xref: /openbmc/linux/net/tipc/socket.c (revision 22246614)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, Ericsson AB
5  * Copyright (c) 2004-2007, 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 <linux/module.h>
38 #include <linux/types.h>
39 #include <linux/net.h>
40 #include <linux/socket.h>
41 #include <linux/errno.h>
42 #include <linux/mm.h>
43 #include <linux/slab.h>
44 #include <linux/poll.h>
45 #include <linux/fcntl.h>
46 #include <asm/string.h>
47 #include <asm/atomic.h>
48 #include <net/sock.h>
49 
50 #include <linux/tipc.h>
51 #include <linux/tipc_config.h>
52 #include <net/tipc/tipc_msg.h>
53 #include <net/tipc/tipc_port.h>
54 
55 #include "core.h"
56 
57 #define SS_LISTENING	-1	/* socket is listening */
58 #define SS_READY	-2	/* socket is connectionless */
59 
60 #define OVERLOAD_LIMIT_BASE	5000
61 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
62 
63 struct tipc_sock {
64 	struct sock sk;
65 	struct tipc_port *p;
66 };
67 
68 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
69 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
70 
71 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
72 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
73 static void wakeupdispatch(struct tipc_port *tport);
74 
75 static const struct proto_ops packet_ops;
76 static const struct proto_ops stream_ops;
77 static const struct proto_ops msg_ops;
78 
79 static struct proto tipc_proto;
80 
81 static int sockets_enabled = 0;
82 
83 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
84 
85 /*
86  * Revised TIPC socket locking policy:
87  *
88  * Most socket operations take the standard socket lock when they start
89  * and hold it until they finish (or until they need to sleep).  Acquiring
90  * this lock grants the owner exclusive access to the fields of the socket
91  * data structures, with the exception of the backlog queue.  A few socket
92  * operations can be done without taking the socket lock because they only
93  * read socket information that never changes during the life of the socket.
94  *
95  * Socket operations may acquire the lock for the associated TIPC port if they
96  * need to perform an operation on the port.  If any routine needs to acquire
97  * both the socket lock and the port lock it must take the socket lock first
98  * to avoid the risk of deadlock.
99  *
100  * The dispatcher handling incoming messages cannot grab the socket lock in
101  * the standard fashion, since invoked it runs at the BH level and cannot block.
102  * Instead, it checks to see if the socket lock is currently owned by someone,
103  * and either handles the message itself or adds it to the socket's backlog
104  * queue; in the latter case the queued message is processed once the process
105  * owning the socket lock releases it.
106  *
107  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
108  * the problem of a blocked socket operation preventing any other operations
109  * from occurring.  However, applications must be careful if they have
110  * multiple threads trying to send (or receive) on the same socket, as these
111  * operations might interfere with each other.  For example, doing a connect
112  * and a receive at the same time might allow the receive to consume the
113  * ACK message meant for the connect.  While additional work could be done
114  * to try and overcome this, it doesn't seem to be worthwhile at the present.
115  *
116  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
117  * that another operation that must be performed in a non-blocking manner is
118  * not delayed for very long because the lock has already been taken.
119  *
120  * NOTE: This code assumes that certain fields of a port/socket pair are
121  * constant over its lifetime; such fields can be examined without taking
122  * the socket lock and/or port lock, and do not need to be re-read even
123  * after resuming processing after waiting.  These fields include:
124  *   - socket type
125  *   - pointer to socket sk structure (aka tipc_sock structure)
126  *   - pointer to port structure
127  *   - port reference
128  */
129 
130 /**
131  * advance_rx_queue - discard first buffer in socket receive queue
132  *
133  * Caller must hold socket lock
134  */
135 
136 static void advance_rx_queue(struct sock *sk)
137 {
138 	buf_discard(__skb_dequeue(&sk->sk_receive_queue));
139 	atomic_dec(&tipc_queue_size);
140 }
141 
142 /**
143  * discard_rx_queue - discard all buffers in socket receive queue
144  *
145  * Caller must hold socket lock
146  */
147 
148 static void discard_rx_queue(struct sock *sk)
149 {
150 	struct sk_buff *buf;
151 
152 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
153 		atomic_dec(&tipc_queue_size);
154 		buf_discard(buf);
155 	}
156 }
157 
158 /**
159  * reject_rx_queue - reject all buffers in socket receive queue
160  *
161  * Caller must hold socket lock
162  */
163 
164 static void reject_rx_queue(struct sock *sk)
165 {
166 	struct sk_buff *buf;
167 
168 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
169 		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
170 		atomic_dec(&tipc_queue_size);
171 	}
172 }
173 
174 /**
175  * tipc_create - create a TIPC socket
176  * @net: network namespace (must be default network)
177  * @sock: pre-allocated socket structure
178  * @protocol: protocol indicator (must be 0)
179  *
180  * This routine creates additional data structures used by the TIPC socket,
181  * initializes them, and links them together.
182  *
183  * Returns 0 on success, errno otherwise
184  */
185 
186 static int tipc_create(struct net *net, struct socket *sock, int protocol)
187 {
188 	const struct proto_ops *ops;
189 	socket_state state;
190 	struct sock *sk;
191 	u32 portref;
192 
193 	/* Validate arguments */
194 
195 	if (net != &init_net)
196 		return -EAFNOSUPPORT;
197 
198 	if (unlikely(protocol != 0))
199 		return -EPROTONOSUPPORT;
200 
201 	switch (sock->type) {
202 	case SOCK_STREAM:
203 		ops = &stream_ops;
204 		state = SS_UNCONNECTED;
205 		break;
206 	case SOCK_SEQPACKET:
207 		ops = &packet_ops;
208 		state = SS_UNCONNECTED;
209 		break;
210 	case SOCK_DGRAM:
211 	case SOCK_RDM:
212 		ops = &msg_ops;
213 		state = SS_READY;
214 		break;
215 	default:
216 		return -EPROTOTYPE;
217 	}
218 
219 	/* Allocate socket's protocol area */
220 
221 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
222 	if (sk == NULL)
223 		return -ENOMEM;
224 
225 	/* Allocate TIPC port for socket to use */
226 
227 	portref = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
228 				      TIPC_LOW_IMPORTANCE);
229 	if (unlikely(portref == 0)) {
230 		sk_free(sk);
231 		return -ENOMEM;
232 	}
233 
234 	/* Finish initializing socket data structures */
235 
236 	sock->ops = ops;
237 	sock->state = state;
238 
239 	sock_init_data(sock, sk);
240 	sk->sk_rcvtimeo = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
241 	sk->sk_backlog_rcv = backlog_rcv;
242 	tipc_sk(sk)->p = tipc_get_port(portref);
243 
244 	if (sock->state == SS_READY) {
245 		tipc_set_portunreturnable(portref, 1);
246 		if (sock->type == SOCK_DGRAM)
247 			tipc_set_portunreliable(portref, 1);
248 	}
249 
250 	atomic_inc(&tipc_user_count);
251 	return 0;
252 }
253 
254 /**
255  * release - destroy a TIPC socket
256  * @sock: socket to destroy
257  *
258  * This routine cleans up any messages that are still queued on the socket.
259  * For DGRAM and RDM socket types, all queued messages are rejected.
260  * For SEQPACKET and STREAM socket types, the first message is rejected
261  * and any others are discarded.  (If the first message on a STREAM socket
262  * is partially-read, it is discarded and the next one is rejected instead.)
263  *
264  * NOTE: Rejected messages are not necessarily returned to the sender!  They
265  * are returned or discarded according to the "destination droppable" setting
266  * specified for the message by the sender.
267  *
268  * Returns 0 on success, errno otherwise
269  */
270 
271 static int release(struct socket *sock)
272 {
273 	struct sock *sk = sock->sk;
274 	struct tipc_port *tport;
275 	struct sk_buff *buf;
276 	int res;
277 
278 	/*
279 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
280 	 * releases a pre-allocated child socket that was never used)
281 	 */
282 
283 	if (sk == NULL)
284 		return 0;
285 
286 	tport = tipc_sk_port(sk);
287 	lock_sock(sk);
288 
289 	/*
290 	 * Reject all unreceived messages, except on an active connection
291 	 * (which disconnects locally & sends a 'FIN+' to peer)
292 	 */
293 
294 	while (sock->state != SS_DISCONNECTING) {
295 		buf = __skb_dequeue(&sk->sk_receive_queue);
296 		if (buf == NULL)
297 			break;
298 		atomic_dec(&tipc_queue_size);
299 		if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
300 			buf_discard(buf);
301 		else {
302 			if ((sock->state == SS_CONNECTING) ||
303 			    (sock->state == SS_CONNECTED)) {
304 				sock->state = SS_DISCONNECTING;
305 				tipc_disconnect(tport->ref);
306 			}
307 			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
308 		}
309 	}
310 
311 	/*
312 	 * Delete TIPC port; this ensures no more messages are queued
313 	 * (also disconnects an active connection & sends a 'FIN-' to peer)
314 	 */
315 
316 	res = tipc_deleteport(tport->ref);
317 
318 	/* Discard any remaining (connection-based) messages in receive queue */
319 
320 	discard_rx_queue(sk);
321 
322 	/* Reject any messages that accumulated in backlog queue */
323 
324 	sock->state = SS_DISCONNECTING;
325 	release_sock(sk);
326 
327 	sock_put(sk);
328 	sock->sk = NULL;
329 
330 	atomic_dec(&tipc_user_count);
331 	return res;
332 }
333 
334 /**
335  * bind - associate or disassocate TIPC name(s) with a socket
336  * @sock: socket structure
337  * @uaddr: socket address describing name(s) and desired operation
338  * @uaddr_len: size of socket address data structure
339  *
340  * Name and name sequence binding is indicated using a positive scope value;
341  * a negative scope value unbinds the specified name.  Specifying no name
342  * (i.e. a socket address length of 0) unbinds all names from the socket.
343  *
344  * Returns 0 on success, errno otherwise
345  *
346  * NOTE: This routine doesn't need to take the socket lock since it doesn't
347  *       access any non-constant socket information.
348  */
349 
350 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
351 {
352 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
353 	u32 portref = tipc_sk_port(sock->sk)->ref;
354 
355 	if (unlikely(!uaddr_len))
356 		return tipc_withdraw(portref, 0, NULL);
357 
358 	if (uaddr_len < sizeof(struct sockaddr_tipc))
359 		return -EINVAL;
360 	if (addr->family != AF_TIPC)
361 		return -EAFNOSUPPORT;
362 
363 	if (addr->addrtype == TIPC_ADDR_NAME)
364 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
365 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
366 		return -EAFNOSUPPORT;
367 
368 	return (addr->scope > 0) ?
369 		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
370 		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
371 }
372 
373 /**
374  * get_name - get port ID of socket or peer socket
375  * @sock: socket structure
376  * @uaddr: area for returned socket address
377  * @uaddr_len: area for returned length of socket address
378  * @peer: 0 to obtain socket name, 1 to obtain peer socket name
379  *
380  * Returns 0 on success, errno otherwise
381  *
382  * NOTE: This routine doesn't need to take the socket lock since it doesn't
383  *       access any non-constant socket information.
384  */
385 
386 static int get_name(struct socket *sock, struct sockaddr *uaddr,
387 		    int *uaddr_len, int peer)
388 {
389 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
390 	u32 portref = tipc_sk_port(sock->sk)->ref;
391 	u32 res;
392 
393 	if (peer) {
394 		res = tipc_peer(portref, &addr->addr.id);
395 		if (res)
396 			return res;
397 	} else {
398 		tipc_ownidentity(portref, &addr->addr.id);
399 	}
400 
401 	*uaddr_len = sizeof(*addr);
402 	addr->addrtype = TIPC_ADDR_ID;
403 	addr->family = AF_TIPC;
404 	addr->scope = 0;
405 	addr->addr.name.domain = 0;
406 
407 	return 0;
408 }
409 
410 /**
411  * poll - read and possibly block on pollmask
412  * @file: file structure associated with the socket
413  * @sock: socket for which to calculate the poll bits
414  * @wait: ???
415  *
416  * Returns pollmask value
417  *
418  * COMMENTARY:
419  * It appears that the usual socket locking mechanisms are not useful here
420  * since the pollmask info is potentially out-of-date the moment this routine
421  * exits.  TCP and other protocols seem to rely on higher level poll routines
422  * to handle any preventable race conditions, so TIPC will do the same ...
423  *
424  * TIPC sets the returned events as follows:
425  * a) POLLRDNORM and POLLIN are set if the socket's receive queue is non-empty
426  *    or if a connection-oriented socket is does not have an active connection
427  *    (i.e. a read operation will not block).
428  * b) POLLOUT is set except when a socket's connection has been terminated
429  *    (i.e. a write operation will not block).
430  * c) POLLHUP is set when a socket's connection has been terminated.
431  *
432  * IMPORTANT: The fact that a read or write operation will not block does NOT
433  * imply that the operation will succeed!
434  */
435 
436 static unsigned int poll(struct file *file, struct socket *sock,
437 			 poll_table *wait)
438 {
439 	struct sock *sk = sock->sk;
440 	u32 mask;
441 
442 	poll_wait(file, sk->sk_sleep, wait);
443 
444 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
445 	    (sock->state == SS_UNCONNECTED) ||
446 	    (sock->state == SS_DISCONNECTING))
447 		mask = (POLLRDNORM | POLLIN);
448 	else
449 		mask = 0;
450 
451 	if (sock->state == SS_DISCONNECTING)
452 		mask |= POLLHUP;
453 	else
454 		mask |= POLLOUT;
455 
456 	return mask;
457 }
458 
459 /**
460  * dest_name_check - verify user is permitted to send to specified port name
461  * @dest: destination address
462  * @m: descriptor for message to be sent
463  *
464  * Prevents restricted configuration commands from being issued by
465  * unauthorized users.
466  *
467  * Returns 0 if permission is granted, otherwise errno
468  */
469 
470 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
471 {
472 	struct tipc_cfg_msg_hdr hdr;
473 
474 	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
475 		return 0;
476 	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
477 		return 0;
478 	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
479 		return -EACCES;
480 
481 	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
482 		return -EFAULT;
483 	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
484 		return -EACCES;
485 
486 	return 0;
487 }
488 
489 /**
490  * send_msg - send message in connectionless manner
491  * @iocb: if NULL, indicates that socket lock is already held
492  * @sock: socket structure
493  * @m: message to send
494  * @total_len: length of message
495  *
496  * Message must have an destination specified explicitly.
497  * Used for SOCK_RDM and SOCK_DGRAM messages,
498  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
499  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
500  *
501  * Returns the number of bytes sent on success, or errno otherwise
502  */
503 
504 static int send_msg(struct kiocb *iocb, struct socket *sock,
505 		    struct msghdr *m, size_t total_len)
506 {
507 	struct sock *sk = sock->sk;
508 	struct tipc_port *tport = tipc_sk_port(sk);
509 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
510 	int needs_conn;
511 	int res = -EINVAL;
512 
513 	if (unlikely(!dest))
514 		return -EDESTADDRREQ;
515 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
516 		     (dest->family != AF_TIPC)))
517 		return -EINVAL;
518 
519 	if (iocb)
520 		lock_sock(sk);
521 
522 	needs_conn = (sock->state != SS_READY);
523 	if (unlikely(needs_conn)) {
524 		if (sock->state == SS_LISTENING) {
525 			res = -EPIPE;
526 			goto exit;
527 		}
528 		if (sock->state != SS_UNCONNECTED) {
529 			res = -EISCONN;
530 			goto exit;
531 		}
532 		if ((tport->published) ||
533 		    ((sock->type == SOCK_STREAM) && (total_len != 0))) {
534 			res = -EOPNOTSUPP;
535 			goto exit;
536 		}
537 		if (dest->addrtype == TIPC_ADDR_NAME) {
538 			tport->conn_type = dest->addr.name.name.type;
539 			tport->conn_instance = dest->addr.name.name.instance;
540 		}
541 
542 		/* Abort any pending connection attempts (very unlikely) */
543 
544 		reject_rx_queue(sk);
545 	}
546 
547 	do {
548 		if (dest->addrtype == TIPC_ADDR_NAME) {
549 			if ((res = dest_name_check(dest, m)))
550 				break;
551 			res = tipc_send2name(tport->ref,
552 					     &dest->addr.name.name,
553 					     dest->addr.name.domain,
554 					     m->msg_iovlen,
555 					     m->msg_iov);
556 		}
557 		else if (dest->addrtype == TIPC_ADDR_ID) {
558 			res = tipc_send2port(tport->ref,
559 					     &dest->addr.id,
560 					     m->msg_iovlen,
561 					     m->msg_iov);
562 		}
563 		else if (dest->addrtype == TIPC_ADDR_MCAST) {
564 			if (needs_conn) {
565 				res = -EOPNOTSUPP;
566 				break;
567 			}
568 			if ((res = dest_name_check(dest, m)))
569 				break;
570 			res = tipc_multicast(tport->ref,
571 					     &dest->addr.nameseq,
572 					     0,
573 					     m->msg_iovlen,
574 					     m->msg_iov);
575 		}
576 		if (likely(res != -ELINKCONG)) {
577 			if (needs_conn && (res >= 0)) {
578 				sock->state = SS_CONNECTING;
579 			}
580 			break;
581 		}
582 		if (m->msg_flags & MSG_DONTWAIT) {
583 			res = -EWOULDBLOCK;
584 			break;
585 		}
586 		release_sock(sk);
587 		res = wait_event_interruptible(*sk->sk_sleep,
588 					       !tport->congested);
589 		lock_sock(sk);
590 		if (res)
591 			break;
592 	} while (1);
593 
594 exit:
595 	if (iocb)
596 		release_sock(sk);
597 	return res;
598 }
599 
600 /**
601  * send_packet - send a connection-oriented message
602  * @iocb: if NULL, indicates that socket lock is already held
603  * @sock: socket structure
604  * @m: message to send
605  * @total_len: length of message
606  *
607  * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
608  *
609  * Returns the number of bytes sent on success, or errno otherwise
610  */
611 
612 static int send_packet(struct kiocb *iocb, struct socket *sock,
613 		       struct msghdr *m, size_t total_len)
614 {
615 	struct sock *sk = sock->sk;
616 	struct tipc_port *tport = tipc_sk_port(sk);
617 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
618 	int res;
619 
620 	/* Handle implied connection establishment */
621 
622 	if (unlikely(dest))
623 		return send_msg(iocb, sock, m, total_len);
624 
625 	if (iocb)
626 		lock_sock(sk);
627 
628 	do {
629 		if (unlikely(sock->state != SS_CONNECTED)) {
630 			if (sock->state == SS_DISCONNECTING)
631 				res = -EPIPE;
632 			else
633 				res = -ENOTCONN;
634 			break;
635 		}
636 
637 		res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
638 		if (likely(res != -ELINKCONG)) {
639 			break;
640 		}
641 		if (m->msg_flags & MSG_DONTWAIT) {
642 			res = -EWOULDBLOCK;
643 			break;
644 		}
645 		release_sock(sk);
646 		res = wait_event_interruptible(*sk->sk_sleep,
647 			(!tport->congested || !tport->connected));
648 		lock_sock(sk);
649 		if (res)
650 			break;
651 	} while (1);
652 
653 	if (iocb)
654 		release_sock(sk);
655 	return res;
656 }
657 
658 /**
659  * send_stream - send stream-oriented data
660  * @iocb: (unused)
661  * @sock: socket structure
662  * @m: data to send
663  * @total_len: total length of data to be sent
664  *
665  * Used for SOCK_STREAM data.
666  *
667  * Returns the number of bytes sent on success (or partial success),
668  * or errno if no data sent
669  */
670 
671 static int send_stream(struct kiocb *iocb, struct socket *sock,
672 		       struct msghdr *m, size_t total_len)
673 {
674 	struct sock *sk = sock->sk;
675 	struct tipc_port *tport = tipc_sk_port(sk);
676 	struct msghdr my_msg;
677 	struct iovec my_iov;
678 	struct iovec *curr_iov;
679 	int curr_iovlen;
680 	char __user *curr_start;
681 	u32 hdr_size;
682 	int curr_left;
683 	int bytes_to_send;
684 	int bytes_sent;
685 	int res;
686 
687 	lock_sock(sk);
688 
689 	/* Handle special cases where there is no connection */
690 
691 	if (unlikely(sock->state != SS_CONNECTED)) {
692 		if (sock->state == SS_UNCONNECTED) {
693 			res = send_packet(NULL, sock, m, total_len);
694 			goto exit;
695 		} else if (sock->state == SS_DISCONNECTING) {
696 			res = -EPIPE;
697 			goto exit;
698 		} else {
699 			res = -ENOTCONN;
700 			goto exit;
701 		}
702 	}
703 
704 	if (unlikely(m->msg_name)) {
705 		res = -EISCONN;
706 		goto exit;
707 	}
708 
709 	/*
710 	 * Send each iovec entry using one or more messages
711 	 *
712 	 * Note: This algorithm is good for the most likely case
713 	 * (i.e. one large iovec entry), but could be improved to pass sets
714 	 * of small iovec entries into send_packet().
715 	 */
716 
717 	curr_iov = m->msg_iov;
718 	curr_iovlen = m->msg_iovlen;
719 	my_msg.msg_iov = &my_iov;
720 	my_msg.msg_iovlen = 1;
721 	my_msg.msg_flags = m->msg_flags;
722 	my_msg.msg_name = NULL;
723 	bytes_sent = 0;
724 
725 	hdr_size = msg_hdr_sz(&tport->phdr);
726 
727 	while (curr_iovlen--) {
728 		curr_start = curr_iov->iov_base;
729 		curr_left = curr_iov->iov_len;
730 
731 		while (curr_left) {
732 			bytes_to_send = tport->max_pkt - hdr_size;
733 			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
734 				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
735 			if (curr_left < bytes_to_send)
736 				bytes_to_send = curr_left;
737 			my_iov.iov_base = curr_start;
738 			my_iov.iov_len = bytes_to_send;
739 			if ((res = send_packet(NULL, sock, &my_msg, 0)) < 0) {
740 				if (bytes_sent)
741 					res = bytes_sent;
742 				goto exit;
743 			}
744 			curr_left -= bytes_to_send;
745 			curr_start += bytes_to_send;
746 			bytes_sent += bytes_to_send;
747 		}
748 
749 		curr_iov++;
750 	}
751 	res = bytes_sent;
752 exit:
753 	release_sock(sk);
754 	return res;
755 }
756 
757 /**
758  * auto_connect - complete connection setup to a remote port
759  * @sock: socket structure
760  * @msg: peer's response message
761  *
762  * Returns 0 on success, errno otherwise
763  */
764 
765 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
766 {
767 	struct tipc_port *tport = tipc_sk_port(sock->sk);
768 	struct tipc_portid peer;
769 
770 	if (msg_errcode(msg)) {
771 		sock->state = SS_DISCONNECTING;
772 		return -ECONNREFUSED;
773 	}
774 
775 	peer.ref = msg_origport(msg);
776 	peer.node = msg_orignode(msg);
777 	tipc_connect2port(tport->ref, &peer);
778 	tipc_set_portimportance(tport->ref, msg_importance(msg));
779 	sock->state = SS_CONNECTED;
780 	return 0;
781 }
782 
783 /**
784  * set_orig_addr - capture sender's address for received message
785  * @m: descriptor for message info
786  * @msg: received message header
787  *
788  * Note: Address is not captured if not requested by receiver.
789  */
790 
791 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
792 {
793 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
794 
795 	if (addr) {
796 		addr->family = AF_TIPC;
797 		addr->addrtype = TIPC_ADDR_ID;
798 		addr->addr.id.ref = msg_origport(msg);
799 		addr->addr.id.node = msg_orignode(msg);
800 		addr->addr.name.domain = 0;   	/* could leave uninitialized */
801 		addr->scope = 0;   		/* could leave uninitialized */
802 		m->msg_namelen = sizeof(struct sockaddr_tipc);
803 	}
804 }
805 
806 /**
807  * anc_data_recv - optionally capture ancillary data for received message
808  * @m: descriptor for message info
809  * @msg: received message header
810  * @tport: TIPC port associated with message
811  *
812  * Note: Ancillary data is not captured if not requested by receiver.
813  *
814  * Returns 0 if successful, otherwise errno
815  */
816 
817 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
818 				struct tipc_port *tport)
819 {
820 	u32 anc_data[3];
821 	u32 err;
822 	u32 dest_type;
823 	int has_name;
824 	int res;
825 
826 	if (likely(m->msg_controllen == 0))
827 		return 0;
828 
829 	/* Optionally capture errored message object(s) */
830 
831 	err = msg ? msg_errcode(msg) : 0;
832 	if (unlikely(err)) {
833 		anc_data[0] = err;
834 		anc_data[1] = msg_data_sz(msg);
835 		if ((res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data)))
836 			return res;
837 		if (anc_data[1] &&
838 		    (res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
839 				    msg_data(msg))))
840 			return res;
841 	}
842 
843 	/* Optionally capture message destination object */
844 
845 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
846 	switch (dest_type) {
847 	case TIPC_NAMED_MSG:
848 		has_name = 1;
849 		anc_data[0] = msg_nametype(msg);
850 		anc_data[1] = msg_namelower(msg);
851 		anc_data[2] = msg_namelower(msg);
852 		break;
853 	case TIPC_MCAST_MSG:
854 		has_name = 1;
855 		anc_data[0] = msg_nametype(msg);
856 		anc_data[1] = msg_namelower(msg);
857 		anc_data[2] = msg_nameupper(msg);
858 		break;
859 	case TIPC_CONN_MSG:
860 		has_name = (tport->conn_type != 0);
861 		anc_data[0] = tport->conn_type;
862 		anc_data[1] = tport->conn_instance;
863 		anc_data[2] = tport->conn_instance;
864 		break;
865 	default:
866 		has_name = 0;
867 	}
868 	if (has_name &&
869 	    (res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data)))
870 		return res;
871 
872 	return 0;
873 }
874 
875 /**
876  * recv_msg - receive packet-oriented message
877  * @iocb: (unused)
878  * @m: descriptor for message info
879  * @buf_len: total size of user buffer area
880  * @flags: receive flags
881  *
882  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
883  * If the complete message doesn't fit in user area, truncate it.
884  *
885  * Returns size of returned message data, errno otherwise
886  */
887 
888 static int recv_msg(struct kiocb *iocb, struct socket *sock,
889 		    struct msghdr *m, size_t buf_len, int flags)
890 {
891 	struct sock *sk = sock->sk;
892 	struct tipc_port *tport = tipc_sk_port(sk);
893 	struct sk_buff *buf;
894 	struct tipc_msg *msg;
895 	unsigned int sz;
896 	u32 err;
897 	int res;
898 
899 	/* Catch invalid receive requests */
900 
901 	if (m->msg_iovlen != 1)
902 		return -EOPNOTSUPP;   /* Don't do multiple iovec entries yet */
903 
904 	if (unlikely(!buf_len))
905 		return -EINVAL;
906 
907 	lock_sock(sk);
908 
909 	if (unlikely(sock->state == SS_UNCONNECTED)) {
910 		res = -ENOTCONN;
911 		goto exit;
912 	}
913 
914 restart:
915 
916 	/* Look for a message in receive queue; wait if necessary */
917 
918 	while (skb_queue_empty(&sk->sk_receive_queue)) {
919 		if (sock->state == SS_DISCONNECTING) {
920 			res = -ENOTCONN;
921 			goto exit;
922 		}
923 		if (flags & MSG_DONTWAIT) {
924 			res = -EWOULDBLOCK;
925 			goto exit;
926 		}
927 		release_sock(sk);
928 		res = wait_event_interruptible(*sk->sk_sleep,
929 			(!skb_queue_empty(&sk->sk_receive_queue) ||
930 			 (sock->state == SS_DISCONNECTING)));
931 		lock_sock(sk);
932 		if (res)
933 			goto exit;
934 	}
935 
936 	/* Look at first message in receive queue */
937 
938 	buf = skb_peek(&sk->sk_receive_queue);
939 	msg = buf_msg(buf);
940 	sz = msg_data_sz(msg);
941 	err = msg_errcode(msg);
942 
943 	/* Complete connection setup for an implied connect */
944 
945 	if (unlikely(sock->state == SS_CONNECTING)) {
946 		res = auto_connect(sock, msg);
947 		if (res)
948 			goto exit;
949 	}
950 
951 	/* Discard an empty non-errored message & try again */
952 
953 	if ((!sz) && (!err)) {
954 		advance_rx_queue(sk);
955 		goto restart;
956 	}
957 
958 	/* Capture sender's address (optional) */
959 
960 	set_orig_addr(m, msg);
961 
962 	/* Capture ancillary data (optional) */
963 
964 	res = anc_data_recv(m, msg, tport);
965 	if (res)
966 		goto exit;
967 
968 	/* Capture message data (if valid) & compute return value (always) */
969 
970 	if (!err) {
971 		if (unlikely(buf_len < sz)) {
972 			sz = buf_len;
973 			m->msg_flags |= MSG_TRUNC;
974 		}
975 		if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
976 					  sz))) {
977 			res = -EFAULT;
978 			goto exit;
979 		}
980 		res = sz;
981 	} else {
982 		if ((sock->state == SS_READY) ||
983 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
984 			res = 0;
985 		else
986 			res = -ECONNRESET;
987 	}
988 
989 	/* Consume received message (optional) */
990 
991 	if (likely(!(flags & MSG_PEEK))) {
992 		if ((sock->state != SS_READY) &&
993 		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
994 			tipc_acknowledge(tport->ref, tport->conn_unacked);
995 		advance_rx_queue(sk);
996 	}
997 exit:
998 	release_sock(sk);
999 	return res;
1000 }
1001 
1002 /**
1003  * recv_stream - receive stream-oriented data
1004  * @iocb: (unused)
1005  * @m: descriptor for message info
1006  * @buf_len: total size of user buffer area
1007  * @flags: receive flags
1008  *
1009  * Used for SOCK_STREAM messages only.  If not enough data is available
1010  * will optionally wait for more; never truncates data.
1011  *
1012  * Returns size of returned message data, errno otherwise
1013  */
1014 
1015 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1016 		       struct msghdr *m, size_t buf_len, int flags)
1017 {
1018 	struct sock *sk = sock->sk;
1019 	struct tipc_port *tport = tipc_sk_port(sk);
1020 	struct sk_buff *buf;
1021 	struct tipc_msg *msg;
1022 	unsigned int sz;
1023 	int sz_to_copy;
1024 	int sz_copied = 0;
1025 	int needed;
1026 	char __user *crs = m->msg_iov->iov_base;
1027 	unsigned char *buf_crs;
1028 	u32 err;
1029 	int res = 0;
1030 
1031 	/* Catch invalid receive attempts */
1032 
1033 	if (m->msg_iovlen != 1)
1034 		return -EOPNOTSUPP;   /* Don't do multiple iovec entries yet */
1035 
1036 	if (unlikely(!buf_len))
1037 		return -EINVAL;
1038 
1039 	lock_sock(sk);
1040 
1041 	if (unlikely((sock->state == SS_UNCONNECTED) ||
1042 		     (sock->state == SS_CONNECTING))) {
1043 		res = -ENOTCONN;
1044 		goto exit;
1045 	}
1046 
1047 restart:
1048 
1049 	/* Look for a message in receive queue; wait if necessary */
1050 
1051 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1052 		if (sock->state == SS_DISCONNECTING) {
1053 			res = -ENOTCONN;
1054 			goto exit;
1055 		}
1056 		if (flags & MSG_DONTWAIT) {
1057 			res = -EWOULDBLOCK;
1058 			goto exit;
1059 		}
1060 		release_sock(sk);
1061 		res = wait_event_interruptible(*sk->sk_sleep,
1062 			(!skb_queue_empty(&sk->sk_receive_queue) ||
1063 			 (sock->state == SS_DISCONNECTING)));
1064 		lock_sock(sk);
1065 		if (res)
1066 			goto exit;
1067 	}
1068 
1069 	/* Look at first message in receive queue */
1070 
1071 	buf = skb_peek(&sk->sk_receive_queue);
1072 	msg = buf_msg(buf);
1073 	sz = msg_data_sz(msg);
1074 	err = msg_errcode(msg);
1075 
1076 	/* Discard an empty non-errored message & try again */
1077 
1078 	if ((!sz) && (!err)) {
1079 		advance_rx_queue(sk);
1080 		goto restart;
1081 	}
1082 
1083 	/* Optionally capture sender's address & ancillary data of first msg */
1084 
1085 	if (sz_copied == 0) {
1086 		set_orig_addr(m, msg);
1087 		res = anc_data_recv(m, msg, tport);
1088 		if (res)
1089 			goto exit;
1090 	}
1091 
1092 	/* Capture message data (if valid) & compute return value (always) */
1093 
1094 	if (!err) {
1095 		buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1096 		sz = (unsigned char *)msg + msg_size(msg) - buf_crs;
1097 
1098 		needed = (buf_len - sz_copied);
1099 		sz_to_copy = (sz <= needed) ? sz : needed;
1100 		if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1101 			res = -EFAULT;
1102 			goto exit;
1103 		}
1104 		sz_copied += sz_to_copy;
1105 
1106 		if (sz_to_copy < sz) {
1107 			if (!(flags & MSG_PEEK))
1108 				TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1109 			goto exit;
1110 		}
1111 
1112 		crs += sz_to_copy;
1113 	} else {
1114 		if (sz_copied != 0)
1115 			goto exit; /* can't add error msg to valid data */
1116 
1117 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1118 			res = 0;
1119 		else
1120 			res = -ECONNRESET;
1121 	}
1122 
1123 	/* Consume received message (optional) */
1124 
1125 	if (likely(!(flags & MSG_PEEK))) {
1126 		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1127 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1128 		advance_rx_queue(sk);
1129 	}
1130 
1131 	/* Loop around if more data is required */
1132 
1133 	if ((sz_copied < buf_len)    /* didn't get all requested data */
1134 	    && (!skb_queue_empty(&sock->sk->sk_receive_queue) ||
1135 		(flags & MSG_WAITALL))
1136 				     /* ... and more is ready or required */
1137 	    && (!(flags & MSG_PEEK)) /* ... and aren't just peeking at data */
1138 	    && (!err)                /* ... and haven't reached a FIN */
1139 	    )
1140 		goto restart;
1141 
1142 exit:
1143 	release_sock(sk);
1144 	return sz_copied ? sz_copied : res;
1145 }
1146 
1147 /**
1148  * rx_queue_full - determine if receive queue can accept another message
1149  * @msg: message to be added to queue
1150  * @queue_size: current size of queue
1151  * @base: nominal maximum size of queue
1152  *
1153  * Returns 1 if queue is unable to accept message, 0 otherwise
1154  */
1155 
1156 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1157 {
1158 	u32 threshold;
1159 	u32 imp = msg_importance(msg);
1160 
1161 	if (imp == TIPC_LOW_IMPORTANCE)
1162 		threshold = base;
1163 	else if (imp == TIPC_MEDIUM_IMPORTANCE)
1164 		threshold = base * 2;
1165 	else if (imp == TIPC_HIGH_IMPORTANCE)
1166 		threshold = base * 100;
1167 	else
1168 		return 0;
1169 
1170 	if (msg_connected(msg))
1171 		threshold *= 4;
1172 
1173 	return (queue_size >= threshold);
1174 }
1175 
1176 /**
1177  * filter_rcv - validate incoming message
1178  * @sk: socket
1179  * @buf: message
1180  *
1181  * Enqueues message on receive queue if acceptable; optionally handles
1182  * disconnect indication for a connected socket.
1183  *
1184  * Called with socket lock already taken; port lock may also be taken.
1185  *
1186  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1187  */
1188 
1189 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1190 {
1191 	struct socket *sock = sk->sk_socket;
1192 	struct tipc_msg *msg = buf_msg(buf);
1193 	u32 recv_q_len;
1194 
1195 	/* Reject message if it is wrong sort of message for socket */
1196 
1197 	/*
1198 	 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1199 	 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1200 	 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1201 	 */
1202 
1203 	if (sock->state == SS_READY) {
1204 		if (msg_connected(msg)) {
1205 			msg_dbg(msg, "dispatch filter 1\n");
1206 			return TIPC_ERR_NO_PORT;
1207 		}
1208 	} else {
1209 		if (msg_mcast(msg)) {
1210 			msg_dbg(msg, "dispatch filter 2\n");
1211 			return TIPC_ERR_NO_PORT;
1212 		}
1213 		if (sock->state == SS_CONNECTED) {
1214 			if (!msg_connected(msg)) {
1215 				msg_dbg(msg, "dispatch filter 3\n");
1216 				return TIPC_ERR_NO_PORT;
1217 			}
1218 		}
1219 		else if (sock->state == SS_CONNECTING) {
1220 			if (!msg_connected(msg) && (msg_errcode(msg) == 0)) {
1221 				msg_dbg(msg, "dispatch filter 4\n");
1222 				return TIPC_ERR_NO_PORT;
1223 			}
1224 		}
1225 		else if (sock->state == SS_LISTENING) {
1226 			if (msg_connected(msg) || msg_errcode(msg)) {
1227 				msg_dbg(msg, "dispatch filter 5\n");
1228 				return TIPC_ERR_NO_PORT;
1229 			}
1230 		}
1231 		else if (sock->state == SS_DISCONNECTING) {
1232 			msg_dbg(msg, "dispatch filter 6\n");
1233 			return TIPC_ERR_NO_PORT;
1234 		}
1235 		else /* (sock->state == SS_UNCONNECTED) */ {
1236 			if (msg_connected(msg) || msg_errcode(msg)) {
1237 				msg_dbg(msg, "dispatch filter 7\n");
1238 				return TIPC_ERR_NO_PORT;
1239 			}
1240 		}
1241 	}
1242 
1243 	/* Reject message if there isn't room to queue it */
1244 
1245 	recv_q_len = (u32)atomic_read(&tipc_queue_size);
1246 	if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1247 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1248 			return TIPC_ERR_OVERLOAD;
1249 	}
1250 	recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1251 	if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1252 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1253 			return TIPC_ERR_OVERLOAD;
1254 	}
1255 
1256 	/* Enqueue message (finally!) */
1257 
1258 	msg_dbg(msg, "<DISP<: ");
1259 	TIPC_SKB_CB(buf)->handle = msg_data(msg);
1260 	atomic_inc(&tipc_queue_size);
1261 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1262 
1263 	/* Initiate connection termination for an incoming 'FIN' */
1264 
1265 	if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1266 		sock->state = SS_DISCONNECTING;
1267 		tipc_disconnect_port(tipc_sk_port(sk));
1268 	}
1269 
1270 	if (waitqueue_active(sk->sk_sleep))
1271 		wake_up_interruptible(sk->sk_sleep);
1272 	return TIPC_OK;
1273 }
1274 
1275 /**
1276  * backlog_rcv - handle incoming message from backlog queue
1277  * @sk: socket
1278  * @buf: message
1279  *
1280  * Caller must hold socket lock, but not port lock.
1281  *
1282  * Returns 0
1283  */
1284 
1285 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1286 {
1287 	u32 res;
1288 
1289 	res = filter_rcv(sk, buf);
1290 	if (res)
1291 		tipc_reject_msg(buf, res);
1292 	return 0;
1293 }
1294 
1295 /**
1296  * dispatch - handle incoming message
1297  * @tport: TIPC port that received message
1298  * @buf: message
1299  *
1300  * Called with port lock already taken.
1301  *
1302  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1303  */
1304 
1305 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1306 {
1307 	struct sock *sk = (struct sock *)tport->usr_handle;
1308 	u32 res;
1309 
1310 	/*
1311 	 * Process message if socket is unlocked; otherwise add to backlog queue
1312 	 *
1313 	 * This code is based on sk_receive_skb(), but must be distinct from it
1314 	 * since a TIPC-specific filter/reject mechanism is utilized
1315 	 */
1316 
1317 	bh_lock_sock(sk);
1318 	if (!sock_owned_by_user(sk)) {
1319 		res = filter_rcv(sk, buf);
1320 	} else {
1321 		sk_add_backlog(sk, buf);
1322 		res = TIPC_OK;
1323 	}
1324 	bh_unlock_sock(sk);
1325 
1326 	return res;
1327 }
1328 
1329 /**
1330  * wakeupdispatch - wake up port after congestion
1331  * @tport: port to wakeup
1332  *
1333  * Called with port lock already taken.
1334  */
1335 
1336 static void wakeupdispatch(struct tipc_port *tport)
1337 {
1338 	struct sock *sk = (struct sock *)tport->usr_handle;
1339 
1340 	if (waitqueue_active(sk->sk_sleep))
1341 		wake_up_interruptible(sk->sk_sleep);
1342 }
1343 
1344 /**
1345  * connect - establish a connection to another TIPC port
1346  * @sock: socket structure
1347  * @dest: socket address for destination port
1348  * @destlen: size of socket address data structure
1349  * @flags: file-related flags associated with socket
1350  *
1351  * Returns 0 on success, errno otherwise
1352  */
1353 
1354 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1355 		   int flags)
1356 {
1357 	struct sock *sk = sock->sk;
1358 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1359 	struct msghdr m = {NULL,};
1360 	struct sk_buff *buf;
1361 	struct tipc_msg *msg;
1362 	int res;
1363 
1364 	lock_sock(sk);
1365 
1366 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1367 
1368 	if (sock->state == SS_READY) {
1369 		res = -EOPNOTSUPP;
1370 		goto exit;
1371 	}
1372 
1373 	/* For now, TIPC does not support the non-blocking form of connect() */
1374 
1375 	if (flags & O_NONBLOCK) {
1376 		res = -EWOULDBLOCK;
1377 		goto exit;
1378 	}
1379 
1380 	/* Issue Posix-compliant error code if socket is in the wrong state */
1381 
1382 	if (sock->state == SS_LISTENING) {
1383 		res = -EOPNOTSUPP;
1384 		goto exit;
1385 	}
1386 	if (sock->state == SS_CONNECTING) {
1387 		res = -EALREADY;
1388 		goto exit;
1389 	}
1390 	if (sock->state != SS_UNCONNECTED) {
1391 		res = -EISCONN;
1392 		goto exit;
1393 	}
1394 
1395 	/*
1396 	 * Reject connection attempt using multicast address
1397 	 *
1398 	 * Note: send_msg() validates the rest of the address fields,
1399 	 *       so there's no need to do it here
1400 	 */
1401 
1402 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1403 		res = -EINVAL;
1404 		goto exit;
1405 	}
1406 
1407 	/* Reject any messages already in receive queue (very unlikely) */
1408 
1409 	reject_rx_queue(sk);
1410 
1411 	/* Send a 'SYN-' to destination */
1412 
1413 	m.msg_name = dest;
1414 	m.msg_namelen = destlen;
1415 	res = send_msg(NULL, sock, &m, 0);
1416 	if (res < 0) {
1417 		goto exit;
1418 	}
1419 
1420 	/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1421 
1422 	release_sock(sk);
1423 	res = wait_event_interruptible_timeout(*sk->sk_sleep,
1424 			(!skb_queue_empty(&sk->sk_receive_queue) ||
1425 			(sock->state != SS_CONNECTING)),
1426 			sk->sk_rcvtimeo);
1427 	lock_sock(sk);
1428 
1429 	if (res > 0) {
1430 		buf = skb_peek(&sk->sk_receive_queue);
1431 		if (buf != NULL) {
1432 			msg = buf_msg(buf);
1433 			res = auto_connect(sock, msg);
1434 			if (!res) {
1435 				if (!msg_data_sz(msg))
1436 					advance_rx_queue(sk);
1437 			}
1438 		} else {
1439 			if (sock->state == SS_CONNECTED) {
1440 				res = -EISCONN;
1441 			} else {
1442 				res = -ECONNREFUSED;
1443 			}
1444 		}
1445 	} else {
1446 		if (res == 0)
1447 			res = -ETIMEDOUT;
1448 		else
1449 			; /* leave "res" unchanged */
1450 		sock->state = SS_DISCONNECTING;
1451 	}
1452 
1453 exit:
1454 	release_sock(sk);
1455 	return res;
1456 }
1457 
1458 /**
1459  * listen - allow socket to listen for incoming connections
1460  * @sock: socket structure
1461  * @len: (unused)
1462  *
1463  * Returns 0 on success, errno otherwise
1464  */
1465 
1466 static int listen(struct socket *sock, int len)
1467 {
1468 	struct sock *sk = sock->sk;
1469 	int res;
1470 
1471 	lock_sock(sk);
1472 
1473 	if (sock->state == SS_READY)
1474 		res = -EOPNOTSUPP;
1475 	else if (sock->state != SS_UNCONNECTED)
1476 		res = -EINVAL;
1477 	else {
1478 		sock->state = SS_LISTENING;
1479 		res = 0;
1480 	}
1481 
1482 	release_sock(sk);
1483 	return res;
1484 }
1485 
1486 /**
1487  * accept - wait for connection request
1488  * @sock: listening socket
1489  * @newsock: new socket that is to be connected
1490  * @flags: file-related flags associated with socket
1491  *
1492  * Returns 0 on success, errno otherwise
1493  */
1494 
1495 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1496 {
1497 	struct sock *sk = sock->sk;
1498 	struct sk_buff *buf;
1499 	int res;
1500 
1501 	lock_sock(sk);
1502 
1503 	if (sock->state == SS_READY) {
1504 		res = -EOPNOTSUPP;
1505 		goto exit;
1506 	}
1507 	if (sock->state != SS_LISTENING) {
1508 		res = -EINVAL;
1509 		goto exit;
1510 	}
1511 
1512 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1513 		if (flags & O_NONBLOCK) {
1514 			res = -EWOULDBLOCK;
1515 			goto exit;
1516 		}
1517 		release_sock(sk);
1518 		res = wait_event_interruptible(*sk->sk_sleep,
1519 				(!skb_queue_empty(&sk->sk_receive_queue)));
1520 		lock_sock(sk);
1521 		if (res)
1522 			goto exit;
1523 	}
1524 
1525 	buf = skb_peek(&sk->sk_receive_queue);
1526 
1527 	res = tipc_create(sock_net(sock->sk), new_sock, 0);
1528 	if (!res) {
1529 		struct sock *new_sk = new_sock->sk;
1530 		struct tipc_port *new_tport = tipc_sk_port(new_sk);
1531 		u32 new_ref = new_tport->ref;
1532 		struct tipc_portid id;
1533 		struct tipc_msg *msg = buf_msg(buf);
1534 
1535 		lock_sock(new_sk);
1536 
1537 		/*
1538 		 * Reject any stray messages received by new socket
1539 		 * before the socket lock was taken (very, very unlikely)
1540 		 */
1541 
1542 		reject_rx_queue(new_sk);
1543 
1544 		/* Connect new socket to it's peer */
1545 
1546 		id.ref = msg_origport(msg);
1547 		id.node = msg_orignode(msg);
1548 		tipc_connect2port(new_ref, &id);
1549 		new_sock->state = SS_CONNECTED;
1550 
1551 		tipc_set_portimportance(new_ref, msg_importance(msg));
1552 		if (msg_named(msg)) {
1553 			new_tport->conn_type = msg_nametype(msg);
1554 			new_tport->conn_instance = msg_nameinst(msg);
1555 		}
1556 
1557 		/*
1558 		 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1559 		 * Respond to 'SYN+' by queuing it on new socket.
1560 		 */
1561 
1562 		msg_dbg(msg,"<ACC<: ");
1563 		if (!msg_data_sz(msg)) {
1564 			struct msghdr m = {NULL,};
1565 
1566 			advance_rx_queue(sk);
1567 			send_packet(NULL, new_sock, &m, 0);
1568 		} else {
1569 			__skb_dequeue(&sk->sk_receive_queue);
1570 			__skb_queue_head(&new_sk->sk_receive_queue, buf);
1571 		}
1572 		release_sock(new_sk);
1573 	}
1574 exit:
1575 	release_sock(sk);
1576 	return res;
1577 }
1578 
1579 /**
1580  * shutdown - shutdown socket connection
1581  * @sock: socket structure
1582  * @how: direction to close (must be SHUT_RDWR)
1583  *
1584  * Terminates connection (if necessary), then purges socket's receive queue.
1585  *
1586  * Returns 0 on success, errno otherwise
1587  */
1588 
1589 static int shutdown(struct socket *sock, int how)
1590 {
1591 	struct sock *sk = sock->sk;
1592 	struct tipc_port *tport = tipc_sk_port(sk);
1593 	struct sk_buff *buf;
1594 	int res;
1595 
1596 	if (how != SHUT_RDWR)
1597 		return -EINVAL;
1598 
1599 	lock_sock(sk);
1600 
1601 	switch (sock->state) {
1602 	case SS_CONNECTING:
1603 	case SS_CONNECTED:
1604 
1605 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1606 restart:
1607 		buf = __skb_dequeue(&sk->sk_receive_queue);
1608 		if (buf) {
1609 			atomic_dec(&tipc_queue_size);
1610 			if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1611 				buf_discard(buf);
1612 				goto restart;
1613 			}
1614 			tipc_disconnect(tport->ref);
1615 			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1616 		} else {
1617 			tipc_shutdown(tport->ref);
1618 		}
1619 
1620 		sock->state = SS_DISCONNECTING;
1621 
1622 		/* fall through */
1623 
1624 	case SS_DISCONNECTING:
1625 
1626 		/* Discard any unreceived messages; wake up sleeping tasks */
1627 
1628 		discard_rx_queue(sk);
1629 		if (waitqueue_active(sk->sk_sleep))
1630 			wake_up_interruptible(sk->sk_sleep);
1631 		res = 0;
1632 		break;
1633 
1634 	default:
1635 		res = -ENOTCONN;
1636 	}
1637 
1638 	release_sock(sk);
1639 	return res;
1640 }
1641 
1642 /**
1643  * setsockopt - set socket option
1644  * @sock: socket structure
1645  * @lvl: option level
1646  * @opt: option identifier
1647  * @ov: pointer to new option value
1648  * @ol: length of option value
1649  *
1650  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1651  * (to ease compatibility).
1652  *
1653  * Returns 0 on success, errno otherwise
1654  */
1655 
1656 static int setsockopt(struct socket *sock,
1657 		      int lvl, int opt, char __user *ov, int ol)
1658 {
1659 	struct sock *sk = sock->sk;
1660 	struct tipc_port *tport = tipc_sk_port(sk);
1661 	u32 value;
1662 	int res;
1663 
1664 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1665 		return 0;
1666 	if (lvl != SOL_TIPC)
1667 		return -ENOPROTOOPT;
1668 	if (ol < sizeof(value))
1669 		return -EINVAL;
1670 	if ((res = get_user(value, (u32 __user *)ov)))
1671 		return res;
1672 
1673 	lock_sock(sk);
1674 
1675 	switch (opt) {
1676 	case TIPC_IMPORTANCE:
1677 		res = tipc_set_portimportance(tport->ref, value);
1678 		break;
1679 	case TIPC_SRC_DROPPABLE:
1680 		if (sock->type != SOCK_STREAM)
1681 			res = tipc_set_portunreliable(tport->ref, value);
1682 		else
1683 			res = -ENOPROTOOPT;
1684 		break;
1685 	case TIPC_DEST_DROPPABLE:
1686 		res = tipc_set_portunreturnable(tport->ref, value);
1687 		break;
1688 	case TIPC_CONN_TIMEOUT:
1689 		sk->sk_rcvtimeo = msecs_to_jiffies(value);
1690 		/* no need to set "res", since already 0 at this point */
1691 		break;
1692 	default:
1693 		res = -EINVAL;
1694 	}
1695 
1696 	release_sock(sk);
1697 
1698 	return res;
1699 }
1700 
1701 /**
1702  * getsockopt - get socket option
1703  * @sock: socket structure
1704  * @lvl: option level
1705  * @opt: option identifier
1706  * @ov: receptacle for option value
1707  * @ol: receptacle for length of option value
1708  *
1709  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1710  * (to ease compatibility).
1711  *
1712  * Returns 0 on success, errno otherwise
1713  */
1714 
1715 static int getsockopt(struct socket *sock,
1716 		      int lvl, int opt, char __user *ov, int __user *ol)
1717 {
1718 	struct sock *sk = sock->sk;
1719 	struct tipc_port *tport = tipc_sk_port(sk);
1720 	int len;
1721 	u32 value;
1722 	int res;
1723 
1724 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1725 		return put_user(0, ol);
1726 	if (lvl != SOL_TIPC)
1727 		return -ENOPROTOOPT;
1728 	if ((res = get_user(len, ol)))
1729 		return res;
1730 
1731 	lock_sock(sk);
1732 
1733 	switch (opt) {
1734 	case TIPC_IMPORTANCE:
1735 		res = tipc_portimportance(tport->ref, &value);
1736 		break;
1737 	case TIPC_SRC_DROPPABLE:
1738 		res = tipc_portunreliable(tport->ref, &value);
1739 		break;
1740 	case TIPC_DEST_DROPPABLE:
1741 		res = tipc_portunreturnable(tport->ref, &value);
1742 		break;
1743 	case TIPC_CONN_TIMEOUT:
1744 		value = jiffies_to_msecs(sk->sk_rcvtimeo);
1745 		/* no need to set "res", since already 0 at this point */
1746 		break;
1747 	default:
1748 		res = -EINVAL;
1749 	}
1750 
1751 	release_sock(sk);
1752 
1753 	if (res) {
1754 		/* "get" failed */
1755 	}
1756 	else if (len < sizeof(value)) {
1757 		res = -EINVAL;
1758 	}
1759 	else if (copy_to_user(ov, &value, sizeof(value))) {
1760 		res = -EFAULT;
1761 	}
1762 	else {
1763 		res = put_user(sizeof(value), ol);
1764 	}
1765 
1766 	return res;
1767 }
1768 
1769 /**
1770  * Protocol switches for the various types of TIPC sockets
1771  */
1772 
1773 static const struct proto_ops msg_ops = {
1774 	.owner 		= THIS_MODULE,
1775 	.family		= AF_TIPC,
1776 	.release	= release,
1777 	.bind		= bind,
1778 	.connect	= connect,
1779 	.socketpair	= sock_no_socketpair,
1780 	.accept		= accept,
1781 	.getname	= get_name,
1782 	.poll		= poll,
1783 	.ioctl		= sock_no_ioctl,
1784 	.listen		= listen,
1785 	.shutdown	= shutdown,
1786 	.setsockopt	= setsockopt,
1787 	.getsockopt	= getsockopt,
1788 	.sendmsg	= send_msg,
1789 	.recvmsg	= recv_msg,
1790 	.mmap		= sock_no_mmap,
1791 	.sendpage	= sock_no_sendpage
1792 };
1793 
1794 static const struct proto_ops packet_ops = {
1795 	.owner 		= THIS_MODULE,
1796 	.family		= AF_TIPC,
1797 	.release	= release,
1798 	.bind		= bind,
1799 	.connect	= connect,
1800 	.socketpair	= sock_no_socketpair,
1801 	.accept		= accept,
1802 	.getname	= get_name,
1803 	.poll		= poll,
1804 	.ioctl		= sock_no_ioctl,
1805 	.listen		= listen,
1806 	.shutdown	= shutdown,
1807 	.setsockopt	= setsockopt,
1808 	.getsockopt	= getsockopt,
1809 	.sendmsg	= send_packet,
1810 	.recvmsg	= recv_msg,
1811 	.mmap		= sock_no_mmap,
1812 	.sendpage	= sock_no_sendpage
1813 };
1814 
1815 static const struct proto_ops stream_ops = {
1816 	.owner 		= THIS_MODULE,
1817 	.family		= AF_TIPC,
1818 	.release	= release,
1819 	.bind		= bind,
1820 	.connect	= connect,
1821 	.socketpair	= sock_no_socketpair,
1822 	.accept		= accept,
1823 	.getname	= get_name,
1824 	.poll		= poll,
1825 	.ioctl		= sock_no_ioctl,
1826 	.listen		= listen,
1827 	.shutdown	= shutdown,
1828 	.setsockopt	= setsockopt,
1829 	.getsockopt	= getsockopt,
1830 	.sendmsg	= send_stream,
1831 	.recvmsg	= recv_stream,
1832 	.mmap		= sock_no_mmap,
1833 	.sendpage	= sock_no_sendpage
1834 };
1835 
1836 static const struct net_proto_family tipc_family_ops = {
1837 	.owner 		= THIS_MODULE,
1838 	.family		= AF_TIPC,
1839 	.create		= tipc_create
1840 };
1841 
1842 static struct proto tipc_proto = {
1843 	.name		= "TIPC",
1844 	.owner		= THIS_MODULE,
1845 	.obj_size	= sizeof(struct tipc_sock)
1846 };
1847 
1848 /**
1849  * tipc_socket_init - initialize TIPC socket interface
1850  *
1851  * Returns 0 on success, errno otherwise
1852  */
1853 int tipc_socket_init(void)
1854 {
1855 	int res;
1856 
1857 	res = proto_register(&tipc_proto, 1);
1858 	if (res) {
1859 		err("Failed to register TIPC protocol type\n");
1860 		goto out;
1861 	}
1862 
1863 	res = sock_register(&tipc_family_ops);
1864 	if (res) {
1865 		err("Failed to register TIPC socket type\n");
1866 		proto_unregister(&tipc_proto);
1867 		goto out;
1868 	}
1869 
1870 	sockets_enabled = 1;
1871  out:
1872 	return res;
1873 }
1874 
1875 /**
1876  * tipc_socket_stop - stop TIPC socket interface
1877  */
1878 
1879 void tipc_socket_stop(void)
1880 {
1881 	if (!sockets_enabled)
1882 		return;
1883 
1884 	sockets_enabled = 0;
1885 	sock_unregister(tipc_family_ops.family);
1886 	proto_unregister(&tipc_proto);
1887 }
1888 
1889