xref: /openbmc/linux/net/tipc/socket.c (revision 9ac8d3fb)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, Ericsson AB
5  * Copyright (c) 2004-2008, 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 	struct tipc_portid peer_name;
67 };
68 
69 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
70 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
71 
72 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
73 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
74 static void wakeupdispatch(struct tipc_port *tport);
75 
76 static const struct proto_ops packet_ops;
77 static const struct proto_ops stream_ops;
78 static const struct proto_ops msg_ops;
79 
80 static struct proto tipc_proto;
81 
82 static int sockets_enabled = 0;
83 
84 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
85 
86 /*
87  * Revised TIPC socket locking policy:
88  *
89  * Most socket operations take the standard socket lock when they start
90  * and hold it until they finish (or until they need to sleep).  Acquiring
91  * this lock grants the owner exclusive access to the fields of the socket
92  * data structures, with the exception of the backlog queue.  A few socket
93  * operations can be done without taking the socket lock because they only
94  * read socket information that never changes during the life of the socket.
95  *
96  * Socket operations may acquire the lock for the associated TIPC port if they
97  * need to perform an operation on the port.  If any routine needs to acquire
98  * both the socket lock and the port lock it must take the socket lock first
99  * to avoid the risk of deadlock.
100  *
101  * The dispatcher handling incoming messages cannot grab the socket lock in
102  * the standard fashion, since invoked it runs at the BH level and cannot block.
103  * Instead, it checks to see if the socket lock is currently owned by someone,
104  * and either handles the message itself or adds it to the socket's backlog
105  * queue; in the latter case the queued message is processed once the process
106  * owning the socket lock releases it.
107  *
108  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
109  * the problem of a blocked socket operation preventing any other operations
110  * from occurring.  However, applications must be careful if they have
111  * multiple threads trying to send (or receive) on the same socket, as these
112  * operations might interfere with each other.  For example, doing a connect
113  * and a receive at the same time might allow the receive to consume the
114  * ACK message meant for the connect.  While additional work could be done
115  * to try and overcome this, it doesn't seem to be worthwhile at the present.
116  *
117  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
118  * that another operation that must be performed in a non-blocking manner is
119  * not delayed for very long because the lock has already been taken.
120  *
121  * NOTE: This code assumes that certain fields of a port/socket pair are
122  * constant over its lifetime; such fields can be examined without taking
123  * the socket lock and/or port lock, and do not need to be re-read even
124  * after resuming processing after waiting.  These fields include:
125  *   - socket type
126  *   - pointer to socket sk structure (aka tipc_sock structure)
127  *   - pointer to port structure
128  *   - port reference
129  */
130 
131 /**
132  * advance_rx_queue - discard first buffer in socket receive queue
133  *
134  * Caller must hold socket lock
135  */
136 
137 static void advance_rx_queue(struct sock *sk)
138 {
139 	buf_discard(__skb_dequeue(&sk->sk_receive_queue));
140 	atomic_dec(&tipc_queue_size);
141 }
142 
143 /**
144  * discard_rx_queue - discard all buffers in socket receive queue
145  *
146  * Caller must hold socket lock
147  */
148 
149 static void discard_rx_queue(struct sock *sk)
150 {
151 	struct sk_buff *buf;
152 
153 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
154 		atomic_dec(&tipc_queue_size);
155 		buf_discard(buf);
156 	}
157 }
158 
159 /**
160  * reject_rx_queue - reject all buffers in socket receive queue
161  *
162  * Caller must hold socket lock
163  */
164 
165 static void reject_rx_queue(struct sock *sk)
166 {
167 	struct sk_buff *buf;
168 
169 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
170 		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
171 		atomic_dec(&tipc_queue_size);
172 	}
173 }
174 
175 /**
176  * tipc_create - create a TIPC socket
177  * @net: network namespace (must be default network)
178  * @sock: pre-allocated socket structure
179  * @protocol: protocol indicator (must be 0)
180  *
181  * This routine creates additional data structures used by the TIPC socket,
182  * initializes them, and links them together.
183  *
184  * Returns 0 on success, errno otherwise
185  */
186 
187 static int tipc_create(struct net *net, struct socket *sock, int protocol)
188 {
189 	const struct proto_ops *ops;
190 	socket_state state;
191 	struct sock *sk;
192 	struct tipc_port *tp_ptr;
193 
194 	/* Validate arguments */
195 
196 	if (net != &init_net)
197 		return -EAFNOSUPPORT;
198 
199 	if (unlikely(protocol != 0))
200 		return -EPROTONOSUPPORT;
201 
202 	switch (sock->type) {
203 	case SOCK_STREAM:
204 		ops = &stream_ops;
205 		state = SS_UNCONNECTED;
206 		break;
207 	case SOCK_SEQPACKET:
208 		ops = &packet_ops;
209 		state = SS_UNCONNECTED;
210 		break;
211 	case SOCK_DGRAM:
212 	case SOCK_RDM:
213 		ops = &msg_ops;
214 		state = SS_READY;
215 		break;
216 	default:
217 		return -EPROTOTYPE;
218 	}
219 
220 	/* Allocate socket's protocol area */
221 
222 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
223 	if (sk == NULL)
224 		return -ENOMEM;
225 
226 	/* Allocate TIPC port for socket to use */
227 
228 	tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
229 				     TIPC_LOW_IMPORTANCE);
230 	if (unlikely(!tp_ptr)) {
231 		sk_free(sk);
232 		return -ENOMEM;
233 	}
234 
235 	/* Finish initializing socket data structures */
236 
237 	sock->ops = ops;
238 	sock->state = state;
239 
240 	sock_init_data(sock, sk);
241 	sk->sk_rcvtimeo = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
242 	sk->sk_backlog_rcv = backlog_rcv;
243 	tipc_sk(sk)->p = tp_ptr;
244 
245 	spin_unlock_bh(tp_ptr->lock);
246 
247 	if (sock->state == SS_READY) {
248 		tipc_set_portunreturnable(tp_ptr->ref, 1);
249 		if (sock->type == SOCK_DGRAM)
250 			tipc_set_portunreliable(tp_ptr->ref, 1);
251 	}
252 
253 	atomic_inc(&tipc_user_count);
254 	return 0;
255 }
256 
257 /**
258  * release - destroy a TIPC socket
259  * @sock: socket to destroy
260  *
261  * This routine cleans up any messages that are still queued on the socket.
262  * For DGRAM and RDM socket types, all queued messages are rejected.
263  * For SEQPACKET and STREAM socket types, the first message is rejected
264  * and any others are discarded.  (If the first message on a STREAM socket
265  * is partially-read, it is discarded and the next one is rejected instead.)
266  *
267  * NOTE: Rejected messages are not necessarily returned to the sender!  They
268  * are returned or discarded according to the "destination droppable" setting
269  * specified for the message by the sender.
270  *
271  * Returns 0 on success, errno otherwise
272  */
273 
274 static int release(struct socket *sock)
275 {
276 	struct sock *sk = sock->sk;
277 	struct tipc_port *tport;
278 	struct sk_buff *buf;
279 	int res;
280 
281 	/*
282 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
283 	 * releases a pre-allocated child socket that was never used)
284 	 */
285 
286 	if (sk == NULL)
287 		return 0;
288 
289 	tport = tipc_sk_port(sk);
290 	lock_sock(sk);
291 
292 	/*
293 	 * Reject all unreceived messages, except on an active connection
294 	 * (which disconnects locally & sends a 'FIN+' to peer)
295 	 */
296 
297 	while (sock->state != SS_DISCONNECTING) {
298 		buf = __skb_dequeue(&sk->sk_receive_queue);
299 		if (buf == NULL)
300 			break;
301 		atomic_dec(&tipc_queue_size);
302 		if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
303 			buf_discard(buf);
304 		else {
305 			if ((sock->state == SS_CONNECTING) ||
306 			    (sock->state == SS_CONNECTED)) {
307 				sock->state = SS_DISCONNECTING;
308 				tipc_disconnect(tport->ref);
309 			}
310 			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
311 		}
312 	}
313 
314 	/*
315 	 * Delete TIPC port; this ensures no more messages are queued
316 	 * (also disconnects an active connection & sends a 'FIN-' to peer)
317 	 */
318 
319 	res = tipc_deleteport(tport->ref);
320 
321 	/* Discard any remaining (connection-based) messages in receive queue */
322 
323 	discard_rx_queue(sk);
324 
325 	/* Reject any messages that accumulated in backlog queue */
326 
327 	sock->state = SS_DISCONNECTING;
328 	release_sock(sk);
329 
330 	sock_put(sk);
331 	sock->sk = NULL;
332 
333 	atomic_dec(&tipc_user_count);
334 	return res;
335 }
336 
337 /**
338  * bind - associate or disassocate TIPC name(s) with a socket
339  * @sock: socket structure
340  * @uaddr: socket address describing name(s) and desired operation
341  * @uaddr_len: size of socket address data structure
342  *
343  * Name and name sequence binding is indicated using a positive scope value;
344  * a negative scope value unbinds the specified name.  Specifying no name
345  * (i.e. a socket address length of 0) unbinds all names from the socket.
346  *
347  * Returns 0 on success, errno otherwise
348  *
349  * NOTE: This routine doesn't need to take the socket lock since it doesn't
350  *       access any non-constant socket information.
351  */
352 
353 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
354 {
355 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
356 	u32 portref = tipc_sk_port(sock->sk)->ref;
357 
358 	if (unlikely(!uaddr_len))
359 		return tipc_withdraw(portref, 0, NULL);
360 
361 	if (uaddr_len < sizeof(struct sockaddr_tipc))
362 		return -EINVAL;
363 	if (addr->family != AF_TIPC)
364 		return -EAFNOSUPPORT;
365 
366 	if (addr->addrtype == TIPC_ADDR_NAME)
367 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
368 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
369 		return -EAFNOSUPPORT;
370 
371 	return (addr->scope > 0) ?
372 		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
373 		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
374 }
375 
376 /**
377  * get_name - get port ID of socket or peer socket
378  * @sock: socket structure
379  * @uaddr: area for returned socket address
380  * @uaddr_len: area for returned length of socket address
381  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
382  *
383  * Returns 0 on success, errno otherwise
384  *
385  * NOTE: This routine doesn't need to take the socket lock since it only
386  *       accesses socket information that is unchanging (or which changes in
387  * 	 a completely predictable manner).
388  */
389 
390 static int get_name(struct socket *sock, struct sockaddr *uaddr,
391 		    int *uaddr_len, int peer)
392 {
393 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
394 	struct tipc_sock *tsock = tipc_sk(sock->sk);
395 
396 	if (peer) {
397 		if ((sock->state != SS_CONNECTED) &&
398 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
399 			return -ENOTCONN;
400 		addr->addr.id.ref = tsock->peer_name.ref;
401 		addr->addr.id.node = tsock->peer_name.node;
402 	} else {
403 		tipc_ownidentity(tsock->p->ref, &addr->addr.id);
404 	}
405 
406 	*uaddr_len = sizeof(*addr);
407 	addr->addrtype = TIPC_ADDR_ID;
408 	addr->family = AF_TIPC;
409 	addr->scope = 0;
410 	addr->addr.name.domain = 0;
411 
412 	return 0;
413 }
414 
415 /**
416  * poll - read and possibly block on pollmask
417  * @file: file structure associated with the socket
418  * @sock: socket for which to calculate the poll bits
419  * @wait: ???
420  *
421  * Returns pollmask value
422  *
423  * COMMENTARY:
424  * It appears that the usual socket locking mechanisms are not useful here
425  * since the pollmask info is potentially out-of-date the moment this routine
426  * exits.  TCP and other protocols seem to rely on higher level poll routines
427  * to handle any preventable race conditions, so TIPC will do the same ...
428  *
429  * TIPC sets the returned events as follows:
430  * a) POLLRDNORM and POLLIN are set if the socket's receive queue is non-empty
431  *    or if a connection-oriented socket is does not have an active connection
432  *    (i.e. a read operation will not block).
433  * b) POLLOUT is set except when a socket's connection has been terminated
434  *    (i.e. a write operation will not block).
435  * c) POLLHUP is set when a socket's connection has been terminated.
436  *
437  * IMPORTANT: The fact that a read or write operation will not block does NOT
438  * imply that the operation will succeed!
439  */
440 
441 static unsigned int poll(struct file *file, struct socket *sock,
442 			 poll_table *wait)
443 {
444 	struct sock *sk = sock->sk;
445 	u32 mask;
446 
447 	poll_wait(file, sk->sk_sleep, wait);
448 
449 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
450 	    (sock->state == SS_UNCONNECTED) ||
451 	    (sock->state == SS_DISCONNECTING))
452 		mask = (POLLRDNORM | POLLIN);
453 	else
454 		mask = 0;
455 
456 	if (sock->state == SS_DISCONNECTING)
457 		mask |= POLLHUP;
458 	else
459 		mask |= POLLOUT;
460 
461 	return mask;
462 }
463 
464 /**
465  * dest_name_check - verify user is permitted to send to specified port name
466  * @dest: destination address
467  * @m: descriptor for message to be sent
468  *
469  * Prevents restricted configuration commands from being issued by
470  * unauthorized users.
471  *
472  * Returns 0 if permission is granted, otherwise errno
473  */
474 
475 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
476 {
477 	struct tipc_cfg_msg_hdr hdr;
478 
479 	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
480 		return 0;
481 	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
482 		return 0;
483 	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
484 		return -EACCES;
485 
486 	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
487 		return -EFAULT;
488 	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
489 		return -EACCES;
490 
491 	return 0;
492 }
493 
494 /**
495  * send_msg - send message in connectionless manner
496  * @iocb: if NULL, indicates that socket lock is already held
497  * @sock: socket structure
498  * @m: message to send
499  * @total_len: length of message
500  *
501  * Message must have an destination specified explicitly.
502  * Used for SOCK_RDM and SOCK_DGRAM messages,
503  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
504  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
505  *
506  * Returns the number of bytes sent on success, or errno otherwise
507  */
508 
509 static int send_msg(struct kiocb *iocb, struct socket *sock,
510 		    struct msghdr *m, size_t total_len)
511 {
512 	struct sock *sk = sock->sk;
513 	struct tipc_port *tport = tipc_sk_port(sk);
514 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
515 	int needs_conn;
516 	int res = -EINVAL;
517 
518 	if (unlikely(!dest))
519 		return -EDESTADDRREQ;
520 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
521 		     (dest->family != AF_TIPC)))
522 		return -EINVAL;
523 
524 	if (iocb)
525 		lock_sock(sk);
526 
527 	needs_conn = (sock->state != SS_READY);
528 	if (unlikely(needs_conn)) {
529 		if (sock->state == SS_LISTENING) {
530 			res = -EPIPE;
531 			goto exit;
532 		}
533 		if (sock->state != SS_UNCONNECTED) {
534 			res = -EISCONN;
535 			goto exit;
536 		}
537 		if ((tport->published) ||
538 		    ((sock->type == SOCK_STREAM) && (total_len != 0))) {
539 			res = -EOPNOTSUPP;
540 			goto exit;
541 		}
542 		if (dest->addrtype == TIPC_ADDR_NAME) {
543 			tport->conn_type = dest->addr.name.name.type;
544 			tport->conn_instance = dest->addr.name.name.instance;
545 		}
546 
547 		/* Abort any pending connection attempts (very unlikely) */
548 
549 		reject_rx_queue(sk);
550 	}
551 
552 	do {
553 		if (dest->addrtype == TIPC_ADDR_NAME) {
554 			if ((res = dest_name_check(dest, m)))
555 				break;
556 			res = tipc_send2name(tport->ref,
557 					     &dest->addr.name.name,
558 					     dest->addr.name.domain,
559 					     m->msg_iovlen,
560 					     m->msg_iov);
561 		}
562 		else if (dest->addrtype == TIPC_ADDR_ID) {
563 			res = tipc_send2port(tport->ref,
564 					     &dest->addr.id,
565 					     m->msg_iovlen,
566 					     m->msg_iov);
567 		}
568 		else if (dest->addrtype == TIPC_ADDR_MCAST) {
569 			if (needs_conn) {
570 				res = -EOPNOTSUPP;
571 				break;
572 			}
573 			if ((res = dest_name_check(dest, m)))
574 				break;
575 			res = tipc_multicast(tport->ref,
576 					     &dest->addr.nameseq,
577 					     0,
578 					     m->msg_iovlen,
579 					     m->msg_iov);
580 		}
581 		if (likely(res != -ELINKCONG)) {
582 			if (needs_conn && (res >= 0)) {
583 				sock->state = SS_CONNECTING;
584 			}
585 			break;
586 		}
587 		if (m->msg_flags & MSG_DONTWAIT) {
588 			res = -EWOULDBLOCK;
589 			break;
590 		}
591 		release_sock(sk);
592 		res = wait_event_interruptible(*sk->sk_sleep,
593 					       !tport->congested);
594 		lock_sock(sk);
595 		if (res)
596 			break;
597 	} while (1);
598 
599 exit:
600 	if (iocb)
601 		release_sock(sk);
602 	return res;
603 }
604 
605 /**
606  * send_packet - send a connection-oriented message
607  * @iocb: if NULL, indicates that socket lock is already held
608  * @sock: socket structure
609  * @m: message to send
610  * @total_len: length of message
611  *
612  * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
613  *
614  * Returns the number of bytes sent on success, or errno otherwise
615  */
616 
617 static int send_packet(struct kiocb *iocb, struct socket *sock,
618 		       struct msghdr *m, size_t total_len)
619 {
620 	struct sock *sk = sock->sk;
621 	struct tipc_port *tport = tipc_sk_port(sk);
622 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
623 	int res;
624 
625 	/* Handle implied connection establishment */
626 
627 	if (unlikely(dest))
628 		return send_msg(iocb, sock, m, total_len);
629 
630 	if (iocb)
631 		lock_sock(sk);
632 
633 	do {
634 		if (unlikely(sock->state != SS_CONNECTED)) {
635 			if (sock->state == SS_DISCONNECTING)
636 				res = -EPIPE;
637 			else
638 				res = -ENOTCONN;
639 			break;
640 		}
641 
642 		res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
643 		if (likely(res != -ELINKCONG)) {
644 			break;
645 		}
646 		if (m->msg_flags & MSG_DONTWAIT) {
647 			res = -EWOULDBLOCK;
648 			break;
649 		}
650 		release_sock(sk);
651 		res = wait_event_interruptible(*sk->sk_sleep,
652 			(!tport->congested || !tport->connected));
653 		lock_sock(sk);
654 		if (res)
655 			break;
656 	} while (1);
657 
658 	if (iocb)
659 		release_sock(sk);
660 	return res;
661 }
662 
663 /**
664  * send_stream - send stream-oriented data
665  * @iocb: (unused)
666  * @sock: socket structure
667  * @m: data to send
668  * @total_len: total length of data to be sent
669  *
670  * Used for SOCK_STREAM data.
671  *
672  * Returns the number of bytes sent on success (or partial success),
673  * or errno if no data sent
674  */
675 
676 static int send_stream(struct kiocb *iocb, struct socket *sock,
677 		       struct msghdr *m, size_t total_len)
678 {
679 	struct sock *sk = sock->sk;
680 	struct tipc_port *tport = tipc_sk_port(sk);
681 	struct msghdr my_msg;
682 	struct iovec my_iov;
683 	struct iovec *curr_iov;
684 	int curr_iovlen;
685 	char __user *curr_start;
686 	u32 hdr_size;
687 	int curr_left;
688 	int bytes_to_send;
689 	int bytes_sent;
690 	int res;
691 
692 	lock_sock(sk);
693 
694 	/* Handle special cases where there is no connection */
695 
696 	if (unlikely(sock->state != SS_CONNECTED)) {
697 		if (sock->state == SS_UNCONNECTED) {
698 			res = send_packet(NULL, sock, m, total_len);
699 			goto exit;
700 		} else if (sock->state == SS_DISCONNECTING) {
701 			res = -EPIPE;
702 			goto exit;
703 		} else {
704 			res = -ENOTCONN;
705 			goto exit;
706 		}
707 	}
708 
709 	if (unlikely(m->msg_name)) {
710 		res = -EISCONN;
711 		goto exit;
712 	}
713 
714 	/*
715 	 * Send each iovec entry using one or more messages
716 	 *
717 	 * Note: This algorithm is good for the most likely case
718 	 * (i.e. one large iovec entry), but could be improved to pass sets
719 	 * of small iovec entries into send_packet().
720 	 */
721 
722 	curr_iov = m->msg_iov;
723 	curr_iovlen = m->msg_iovlen;
724 	my_msg.msg_iov = &my_iov;
725 	my_msg.msg_iovlen = 1;
726 	my_msg.msg_flags = m->msg_flags;
727 	my_msg.msg_name = NULL;
728 	bytes_sent = 0;
729 
730 	hdr_size = msg_hdr_sz(&tport->phdr);
731 
732 	while (curr_iovlen--) {
733 		curr_start = curr_iov->iov_base;
734 		curr_left = curr_iov->iov_len;
735 
736 		while (curr_left) {
737 			bytes_to_send = tport->max_pkt - hdr_size;
738 			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
739 				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
740 			if (curr_left < bytes_to_send)
741 				bytes_to_send = curr_left;
742 			my_iov.iov_base = curr_start;
743 			my_iov.iov_len = bytes_to_send;
744 			if ((res = send_packet(NULL, sock, &my_msg, 0)) < 0) {
745 				if (bytes_sent)
746 					res = bytes_sent;
747 				goto exit;
748 			}
749 			curr_left -= bytes_to_send;
750 			curr_start += bytes_to_send;
751 			bytes_sent += bytes_to_send;
752 		}
753 
754 		curr_iov++;
755 	}
756 	res = bytes_sent;
757 exit:
758 	release_sock(sk);
759 	return res;
760 }
761 
762 /**
763  * auto_connect - complete connection setup to a remote port
764  * @sock: socket structure
765  * @msg: peer's response message
766  *
767  * Returns 0 on success, errno otherwise
768  */
769 
770 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
771 {
772 	struct tipc_sock *tsock = tipc_sk(sock->sk);
773 
774 	if (msg_errcode(msg)) {
775 		sock->state = SS_DISCONNECTING;
776 		return -ECONNREFUSED;
777 	}
778 
779 	tsock->peer_name.ref = msg_origport(msg);
780 	tsock->peer_name.node = msg_orignode(msg);
781 	tipc_connect2port(tsock->p->ref, &tsock->peer_name);
782 	tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
783 	sock->state = SS_CONNECTED;
784 	return 0;
785 }
786 
787 /**
788  * set_orig_addr - capture sender's address for received message
789  * @m: descriptor for message info
790  * @msg: received message header
791  *
792  * Note: Address is not captured if not requested by receiver.
793  */
794 
795 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
796 {
797 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
798 
799 	if (addr) {
800 		addr->family = AF_TIPC;
801 		addr->addrtype = TIPC_ADDR_ID;
802 		addr->addr.id.ref = msg_origport(msg);
803 		addr->addr.id.node = msg_orignode(msg);
804 		addr->addr.name.domain = 0;   	/* could leave uninitialized */
805 		addr->scope = 0;   		/* could leave uninitialized */
806 		m->msg_namelen = sizeof(struct sockaddr_tipc);
807 	}
808 }
809 
810 /**
811  * anc_data_recv - optionally capture ancillary data for received message
812  * @m: descriptor for message info
813  * @msg: received message header
814  * @tport: TIPC port associated with message
815  *
816  * Note: Ancillary data is not captured if not requested by receiver.
817  *
818  * Returns 0 if successful, otherwise errno
819  */
820 
821 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
822 				struct tipc_port *tport)
823 {
824 	u32 anc_data[3];
825 	u32 err;
826 	u32 dest_type;
827 	int has_name;
828 	int res;
829 
830 	if (likely(m->msg_controllen == 0))
831 		return 0;
832 
833 	/* Optionally capture errored message object(s) */
834 
835 	err = msg ? msg_errcode(msg) : 0;
836 	if (unlikely(err)) {
837 		anc_data[0] = err;
838 		anc_data[1] = msg_data_sz(msg);
839 		if ((res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data)))
840 			return res;
841 		if (anc_data[1] &&
842 		    (res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
843 				    msg_data(msg))))
844 			return res;
845 	}
846 
847 	/* Optionally capture message destination object */
848 
849 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
850 	switch (dest_type) {
851 	case TIPC_NAMED_MSG:
852 		has_name = 1;
853 		anc_data[0] = msg_nametype(msg);
854 		anc_data[1] = msg_namelower(msg);
855 		anc_data[2] = msg_namelower(msg);
856 		break;
857 	case TIPC_MCAST_MSG:
858 		has_name = 1;
859 		anc_data[0] = msg_nametype(msg);
860 		anc_data[1] = msg_namelower(msg);
861 		anc_data[2] = msg_nameupper(msg);
862 		break;
863 	case TIPC_CONN_MSG:
864 		has_name = (tport->conn_type != 0);
865 		anc_data[0] = tport->conn_type;
866 		anc_data[1] = tport->conn_instance;
867 		anc_data[2] = tport->conn_instance;
868 		break;
869 	default:
870 		has_name = 0;
871 	}
872 	if (has_name &&
873 	    (res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data)))
874 		return res;
875 
876 	return 0;
877 }
878 
879 /**
880  * recv_msg - receive packet-oriented message
881  * @iocb: (unused)
882  * @m: descriptor for message info
883  * @buf_len: total size of user buffer area
884  * @flags: receive flags
885  *
886  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
887  * If the complete message doesn't fit in user area, truncate it.
888  *
889  * Returns size of returned message data, errno otherwise
890  */
891 
892 static int recv_msg(struct kiocb *iocb, struct socket *sock,
893 		    struct msghdr *m, size_t buf_len, int flags)
894 {
895 	struct sock *sk = sock->sk;
896 	struct tipc_port *tport = tipc_sk_port(sk);
897 	struct sk_buff *buf;
898 	struct tipc_msg *msg;
899 	unsigned int sz;
900 	u32 err;
901 	int res;
902 
903 	/* Catch invalid receive requests */
904 
905 	if (m->msg_iovlen != 1)
906 		return -EOPNOTSUPP;   /* Don't do multiple iovec entries yet */
907 
908 	if (unlikely(!buf_len))
909 		return -EINVAL;
910 
911 	lock_sock(sk);
912 
913 	if (unlikely(sock->state == SS_UNCONNECTED)) {
914 		res = -ENOTCONN;
915 		goto exit;
916 	}
917 
918 restart:
919 
920 	/* Look for a message in receive queue; wait if necessary */
921 
922 	while (skb_queue_empty(&sk->sk_receive_queue)) {
923 		if (sock->state == SS_DISCONNECTING) {
924 			res = -ENOTCONN;
925 			goto exit;
926 		}
927 		if (flags & MSG_DONTWAIT) {
928 			res = -EWOULDBLOCK;
929 			goto exit;
930 		}
931 		release_sock(sk);
932 		res = wait_event_interruptible(*sk->sk_sleep,
933 			(!skb_queue_empty(&sk->sk_receive_queue) ||
934 			 (sock->state == SS_DISCONNECTING)));
935 		lock_sock(sk);
936 		if (res)
937 			goto exit;
938 	}
939 
940 	/* Look at first message in receive queue */
941 
942 	buf = skb_peek(&sk->sk_receive_queue);
943 	msg = buf_msg(buf);
944 	sz = msg_data_sz(msg);
945 	err = msg_errcode(msg);
946 
947 	/* Complete connection setup for an implied connect */
948 
949 	if (unlikely(sock->state == SS_CONNECTING)) {
950 		res = auto_connect(sock, msg);
951 		if (res)
952 			goto exit;
953 	}
954 
955 	/* Discard an empty non-errored message & try again */
956 
957 	if ((!sz) && (!err)) {
958 		advance_rx_queue(sk);
959 		goto restart;
960 	}
961 
962 	/* Capture sender's address (optional) */
963 
964 	set_orig_addr(m, msg);
965 
966 	/* Capture ancillary data (optional) */
967 
968 	res = anc_data_recv(m, msg, tport);
969 	if (res)
970 		goto exit;
971 
972 	/* Capture message data (if valid) & compute return value (always) */
973 
974 	if (!err) {
975 		if (unlikely(buf_len < sz)) {
976 			sz = buf_len;
977 			m->msg_flags |= MSG_TRUNC;
978 		}
979 		if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
980 					  sz))) {
981 			res = -EFAULT;
982 			goto exit;
983 		}
984 		res = sz;
985 	} else {
986 		if ((sock->state == SS_READY) ||
987 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
988 			res = 0;
989 		else
990 			res = -ECONNRESET;
991 	}
992 
993 	/* Consume received message (optional) */
994 
995 	if (likely(!(flags & MSG_PEEK))) {
996 		if ((sock->state != SS_READY) &&
997 		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
998 			tipc_acknowledge(tport->ref, tport->conn_unacked);
999 		advance_rx_queue(sk);
1000 	}
1001 exit:
1002 	release_sock(sk);
1003 	return res;
1004 }
1005 
1006 /**
1007  * recv_stream - receive stream-oriented data
1008  * @iocb: (unused)
1009  * @m: descriptor for message info
1010  * @buf_len: total size of user buffer area
1011  * @flags: receive flags
1012  *
1013  * Used for SOCK_STREAM messages only.  If not enough data is available
1014  * will optionally wait for more; never truncates data.
1015  *
1016  * Returns size of returned message data, errno otherwise
1017  */
1018 
1019 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1020 		       struct msghdr *m, size_t buf_len, int flags)
1021 {
1022 	struct sock *sk = sock->sk;
1023 	struct tipc_port *tport = tipc_sk_port(sk);
1024 	struct sk_buff *buf;
1025 	struct tipc_msg *msg;
1026 	unsigned int sz;
1027 	int sz_to_copy;
1028 	int sz_copied = 0;
1029 	int needed;
1030 	char __user *crs = m->msg_iov->iov_base;
1031 	unsigned char *buf_crs;
1032 	u32 err;
1033 	int res = 0;
1034 
1035 	/* Catch invalid receive attempts */
1036 
1037 	if (m->msg_iovlen != 1)
1038 		return -EOPNOTSUPP;   /* Don't do multiple iovec entries yet */
1039 
1040 	if (unlikely(!buf_len))
1041 		return -EINVAL;
1042 
1043 	lock_sock(sk);
1044 
1045 	if (unlikely((sock->state == SS_UNCONNECTED) ||
1046 		     (sock->state == SS_CONNECTING))) {
1047 		res = -ENOTCONN;
1048 		goto exit;
1049 	}
1050 
1051 restart:
1052 
1053 	/* Look for a message in receive queue; wait if necessary */
1054 
1055 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1056 		if (sock->state == SS_DISCONNECTING) {
1057 			res = -ENOTCONN;
1058 			goto exit;
1059 		}
1060 		if (flags & MSG_DONTWAIT) {
1061 			res = -EWOULDBLOCK;
1062 			goto exit;
1063 		}
1064 		release_sock(sk);
1065 		res = wait_event_interruptible(*sk->sk_sleep,
1066 			(!skb_queue_empty(&sk->sk_receive_queue) ||
1067 			 (sock->state == SS_DISCONNECTING)));
1068 		lock_sock(sk);
1069 		if (res)
1070 			goto exit;
1071 	}
1072 
1073 	/* Look at first message in receive queue */
1074 
1075 	buf = skb_peek(&sk->sk_receive_queue);
1076 	msg = buf_msg(buf);
1077 	sz = msg_data_sz(msg);
1078 	err = msg_errcode(msg);
1079 
1080 	/* Discard an empty non-errored message & try again */
1081 
1082 	if ((!sz) && (!err)) {
1083 		advance_rx_queue(sk);
1084 		goto restart;
1085 	}
1086 
1087 	/* Optionally capture sender's address & ancillary data of first msg */
1088 
1089 	if (sz_copied == 0) {
1090 		set_orig_addr(m, msg);
1091 		res = anc_data_recv(m, msg, tport);
1092 		if (res)
1093 			goto exit;
1094 	}
1095 
1096 	/* Capture message data (if valid) & compute return value (always) */
1097 
1098 	if (!err) {
1099 		buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1100 		sz = (unsigned char *)msg + msg_size(msg) - buf_crs;
1101 
1102 		needed = (buf_len - sz_copied);
1103 		sz_to_copy = (sz <= needed) ? sz : needed;
1104 		if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1105 			res = -EFAULT;
1106 			goto exit;
1107 		}
1108 		sz_copied += sz_to_copy;
1109 
1110 		if (sz_to_copy < sz) {
1111 			if (!(flags & MSG_PEEK))
1112 				TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1113 			goto exit;
1114 		}
1115 
1116 		crs += sz_to_copy;
1117 	} else {
1118 		if (sz_copied != 0)
1119 			goto exit; /* can't add error msg to valid data */
1120 
1121 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1122 			res = 0;
1123 		else
1124 			res = -ECONNRESET;
1125 	}
1126 
1127 	/* Consume received message (optional) */
1128 
1129 	if (likely(!(flags & MSG_PEEK))) {
1130 		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1131 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1132 		advance_rx_queue(sk);
1133 	}
1134 
1135 	/* Loop around if more data is required */
1136 
1137 	if ((sz_copied < buf_len)    /* didn't get all requested data */
1138 	    && (!skb_queue_empty(&sk->sk_receive_queue) ||
1139 		(flags & MSG_WAITALL))
1140 				     /* ... and more is ready or required */
1141 	    && (!(flags & MSG_PEEK)) /* ... and aren't just peeking at data */
1142 	    && (!err)                /* ... and haven't reached a FIN */
1143 	    )
1144 		goto restart;
1145 
1146 exit:
1147 	release_sock(sk);
1148 	return sz_copied ? sz_copied : res;
1149 }
1150 
1151 /**
1152  * rx_queue_full - determine if receive queue can accept another message
1153  * @msg: message to be added to queue
1154  * @queue_size: current size of queue
1155  * @base: nominal maximum size of queue
1156  *
1157  * Returns 1 if queue is unable to accept message, 0 otherwise
1158  */
1159 
1160 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1161 {
1162 	u32 threshold;
1163 	u32 imp = msg_importance(msg);
1164 
1165 	if (imp == TIPC_LOW_IMPORTANCE)
1166 		threshold = base;
1167 	else if (imp == TIPC_MEDIUM_IMPORTANCE)
1168 		threshold = base * 2;
1169 	else if (imp == TIPC_HIGH_IMPORTANCE)
1170 		threshold = base * 100;
1171 	else
1172 		return 0;
1173 
1174 	if (msg_connected(msg))
1175 		threshold *= 4;
1176 
1177 	return (queue_size >= threshold);
1178 }
1179 
1180 /**
1181  * filter_rcv - validate incoming message
1182  * @sk: socket
1183  * @buf: message
1184  *
1185  * Enqueues message on receive queue if acceptable; optionally handles
1186  * disconnect indication for a connected socket.
1187  *
1188  * Called with socket lock already taken; port lock may also be taken.
1189  *
1190  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1191  */
1192 
1193 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1194 {
1195 	struct socket *sock = sk->sk_socket;
1196 	struct tipc_msg *msg = buf_msg(buf);
1197 	u32 recv_q_len;
1198 
1199 	/* Reject message if it is wrong sort of message for socket */
1200 
1201 	/*
1202 	 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1203 	 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1204 	 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1205 	 */
1206 
1207 	if (sock->state == SS_READY) {
1208 		if (msg_connected(msg)) {
1209 			msg_dbg(msg, "dispatch filter 1\n");
1210 			return TIPC_ERR_NO_PORT;
1211 		}
1212 	} else {
1213 		if (msg_mcast(msg)) {
1214 			msg_dbg(msg, "dispatch filter 2\n");
1215 			return TIPC_ERR_NO_PORT;
1216 		}
1217 		if (sock->state == SS_CONNECTED) {
1218 			if (!msg_connected(msg)) {
1219 				msg_dbg(msg, "dispatch filter 3\n");
1220 				return TIPC_ERR_NO_PORT;
1221 			}
1222 		}
1223 		else if (sock->state == SS_CONNECTING) {
1224 			if (!msg_connected(msg) && (msg_errcode(msg) == 0)) {
1225 				msg_dbg(msg, "dispatch filter 4\n");
1226 				return TIPC_ERR_NO_PORT;
1227 			}
1228 		}
1229 		else if (sock->state == SS_LISTENING) {
1230 			if (msg_connected(msg) || msg_errcode(msg)) {
1231 				msg_dbg(msg, "dispatch filter 5\n");
1232 				return TIPC_ERR_NO_PORT;
1233 			}
1234 		}
1235 		else if (sock->state == SS_DISCONNECTING) {
1236 			msg_dbg(msg, "dispatch filter 6\n");
1237 			return TIPC_ERR_NO_PORT;
1238 		}
1239 		else /* (sock->state == SS_UNCONNECTED) */ {
1240 			if (msg_connected(msg) || msg_errcode(msg)) {
1241 				msg_dbg(msg, "dispatch filter 7\n");
1242 				return TIPC_ERR_NO_PORT;
1243 			}
1244 		}
1245 	}
1246 
1247 	/* Reject message if there isn't room to queue it */
1248 
1249 	recv_q_len = (u32)atomic_read(&tipc_queue_size);
1250 	if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1251 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1252 			return TIPC_ERR_OVERLOAD;
1253 	}
1254 	recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1255 	if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1256 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1257 			return TIPC_ERR_OVERLOAD;
1258 	}
1259 
1260 	/* Enqueue message (finally!) */
1261 
1262 	msg_dbg(msg, "<DISP<: ");
1263 	TIPC_SKB_CB(buf)->handle = msg_data(msg);
1264 	atomic_inc(&tipc_queue_size);
1265 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1266 
1267 	/* Initiate connection termination for an incoming 'FIN' */
1268 
1269 	if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1270 		sock->state = SS_DISCONNECTING;
1271 		tipc_disconnect_port(tipc_sk_port(sk));
1272 	}
1273 
1274 	if (waitqueue_active(sk->sk_sleep))
1275 		wake_up_interruptible(sk->sk_sleep);
1276 	return TIPC_OK;
1277 }
1278 
1279 /**
1280  * backlog_rcv - handle incoming message from backlog queue
1281  * @sk: socket
1282  * @buf: message
1283  *
1284  * Caller must hold socket lock, but not port lock.
1285  *
1286  * Returns 0
1287  */
1288 
1289 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1290 {
1291 	u32 res;
1292 
1293 	res = filter_rcv(sk, buf);
1294 	if (res)
1295 		tipc_reject_msg(buf, res);
1296 	return 0;
1297 }
1298 
1299 /**
1300  * dispatch - handle incoming message
1301  * @tport: TIPC port that received message
1302  * @buf: message
1303  *
1304  * Called with port lock already taken.
1305  *
1306  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1307  */
1308 
1309 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1310 {
1311 	struct sock *sk = (struct sock *)tport->usr_handle;
1312 	u32 res;
1313 
1314 	/*
1315 	 * Process message if socket is unlocked; otherwise add to backlog queue
1316 	 *
1317 	 * This code is based on sk_receive_skb(), but must be distinct from it
1318 	 * since a TIPC-specific filter/reject mechanism is utilized
1319 	 */
1320 
1321 	bh_lock_sock(sk);
1322 	if (!sock_owned_by_user(sk)) {
1323 		res = filter_rcv(sk, buf);
1324 	} else {
1325 		sk_add_backlog(sk, buf);
1326 		res = TIPC_OK;
1327 	}
1328 	bh_unlock_sock(sk);
1329 
1330 	return res;
1331 }
1332 
1333 /**
1334  * wakeupdispatch - wake up port after congestion
1335  * @tport: port to wakeup
1336  *
1337  * Called with port lock already taken.
1338  */
1339 
1340 static void wakeupdispatch(struct tipc_port *tport)
1341 {
1342 	struct sock *sk = (struct sock *)tport->usr_handle;
1343 
1344 	if (waitqueue_active(sk->sk_sleep))
1345 		wake_up_interruptible(sk->sk_sleep);
1346 }
1347 
1348 /**
1349  * connect - establish a connection to another TIPC port
1350  * @sock: socket structure
1351  * @dest: socket address for destination port
1352  * @destlen: size of socket address data structure
1353  * @flags: file-related flags associated with socket
1354  *
1355  * Returns 0 on success, errno otherwise
1356  */
1357 
1358 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1359 		   int flags)
1360 {
1361 	struct sock *sk = sock->sk;
1362 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1363 	struct msghdr m = {NULL,};
1364 	struct sk_buff *buf;
1365 	struct tipc_msg *msg;
1366 	int res;
1367 
1368 	lock_sock(sk);
1369 
1370 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1371 
1372 	if (sock->state == SS_READY) {
1373 		res = -EOPNOTSUPP;
1374 		goto exit;
1375 	}
1376 
1377 	/* For now, TIPC does not support the non-blocking form of connect() */
1378 
1379 	if (flags & O_NONBLOCK) {
1380 		res = -EWOULDBLOCK;
1381 		goto exit;
1382 	}
1383 
1384 	/* Issue Posix-compliant error code if socket is in the wrong state */
1385 
1386 	if (sock->state == SS_LISTENING) {
1387 		res = -EOPNOTSUPP;
1388 		goto exit;
1389 	}
1390 	if (sock->state == SS_CONNECTING) {
1391 		res = -EALREADY;
1392 		goto exit;
1393 	}
1394 	if (sock->state != SS_UNCONNECTED) {
1395 		res = -EISCONN;
1396 		goto exit;
1397 	}
1398 
1399 	/*
1400 	 * Reject connection attempt using multicast address
1401 	 *
1402 	 * Note: send_msg() validates the rest of the address fields,
1403 	 *       so there's no need to do it here
1404 	 */
1405 
1406 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1407 		res = -EINVAL;
1408 		goto exit;
1409 	}
1410 
1411 	/* Reject any messages already in receive queue (very unlikely) */
1412 
1413 	reject_rx_queue(sk);
1414 
1415 	/* Send a 'SYN-' to destination */
1416 
1417 	m.msg_name = dest;
1418 	m.msg_namelen = destlen;
1419 	res = send_msg(NULL, sock, &m, 0);
1420 	if (res < 0) {
1421 		goto exit;
1422 	}
1423 
1424 	/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1425 
1426 	release_sock(sk);
1427 	res = wait_event_interruptible_timeout(*sk->sk_sleep,
1428 			(!skb_queue_empty(&sk->sk_receive_queue) ||
1429 			(sock->state != SS_CONNECTING)),
1430 			sk->sk_rcvtimeo);
1431 	lock_sock(sk);
1432 
1433 	if (res > 0) {
1434 		buf = skb_peek(&sk->sk_receive_queue);
1435 		if (buf != NULL) {
1436 			msg = buf_msg(buf);
1437 			res = auto_connect(sock, msg);
1438 			if (!res) {
1439 				if (!msg_data_sz(msg))
1440 					advance_rx_queue(sk);
1441 			}
1442 		} else {
1443 			if (sock->state == SS_CONNECTED) {
1444 				res = -EISCONN;
1445 			} else {
1446 				res = -ECONNREFUSED;
1447 			}
1448 		}
1449 	} else {
1450 		if (res == 0)
1451 			res = -ETIMEDOUT;
1452 		else
1453 			; /* leave "res" unchanged */
1454 		sock->state = SS_DISCONNECTING;
1455 	}
1456 
1457 exit:
1458 	release_sock(sk);
1459 	return res;
1460 }
1461 
1462 /**
1463  * listen - allow socket to listen for incoming connections
1464  * @sock: socket structure
1465  * @len: (unused)
1466  *
1467  * Returns 0 on success, errno otherwise
1468  */
1469 
1470 static int listen(struct socket *sock, int len)
1471 {
1472 	struct sock *sk = sock->sk;
1473 	int res;
1474 
1475 	lock_sock(sk);
1476 
1477 	if (sock->state == SS_READY)
1478 		res = -EOPNOTSUPP;
1479 	else if (sock->state != SS_UNCONNECTED)
1480 		res = -EINVAL;
1481 	else {
1482 		sock->state = SS_LISTENING;
1483 		res = 0;
1484 	}
1485 
1486 	release_sock(sk);
1487 	return res;
1488 }
1489 
1490 /**
1491  * accept - wait for connection request
1492  * @sock: listening socket
1493  * @newsock: new socket that is to be connected
1494  * @flags: file-related flags associated with socket
1495  *
1496  * Returns 0 on success, errno otherwise
1497  */
1498 
1499 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1500 {
1501 	struct sock *sk = sock->sk;
1502 	struct sk_buff *buf;
1503 	int res;
1504 
1505 	lock_sock(sk);
1506 
1507 	if (sock->state == SS_READY) {
1508 		res = -EOPNOTSUPP;
1509 		goto exit;
1510 	}
1511 	if (sock->state != SS_LISTENING) {
1512 		res = -EINVAL;
1513 		goto exit;
1514 	}
1515 
1516 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1517 		if (flags & O_NONBLOCK) {
1518 			res = -EWOULDBLOCK;
1519 			goto exit;
1520 		}
1521 		release_sock(sk);
1522 		res = wait_event_interruptible(*sk->sk_sleep,
1523 				(!skb_queue_empty(&sk->sk_receive_queue)));
1524 		lock_sock(sk);
1525 		if (res)
1526 			goto exit;
1527 	}
1528 
1529 	buf = skb_peek(&sk->sk_receive_queue);
1530 
1531 	res = tipc_create(sock_net(sock->sk), new_sock, 0);
1532 	if (!res) {
1533 		struct sock *new_sk = new_sock->sk;
1534 		struct tipc_sock *new_tsock = tipc_sk(new_sk);
1535 		struct tipc_port *new_tport = new_tsock->p;
1536 		u32 new_ref = new_tport->ref;
1537 		struct tipc_msg *msg = buf_msg(buf);
1538 
1539 		lock_sock(new_sk);
1540 
1541 		/*
1542 		 * Reject any stray messages received by new socket
1543 		 * before the socket lock was taken (very, very unlikely)
1544 		 */
1545 
1546 		reject_rx_queue(new_sk);
1547 
1548 		/* Connect new socket to it's peer */
1549 
1550 		new_tsock->peer_name.ref = msg_origport(msg);
1551 		new_tsock->peer_name.node = msg_orignode(msg);
1552 		tipc_connect2port(new_ref, &new_tsock->peer_name);
1553 		new_sock->state = SS_CONNECTED;
1554 
1555 		tipc_set_portimportance(new_ref, msg_importance(msg));
1556 		if (msg_named(msg)) {
1557 			new_tport->conn_type = msg_nametype(msg);
1558 			new_tport->conn_instance = msg_nameinst(msg);
1559 		}
1560 
1561 		/*
1562 		 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1563 		 * Respond to 'SYN+' by queuing it on new socket.
1564 		 */
1565 
1566 		msg_dbg(msg,"<ACC<: ");
1567 		if (!msg_data_sz(msg)) {
1568 			struct msghdr m = {NULL,};
1569 
1570 			advance_rx_queue(sk);
1571 			send_packet(NULL, new_sock, &m, 0);
1572 		} else {
1573 			__skb_dequeue(&sk->sk_receive_queue);
1574 			__skb_queue_head(&new_sk->sk_receive_queue, buf);
1575 		}
1576 		release_sock(new_sk);
1577 	}
1578 exit:
1579 	release_sock(sk);
1580 	return res;
1581 }
1582 
1583 /**
1584  * shutdown - shutdown socket connection
1585  * @sock: socket structure
1586  * @how: direction to close (must be SHUT_RDWR)
1587  *
1588  * Terminates connection (if necessary), then purges socket's receive queue.
1589  *
1590  * Returns 0 on success, errno otherwise
1591  */
1592 
1593 static int shutdown(struct socket *sock, int how)
1594 {
1595 	struct sock *sk = sock->sk;
1596 	struct tipc_port *tport = tipc_sk_port(sk);
1597 	struct sk_buff *buf;
1598 	int res;
1599 
1600 	if (how != SHUT_RDWR)
1601 		return -EINVAL;
1602 
1603 	lock_sock(sk);
1604 
1605 	switch (sock->state) {
1606 	case SS_CONNECTING:
1607 	case SS_CONNECTED:
1608 
1609 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1610 restart:
1611 		buf = __skb_dequeue(&sk->sk_receive_queue);
1612 		if (buf) {
1613 			atomic_dec(&tipc_queue_size);
1614 			if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1615 				buf_discard(buf);
1616 				goto restart;
1617 			}
1618 			tipc_disconnect(tport->ref);
1619 			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1620 		} else {
1621 			tipc_shutdown(tport->ref);
1622 		}
1623 
1624 		sock->state = SS_DISCONNECTING;
1625 
1626 		/* fall through */
1627 
1628 	case SS_DISCONNECTING:
1629 
1630 		/* Discard any unreceived messages; wake up sleeping tasks */
1631 
1632 		discard_rx_queue(sk);
1633 		if (waitqueue_active(sk->sk_sleep))
1634 			wake_up_interruptible(sk->sk_sleep);
1635 		res = 0;
1636 		break;
1637 
1638 	default:
1639 		res = -ENOTCONN;
1640 	}
1641 
1642 	release_sock(sk);
1643 	return res;
1644 }
1645 
1646 /**
1647  * setsockopt - set socket option
1648  * @sock: socket structure
1649  * @lvl: option level
1650  * @opt: option identifier
1651  * @ov: pointer to new option value
1652  * @ol: length of option value
1653  *
1654  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1655  * (to ease compatibility).
1656  *
1657  * Returns 0 on success, errno otherwise
1658  */
1659 
1660 static int setsockopt(struct socket *sock,
1661 		      int lvl, int opt, char __user *ov, int ol)
1662 {
1663 	struct sock *sk = sock->sk;
1664 	struct tipc_port *tport = tipc_sk_port(sk);
1665 	u32 value;
1666 	int res;
1667 
1668 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1669 		return 0;
1670 	if (lvl != SOL_TIPC)
1671 		return -ENOPROTOOPT;
1672 	if (ol < sizeof(value))
1673 		return -EINVAL;
1674 	if ((res = get_user(value, (u32 __user *)ov)))
1675 		return res;
1676 
1677 	lock_sock(sk);
1678 
1679 	switch (opt) {
1680 	case TIPC_IMPORTANCE:
1681 		res = tipc_set_portimportance(tport->ref, value);
1682 		break;
1683 	case TIPC_SRC_DROPPABLE:
1684 		if (sock->type != SOCK_STREAM)
1685 			res = tipc_set_portunreliable(tport->ref, value);
1686 		else
1687 			res = -ENOPROTOOPT;
1688 		break;
1689 	case TIPC_DEST_DROPPABLE:
1690 		res = tipc_set_portunreturnable(tport->ref, value);
1691 		break;
1692 	case TIPC_CONN_TIMEOUT:
1693 		sk->sk_rcvtimeo = msecs_to_jiffies(value);
1694 		/* no need to set "res", since already 0 at this point */
1695 		break;
1696 	default:
1697 		res = -EINVAL;
1698 	}
1699 
1700 	release_sock(sk);
1701 
1702 	return res;
1703 }
1704 
1705 /**
1706  * getsockopt - get socket option
1707  * @sock: socket structure
1708  * @lvl: option level
1709  * @opt: option identifier
1710  * @ov: receptacle for option value
1711  * @ol: receptacle for length of option value
1712  *
1713  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1714  * (to ease compatibility).
1715  *
1716  * Returns 0 on success, errno otherwise
1717  */
1718 
1719 static int getsockopt(struct socket *sock,
1720 		      int lvl, int opt, char __user *ov, int __user *ol)
1721 {
1722 	struct sock *sk = sock->sk;
1723 	struct tipc_port *tport = tipc_sk_port(sk);
1724 	int len;
1725 	u32 value;
1726 	int res;
1727 
1728 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1729 		return put_user(0, ol);
1730 	if (lvl != SOL_TIPC)
1731 		return -ENOPROTOOPT;
1732 	if ((res = get_user(len, ol)))
1733 		return res;
1734 
1735 	lock_sock(sk);
1736 
1737 	switch (opt) {
1738 	case TIPC_IMPORTANCE:
1739 		res = tipc_portimportance(tport->ref, &value);
1740 		break;
1741 	case TIPC_SRC_DROPPABLE:
1742 		res = tipc_portunreliable(tport->ref, &value);
1743 		break;
1744 	case TIPC_DEST_DROPPABLE:
1745 		res = tipc_portunreturnable(tport->ref, &value);
1746 		break;
1747 	case TIPC_CONN_TIMEOUT:
1748 		value = jiffies_to_msecs(sk->sk_rcvtimeo);
1749 		/* no need to set "res", since already 0 at this point */
1750 		break;
1751 	default:
1752 		res = -EINVAL;
1753 	}
1754 
1755 	release_sock(sk);
1756 
1757 	if (res) {
1758 		/* "get" failed */
1759 	}
1760 	else if (len < sizeof(value)) {
1761 		res = -EINVAL;
1762 	}
1763 	else if (copy_to_user(ov, &value, sizeof(value))) {
1764 		res = -EFAULT;
1765 	}
1766 	else {
1767 		res = put_user(sizeof(value), ol);
1768 	}
1769 
1770 	return res;
1771 }
1772 
1773 /**
1774  * Protocol switches for the various types of TIPC sockets
1775  */
1776 
1777 static const struct proto_ops msg_ops = {
1778 	.owner 		= THIS_MODULE,
1779 	.family		= AF_TIPC,
1780 	.release	= release,
1781 	.bind		= bind,
1782 	.connect	= connect,
1783 	.socketpair	= sock_no_socketpair,
1784 	.accept		= accept,
1785 	.getname	= get_name,
1786 	.poll		= poll,
1787 	.ioctl		= sock_no_ioctl,
1788 	.listen		= listen,
1789 	.shutdown	= shutdown,
1790 	.setsockopt	= setsockopt,
1791 	.getsockopt	= getsockopt,
1792 	.sendmsg	= send_msg,
1793 	.recvmsg	= recv_msg,
1794 	.mmap		= sock_no_mmap,
1795 	.sendpage	= sock_no_sendpage
1796 };
1797 
1798 static const struct proto_ops packet_ops = {
1799 	.owner 		= THIS_MODULE,
1800 	.family		= AF_TIPC,
1801 	.release	= release,
1802 	.bind		= bind,
1803 	.connect	= connect,
1804 	.socketpair	= sock_no_socketpair,
1805 	.accept		= accept,
1806 	.getname	= get_name,
1807 	.poll		= poll,
1808 	.ioctl		= sock_no_ioctl,
1809 	.listen		= listen,
1810 	.shutdown	= shutdown,
1811 	.setsockopt	= setsockopt,
1812 	.getsockopt	= getsockopt,
1813 	.sendmsg	= send_packet,
1814 	.recvmsg	= recv_msg,
1815 	.mmap		= sock_no_mmap,
1816 	.sendpage	= sock_no_sendpage
1817 };
1818 
1819 static const struct proto_ops stream_ops = {
1820 	.owner 		= THIS_MODULE,
1821 	.family		= AF_TIPC,
1822 	.release	= release,
1823 	.bind		= bind,
1824 	.connect	= connect,
1825 	.socketpair	= sock_no_socketpair,
1826 	.accept		= accept,
1827 	.getname	= get_name,
1828 	.poll		= poll,
1829 	.ioctl		= sock_no_ioctl,
1830 	.listen		= listen,
1831 	.shutdown	= shutdown,
1832 	.setsockopt	= setsockopt,
1833 	.getsockopt	= getsockopt,
1834 	.sendmsg	= send_stream,
1835 	.recvmsg	= recv_stream,
1836 	.mmap		= sock_no_mmap,
1837 	.sendpage	= sock_no_sendpage
1838 };
1839 
1840 static const struct net_proto_family tipc_family_ops = {
1841 	.owner 		= THIS_MODULE,
1842 	.family		= AF_TIPC,
1843 	.create		= tipc_create
1844 };
1845 
1846 static struct proto tipc_proto = {
1847 	.name		= "TIPC",
1848 	.owner		= THIS_MODULE,
1849 	.obj_size	= sizeof(struct tipc_sock)
1850 };
1851 
1852 /**
1853  * tipc_socket_init - initialize TIPC socket interface
1854  *
1855  * Returns 0 on success, errno otherwise
1856  */
1857 int tipc_socket_init(void)
1858 {
1859 	int res;
1860 
1861 	res = proto_register(&tipc_proto, 1);
1862 	if (res) {
1863 		err("Failed to register TIPC protocol type\n");
1864 		goto out;
1865 	}
1866 
1867 	res = sock_register(&tipc_family_ops);
1868 	if (res) {
1869 		err("Failed to register TIPC socket type\n");
1870 		proto_unregister(&tipc_proto);
1871 		goto out;
1872 	}
1873 
1874 	sockets_enabled = 1;
1875  out:
1876 	return res;
1877 }
1878 
1879 /**
1880  * tipc_socket_stop - stop TIPC socket interface
1881  */
1882 
1883 void tipc_socket_stop(void)
1884 {
1885 	if (!sockets_enabled)
1886 		return;
1887 
1888 	sockets_enabled = 0;
1889 	sock_unregister(tipc_family_ops.family);
1890 	proto_unregister(&tipc_proto);
1891 }
1892 
1893