xref: /linux/net/tipc/socket.c (revision 0d456bad36d42d16022be045c8a53ddbb59ee478)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012 Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2012, 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 "core.h"
38 #include "port.h"
39 
40 #include <linux/export.h>
41 #include <net/sock.h>
42 
43 #define SS_LISTENING	-1	/* socket is listening */
44 #define SS_READY	-2	/* socket is connectionless */
45 
46 #define OVERLOAD_LIMIT_BASE	10000
47 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
48 
49 struct tipc_sock {
50 	struct sock sk;
51 	struct tipc_port *p;
52 	struct tipc_portid peer_name;
53 	unsigned int conn_timeout;
54 };
55 
56 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
57 #define tipc_sk_port(sk) (tipc_sk(sk)->p)
58 
59 #define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
60 			(sock->state == SS_DISCONNECTING))
61 
62 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
63 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
64 static void wakeupdispatch(struct tipc_port *tport);
65 static void tipc_data_ready(struct sock *sk, int len);
66 static void tipc_write_space(struct sock *sk);
67 
68 static const struct proto_ops packet_ops;
69 static const struct proto_ops stream_ops;
70 static const struct proto_ops msg_ops;
71 
72 static struct proto tipc_proto;
73 
74 static int sockets_enabled;
75 
76 /*
77  * Revised TIPC socket locking policy:
78  *
79  * Most socket operations take the standard socket lock when they start
80  * and hold it until they finish (or until they need to sleep).  Acquiring
81  * this lock grants the owner exclusive access to the fields of the socket
82  * data structures, with the exception of the backlog queue.  A few socket
83  * operations can be done without taking the socket lock because they only
84  * read socket information that never changes during the life of the socket.
85  *
86  * Socket operations may acquire the lock for the associated TIPC port if they
87  * need to perform an operation on the port.  If any routine needs to acquire
88  * both the socket lock and the port lock it must take the socket lock first
89  * to avoid the risk of deadlock.
90  *
91  * The dispatcher handling incoming messages cannot grab the socket lock in
92  * the standard fashion, since invoked it runs at the BH level and cannot block.
93  * Instead, it checks to see if the socket lock is currently owned by someone,
94  * and either handles the message itself or adds it to the socket's backlog
95  * queue; in the latter case the queued message is processed once the process
96  * owning the socket lock releases it.
97  *
98  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
99  * the problem of a blocked socket operation preventing any other operations
100  * from occurring.  However, applications must be careful if they have
101  * multiple threads trying to send (or receive) on the same socket, as these
102  * operations might interfere with each other.  For example, doing a connect
103  * and a receive at the same time might allow the receive to consume the
104  * ACK message meant for the connect.  While additional work could be done
105  * to try and overcome this, it doesn't seem to be worthwhile at the present.
106  *
107  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
108  * that another operation that must be performed in a non-blocking manner is
109  * not delayed for very long because the lock has already been taken.
110  *
111  * NOTE: This code assumes that certain fields of a port/socket pair are
112  * constant over its lifetime; such fields can be examined without taking
113  * the socket lock and/or port lock, and do not need to be re-read even
114  * after resuming processing after waiting.  These fields include:
115  *   - socket type
116  *   - pointer to socket sk structure (aka tipc_sock structure)
117  *   - pointer to port structure
118  *   - port reference
119  */
120 
121 /**
122  * advance_rx_queue - discard first buffer in socket receive queue
123  *
124  * Caller must hold socket lock
125  */
126 static void advance_rx_queue(struct sock *sk)
127 {
128 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
129 }
130 
131 /**
132  * discard_rx_queue - discard all buffers in socket receive queue
133  *
134  * Caller must hold socket lock
135  */
136 static void discard_rx_queue(struct sock *sk)
137 {
138 	struct sk_buff *buf;
139 
140 	while ((buf = __skb_dequeue(&sk->sk_receive_queue)))
141 		kfree_skb(buf);
142 }
143 
144 /**
145  * reject_rx_queue - reject all buffers in socket receive queue
146  *
147  * Caller must hold socket lock
148  */
149 static void reject_rx_queue(struct sock *sk)
150 {
151 	struct sk_buff *buf;
152 
153 	while ((buf = __skb_dequeue(&sk->sk_receive_queue)))
154 		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
155 }
156 
157 /**
158  * tipc_create - create a TIPC socket
159  * @net: network namespace (must be default network)
160  * @sock: pre-allocated socket structure
161  * @protocol: protocol indicator (must be 0)
162  * @kern: caused by kernel or by userspace?
163  *
164  * This routine creates additional data structures used by the TIPC socket,
165  * initializes them, and links them together.
166  *
167  * Returns 0 on success, errno otherwise
168  */
169 static int tipc_create(struct net *net, struct socket *sock, int protocol,
170 		       int kern)
171 {
172 	const struct proto_ops *ops;
173 	socket_state state;
174 	struct sock *sk;
175 	struct tipc_port *tp_ptr;
176 
177 	/* Validate arguments */
178 	if (unlikely(protocol != 0))
179 		return -EPROTONOSUPPORT;
180 
181 	switch (sock->type) {
182 	case SOCK_STREAM:
183 		ops = &stream_ops;
184 		state = SS_UNCONNECTED;
185 		break;
186 	case SOCK_SEQPACKET:
187 		ops = &packet_ops;
188 		state = SS_UNCONNECTED;
189 		break;
190 	case SOCK_DGRAM:
191 	case SOCK_RDM:
192 		ops = &msg_ops;
193 		state = SS_READY;
194 		break;
195 	default:
196 		return -EPROTOTYPE;
197 	}
198 
199 	/* Allocate socket's protocol area */
200 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
201 	if (sk == NULL)
202 		return -ENOMEM;
203 
204 	/* Allocate TIPC port for socket to use */
205 	tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
206 				     TIPC_LOW_IMPORTANCE);
207 	if (unlikely(!tp_ptr)) {
208 		sk_free(sk);
209 		return -ENOMEM;
210 	}
211 
212 	/* Finish initializing socket data structures */
213 	sock->ops = ops;
214 	sock->state = state;
215 
216 	sock_init_data(sock, sk);
217 	sk->sk_backlog_rcv = backlog_rcv;
218 	sk->sk_rcvbuf = TIPC_FLOW_CONTROL_WIN * 2 * TIPC_MAX_USER_MSG_SIZE * 2;
219 	sk->sk_data_ready = tipc_data_ready;
220 	sk->sk_write_space = tipc_write_space;
221 	tipc_sk(sk)->p = tp_ptr;
222 	tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT;
223 
224 	spin_unlock_bh(tp_ptr->lock);
225 
226 	if (sock->state == SS_READY) {
227 		tipc_set_portunreturnable(tp_ptr->ref, 1);
228 		if (sock->type == SOCK_DGRAM)
229 			tipc_set_portunreliable(tp_ptr->ref, 1);
230 	}
231 
232 	return 0;
233 }
234 
235 /**
236  * release - destroy a TIPC socket
237  * @sock: socket to destroy
238  *
239  * This routine cleans up any messages that are still queued on the socket.
240  * For DGRAM and RDM socket types, all queued messages are rejected.
241  * For SEQPACKET and STREAM socket types, the first message is rejected
242  * and any others are discarded.  (If the first message on a STREAM socket
243  * is partially-read, it is discarded and the next one is rejected instead.)
244  *
245  * NOTE: Rejected messages are not necessarily returned to the sender!  They
246  * are returned or discarded according to the "destination droppable" setting
247  * specified for the message by the sender.
248  *
249  * Returns 0 on success, errno otherwise
250  */
251 static int release(struct socket *sock)
252 {
253 	struct sock *sk = sock->sk;
254 	struct tipc_port *tport;
255 	struct sk_buff *buf;
256 	int res;
257 
258 	/*
259 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
260 	 * releases a pre-allocated child socket that was never used)
261 	 */
262 	if (sk == NULL)
263 		return 0;
264 
265 	tport = tipc_sk_port(sk);
266 	lock_sock(sk);
267 
268 	/*
269 	 * Reject all unreceived messages, except on an active connection
270 	 * (which disconnects locally & sends a 'FIN+' to peer)
271 	 */
272 	while (sock->state != SS_DISCONNECTING) {
273 		buf = __skb_dequeue(&sk->sk_receive_queue);
274 		if (buf == NULL)
275 			break;
276 		if (TIPC_SKB_CB(buf)->handle != 0)
277 			kfree_skb(buf);
278 		else {
279 			if ((sock->state == SS_CONNECTING) ||
280 			    (sock->state == SS_CONNECTED)) {
281 				sock->state = SS_DISCONNECTING;
282 				tipc_disconnect(tport->ref);
283 			}
284 			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
285 		}
286 	}
287 
288 	/*
289 	 * Delete TIPC port; this ensures no more messages are queued
290 	 * (also disconnects an active connection & sends a 'FIN-' to peer)
291 	 */
292 	res = tipc_deleteport(tport->ref);
293 
294 	/* Discard any remaining (connection-based) messages in receive queue */
295 	discard_rx_queue(sk);
296 
297 	/* Reject any messages that accumulated in backlog queue */
298 	sock->state = SS_DISCONNECTING;
299 	release_sock(sk);
300 
301 	sock_put(sk);
302 	sock->sk = NULL;
303 
304 	return res;
305 }
306 
307 /**
308  * bind - associate or disassocate TIPC name(s) with a socket
309  * @sock: socket structure
310  * @uaddr: socket address describing name(s) and desired operation
311  * @uaddr_len: size of socket address data structure
312  *
313  * Name and name sequence binding is indicated using a positive scope value;
314  * a negative scope value unbinds the specified name.  Specifying no name
315  * (i.e. a socket address length of 0) unbinds all names from the socket.
316  *
317  * Returns 0 on success, errno otherwise
318  *
319  * NOTE: This routine doesn't need to take the socket lock since it doesn't
320  *       access any non-constant socket information.
321  */
322 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
323 {
324 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
325 	u32 portref = tipc_sk_port(sock->sk)->ref;
326 
327 	if (unlikely(!uaddr_len))
328 		return tipc_withdraw(portref, 0, NULL);
329 
330 	if (uaddr_len < sizeof(struct sockaddr_tipc))
331 		return -EINVAL;
332 	if (addr->family != AF_TIPC)
333 		return -EAFNOSUPPORT;
334 
335 	if (addr->addrtype == TIPC_ADDR_NAME)
336 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
337 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
338 		return -EAFNOSUPPORT;
339 
340 	if (addr->addr.nameseq.type < TIPC_RESERVED_TYPES)
341 		return -EACCES;
342 
343 	return (addr->scope > 0) ?
344 		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
345 		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
346 }
347 
348 /**
349  * get_name - get port ID of socket or peer socket
350  * @sock: socket structure
351  * @uaddr: area for returned socket address
352  * @uaddr_len: area for returned length of socket address
353  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
354  *
355  * Returns 0 on success, errno otherwise
356  *
357  * NOTE: This routine doesn't need to take the socket lock since it only
358  *       accesses socket information that is unchanging (or which changes in
359  *       a completely predictable manner).
360  */
361 static int get_name(struct socket *sock, struct sockaddr *uaddr,
362 		    int *uaddr_len, int peer)
363 {
364 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
365 	struct tipc_sock *tsock = tipc_sk(sock->sk);
366 
367 	memset(addr, 0, sizeof(*addr));
368 	if (peer) {
369 		if ((sock->state != SS_CONNECTED) &&
370 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
371 			return -ENOTCONN;
372 		addr->addr.id.ref = tsock->peer_name.ref;
373 		addr->addr.id.node = tsock->peer_name.node;
374 	} else {
375 		addr->addr.id.ref = tsock->p->ref;
376 		addr->addr.id.node = tipc_own_addr;
377 	}
378 
379 	*uaddr_len = sizeof(*addr);
380 	addr->addrtype = TIPC_ADDR_ID;
381 	addr->family = AF_TIPC;
382 	addr->scope = 0;
383 	addr->addr.name.domain = 0;
384 
385 	return 0;
386 }
387 
388 /**
389  * poll - read and possibly block on pollmask
390  * @file: file structure associated with the socket
391  * @sock: socket for which to calculate the poll bits
392  * @wait: ???
393  *
394  * Returns pollmask value
395  *
396  * COMMENTARY:
397  * It appears that the usual socket locking mechanisms are not useful here
398  * since the pollmask info is potentially out-of-date the moment this routine
399  * exits.  TCP and other protocols seem to rely on higher level poll routines
400  * to handle any preventable race conditions, so TIPC will do the same ...
401  *
402  * TIPC sets the returned events as follows:
403  *
404  * socket state		flags set
405  * ------------		---------
406  * unconnected		no read flags
407  *			POLLOUT if port is not congested
408  *
409  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
410  *			no write flags
411  *
412  * connected		POLLIN/POLLRDNORM if data in rx queue
413  *			POLLOUT if port is not congested
414  *
415  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
416  *			no write flags
417  *
418  * listening		POLLIN if SYN in rx queue
419  *			no write flags
420  *
421  * ready		POLLIN/POLLRDNORM if data in rx queue
422  * [connectionless]	POLLOUT (since port cannot be congested)
423  *
424  * IMPORTANT: The fact that a read or write operation is indicated does NOT
425  * imply that the operation will succeed, merely that it should be performed
426  * and will not block.
427  */
428 static unsigned int poll(struct file *file, struct socket *sock,
429 			 poll_table *wait)
430 {
431 	struct sock *sk = sock->sk;
432 	u32 mask = 0;
433 
434 	sock_poll_wait(file, sk_sleep(sk), wait);
435 
436 	switch ((int)sock->state) {
437 	case SS_UNCONNECTED:
438 		if (!tipc_sk_port(sk)->congested)
439 			mask |= POLLOUT;
440 		break;
441 	case SS_READY:
442 	case SS_CONNECTED:
443 		if (!tipc_sk_port(sk)->congested)
444 			mask |= POLLOUT;
445 		/* fall thru' */
446 	case SS_CONNECTING:
447 	case SS_LISTENING:
448 		if (!skb_queue_empty(&sk->sk_receive_queue))
449 			mask |= (POLLIN | POLLRDNORM);
450 		break;
451 	case SS_DISCONNECTING:
452 		mask = (POLLIN | POLLRDNORM | POLLHUP);
453 		break;
454 	}
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 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
470 {
471 	struct tipc_cfg_msg_hdr hdr;
472 
473 	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
474 		return 0;
475 	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
476 		return 0;
477 	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
478 		return -EACCES;
479 
480 	if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
481 		return -EMSGSIZE;
482 	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
483 		return -EFAULT;
484 	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
485 		return -EACCES;
486 
487 	return 0;
488 }
489 
490 /**
491  * send_msg - send message in connectionless manner
492  * @iocb: if NULL, indicates that socket lock is already held
493  * @sock: socket structure
494  * @m: message to send
495  * @total_len: length of message
496  *
497  * Message must have an destination specified explicitly.
498  * Used for SOCK_RDM and SOCK_DGRAM messages,
499  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
500  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
501  *
502  * Returns the number of bytes sent on success, or errno otherwise
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 	long timeout_val;
512 	int res = -EINVAL;
513 
514 	if (unlikely(!dest))
515 		return -EDESTADDRREQ;
516 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
517 		     (dest->family != AF_TIPC)))
518 		return -EINVAL;
519 	if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
520 	    (m->msg_iovlen > (unsigned int)INT_MAX))
521 		return -EMSGSIZE;
522 
523 	if (iocb)
524 		lock_sock(sk);
525 
526 	needs_conn = (sock->state != SS_READY);
527 	if (unlikely(needs_conn)) {
528 		if (sock->state == SS_LISTENING) {
529 			res = -EPIPE;
530 			goto exit;
531 		}
532 		if (sock->state != SS_UNCONNECTED) {
533 			res = -EISCONN;
534 			goto exit;
535 		}
536 		if ((tport->published) ||
537 		    ((sock->type == SOCK_STREAM) && (total_len != 0))) {
538 			res = -EOPNOTSUPP;
539 			goto exit;
540 		}
541 		if (dest->addrtype == TIPC_ADDR_NAME) {
542 			tport->conn_type = dest->addr.name.name.type;
543 			tport->conn_instance = dest->addr.name.name.instance;
544 		}
545 
546 		/* Abort any pending connection attempts (very unlikely) */
547 		reject_rx_queue(sk);
548 	}
549 
550 	timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
551 
552 	do {
553 		if (dest->addrtype == TIPC_ADDR_NAME) {
554 			res = dest_name_check(dest, m);
555 			if (res)
556 				break;
557 			res = tipc_send2name(tport->ref,
558 					     &dest->addr.name.name,
559 					     dest->addr.name.domain,
560 					     m->msg_iovlen,
561 					     m->msg_iov,
562 					     total_len);
563 		} else if (dest->addrtype == TIPC_ADDR_ID) {
564 			res = tipc_send2port(tport->ref,
565 					     &dest->addr.id,
566 					     m->msg_iovlen,
567 					     m->msg_iov,
568 					     total_len);
569 		} else if (dest->addrtype == TIPC_ADDR_MCAST) {
570 			if (needs_conn) {
571 				res = -EOPNOTSUPP;
572 				break;
573 			}
574 			res = dest_name_check(dest, m);
575 			if (res)
576 				break;
577 			res = tipc_multicast(tport->ref,
578 					     &dest->addr.nameseq,
579 					     m->msg_iovlen,
580 					     m->msg_iov,
581 					     total_len);
582 		}
583 		if (likely(res != -ELINKCONG)) {
584 			if (needs_conn && (res >= 0))
585 				sock->state = SS_CONNECTING;
586 			break;
587 		}
588 		if (timeout_val <= 0L) {
589 			res = timeout_val ? timeout_val : -EWOULDBLOCK;
590 			break;
591 		}
592 		release_sock(sk);
593 		timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
594 					       !tport->congested, timeout_val);
595 		lock_sock(sk);
596 	} while (1);
597 
598 exit:
599 	if (iocb)
600 		release_sock(sk);
601 	return res;
602 }
603 
604 /**
605  * send_packet - send a connection-oriented message
606  * @iocb: if NULL, indicates that socket lock is already held
607  * @sock: socket structure
608  * @m: message to send
609  * @total_len: length of message
610  *
611  * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
612  *
613  * Returns the number of bytes sent on success, or errno otherwise
614  */
615 static int send_packet(struct kiocb *iocb, struct socket *sock,
616 		       struct msghdr *m, size_t total_len)
617 {
618 	struct sock *sk = sock->sk;
619 	struct tipc_port *tport = tipc_sk_port(sk);
620 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
621 	long timeout_val;
622 	int res;
623 
624 	/* Handle implied connection establishment */
625 	if (unlikely(dest))
626 		return send_msg(iocb, sock, m, total_len);
627 
628 	if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
629 	    (m->msg_iovlen > (unsigned int)INT_MAX))
630 		return -EMSGSIZE;
631 
632 	if (iocb)
633 		lock_sock(sk);
634 
635 	timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
636 
637 	do {
638 		if (unlikely(sock->state != SS_CONNECTED)) {
639 			if (sock->state == SS_DISCONNECTING)
640 				res = -EPIPE;
641 			else
642 				res = -ENOTCONN;
643 			break;
644 		}
645 
646 		res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
647 				total_len);
648 		if (likely(res != -ELINKCONG))
649 			break;
650 		if (timeout_val <= 0L) {
651 			res = timeout_val ? timeout_val : -EWOULDBLOCK;
652 			break;
653 		}
654 		release_sock(sk);
655 		timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
656 			(!tport->congested || !tport->connected), timeout_val);
657 		lock_sock(sk);
658 	} while (1);
659 
660 	if (iocb)
661 		release_sock(sk);
662 	return res;
663 }
664 
665 /**
666  * send_stream - send stream-oriented data
667  * @iocb: (unused)
668  * @sock: socket structure
669  * @m: data to send
670  * @total_len: total length of data to be sent
671  *
672  * Used for SOCK_STREAM data.
673  *
674  * Returns the number of bytes sent on success (or partial success),
675  * or errno if no data sent
676  */
677 static int send_stream(struct kiocb *iocb, struct socket *sock,
678 		       struct msghdr *m, size_t total_len)
679 {
680 	struct sock *sk = sock->sk;
681 	struct tipc_port *tport = tipc_sk_port(sk);
682 	struct msghdr my_msg;
683 	struct iovec my_iov;
684 	struct iovec *curr_iov;
685 	int curr_iovlen;
686 	char __user *curr_start;
687 	u32 hdr_size;
688 	int curr_left;
689 	int bytes_to_send;
690 	int bytes_sent;
691 	int res;
692 
693 	lock_sock(sk);
694 
695 	/* Handle special cases where there is no connection */
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 	if ((total_len > (unsigned int)INT_MAX) ||
715 	    (m->msg_iovlen > (unsigned int)INT_MAX)) {
716 		res = -EMSGSIZE;
717 		goto exit;
718 	}
719 
720 	/*
721 	 * Send each iovec entry using one or more messages
722 	 *
723 	 * Note: This algorithm is good for the most likely case
724 	 * (i.e. one large iovec entry), but could be improved to pass sets
725 	 * of small iovec entries into send_packet().
726 	 */
727 	curr_iov = m->msg_iov;
728 	curr_iovlen = m->msg_iovlen;
729 	my_msg.msg_iov = &my_iov;
730 	my_msg.msg_iovlen = 1;
731 	my_msg.msg_flags = m->msg_flags;
732 	my_msg.msg_name = NULL;
733 	bytes_sent = 0;
734 
735 	hdr_size = msg_hdr_sz(&tport->phdr);
736 
737 	while (curr_iovlen--) {
738 		curr_start = curr_iov->iov_base;
739 		curr_left = curr_iov->iov_len;
740 
741 		while (curr_left) {
742 			bytes_to_send = tport->max_pkt - hdr_size;
743 			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
744 				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
745 			if (curr_left < bytes_to_send)
746 				bytes_to_send = curr_left;
747 			my_iov.iov_base = curr_start;
748 			my_iov.iov_len = bytes_to_send;
749 			res = send_packet(NULL, sock, &my_msg, bytes_to_send);
750 			if (res < 0) {
751 				if (bytes_sent)
752 					res = bytes_sent;
753 				goto exit;
754 			}
755 			curr_left -= bytes_to_send;
756 			curr_start += bytes_to_send;
757 			bytes_sent += bytes_to_send;
758 		}
759 
760 		curr_iov++;
761 	}
762 	res = bytes_sent;
763 exit:
764 	release_sock(sk);
765 	return res;
766 }
767 
768 /**
769  * auto_connect - complete connection setup to a remote port
770  * @sock: socket structure
771  * @msg: peer's response message
772  *
773  * Returns 0 on success, errno otherwise
774  */
775 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
776 {
777 	struct tipc_sock *tsock = tipc_sk(sock->sk);
778 	struct tipc_port *p_ptr;
779 
780 	tsock->peer_name.ref = msg_origport(msg);
781 	tsock->peer_name.node = msg_orignode(msg);
782 	p_ptr = tipc_port_deref(tsock->p->ref);
783 	if (!p_ptr)
784 		return -EINVAL;
785 
786 	__tipc_connect(tsock->p->ref, p_ptr, &tsock->peer_name);
787 
788 	if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)
789 		return -EINVAL;
790 	msg_set_importance(&p_ptr->phdr, (u32)msg_importance(msg));
791 	sock->state = SS_CONNECTED;
792 	return 0;
793 }
794 
795 /**
796  * set_orig_addr - capture sender's address for received message
797  * @m: descriptor for message info
798  * @msg: received message header
799  *
800  * Note: Address is not captured if not requested by receiver.
801  */
802 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
803 {
804 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
805 
806 	if (addr) {
807 		addr->family = AF_TIPC;
808 		addr->addrtype = TIPC_ADDR_ID;
809 		addr->addr.id.ref = msg_origport(msg);
810 		addr->addr.id.node = msg_orignode(msg);
811 		addr->addr.name.domain = 0;	/* could leave uninitialized */
812 		addr->scope = 0;		/* could leave uninitialized */
813 		m->msg_namelen = sizeof(struct sockaddr_tipc);
814 	}
815 }
816 
817 /**
818  * anc_data_recv - optionally capture ancillary data for received message
819  * @m: descriptor for message info
820  * @msg: received message header
821  * @tport: TIPC port associated with message
822  *
823  * Note: Ancillary data is not captured if not requested by receiver.
824  *
825  * Returns 0 if successful, otherwise errno
826  */
827 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
828 				struct tipc_port *tport)
829 {
830 	u32 anc_data[3];
831 	u32 err;
832 	u32 dest_type;
833 	int has_name;
834 	int res;
835 
836 	if (likely(m->msg_controllen == 0))
837 		return 0;
838 
839 	/* Optionally capture errored message object(s) */
840 	err = msg ? msg_errcode(msg) : 0;
841 	if (unlikely(err)) {
842 		anc_data[0] = err;
843 		anc_data[1] = msg_data_sz(msg);
844 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
845 		if (res)
846 			return res;
847 		if (anc_data[1]) {
848 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
849 				       msg_data(msg));
850 			if (res)
851 				return res;
852 		}
853 	}
854 
855 	/* Optionally capture message destination object */
856 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
857 	switch (dest_type) {
858 	case TIPC_NAMED_MSG:
859 		has_name = 1;
860 		anc_data[0] = msg_nametype(msg);
861 		anc_data[1] = msg_namelower(msg);
862 		anc_data[2] = msg_namelower(msg);
863 		break;
864 	case TIPC_MCAST_MSG:
865 		has_name = 1;
866 		anc_data[0] = msg_nametype(msg);
867 		anc_data[1] = msg_namelower(msg);
868 		anc_data[2] = msg_nameupper(msg);
869 		break;
870 	case TIPC_CONN_MSG:
871 		has_name = (tport->conn_type != 0);
872 		anc_data[0] = tport->conn_type;
873 		anc_data[1] = tport->conn_instance;
874 		anc_data[2] = tport->conn_instance;
875 		break;
876 	default:
877 		has_name = 0;
878 	}
879 	if (has_name) {
880 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
881 		if (res)
882 			return res;
883 	}
884 
885 	return 0;
886 }
887 
888 /**
889  * recv_msg - receive packet-oriented message
890  * @iocb: (unused)
891  * @m: descriptor for message info
892  * @buf_len: total size of user buffer area
893  * @flags: receive flags
894  *
895  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
896  * If the complete message doesn't fit in user area, truncate it.
897  *
898  * Returns size of returned message data, errno otherwise
899  */
900 static int recv_msg(struct kiocb *iocb, struct socket *sock,
901 		    struct msghdr *m, size_t buf_len, int flags)
902 {
903 	struct sock *sk = sock->sk;
904 	struct tipc_port *tport = tipc_sk_port(sk);
905 	struct sk_buff *buf;
906 	struct tipc_msg *msg;
907 	long timeout;
908 	unsigned int sz;
909 	u32 err;
910 	int res;
911 
912 	/* Catch invalid receive requests */
913 	if (unlikely(!buf_len))
914 		return -EINVAL;
915 
916 	lock_sock(sk);
917 
918 	if (unlikely(sock->state == SS_UNCONNECTED)) {
919 		res = -ENOTCONN;
920 		goto exit;
921 	}
922 
923 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
924 restart:
925 
926 	/* Look for a message in receive queue; wait if necessary */
927 	while (skb_queue_empty(&sk->sk_receive_queue)) {
928 		if (sock->state == SS_DISCONNECTING) {
929 			res = -ENOTCONN;
930 			goto exit;
931 		}
932 		if (timeout <= 0L) {
933 			res = timeout ? timeout : -EWOULDBLOCK;
934 			goto exit;
935 		}
936 		release_sock(sk);
937 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
938 							   tipc_rx_ready(sock),
939 							   timeout);
940 		lock_sock(sk);
941 	}
942 
943 	/* Look at first message in receive queue */
944 	buf = skb_peek(&sk->sk_receive_queue);
945 	msg = buf_msg(buf);
946 	sz = msg_data_sz(msg);
947 	err = msg_errcode(msg);
948 
949 	/* Discard an empty non-errored message & try again */
950 	if ((!sz) && (!err)) {
951 		advance_rx_queue(sk);
952 		goto restart;
953 	}
954 
955 	/* Capture sender's address (optional) */
956 	set_orig_addr(m, msg);
957 
958 	/* Capture ancillary data (optional) */
959 	res = anc_data_recv(m, msg, tport);
960 	if (res)
961 		goto exit;
962 
963 	/* Capture message data (if valid) & compute return value (always) */
964 	if (!err) {
965 		if (unlikely(buf_len < sz)) {
966 			sz = buf_len;
967 			m->msg_flags |= MSG_TRUNC;
968 		}
969 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
970 					      m->msg_iov, sz);
971 		if (res)
972 			goto exit;
973 		res = sz;
974 	} else {
975 		if ((sock->state == SS_READY) ||
976 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
977 			res = 0;
978 		else
979 			res = -ECONNRESET;
980 	}
981 
982 	/* Consume received message (optional) */
983 	if (likely(!(flags & MSG_PEEK))) {
984 		if ((sock->state != SS_READY) &&
985 		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
986 			tipc_acknowledge(tport->ref, tport->conn_unacked);
987 		advance_rx_queue(sk);
988 	}
989 exit:
990 	release_sock(sk);
991 	return res;
992 }
993 
994 /**
995  * recv_stream - receive stream-oriented data
996  * @iocb: (unused)
997  * @m: descriptor for message info
998  * @buf_len: total size of user buffer area
999  * @flags: receive flags
1000  *
1001  * Used for SOCK_STREAM messages only.  If not enough data is available
1002  * will optionally wait for more; never truncates data.
1003  *
1004  * Returns size of returned message data, errno otherwise
1005  */
1006 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1007 		       struct msghdr *m, size_t buf_len, int flags)
1008 {
1009 	struct sock *sk = sock->sk;
1010 	struct tipc_port *tport = tipc_sk_port(sk);
1011 	struct sk_buff *buf;
1012 	struct tipc_msg *msg;
1013 	long timeout;
1014 	unsigned int sz;
1015 	int sz_to_copy, target, needed;
1016 	int sz_copied = 0;
1017 	u32 err;
1018 	int res = 0;
1019 
1020 	/* Catch invalid receive attempts */
1021 	if (unlikely(!buf_len))
1022 		return -EINVAL;
1023 
1024 	lock_sock(sk);
1025 
1026 	if (unlikely((sock->state == SS_UNCONNECTED) ||
1027 		     (sock->state == SS_CONNECTING))) {
1028 		res = -ENOTCONN;
1029 		goto exit;
1030 	}
1031 
1032 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1033 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1034 
1035 restart:
1036 	/* Look for a message in receive queue; wait if necessary */
1037 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1038 		if (sock->state == SS_DISCONNECTING) {
1039 			res = -ENOTCONN;
1040 			goto exit;
1041 		}
1042 		if (timeout <= 0L) {
1043 			res = timeout ? timeout : -EWOULDBLOCK;
1044 			goto exit;
1045 		}
1046 		release_sock(sk);
1047 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1048 							   tipc_rx_ready(sock),
1049 							   timeout);
1050 		lock_sock(sk);
1051 	}
1052 
1053 	/* Look at first message in receive queue */
1054 	buf = skb_peek(&sk->sk_receive_queue);
1055 	msg = buf_msg(buf);
1056 	sz = msg_data_sz(msg);
1057 	err = msg_errcode(msg);
1058 
1059 	/* Discard an empty non-errored message & try again */
1060 	if ((!sz) && (!err)) {
1061 		advance_rx_queue(sk);
1062 		goto restart;
1063 	}
1064 
1065 	/* Optionally capture sender's address & ancillary data of first msg */
1066 	if (sz_copied == 0) {
1067 		set_orig_addr(m, msg);
1068 		res = anc_data_recv(m, msg, tport);
1069 		if (res)
1070 			goto exit;
1071 	}
1072 
1073 	/* Capture message data (if valid) & compute return value (always) */
1074 	if (!err) {
1075 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1076 
1077 		sz -= offset;
1078 		needed = (buf_len - sz_copied);
1079 		sz_to_copy = (sz <= needed) ? sz : needed;
1080 
1081 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1082 					      m->msg_iov, sz_to_copy);
1083 		if (res)
1084 			goto exit;
1085 
1086 		sz_copied += sz_to_copy;
1087 
1088 		if (sz_to_copy < sz) {
1089 			if (!(flags & MSG_PEEK))
1090 				TIPC_SKB_CB(buf)->handle =
1091 				(void *)(unsigned long)(offset + sz_to_copy);
1092 			goto exit;
1093 		}
1094 	} else {
1095 		if (sz_copied != 0)
1096 			goto exit; /* can't add error msg to valid data */
1097 
1098 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1099 			res = 0;
1100 		else
1101 			res = -ECONNRESET;
1102 	}
1103 
1104 	/* Consume received message (optional) */
1105 	if (likely(!(flags & MSG_PEEK))) {
1106 		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1107 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1108 		advance_rx_queue(sk);
1109 	}
1110 
1111 	/* Loop around if more data is required */
1112 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1113 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1114 	    (sz_copied < target)) &&	/* and more is ready or required */
1115 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1116 	    (!err))			/* and haven't reached a FIN */
1117 		goto restart;
1118 
1119 exit:
1120 	release_sock(sk);
1121 	return sz_copied ? sz_copied : res;
1122 }
1123 
1124 /**
1125  * tipc_write_space - wake up thread if port congestion is released
1126  * @sk: socket
1127  */
1128 static void tipc_write_space(struct sock *sk)
1129 {
1130 	struct socket_wq *wq;
1131 
1132 	rcu_read_lock();
1133 	wq = rcu_dereference(sk->sk_wq);
1134 	if (wq_has_sleeper(wq))
1135 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1136 						POLLWRNORM | POLLWRBAND);
1137 	rcu_read_unlock();
1138 }
1139 
1140 /**
1141  * tipc_data_ready - wake up threads to indicate messages have been received
1142  * @sk: socket
1143  * @len: the length of messages
1144  */
1145 static void tipc_data_ready(struct sock *sk, int len)
1146 {
1147 	struct socket_wq *wq;
1148 
1149 	rcu_read_lock();
1150 	wq = rcu_dereference(sk->sk_wq);
1151 	if (wq_has_sleeper(wq))
1152 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1153 						POLLRDNORM | POLLRDBAND);
1154 	rcu_read_unlock();
1155 }
1156 
1157 /**
1158  * rx_queue_full - determine if receive queue can accept another message
1159  * @msg: message to be added to queue
1160  * @queue_size: current size of queue
1161  * @base: nominal maximum size of queue
1162  *
1163  * Returns 1 if queue is unable to accept message, 0 otherwise
1164  */
1165 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1166 {
1167 	u32 threshold;
1168 	u32 imp = msg_importance(msg);
1169 
1170 	if (imp == TIPC_LOW_IMPORTANCE)
1171 		threshold = base;
1172 	else if (imp == TIPC_MEDIUM_IMPORTANCE)
1173 		threshold = base * 2;
1174 	else if (imp == TIPC_HIGH_IMPORTANCE)
1175 		threshold = base * 100;
1176 	else
1177 		return 0;
1178 
1179 	if (msg_connected(msg))
1180 		threshold *= 4;
1181 
1182 	return queue_size >= threshold;
1183 }
1184 
1185 /**
1186  * filter_connect - Handle all incoming messages for a connection-based socket
1187  * @tsock: TIPC socket
1188  * @msg: message
1189  *
1190  * Returns TIPC error status code and socket error status code
1191  * once it encounters some errors
1192  */
1193 static u32 filter_connect(struct tipc_sock *tsock, struct sk_buff **buf)
1194 {
1195 	struct socket *sock = tsock->sk.sk_socket;
1196 	struct tipc_msg *msg = buf_msg(*buf);
1197 	struct sock *sk = &tsock->sk;
1198 	u32 retval = TIPC_ERR_NO_PORT;
1199 	int res;
1200 
1201 	if (msg_mcast(msg))
1202 		return retval;
1203 
1204 	switch ((int)sock->state) {
1205 	case SS_CONNECTED:
1206 		/* Accept only connection-based messages sent by peer */
1207 		if (msg_connected(msg) && tipc_port_peer_msg(tsock->p, msg)) {
1208 			if (unlikely(msg_errcode(msg))) {
1209 				sock->state = SS_DISCONNECTING;
1210 				__tipc_disconnect(tsock->p);
1211 			}
1212 			retval = TIPC_OK;
1213 		}
1214 		break;
1215 	case SS_CONNECTING:
1216 		/* Accept only ACK or NACK message */
1217 		if (unlikely(msg_errcode(msg))) {
1218 			sock->state = SS_DISCONNECTING;
1219 			sk->sk_err = -ECONNREFUSED;
1220 			retval = TIPC_OK;
1221 			break;
1222 		}
1223 
1224 		if (unlikely(!msg_connected(msg)))
1225 			break;
1226 
1227 		res = auto_connect(sock, msg);
1228 		if (res) {
1229 			sock->state = SS_DISCONNECTING;
1230 			sk->sk_err = res;
1231 			retval = TIPC_OK;
1232 			break;
1233 		}
1234 
1235 		/* If an incoming message is an 'ACK-', it should be
1236 		 * discarded here because it doesn't contain useful
1237 		 * data. In addition, we should try to wake up
1238 		 * connect() routine if sleeping.
1239 		 */
1240 		if (msg_data_sz(msg) == 0) {
1241 			kfree_skb(*buf);
1242 			*buf = NULL;
1243 			if (waitqueue_active(sk_sleep(sk)))
1244 				wake_up_interruptible(sk_sleep(sk));
1245 		}
1246 		retval = TIPC_OK;
1247 		break;
1248 	case SS_LISTENING:
1249 	case SS_UNCONNECTED:
1250 		/* Accept only SYN message */
1251 		if (!msg_connected(msg) && !(msg_errcode(msg)))
1252 			retval = TIPC_OK;
1253 		break;
1254 	case SS_DISCONNECTING:
1255 		break;
1256 	default:
1257 		pr_err("Unknown socket state %u\n", sock->state);
1258 	}
1259 	return retval;
1260 }
1261 
1262 /**
1263  * filter_rcv - validate incoming message
1264  * @sk: socket
1265  * @buf: message
1266  *
1267  * Enqueues message on receive queue if acceptable; optionally handles
1268  * disconnect indication for a connected socket.
1269  *
1270  * Called with socket lock already taken; port lock may also be taken.
1271  *
1272  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1273  */
1274 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1275 {
1276 	struct socket *sock = sk->sk_socket;
1277 	struct tipc_msg *msg = buf_msg(buf);
1278 	u32 recv_q_len;
1279 	u32 res = TIPC_OK;
1280 
1281 	/* Reject message if it is wrong sort of message for socket */
1282 	if (msg_type(msg) > TIPC_DIRECT_MSG)
1283 		return TIPC_ERR_NO_PORT;
1284 
1285 	if (sock->state == SS_READY) {
1286 		if (msg_connected(msg))
1287 			return TIPC_ERR_NO_PORT;
1288 	} else {
1289 		res = filter_connect(tipc_sk(sk), &buf);
1290 		if (res != TIPC_OK || buf == NULL)
1291 			return res;
1292 	}
1293 
1294 	/* Reject message if there isn't room to queue it */
1295 	recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1296 	if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1297 		if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1298 			return TIPC_ERR_OVERLOAD;
1299 	}
1300 
1301 	/* Enqueue message (finally!) */
1302 	TIPC_SKB_CB(buf)->handle = 0;
1303 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1304 
1305 	sk->sk_data_ready(sk, 0);
1306 	return TIPC_OK;
1307 }
1308 
1309 /**
1310  * backlog_rcv - handle incoming message from backlog queue
1311  * @sk: socket
1312  * @buf: message
1313  *
1314  * Caller must hold socket lock, but not port lock.
1315  *
1316  * Returns 0
1317  */
1318 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1319 {
1320 	u32 res;
1321 
1322 	res = filter_rcv(sk, buf);
1323 	if (res)
1324 		tipc_reject_msg(buf, res);
1325 	return 0;
1326 }
1327 
1328 /**
1329  * dispatch - handle incoming message
1330  * @tport: TIPC port that received message
1331  * @buf: message
1332  *
1333  * Called with port lock already taken.
1334  *
1335  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1336  */
1337 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1338 {
1339 	struct sock *sk = (struct sock *)tport->usr_handle;
1340 	u32 res;
1341 
1342 	/*
1343 	 * Process message if socket is unlocked; otherwise add to backlog queue
1344 	 *
1345 	 * This code is based on sk_receive_skb(), but must be distinct from it
1346 	 * since a TIPC-specific filter/reject mechanism is utilized
1347 	 */
1348 	bh_lock_sock(sk);
1349 	if (!sock_owned_by_user(sk)) {
1350 		res = filter_rcv(sk, buf);
1351 	} else {
1352 		if (sk_add_backlog(sk, buf, sk->sk_rcvbuf))
1353 			res = TIPC_ERR_OVERLOAD;
1354 		else
1355 			res = TIPC_OK;
1356 	}
1357 	bh_unlock_sock(sk);
1358 
1359 	return res;
1360 }
1361 
1362 /**
1363  * wakeupdispatch - wake up port after congestion
1364  * @tport: port to wakeup
1365  *
1366  * Called with port lock already taken.
1367  */
1368 static void wakeupdispatch(struct tipc_port *tport)
1369 {
1370 	struct sock *sk = (struct sock *)tport->usr_handle;
1371 
1372 	sk->sk_write_space(sk);
1373 }
1374 
1375 /**
1376  * connect - establish a connection to another TIPC port
1377  * @sock: socket structure
1378  * @dest: socket address for destination port
1379  * @destlen: size of socket address data structure
1380  * @flags: file-related flags associated with socket
1381  *
1382  * Returns 0 on success, errno otherwise
1383  */
1384 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1385 		   int flags)
1386 {
1387 	struct sock *sk = sock->sk;
1388 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1389 	struct msghdr m = {NULL,};
1390 	unsigned int timeout;
1391 	int res;
1392 
1393 	lock_sock(sk);
1394 
1395 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1396 	if (sock->state == SS_READY) {
1397 		res = -EOPNOTSUPP;
1398 		goto exit;
1399 	}
1400 
1401 	/*
1402 	 * Reject connection attempt using multicast address
1403 	 *
1404 	 * Note: send_msg() validates the rest of the address fields,
1405 	 *       so there's no need to do it here
1406 	 */
1407 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1408 		res = -EINVAL;
1409 		goto exit;
1410 	}
1411 
1412 	timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1413 
1414 	switch (sock->state) {
1415 	case SS_UNCONNECTED:
1416 		/* Send a 'SYN-' to destination */
1417 		m.msg_name = dest;
1418 		m.msg_namelen = destlen;
1419 
1420 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1421 		 * indicate send_msg() is never blocked.
1422 		 */
1423 		if (!timeout)
1424 			m.msg_flags = MSG_DONTWAIT;
1425 
1426 		res = send_msg(NULL, sock, &m, 0);
1427 		if ((res < 0) && (res != -EWOULDBLOCK))
1428 			goto exit;
1429 
1430 		/* Just entered SS_CONNECTING state; the only
1431 		 * difference is that return value in non-blocking
1432 		 * case is EINPROGRESS, rather than EALREADY.
1433 		 */
1434 		res = -EINPROGRESS;
1435 		break;
1436 	case SS_CONNECTING:
1437 		res = -EALREADY;
1438 		break;
1439 	case SS_CONNECTED:
1440 		res = -EISCONN;
1441 		break;
1442 	default:
1443 		res = -EINVAL;
1444 		goto exit;
1445 	}
1446 
1447 	if (sock->state == SS_CONNECTING) {
1448 		if (!timeout)
1449 			goto exit;
1450 
1451 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1452 		release_sock(sk);
1453 		res = wait_event_interruptible_timeout(*sk_sleep(sk),
1454 				sock->state != SS_CONNECTING,
1455 				timeout ? (long)msecs_to_jiffies(timeout)
1456 					: MAX_SCHEDULE_TIMEOUT);
1457 		lock_sock(sk);
1458 		if (res <= 0) {
1459 			if (res == 0)
1460 				res = -ETIMEDOUT;
1461 			else
1462 				; /* leave "res" unchanged */
1463 			goto exit;
1464 		}
1465 	}
1466 
1467 	if (unlikely(sock->state == SS_DISCONNECTING))
1468 		res = sock_error(sk);
1469 	else
1470 		res = 0;
1471 
1472 exit:
1473 	release_sock(sk);
1474 	return res;
1475 }
1476 
1477 /**
1478  * listen - allow socket to listen for incoming connections
1479  * @sock: socket structure
1480  * @len: (unused)
1481  *
1482  * Returns 0 on success, errno otherwise
1483  */
1484 static int listen(struct socket *sock, int len)
1485 {
1486 	struct sock *sk = sock->sk;
1487 	int res;
1488 
1489 	lock_sock(sk);
1490 
1491 	if (sock->state != SS_UNCONNECTED)
1492 		res = -EINVAL;
1493 	else {
1494 		sock->state = SS_LISTENING;
1495 		res = 0;
1496 	}
1497 
1498 	release_sock(sk);
1499 	return res;
1500 }
1501 
1502 /**
1503  * accept - wait for connection request
1504  * @sock: listening socket
1505  * @newsock: new socket that is to be connected
1506  * @flags: file-related flags associated with socket
1507  *
1508  * Returns 0 on success, errno otherwise
1509  */
1510 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1511 {
1512 	struct sock *new_sk, *sk = sock->sk;
1513 	struct sk_buff *buf;
1514 	struct tipc_sock *new_tsock;
1515 	struct tipc_port *new_tport;
1516 	struct tipc_msg *msg;
1517 	u32 new_ref;
1518 
1519 	int res;
1520 
1521 	lock_sock(sk);
1522 
1523 	if (sock->state != SS_LISTENING) {
1524 		res = -EINVAL;
1525 		goto exit;
1526 	}
1527 
1528 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1529 		if (flags & O_NONBLOCK) {
1530 			res = -EWOULDBLOCK;
1531 			goto exit;
1532 		}
1533 		release_sock(sk);
1534 		res = wait_event_interruptible(*sk_sleep(sk),
1535 				(!skb_queue_empty(&sk->sk_receive_queue)));
1536 		lock_sock(sk);
1537 		if (res)
1538 			goto exit;
1539 	}
1540 
1541 	buf = skb_peek(&sk->sk_receive_queue);
1542 
1543 	res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1544 	if (res)
1545 		goto exit;
1546 
1547 	new_sk = new_sock->sk;
1548 	new_tsock = tipc_sk(new_sk);
1549 	new_tport = new_tsock->p;
1550 	new_ref = new_tport->ref;
1551 	msg = buf_msg(buf);
1552 
1553 	/* we lock on new_sk; but lockdep sees the lock on sk */
1554 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1555 
1556 	/*
1557 	 * Reject any stray messages received by new socket
1558 	 * before the socket lock was taken (very, very unlikely)
1559 	 */
1560 	reject_rx_queue(new_sk);
1561 
1562 	/* Connect new socket to it's peer */
1563 	new_tsock->peer_name.ref = msg_origport(msg);
1564 	new_tsock->peer_name.node = msg_orignode(msg);
1565 	tipc_connect(new_ref, &new_tsock->peer_name);
1566 	new_sock->state = SS_CONNECTED;
1567 
1568 	tipc_set_portimportance(new_ref, msg_importance(msg));
1569 	if (msg_named(msg)) {
1570 		new_tport->conn_type = msg_nametype(msg);
1571 		new_tport->conn_instance = msg_nameinst(msg);
1572 	}
1573 
1574 	/*
1575 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1576 	 * Respond to 'SYN+' by queuing it on new socket.
1577 	 */
1578 	if (!msg_data_sz(msg)) {
1579 		struct msghdr m = {NULL,};
1580 
1581 		advance_rx_queue(sk);
1582 		send_packet(NULL, new_sock, &m, 0);
1583 	} else {
1584 		__skb_dequeue(&sk->sk_receive_queue);
1585 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
1586 	}
1587 	release_sock(new_sk);
1588 
1589 exit:
1590 	release_sock(sk);
1591 	return res;
1592 }
1593 
1594 /**
1595  * shutdown - shutdown socket connection
1596  * @sock: socket structure
1597  * @how: direction to close (must be SHUT_RDWR)
1598  *
1599  * Terminates connection (if necessary), then purges socket's receive queue.
1600  *
1601  * Returns 0 on success, errno otherwise
1602  */
1603 static int shutdown(struct socket *sock, int how)
1604 {
1605 	struct sock *sk = sock->sk;
1606 	struct tipc_port *tport = tipc_sk_port(sk);
1607 	struct sk_buff *buf;
1608 	int res;
1609 
1610 	if (how != SHUT_RDWR)
1611 		return -EINVAL;
1612 
1613 	lock_sock(sk);
1614 
1615 	switch (sock->state) {
1616 	case SS_CONNECTING:
1617 	case SS_CONNECTED:
1618 
1619 restart:
1620 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1621 		buf = __skb_dequeue(&sk->sk_receive_queue);
1622 		if (buf) {
1623 			if (TIPC_SKB_CB(buf)->handle != 0) {
1624 				kfree_skb(buf);
1625 				goto restart;
1626 			}
1627 			tipc_disconnect(tport->ref);
1628 			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1629 		} else {
1630 			tipc_shutdown(tport->ref);
1631 		}
1632 
1633 		sock->state = SS_DISCONNECTING;
1634 
1635 		/* fall through */
1636 
1637 	case SS_DISCONNECTING:
1638 
1639 		/* Discard any unreceived messages */
1640 		discard_rx_queue(sk);
1641 
1642 		/* Wake up anyone sleeping in poll */
1643 		sk->sk_state_change(sk);
1644 		res = 0;
1645 		break;
1646 
1647 	default:
1648 		res = -ENOTCONN;
1649 	}
1650 
1651 	release_sock(sk);
1652 	return res;
1653 }
1654 
1655 /**
1656  * setsockopt - set socket option
1657  * @sock: socket structure
1658  * @lvl: option level
1659  * @opt: option identifier
1660  * @ov: pointer to new option value
1661  * @ol: length of option value
1662  *
1663  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1664  * (to ease compatibility).
1665  *
1666  * Returns 0 on success, errno otherwise
1667  */
1668 static int setsockopt(struct socket *sock,
1669 		      int lvl, int opt, char __user *ov, unsigned int ol)
1670 {
1671 	struct sock *sk = sock->sk;
1672 	struct tipc_port *tport = tipc_sk_port(sk);
1673 	u32 value;
1674 	int res;
1675 
1676 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1677 		return 0;
1678 	if (lvl != SOL_TIPC)
1679 		return -ENOPROTOOPT;
1680 	if (ol < sizeof(value))
1681 		return -EINVAL;
1682 	res = get_user(value, (u32 __user *)ov);
1683 	if (res)
1684 		return res;
1685 
1686 	lock_sock(sk);
1687 
1688 	switch (opt) {
1689 	case TIPC_IMPORTANCE:
1690 		res = tipc_set_portimportance(tport->ref, value);
1691 		break;
1692 	case TIPC_SRC_DROPPABLE:
1693 		if (sock->type != SOCK_STREAM)
1694 			res = tipc_set_portunreliable(tport->ref, value);
1695 		else
1696 			res = -ENOPROTOOPT;
1697 		break;
1698 	case TIPC_DEST_DROPPABLE:
1699 		res = tipc_set_portunreturnable(tport->ref, value);
1700 		break;
1701 	case TIPC_CONN_TIMEOUT:
1702 		tipc_sk(sk)->conn_timeout = value;
1703 		/* no need to set "res", since already 0 at this point */
1704 		break;
1705 	default:
1706 		res = -EINVAL;
1707 	}
1708 
1709 	release_sock(sk);
1710 
1711 	return res;
1712 }
1713 
1714 /**
1715  * getsockopt - get socket option
1716  * @sock: socket structure
1717  * @lvl: option level
1718  * @opt: option identifier
1719  * @ov: receptacle for option value
1720  * @ol: receptacle for length of option value
1721  *
1722  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1723  * (to ease compatibility).
1724  *
1725  * Returns 0 on success, errno otherwise
1726  */
1727 static int getsockopt(struct socket *sock,
1728 		      int lvl, int opt, char __user *ov, int __user *ol)
1729 {
1730 	struct sock *sk = sock->sk;
1731 	struct tipc_port *tport = tipc_sk_port(sk);
1732 	int len;
1733 	u32 value;
1734 	int res;
1735 
1736 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1737 		return put_user(0, ol);
1738 	if (lvl != SOL_TIPC)
1739 		return -ENOPROTOOPT;
1740 	res = get_user(len, ol);
1741 	if (res)
1742 		return res;
1743 
1744 	lock_sock(sk);
1745 
1746 	switch (opt) {
1747 	case TIPC_IMPORTANCE:
1748 		res = tipc_portimportance(tport->ref, &value);
1749 		break;
1750 	case TIPC_SRC_DROPPABLE:
1751 		res = tipc_portunreliable(tport->ref, &value);
1752 		break;
1753 	case TIPC_DEST_DROPPABLE:
1754 		res = tipc_portunreturnable(tport->ref, &value);
1755 		break;
1756 	case TIPC_CONN_TIMEOUT:
1757 		value = tipc_sk(sk)->conn_timeout;
1758 		/* no need to set "res", since already 0 at this point */
1759 		break;
1760 	case TIPC_NODE_RECVQ_DEPTH:
1761 		value = 0; /* was tipc_queue_size, now obsolete */
1762 		break;
1763 	case TIPC_SOCK_RECVQ_DEPTH:
1764 		value = skb_queue_len(&sk->sk_receive_queue);
1765 		break;
1766 	default:
1767 		res = -EINVAL;
1768 	}
1769 
1770 	release_sock(sk);
1771 
1772 	if (res)
1773 		return res;	/* "get" failed */
1774 
1775 	if (len < sizeof(value))
1776 		return -EINVAL;
1777 
1778 	if (copy_to_user(ov, &value, sizeof(value)))
1779 		return -EFAULT;
1780 
1781 	return put_user(sizeof(value), ol);
1782 }
1783 
1784 /* Protocol switches for the various types of TIPC sockets */
1785 
1786 static const struct proto_ops msg_ops = {
1787 	.owner		= THIS_MODULE,
1788 	.family		= AF_TIPC,
1789 	.release	= release,
1790 	.bind		= bind,
1791 	.connect	= connect,
1792 	.socketpair	= sock_no_socketpair,
1793 	.accept		= sock_no_accept,
1794 	.getname	= get_name,
1795 	.poll		= poll,
1796 	.ioctl		= sock_no_ioctl,
1797 	.listen		= sock_no_listen,
1798 	.shutdown	= shutdown,
1799 	.setsockopt	= setsockopt,
1800 	.getsockopt	= getsockopt,
1801 	.sendmsg	= send_msg,
1802 	.recvmsg	= recv_msg,
1803 	.mmap		= sock_no_mmap,
1804 	.sendpage	= sock_no_sendpage
1805 };
1806 
1807 static const struct proto_ops packet_ops = {
1808 	.owner		= THIS_MODULE,
1809 	.family		= AF_TIPC,
1810 	.release	= release,
1811 	.bind		= bind,
1812 	.connect	= connect,
1813 	.socketpair	= sock_no_socketpair,
1814 	.accept		= accept,
1815 	.getname	= get_name,
1816 	.poll		= poll,
1817 	.ioctl		= sock_no_ioctl,
1818 	.listen		= listen,
1819 	.shutdown	= shutdown,
1820 	.setsockopt	= setsockopt,
1821 	.getsockopt	= getsockopt,
1822 	.sendmsg	= send_packet,
1823 	.recvmsg	= recv_msg,
1824 	.mmap		= sock_no_mmap,
1825 	.sendpage	= sock_no_sendpage
1826 };
1827 
1828 static const struct proto_ops stream_ops = {
1829 	.owner		= THIS_MODULE,
1830 	.family		= AF_TIPC,
1831 	.release	= release,
1832 	.bind		= bind,
1833 	.connect	= connect,
1834 	.socketpair	= sock_no_socketpair,
1835 	.accept		= accept,
1836 	.getname	= get_name,
1837 	.poll		= poll,
1838 	.ioctl		= sock_no_ioctl,
1839 	.listen		= listen,
1840 	.shutdown	= shutdown,
1841 	.setsockopt	= setsockopt,
1842 	.getsockopt	= getsockopt,
1843 	.sendmsg	= send_stream,
1844 	.recvmsg	= recv_stream,
1845 	.mmap		= sock_no_mmap,
1846 	.sendpage	= sock_no_sendpage
1847 };
1848 
1849 static const struct net_proto_family tipc_family_ops = {
1850 	.owner		= THIS_MODULE,
1851 	.family		= AF_TIPC,
1852 	.create		= tipc_create
1853 };
1854 
1855 static struct proto tipc_proto = {
1856 	.name		= "TIPC",
1857 	.owner		= THIS_MODULE,
1858 	.obj_size	= sizeof(struct tipc_sock)
1859 };
1860 
1861 /**
1862  * tipc_socket_init - initialize TIPC socket interface
1863  *
1864  * Returns 0 on success, errno otherwise
1865  */
1866 int tipc_socket_init(void)
1867 {
1868 	int res;
1869 
1870 	res = proto_register(&tipc_proto, 1);
1871 	if (res) {
1872 		pr_err("Failed to register TIPC protocol type\n");
1873 		goto out;
1874 	}
1875 
1876 	res = sock_register(&tipc_family_ops);
1877 	if (res) {
1878 		pr_err("Failed to register TIPC socket type\n");
1879 		proto_unregister(&tipc_proto);
1880 		goto out;
1881 	}
1882 
1883 	sockets_enabled = 1;
1884  out:
1885 	return res;
1886 }
1887 
1888 /**
1889  * tipc_socket_stop - stop TIPC socket interface
1890  */
1891 void tipc_socket_stop(void)
1892 {
1893 	if (!sockets_enabled)
1894 		return;
1895 
1896 	sockets_enabled = 0;
1897 	sock_unregister(tipc_family_ops.family);
1898 	proto_unregister(&tipc_proto);
1899 }
1900