 		/* fall thru' */
464 	case SS_CONNECTING:
465 	case SS_LISTENING:
466 		if (!skb_queue_empty(&sk->sk_receive_queue))
467 			mask |= (POLLIN | POLLRDNORM);
468 		break;
469 	case SS_DISCONNECTING:
470 		mask = (POLLIN | POLLRDNORM | POLLHUP);
471 		break;
472 	}
473 
474 	return mask;
475 }
476 
477 /**
478  * dest_name_check - verify user is permitted to send to specified port name
479  * @dest: destination address
480  * @m: descriptor for message to be sent
481  *
482  * Prevents restricted configuration commands from being issued by
483  * unauthorized users.
484  *
485  * Returns 0 if permission is granted, otherwise errno
486  */
487 
dest_name_check(struct sockaddr_tipc * dest,struct msghdr * m)488 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
489 {
490 	struct tipc_cfg_msg_hdr hdr;
491 
492 	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
493 		return 0;
494 	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
495 		return 0;
496 	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
497 		return -EACCES;
498 
499 	if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
500 		return -EMSGSIZE;
501 	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
502 		return -EFAULT;
503 	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
504 		return -EACCES;
505 
506 	return 0;
507 }
508 
509 /**
510  * send_msg - send message in connectionless manner
511  * @iocb: if NULL, indicates that socket lock is already held
512  * @sock: socket structure
513  * @m: message to send
514  * @total_len: length of message
515  *
516  * Message must have an destination specified explicitly.
517  * Used for SOCK_RDM and SOCK_DGRAM messages,
518  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
519  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
520  *
521  * Returns the number of bytes sent on success, or errno otherwise
522  */
523 
send_msg(struct kiocb * iocb,struct socket * sock,struct msghdr * m,size_t total_len)524 static int send_msg(struct kiocb *iocb, struct socket *sock,
525 		    struct msghdr *m, size_t total_len)
526 {
527 	struct sock *sk = sock->sk;
528 	struct tipc_port *tport = tipc_sk_port(sk);
529 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
530 	int needs_conn;
531 	int res = -EINVAL;
532 
533 	if (unlikely(!dest))
534 		return -EDESTADDRREQ;
535 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
536 		     (dest->family != AF_TIPC)))
537 		return -EINVAL;
538 
539 	if (iocb)
540 		lock_sock(sk);
541 
542 	needs_conn = (sock->state != SS_READY);
543 	if (unlikely(needs_conn)) {
544 		if (sock->state == SS_LISTENING) {
545 			res = -EPIPE;
546 			goto exit;
547 		}
548 		if (sock->state != SS_UNCONNECTED) {
549 			res = -EISCONN;
550 			goto exit;
551 		}
552 		if ((tport->published) ||
553 		    ((sock->type == SOCK_STREAM) && (total_len != 0))) {
554 			res = -EOPNOTSUPP;
555 			goto exit;
556 		}
557 		if (dest->addrtype == TIPC_ADDR_NAME) {
558 			tport->conn_type = dest->addr.name.name.type;
559 			tport->conn_instance = dest->addr.name.name.instance;
560 		}
561 
562 		/* Abort any pending connection attempts (very unlikely) */
563 
564 		reject_rx_queue(sk);
565 	}
566 
567 	do {
568 		if (dest->addrtype == TIPC_ADDR_NAME) {
569 			res = dest_name_check(dest, m);
570 			if (res)
571 				break;
572 			res = tipc_send2name(tport->ref,
573 					     &dest->addr.name.name,
574 					     dest->addr.name.domain,
575 					     m->msg_iovlen,
576 					     m->msg_iov);
577 		} else if (dest->addrtype == TIPC_ADDR_ID) {
578 			res = tipc_send2port(tport->ref,
579 					     &dest->addr.id,
580 					     m->msg_iovlen,
581 					     m->msg_iov);
582 		} else if (dest->addrtype == TIPC_ADDR_MCAST) {
583 			if (needs_conn) {
584 				res = -EOPNOTSUPP;
585 				break;
586 			}
587 			res = dest_name_check(dest, m);
588 			if (res)
589 				break;
590 			res = tipc_multicast(tport->ref,
591 					     &dest->addr.nameseq,
592 					     m->msg_iovlen,
593 					     m->msg_iov);
594 		}
595 		if (likely(res != -ELINKCONG)) {
596 			if (needs_conn && (res >= 0))
597 				sock->state = SS_CONNECTING;
598 			break;
599 		}
600 		if (m->msg_flags & MSG_DONTWAIT) {
601 			res = -EWOULDBLOCK;
602 			break;
603 		}
604 		release_sock(sk);
605 		res = wait_event_interruptible(*sk_sleep(sk),
606 					       !tport->congested);
607 		lock_sock(sk);
608 		if (res)
609 			break;
610 	} while (1);
611 
612 exit:
613 	if (iocb)
614 		release_sock(sk);
615 	return res;
616 }
617 
618 /**
619  * send_packet - send a connection-oriented message
620  * @iocb: if NULL, indicates that socket lock is already held
621  * @sock: socket structure
622  * @m: message to send
623  * @total_len: length of message
624  *
625  * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
626  *
627  * Returns the number of bytes sent on success, or errno otherwise
628  */
629 
send_packet(struct kiocb * iocb,struct socket * sock,struct msghdr * m,size_t total_len)630 static int send_packet(struct kiocb *iocb, struct socket *sock,
631 		       struct msghdr *m, size_t total_len)
632 {
633 	struct sock *sk = sock->sk;
634 	struct tipc_port *tport = tipc_sk_port(sk);
635 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
636 	int res;
637 
638 	/* Handle implied connection establishment */
639 
640 	if (unlikely(dest))
641 		return send_msg(iocb, sock, m, total_len);
642 
643 	if (iocb)
644 		lock_sock(sk);
645 
646 	do {
647 		if (unlikely(sock->state != SS_CONNECTED)) {
648 			if (sock->state == SS_DISCONNECTING)
649 				res = -EPIPE;
650 			else
651 				res = -ENOTCONN;
652 			break;
653 		}
654 
655 		res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
656 		if (likely(res != -ELINKCONG))
657 			break;
658 		if (m->msg_flags & MSG_DONTWAIT) {
659 			res = -EWOULDBLOCK;
660 			break;
661 		}
662 		release_sock(sk);
663 		res = wait_event_interruptible(*sk_sleep(sk),
664 			(!tport->congested || !tport->connected));
665 		lock_sock(sk);
666 		if (res)
667 			break;
668 	} while (1);
669 
670 	if (iocb)
671 		release_sock(sk);
672 	return res;
673 }
674 
675 /**
676  * send_stream - send stream-oriented data
677  * @iocb: (unused)
678  * @sock: socket structure
679  * @m: data to send
680  * @total_len: total length of data to be sent
681  *
682  * Used for SOCK_STREAM data.
683  *
684  * Returns the number of bytes sent on success (or partial success),
685  * or errno if no data sent
686  */
687 
send_stream(struct kiocb * iocb,struct socket * sock,struct msghdr * m,size_t total_len)688 static int send_stream(struct kiocb *iocb, struct socket *sock,
689 		       struct msghdr *m, size_t total_len)
690 {
691 	struct sock *sk = sock->sk;
692 	struct tipc_port *tport = tipc_sk_port(sk);
693 	struct msghdr my_msg;
694 	struct iovec my_iov;
695 	struct iovec *curr_iov;
696 	int curr_iovlen;
697 	char __user *curr_start;
698 	u32 hdr_size;
699 	int curr_left;
700 	int bytes_to_send;
701 	int bytes_sent;
702 	int res;
703 
704 	lock_sock(sk);
705 
706 	/* Handle special cases where there is no connection */
707 
708 	if (unlikely(sock->state != SS_CONNECTED)) {
709 		if (sock->state == SS_UNCONNECTED) {
710 			res = send_packet(NULL, sock, m, total_len);
711 			goto exit;
712 		} else if (sock->state == SS_DISCONNECTING) {
713 			res = -EPIPE;
714 			goto exit;
715 		} else {
716 			res = -ENOTCONN;
717 			goto exit;
718 		}
719 	}
720 
721 	if (unlikely(m->msg_name)) {
722 		res = -EISCONN;
723 		goto exit;
724 	}
725 
726 	/*
727 	 * Send each iovec entry using one or more messages
728 	 *
729 	 * Note: This algorithm is good for the most likely case
730 	 * (i.e. one large iovec entry), but could be improved to pass sets
731 	 * of small iovec entries into send_packet().
732 	 */
733 
734 	curr_iov = m->msg_iov;
735 	curr_iovlen = m->msg_iovlen;
736 	my_msg.msg_iov = &my_iov;
737 	my_msg.msg_iovlen = 1;
738 	my_msg.msg_flags = m->msg_flags;
739 	my_msg.msg_name = NULL;
740 	bytes_sent = 0;
741 
742 	hdr_size = msg_hdr_sz(&tport->phdr);
743 
744 	while (curr_iovlen--) {
745 		curr_start = curr_iov->iov_base;
746 		curr_left = curr_iov->iov_len;
747 
748 		while (curr_left) {
749 			bytes_to_send = tport->max_pkt - hdr_size;
750 			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
751 				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
752 			if (curr_left < bytes_to_send)
753 				bytes_to_send = curr_left;
754 			my_iov.iov_base = curr_start;
755 			my_iov.iov_len = bytes_to_send;
756 			res = send_packet(NULL, sock, &my_msg, 0);
757 			if (res < 0) {
758 				if (bytes_sent)
759 					res = bytes_sent;
760 				goto exit;
761 			}
762 			curr_left -= bytes_to_send;
763 			curr_start += bytes_to_send;
764 			bytes_sent += bytes_to_send;
765 		}
766 
767 		curr_iov++;
768 	}
769 	res = bytes_sent;
770 exit:
771 	release_sock(sk);
772 	return res;
773 }
774 
775 /**
776  * auto_connect - complete connection setup to a remote port
777  * @sock: socket structure
778  * @msg: peer's response message
779  *
780  * Returns 0 on success, errno otherwise
781  */
782 
auto_connect(struct socket * sock,struct tipc_msg * msg)783 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
784 {
785 	struct tipc_sock *tsock = tipc_sk(sock->sk);
786 
787 	if (msg_errcode(msg)) {
788 		sock->state = SS_DISCONNECTING;
789 		return -ECONNREFUSED;
790 	}
791 
792 	tsock->peer_name.ref = msg_origport(msg);
793 	tsock->peer_name.node = msg_orignode(msg);
794 	tipc_connect2port(tsock->p->ref, &tsock->peer_name);
795 	tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
796 	sock->state = SS_CONNECTED;
797 	return 0;
798 }
799 
800 /**
801  * set_orig_addr - capture sender's address for received message
802  * @m: descriptor for message info
803  * @msg: received message header
804  *
805  * Note: Address is not captured if not requested by receiver.
806  */
807 
set_orig_addr(struct msghdr * m,struct tipc_msg * msg)808 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
809 {
810 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
811 
812 	if (addr) {
813 		addr->family = AF_TIPC;
814 		addr->addrtype = TIPC_ADDR_ID;
815 		addr->addr.id.ref = msg_origport(msg);
816 		addr->addr.id.node = msg_orignode(msg);
817 		addr->addr.name.domain = 0;	/* could leave uninitialized */
818 		addr->scope = 0;		/* could leave uninitialized */
819 		m->msg_namelen = sizeof(struct sockaddr_tipc);
820 	}
821 }
822 
823 /**
824  * anc_data_recv - optionally capture ancillary data for received message
825  * @m: descriptor for message info
826  * @msg: received message header
827  * @tport: TIPC port associated with message
828  *
829  * Note: Ancillary data is not captured if not requested by receiver.
830  *
831  * Returns 0 if successful, otherwise errno
832  */
833 
anc_data_recv(struct msghdr * m,struct tipc_msg * msg,struct tipc_port * tport)834 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
835 				struct tipc_port *tport)
836 {
837 	u32 anc_data[3];
838 	u32 err;
839 	u32 dest_type;
840 	int has_name;
841 	int res;
842 
843 	if (likely(m->msg_controllen == 0))
844 		return 0;
845 
846 	/* Optionally capture errored message object(s) */
847 
848 	err = msg ? msg_errcode(msg) : 0;
849 	if (unlikely(err)) {
850 		anc_data[0] = err;
851 		anc_data[1] = msg_data_sz(msg);
852 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
853 		if (res)
854 			return res;
855 		if (anc_data[1]) {
856 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
857 				       msg_data(msg));
858 			if (res)
859 				return res;
860 		}
861 	}
862 
863 	/* Optionally capture message destination object */
864 
865 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
866 	switch (dest_type) {
867 	case TIPC_NAMED_MSG:
868 		has_name = 1;
869 		anc_data[0] = msg_nametype(msg);
870 		anc_data[1] = msg_namelower(msg);
871 		anc_data[2] = msg_namelower(msg);
872 		break;
873 	case TIPC_MCAST_MSG:
874 		has_name = 1;
875 		anc_data[0] = msg_nametype(msg);
876 		anc_data[1] = msg_namelower(msg);
877 		anc_data[2] = msg_nameupper(msg);
878 		break;
879 	case TIPC_CONN_MSG:
880 		has_name = (tport->conn_type != 0);
881 		anc_data[0] = tport->conn_type;
882 		anc_data[1] = tport->conn_instance;
883 		anc_data[2] = tport->conn_instance;
884 		break;
885 	default:
886 		has_name = 0;
887 	}
888 	if (has_name) {
889 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
890 		if (res)
891 			return res;
892 	}
893 
894 	return 0;
895 }
896 
897 /**
898  * recv_msg - receive packet-oriented message
899  * @iocb: (unused)
900  * @m: descriptor for message info
901  * @buf_len: total size of user buffer area
902  * @flags: receive flags
903  *
904  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
905  * If the complete message doesn't fit in user area, truncate it.
906  *
907  * Returns size of returned message data, errno otherwise
908  */
909 
recv_msg(struct kiocb * iocb,struct socket * sock,struct msghdr * m,size_t buf_len,int flags)910 static int recv_msg(struct kiocb *iocb, struct socket *sock,
911 		    struct msghdr *m, size_t buf_len, int flags)
912 {
913 	struct sock *sk = sock->sk;
914 	struct tipc_port *tport = tipc_sk_port(sk);
915 	struct sk_buff *buf;
916 	struct tipc_msg *msg;
917 	long timeout;
918 	unsigned int sz;
919 	u32 err;
920 	int res;
921 
922 	/* Catch invalid receive requests */
923 
924 	if (unlikely(!buf_len))
925 		return -EINVAL;
926 
927 	lock_sock(sk);
928 
929 	if (unlikely(sock->state == SS_UNCONNECTED)) {
930 		res = -ENOTCONN;
931 		goto exit;
932 	}
933 
934 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
935 restart:
936 
937 	/* Look for a message in receive queue; wait if necessary */
938 
939 	while (skb_queue_empty(&sk->sk_receive_queue)) {
940 		if (sock->state == SS_DISCONNECTING) {
941 			res = -ENOTCONN;
942 			goto exit;
943 		}
944 		if (timeout <= 0L) {
945 			res = timeout ? timeout : -EWOULDBLOCK;
946 			goto exit;
947 		}
948 		release_sock(sk);
949 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
950 							   tipc_rx_ready(sock),
951 							   timeout);
952 		lock_sock(sk);
953 	}
954 
955 	/* Look at first message in receive queue */
956 
957 	buf = skb_peek(&sk->sk_receive_queue);
958 	msg = buf_msg(buf);
959 	sz = msg_data_sz(msg);
960 	err = msg_errcode(msg);
961 
962 	/* Complete connection setup for an implied connect */
963 
964 	if (unlikely(sock->state == SS_CONNECTING)) {
965 		res = auto_connect(sock, msg);
966 		if (res)
967 			goto exit;
968 	}
969 
970 	/* Discard an empty non-errored message & try again */
971 
972 	if ((!sz) && (!err)) {
973 		advance_rx_queue(sk);
974 		goto restart;
975 	}
976 
977 	/* Capture sender's address (optional) */
978 
979 	set_orig_addr(m, msg);
980 
981 	/* Capture ancillary data (optional) */
982 
983 	res = anc_data_recv(m, msg, tport);
984 	if (res)
985 		goto exit;
986 
987 	/* Capture message data (if valid) & compute return value (always) */
988 
989 	if (!err) {
990 		if (unlikely(buf_len < sz)) {
991 			sz = buf_len;
992 			m->msg_flags |= MSG_TRUNC;
993 		}
994 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
995 					      m->msg_iov, sz);
996 		if (res)
997 			goto exit;
998 		res = sz;
999 	} else {
1000 		if ((sock->state == SS_READY) ||
1001 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1002 			res = 0;
1003 		else
1004 			res = -ECONNRESET;
1005 	}
1006 
1007 	/* Consume received message (optional) */
1008 
1009 	if (likely(!(flags & MSG_PEEK))) {
1010 		if ((sock->state != SS_READY) &&
1011 		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1012 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1013 		advance_rx_queue(sk);
1014 	}
1015 exit:
1016 	release_sock(sk);
1017 	return res;
1018 }
1019 
1020 /**
1021  * recv_stream - receive stream-oriented data
1022  * @iocb: (unused)
1023  * @m: descriptor for message info
1024  * @buf_len: total size of user buffer area
1025  * @flags: receive flags
1026  *
1027  * Used for SOCK_STREAM messages only.  If not enough data is available
1028  * will optionally wait for more; never truncates data.
1029  *
1030  * Returns size of returned message data, errno otherwise
1031  */
1032 
recv_stream(struct kiocb * iocb,struct socket * sock,struct msghdr * m,size_t buf_len,int flags)1033 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1034 		       struct msghdr *m, size_t buf_len, int flags)
1035 {
1036 	struct sock *sk = sock->sk;
1037 	struct tipc_port *tport = tipc_sk_port(sk);
1038 	struct sk_buff *buf;
1039 	struct tipc_msg *msg;
1040 	long timeout;
1041 	unsigned int sz;
1042 	int sz_to_copy, target, needed;
1043 	int sz_copied = 0;
1044 	u32 err;
1045 	int res = 0;
1046 
1047 	/* Catch invalid receive attempts */
1048 
1049 	if (unlikely(!buf_len))
1050 		return -EINVAL;
1051 
1052 	lock_sock(sk);
1053 
1054 	if (unlikely((sock->state == SS_UNCONNECTED) ||
1055 		     (sock->state == SS_CONNECTING))) {
1056 		res = -ENOTCONN;
1057 		goto exit;
1058 	}
1059 
1060 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1061 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1062 restart:
1063 
1064 	/* Look for a message in receive queue; wait if necessary */
1065 
1066 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1067 		if (sock->state == SS_DISCONNECTING) {
1068 			res = -ENOTCONN;
1069 			goto exit;
1070 		}
1071 		if (timeout <= 0L) {
1072 			res = timeout ? timeout : -EWOULDBLOCK;
1073 			goto exit;
1074 		}
1075 		release_sock(sk);
1076 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1077 							   tipc_rx_ready(sock),
1078 							   timeout);
1079 		lock_sock(sk);
1080 	}
1081 
1082 	/* Look at first message in receive queue */
1083 
1084 	buf = skb_peek(&sk->sk_receive_queue);
1085 	msg = buf_msg(buf);
1086 	sz = msg_data_sz(msg);
1087 	err = msg_errcode(msg);
1088 
1089 	/* Discard an empty non-errored message & try again */
1090 
1091 	if ((!sz) && (!err)) {
1092 		advance_rx_queue(sk);
1093 		goto restart;
1094 	}
1095 
1096 	/* Optionally capture sender's address & ancillary data of first msg */
1097 
1098 	if (sz_copied == 0) {
1099 		set_orig_addr(m, msg);
1100 		res = anc_data_recv(m, msg, tport);
1101 		if (res)
1102 			goto exit;
1103 	}
1104 
1105 	/* Capture message data (if valid) & compute return value (always) */
1106 
1107 	if (!err) {
1108 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1109 
1110 		sz -= offset;
1111 		needed = (buf_len - sz_copied);
1112 		sz_to_copy = (sz <= needed) ? sz : needed;
1113 
1114 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1115 					      m->msg_iov, sz_to_copy);
1116 		if (res)
1117 			goto exit;
1118 
1119 		sz_copied += sz_to_copy;
1120 
1121 		if (sz_to_copy < sz) {
1122 			if (!(flags & MSG_PEEK))
1123 				TIPC_SKB_CB(buf)->handle =
1124 				(void *)(unsigned long)(offset + sz_to_copy);
1125 			goto exit;
1126 		}
1127 	} else {
1128 		if (sz_copied != 0)
1129 			goto exit; /* can't add error msg to valid data */
1130 
1131 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1132 			res = 0;
1133 		else
1134 			res = -ECONNRESET;
1135 	}
1136 
1137 	/* Consume received message (optional) */
1138 
1139 	if (likely(!(flags & MSG_PEEK))) {
1140 		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1141 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1142 		advance_rx_queue(sk);
1143 	}
1144 
1145 	/* Loop around if more data is required */
1146 
1147 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1148 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1149 	    (sz_copied < target)) &&	/* and more is ready or required */
1150 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1151 	    (!err))			/* and haven't reached a FIN */
1152 		goto restart;
1153 
1154 exit:
1155 	release_sock(sk);
1156 	return sz_copied ? sz_copied : res;
1157 }
1158 
1159 /**
1160  * rx_queue_full - determine if receive queue can accept another message
1161  * @msg: message to be added to queue
1162  * @queue_size: current size of queue
1163  * @base: nominal maximum size of queue
1164  *
1165  * Returns 1 if queue is unable to accept message, 0 otherwise
1166  */
1167 
rx_queue_full(struct tipc_msg * msg,u32 queue_size,u32 base)1168 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1169 {
1170 	u32 threshold;
1171 	u32 imp = msg_importance(msg);
1172 
1173 	if (imp == TIPC_LOW_IMPORTANCE)
1174 		threshold = base;
1175 	else if (imp == TIPC_MEDIUM_IMPORTANCE)
1176 		threshold = base * 2;
1177 	else if (imp == TIPC_HIGH_IMPORTANCE)
1178 		threshold = base * 100;
1179 	else
1180 		return 0;
1181 
1182 	if (msg_connected(msg))
1183 		threshold *= 4;
1184 
1185 	return queue_size >= threshold;
1186 }
1187 
1188 /**
1189  * filter_rcv - validate incoming message
1190  * @sk: socket
1191  * @buf: message
1192  *
1193  * Enqueues message on receive queue if acceptable; optionally handles
1194  * disconnect indication for a connected socket.
1195  *
1196  * Called with socket lock already taken; port lock may also be taken.
1197  *
1198  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1199  */
1200 
filter_rcv(struct sock * sk,struct sk_buff * buf)1201 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1202 {
1203 	struct socket *sock = sk->sk_socket;
1204 	struct tipc_msg *msg = buf_msg(buf);
1205 	u32 recv_q_len;
1206 
1207 	/* Reject message if it is wrong sort of message for socket */
1208 
1209 	/*
1210 	 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1211 	 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1212 	 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1213 	 */
1214 
1215 	if (sock->state == SS_READY) {
1216 		if (msg_connected(msg))
1217 			return TIPC_ERR_NO_PORT;
1218 	} else {
1219 		if (msg_mcast(msg))
1220 			return TIPC_ERR_NO_PORT;
1221 		if (sock->state == SS_CONNECTED) {
1222 			if (!msg_connected(msg))
1223 				return TIPC_ERR_NO_PORT;
1224 		} else if (sock->state == SS_CONNECTING) {
1225 			if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1226 				return TIPC_ERR_NO_PORT;
1227 		} else if (sock->state == SS_LISTENING) {
1228 			if (msg_connected(msg) || msg_errcode(msg))
1229 				return TIPC_ERR_NO_PORT;
1230 		} else if (sock->state == SS_DISCONNECTING) {
1231 			return TIPC_ERR_NO_PORT;
1232 		} else /* (sock->state == SS_UNCONNECTED) */ {
1233 			if (msg_connected(msg) || msg_errcode(msg))
1234 				return TIPC_ERR_NO_PORT;
1235 		}
1236 	}
1237 
1238 	/* Reject message if there isn't room to queue it */
1239 
1240 	recv_q_len = (u32)atomic_read(&tipc_queue_size);
1241 	if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1242 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1243 			return TIPC_ERR_OVERLOAD;
1244 	}
1245 	recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1246 	if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1247 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1248 			return TIPC_ERR_OVERLOAD;
1249 	}
1250 
1251 	/* Enqueue message (finally!) */
1252 
1253 	TIPC_SKB_CB(buf)->handle = 0;
1254 	atomic_inc(&tipc_queue_size);
1255 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1256 
1257 	/* Initiate connection termination for an incoming 'FIN' */
1258 
1259 	if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1260 		sock->state = SS_DISCONNECTING;
1261 		tipc_disconnect_port(tipc_sk_port(sk));
1262 	}
1263 
1264 	if (waitqueue_active(sk_sleep(sk)))
1265 		wake_up_interruptible(sk_sleep(sk));
1266 	return TIPC_OK;
1267 }
1268 
1269 /**
1270  * backlog_rcv - handle incoming message from backlog queue
1271  * @sk: socket
1272  * @buf: message
1273  *
1274  * Caller must hold socket lock, but not port lock.
1275  *
1276  * Returns 0
1277  */
1278 
backlog_rcv(struct sock * sk,struct sk_buff * buf)1279 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1280 {
1281 	u32 res;
1282 
1283 	res = filter_rcv(sk, buf);
1284 	if (res)
1285 		tipc_reject_msg(buf, res);
1286 	return 0;
1287 }
1288 
1289 /**
1290  * dispatch - handle incoming message
1291  * @tport: TIPC port that received message
1292  * @buf: message
1293  *
1294  * Called with port lock already taken.
1295  *
1296  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1297  */
1298 
dispatch(struct tipc_port * tport,struct sk_buff * buf)1299 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1300 {
1301 	struct sock *sk = (struct sock *)tport->usr_handle;
1302 	u32 res;
1303 
1304 	/*
1305 	 * Process message if socket is unlocked; otherwise add to backlog queue
1306 	 *
1307 	 * This code is based on sk_receive_skb(), but must be distinct from it
1308 	 * since a TIPC-specific filter/reject mechanism is utilized
1309 	 */
1310 
1311 	bh_lock_sock(sk);
1312 	if (!sock_owned_by_user(sk)) {
1313 		res = filter_rcv(sk, buf);
1314 	} else {
1315 		if (sk_add_backlog(sk, buf))
1316 			res = TIPC_ERR_OVERLOAD;
1317 		else
1318 			res = TIPC_OK;
1319 	}
1320 	bh_unlock_sock(sk);
1321 
1322 	return res;
1323 }
1324 
1325 /**
1326  * wakeupdispatch - wake up port after congestion
1327  * @tport: port to wakeup
1328  *
1329  * Called with port lock already taken.
1330  */
1331 
wakeupdispatch(struct tipc_port * tport)1332 static void wakeupdispatch(struct tipc_port *tport)
1333 {
1334 	struct sock *sk = (struct sock *)tport->usr_handle;
1335 
1336 	if (waitqueue_active(sk_sleep(sk)))
1337 		wake_up_interruptible(sk_sleep(sk));
1338 }
1339 
1340 /**
1341  * connect - establish a connection to another TIPC port
1342  * @sock: socket structure
1343  * @dest: socket address for destination port
1344  * @destlen: size of socket address data structure
1345  * @flags: file-related flags associated with socket
1346  *
1347  * Returns 0 on success, errno otherwise
1348  */
1349 
connect(struct socket * sock,struct sockaddr * dest,int destlen,int flags)1350 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1351 		   int flags)
1352 {
1353 	struct sock *sk = sock->sk;
1354 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1355 	struct msghdr m = {NULL,};
1356 	struct sk_buff *buf;
1357 	struct tipc_msg *msg;
1358 	long timeout;
1359 	int res;
1360 
1361 	lock_sock(sk);
1362 
1363 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1364 
1365 	if (sock->state == SS_READY) {
1366 		res = -EOPNOTSUPP;
1367 		goto exit;
1368 	}
1369 
1370 	/* For now, TIPC does not support the non-blocking form of connect() */
1371 
1372 	if (flags & O_NONBLOCK) {
1373 		res = -EOPNOTSUPP;
1374 		goto exit;
1375 	}
1376 
1377 	/* Issue Posix-compliant error code if socket is in the wrong state */
1378 
1379 	if (sock->state == SS_LISTENING) {
1380 		res = -EOPNOTSUPP;
1381 		goto exit;
1382 	}
1383 	if (sock->state == SS_CONNECTING) {
1384 		res = -EALREADY;
1385 		goto exit;
1386 	}
1387 	if (sock->state != SS_UNCONNECTED) {
1388 		res = -EISCONN;
1389 		goto exit;
1390 	}
1391 
1392 	/*
1393 	 * Reject connection attempt using multicast address
1394 	 *
1395 	 * Note: send_msg() validates the rest of the address fields,
1396 	 *       so there's no need to do it here
1397 	 */
1398 
1399 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1400 		res = -EINVAL;
1401 		goto exit;
1402 	}
1403 
1404 	/* Reject any messages already in receive queue (very unlikely) */
1405 
1406 	reject_rx_queue(sk);
1407 
1408 	/* Send a 'SYN-' to destination */
1409 
1410 	m.msg_name = dest;
1411 	m.msg_namelen = destlen;
1412 	res = send_msg(NULL, sock, &m, 0);
1413 	if (res < 0)
1414 		goto exit;
1415 
1416 	/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1417 
1418 	timeout = tipc_sk(sk)->conn_timeout;
1419 	release_sock(sk);
1420 	res = wait_event_interruptible_timeout(*sk_sleep(sk),
1421 			(!skb_queue_empty(&sk->sk_receive_queue) ||
1422 			(sock->state != SS_CONNECTING)),
1423 			timeout ? timeout : MAX_SCHEDULE_TIMEOUT);
1424 	lock_sock(sk);
1425 
1426 	if (res > 0) {
1427 		buf = skb_peek(&sk->sk_receive_queue);
1428 		if (buf != NULL) {
1429 			msg = buf_msg(buf);
1430 			res = auto_connect(sock, msg);
1431 			if (!res) {
1432 				if (!msg_data_sz(msg))
1433 					advance_rx_queue(sk);
1434 			}
1435 		} else {
1436 			if (sock->state == SS_CONNECTED)
1437 				res = -EISCONN;
1438 			else
1439 				res = -ECONNREFUSED;
1440 		}
1441 	} else {
1442 		if (res == 0)
1443 			res = -ETIMEDOUT;
1444 		else
1445 			; /* leave "res" unchanged */
1446 		sock->state = SS_DISCONNECTING;
1447 	}
1448 
1449 exit:
1450 	release_sock(sk);
1451 	return res;
1452 }
1453 
1454 /**
1455  * listen - allow socket to listen for incoming connections
1456  * @sock: socket structure
1457  * @len: (unused)
1458  *
1459  * Returns 0 on success, errno otherwise
1460  */
1461 
listen(struct socket * sock,int len)1462 static int listen(struct socket *sock, int len)
1463 {
1464 	struct sock *sk = sock->sk;
1465 	int res;
1466 
1467 	lock_sock(sk);
1468 
1469 	if (sock->state == SS_READY)
1470 		res = -EOPNOTSUPP;
1471 	else if (sock->state != SS_UNCONNECTED)
1472 		res = -EINVAL;
1473 	else {
1474 		sock->state = SS_LISTENING;
1475 		res = 0;
1476 	}
1477 
1478 	release_sock(sk);
1479 	return res;
1480 }
1481 
1482 /**
1483  * accept - wait for connection request
1484  * @sock: listening socket
1485  * @newsock: new socket that is to be connected
1486  * @flags: file-related flags associated with socket
1487  *
1488  * Returns 0 on success, errno otherwise
1489  */
1490 
accept(struct socket * sock,struct socket * new_sock,int flags)1491 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1492 {
1493 	struct sock *sk = sock->sk;
1494 	struct sk_buff *buf;
1495 	int res;
1496 
1497 	lock_sock(sk);
1498 
1499 	if (sock->state == SS_READY) {
1500 		res = -EOPNOTSUPP;
1501 		goto exit;
1502 	}
1503 	if (sock->state != SS_LISTENING) {
1504 		res = -EINVAL;
1505 		goto exit;
1506 	}
1507 
1508 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1509 		if (flags & O_NONBLOCK) {
1510 			res = -EWOULDBLOCK;
1511 			goto exit;
1512 		}
1513 		release_sock(sk);
1514 		res = wait_event_interruptible(*sk_sleep(sk),
1515 				(!skb_queue_empty(&sk->sk_receive_queue)));
1516 		lock_sock(sk);
1517 		if (res)
1518 			goto exit;
1519 	}
1520 
1521 	buf = skb_peek(&sk->sk_receive_queue);
1522 
1523 	res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1524 	if (!res) {
1525 		struct sock *new_sk = new_sock->sk;
1526 		struct tipc_sock *new_tsock = tipc_sk(new_sk);
1527 		struct tipc_port *new_tport = new_tsock->p;
1528 		u32 new_ref = new_tport->ref;
1529 		struct tipc_msg *msg = buf_msg(buf);
1530 
1531 		lock_sock(new_sk);
1532 
1533 		/*
1534 		 * Reject any stray messages received by new socket
1535 		 * before the socket lock was taken (very, very unlikely)
1536 		 */
1537 
1538 		reject_rx_queue(new_sk);
1539 
1540 		/* Connect new socket to it's peer */
1541 
1542 		new_tsock->peer_name.ref = msg_origport(msg);
1543 		new_tsock->peer_name.node = msg_orignode(msg);
1544 		tipc_connect2port(new_ref, &new_tsock->peer_name);
1545 		new_sock->state = SS_CONNECTED;
1546 
1547 		tipc_set_portimportance(new_ref, msg_importance(msg));
1548 		if (msg_named(msg)) {
1549 			new_tport->conn_type = msg_nametype(msg);
1550 			new_tport->conn_instance = msg_nameinst(msg);
1551 		}
1552 
1553 		/*
1554 		 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1555 		 * Respond to 'SYN+' by queuing it on new socket.
1556 		 */
1557 
1558 		if (!msg_data_sz(msg)) {
1559 			struct msghdr m = {NULL,};
1560 
1561 			advance_rx_queue(sk);
1562 			send_packet(NULL, new_sock, &m, 0);
1563 		} else {
1564 			__skb_dequeue(&sk->sk_receive_queue);
1565 			__skb_queue_head(&new_sk->sk_receive_queue, buf);
1566 		}
1567 		release_sock(new_sk);
1568 	}
1569 exit:
1570 	release_sock(sk);
1571 	return res;
1572 }
1573 
1574 /**
1575  * shutdown - shutdown socket connection
1576  * @sock: socket structure
1577  * @how: direction to close (must be SHUT_RDWR)
1578  *
1579  * Terminates connection (if necessary), then purges socket's receive queue.
1580  *
1581  * Returns 0 on success, errno otherwise
1582  */
1583 
shutdown(struct socket * sock,int how)1584 static int shutdown(struct socket *sock, int how)
1585 {
1586 	struct sock *sk = sock->sk;
1587 	struct tipc_port *tport = tipc_sk_port(sk);
1588 	struct sk_buff *buf;
1589 	int res;
1590 
1591 	if (how != SHUT_RDWR)
1592 		return -EINVAL;
1593 
1594 	lock_sock(sk);
1595 
1596 	switch (sock->state) {
1597 	case SS_CONNECTING:
1598 	case SS_CONNECTED:
1599 
1600 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1601 restart:
1602 		buf = __skb_dequeue(&sk->sk_receive_queue);
1603 		if (buf) {
1604 			atomic_dec(&tipc_queue_size);
1605 			if (TIPC_SKB_CB(buf)->handle != 0) {
1606 				buf_discard(buf);
1607 				goto restart;
1608 			}
1609 			tipc_disconnect(tport->ref);
1610 			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1611 		} else {
1612 			tipc_shutdown(tport->ref);
1613 		}
1614 
1615 		sock->state = SS_DISCONNECTING;
1616 
1617 		/* fall through */
1618 
1619 	case SS_DISCONNECTING:
1620 
1621 		/* Discard any unreceived messages; wake up sleeping tasks */
1622 
1623 		discard_rx_queue(sk);
1624 		if (waitqueue_active(sk_sleep(sk)))
1625 			wake_up_interruptible(sk_sleep(sk));
1626 		res = 0;
1627 		break;
1628 
1629 	default:
1630 		res = -ENOTCONN;
1631 	}
1632 
1633 	release_sock(sk);
1634 	return res;
1635 }
1636 
1637 /**
1638  * setsockopt - set socket option
1639  * @sock: socket structure
1640  * @lvl: option level
1641  * @opt: option identifier
1642  * @ov: pointer to new option value
1643  * @ol: length of option value
1644  *
1645  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1646  * (to ease compatibility).
1647  *
1648  * Returns 0 on success, errno otherwise
1649  */
1650 
setsockopt(struct socket * sock,int lvl,int opt,char __user * ov,unsigned int ol)1651 static int setsockopt(struct socket *sock,
1652 		      int lvl, int opt, char __user *ov, unsigned int ol)
1653 {
1654 	struct sock *sk = sock->sk;
1655 	struct tipc_port *tport = tipc_sk_port(sk);
1656 	u32 value;
1657 	int res;
1658 
1659 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1660 		return 0;
1661 	if (lvl != SOL_TIPC)
1662 		return -ENOPROTOOPT;
1663 	if (ol < sizeof(value))
1664 		return -EINVAL;
1665 	res = get_user(value, (u32 __user *)ov);
1666 	if (res)
1667 		return res;
1668 
1669 	lock_sock(sk);
1670 
1671 	switch (opt) {
1672 	case TIPC_IMPORTANCE:
1673 		res = tipc_set_portimportance(tport->ref, value);
1674 		break;
1675 	case TIPC_SRC_DROPPABLE:
1676 		if (sock->type != SOCK_STREAM)
1677 			res = tipc_set_portunreliable(tport->ref, value);
1678 		else
1679 			res = -ENOPROTOOPT;
1680 		break;
1681 	case TIPC_DEST_DROPPABLE:
1682 		res = tipc_set_portunreturnable(tport->ref, value);
1683 		break;
1684 	case TIPC_CONN_TIMEOUT:
1685 		tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value);
1686 		/* no need to set "res", since already 0 at this point */
1687 		break;
1688 	default:
1689 		res = -EINVAL;
1690 	}
1691 
1692 	release_sock(sk);
1693 
1694 	return res;
1695 }
1696 
1697 /**
1698  * getsockopt - get socket option
1699  * @sock: socket structure
1700  * @lvl: option level
1701  * @opt: option identifier
1702  * @ov: receptacle for option value
1703  * @ol: receptacle for length of option value
1704  *
1705  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1706  * (to ease compatibility).
1707  *
1708  * Returns 0 on success, errno otherwise
1709  */
1710 
getsockopt(struct socket * sock,int lvl,int opt,char __user * ov,int __user * ol)1711 static int getsockopt(struct socket *sock,
1712 		      int lvl, int opt, char __user *ov, int __user *ol)
1713 {
1714 	struct sock *sk = sock->sk;
1715 	struct tipc_port *tport = tipc_sk_port(sk);
1716 	int len;
1717 	u32 value;
1718 	int res;
1719 
1720 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1721 		return put_user(0, ol);
1722 	if (lvl != SOL_TIPC)
1723 		return -ENOPROTOOPT;
1724 	res = get_user(len, ol);
1725 	if (res)
1726 		return res;
1727 
1728 	lock_sock(sk);
1729 
1730 	switch (opt) {
1731 	case TIPC_IMPORTANCE:
1732 		res = tipc_portimportance(tport->ref, &value);
1733 		break;
1734 	case TIPC_SRC_DROPPABLE:
1735 		res = tipc_portunreliable(tport->ref, &value);
1736 		break;
1737 	case TIPC_DEST_DROPPABLE:
1738 		res = tipc_portunreturnable(tport->ref, &value);
1739 		break;
1740 	case TIPC_CONN_TIMEOUT:
1741 		value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout);
1742 		/* no need to set "res", since already 0 at this point */
1743 		break;
1744 	case TIPC_NODE_RECVQ_DEPTH:
1745 		value = (u32)atomic_read(&tipc_queue_size);
1746 		break;
1747 	case TIPC_SOCK_RECVQ_DEPTH:
1748 		value = skb_queue_len(&sk->sk_receive_queue);
1749 		break;
1750 	default:
1751 		res = -EINVAL;
1752 	}
1753 
1754 	release_sock(sk);
1755 
1756 	if (res)
1757 		return res;	/* "get" failed */
1758 
1759 	if (len < sizeof(value))
1760 		return -EINVAL;
1761 
1762 	if (copy_to_user(ov, &value, sizeof(value)))
1763 		return -EFAULT;
1764 
1765 	return put_user(sizeof(value), ol);
1766 }
1767 
1768 /**
1769  * Protocol switches for the various types of TIPC sockets
1770  */
1771 
1772 static const struct proto_ops msg_ops = {
1773 	.owner		= THIS_MODULE,
1774 	.family		= AF_TIPC,
1775 	.release	= release,
1776 	.bind		= bind,
1777 	.connect	= connect,
1778 	.socketpair	= sock_no_socketpair,
1779 	.accept		= accept,
1780 	.getname	= get_name,
1781 	.poll		= poll,
1782 	.ioctl		= sock_no_ioctl,
1783 	.listen		= listen,
1784 	.shutdown	= shutdown,
1785 	.setsockopt	= setsockopt,
1786 	.getsockopt	= getsockopt,
1787 	.sendmsg	= send_msg,
1788 	.recvmsg	= recv_msg,
1789 	.mmap		= sock_no_mmap,
1790 	.sendpage	= sock_no_sendpage
1791 };
1792 
1793 static const struct proto_ops packet_ops = {
1794 	.owner		= THIS_MODULE,
1795 	.family		= AF_TIPC,
1796 	.release	= release,
1797 	.bind		= bind,
1798 	.connect	= connect,
1799 	.socketpair	= sock_no_socketpair,
1800 	.accept		= accept,
1801 	.getname	= get_name,
1802 	.poll		= poll,
1803 	.ioctl		= sock_no_ioctl,
1804 	.listen		= listen,
1805 	.shutdown	= shutdown,
1806 	.setsockopt	= setsockopt,
1807 	.getsockopt	= getsockopt,
1808 	.sendmsg	= send_packet,
1809 	.recvmsg	= recv_msg,
1810 	.mmap		= sock_no_mmap,
1811 	.sendpage	= sock_no_sendpage
1812 };
1813 
1814 static const struct proto_ops stream_ops = {
1815 	.owner		= THIS_MODULE,
1816 	.family		= AF_TIPC,
1817 	.release	= release,
1818 	.bind		= bind,
1819 	.connect	= connect,
1820 	.socketpair	= sock_no_socketpair,
1821 	.accept		= accept,
1822 	.getname	= get_name,
1823 	.poll		= poll,
1824 	.ioctl		= sock_no_ioctl,
1825 	.listen		= listen,
1826 	.shutdown	= shutdown,
1827 	.setsockopt	= setsockopt,
1828 	.getsockopt	= getsockopt,
1829 	.sendmsg	= send_stream,
1830 	.recvmsg	= recv_stream,
1831 	.mmap		= sock_no_mmap,
1832 	.sendpage	= sock_no_sendpage
1833 };
1834 
1835 static const struct net_proto_family tipc_family_ops = {
1836 	.owner		= THIS_MODULE,
1837 	.family		= AF_TIPC,
1838 	.create		= tipc_create
1839 };
1840 
1841 static struct proto tipc_proto = {
1842 	.name		= "TIPC",
1843 	.owner		= THIS_MODULE,
1844 	.obj_size	= sizeof(struct tipc_sock)
1845 };
1846 
1847 /**
1848  * tipc_socket_init - initialize TIPC socket interface
1849  *
1850  * Returns 0 on success, errno otherwise
1851  */
tipc_socket_init(void)1852 int tipc_socket_init(void)
1853 {
1854 	int res;
1855 
1856 	res = proto_register(&tipc_proto, 1);
1857 	if (res) {
1858 		err("Failed to register TIPC protocol type\n");
1859 		goto out;
1860 	}
1861 
1862 	res = sock_register(&tipc_family_ops);
1863 	if (res) {
1864 		err("Failed to register TIPC socket type\n");
1865 		proto_unregister(&tipc_proto);
1866 		goto out;
1867 	}
1868 
1869 	sockets_enabled = 1;
1870  out:
1871 	return res;
1872 }
1873 
1874 /**
1875  * tipc_socket_stop - stop TIPC socket interface
1876  */
1877 
tipc_socket_stop(void)1878 void tipc_socket_stop(void)
1879 {
1880 	if (!sockets_enabled)
1881 		return;
1882 
1883 	sockets_enabled = 0;
1884 	sock_unregister(tipc_family_ops.family);
1885 	proto_unregister(&tipc_proto);
1886 }
1887 
1888