xref: /openbmc/linux/net/tipc/socket.c (revision 455f9726)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2014, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, 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 #include "name_table.h"
40 #include "node.h"
41 #include "link.h"
42 #include <linux/export.h>
43 
44 #define SS_LISTENING	-1	/* socket is listening */
45 #define SS_READY	-2	/* socket is connectionless */
46 
47 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
48 #define TIPC_FWD_MSG	        1
49 
50 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
51 static void tipc_data_ready(struct sock *sk);
52 static void tipc_write_space(struct sock *sk);
53 static int tipc_release(struct socket *sock);
54 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
55 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
56 
57 static const struct proto_ops packet_ops;
58 static const struct proto_ops stream_ops;
59 static const struct proto_ops msg_ops;
60 
61 static struct proto tipc_proto;
62 static struct proto tipc_proto_kern;
63 
64 /*
65  * Revised TIPC socket locking policy:
66  *
67  * Most socket operations take the standard socket lock when they start
68  * and hold it until they finish (or until they need to sleep).  Acquiring
69  * this lock grants the owner exclusive access to the fields of the socket
70  * data structures, with the exception of the backlog queue.  A few socket
71  * operations can be done without taking the socket lock because they only
72  * read socket information that never changes during the life of the socket.
73  *
74  * Socket operations may acquire the lock for the associated TIPC port if they
75  * need to perform an operation on the port.  If any routine needs to acquire
76  * both the socket lock and the port lock it must take the socket lock first
77  * to avoid the risk of deadlock.
78  *
79  * The dispatcher handling incoming messages cannot grab the socket lock in
80  * the standard fashion, since invoked it runs at the BH level and cannot block.
81  * Instead, it checks to see if the socket lock is currently owned by someone,
82  * and either handles the message itself or adds it to the socket's backlog
83  * queue; in the latter case the queued message is processed once the process
84  * owning the socket lock releases it.
85  *
86  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
87  * the problem of a blocked socket operation preventing any other operations
88  * from occurring.  However, applications must be careful if they have
89  * multiple threads trying to send (or receive) on the same socket, as these
90  * operations might interfere with each other.  For example, doing a connect
91  * and a receive at the same time might allow the receive to consume the
92  * ACK message meant for the connect.  While additional work could be done
93  * to try and overcome this, it doesn't seem to be worthwhile at the present.
94  *
95  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
96  * that another operation that must be performed in a non-blocking manner is
97  * not delayed for very long because the lock has already been taken.
98  *
99  * NOTE: This code assumes that certain fields of a port/socket pair are
100  * constant over its lifetime; such fields can be examined without taking
101  * the socket lock and/or port lock, and do not need to be re-read even
102  * after resuming processing after waiting.  These fields include:
103  *   - socket type
104  *   - pointer to socket sk structure (aka tipc_sock structure)
105  *   - pointer to port structure
106  *   - port reference
107  */
108 
109 #include "socket.h"
110 
111 /**
112  * advance_rx_queue - discard first buffer in socket receive queue
113  *
114  * Caller must hold socket lock
115  */
116 static void advance_rx_queue(struct sock *sk)
117 {
118 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
119 }
120 
121 /**
122  * reject_rx_queue - reject all buffers in socket receive queue
123  *
124  * Caller must hold socket lock
125  */
126 static void reject_rx_queue(struct sock *sk)
127 {
128 	struct sk_buff *buf;
129 	u32 dnode;
130 
131 	while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
132 		if (tipc_msg_reverse(buf, &dnode, TIPC_ERR_NO_PORT))
133 			tipc_link_xmit(buf, dnode, 0);
134 	}
135 }
136 
137 /**
138  * tipc_sk_create - create a TIPC socket
139  * @net: network namespace (must be default network)
140  * @sock: pre-allocated socket structure
141  * @protocol: protocol indicator (must be 0)
142  * @kern: caused by kernel or by userspace?
143  *
144  * This routine creates additional data structures used by the TIPC socket,
145  * initializes them, and links them together.
146  *
147  * Returns 0 on success, errno otherwise
148  */
149 static int tipc_sk_create(struct net *net, struct socket *sock,
150 			  int protocol, int kern)
151 {
152 	const struct proto_ops *ops;
153 	socket_state state;
154 	struct sock *sk;
155 	struct tipc_sock *tsk;
156 	struct tipc_port *port;
157 	u32 ref;
158 
159 	/* Validate arguments */
160 	if (unlikely(protocol != 0))
161 		return -EPROTONOSUPPORT;
162 
163 	switch (sock->type) {
164 	case SOCK_STREAM:
165 		ops = &stream_ops;
166 		state = SS_UNCONNECTED;
167 		break;
168 	case SOCK_SEQPACKET:
169 		ops = &packet_ops;
170 		state = SS_UNCONNECTED;
171 		break;
172 	case SOCK_DGRAM:
173 	case SOCK_RDM:
174 		ops = &msg_ops;
175 		state = SS_READY;
176 		break;
177 	default:
178 		return -EPROTOTYPE;
179 	}
180 
181 	/* Allocate socket's protocol area */
182 	if (!kern)
183 		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
184 	else
185 		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
186 
187 	if (sk == NULL)
188 		return -ENOMEM;
189 
190 	tsk = tipc_sk(sk);
191 	port = &tsk->port;
192 
193 	ref = tipc_port_init(port, TIPC_LOW_IMPORTANCE);
194 	if (!ref) {
195 		pr_warn("Socket registration failed, ref. table exhausted\n");
196 		sk_free(sk);
197 		return -ENOMEM;
198 	}
199 
200 	/* Finish initializing socket data structures */
201 	sock->ops = ops;
202 	sock->state = state;
203 
204 	sock_init_data(sock, sk);
205 	sk->sk_backlog_rcv = tipc_backlog_rcv;
206 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
207 	sk->sk_data_ready = tipc_data_ready;
208 	sk->sk_write_space = tipc_write_space;
209 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
210 	tsk->sent_unacked = 0;
211 	atomic_set(&tsk->dupl_rcvcnt, 0);
212 	tipc_port_unlock(port);
213 
214 	if (sock->state == SS_READY) {
215 		tipc_port_set_unreturnable(port, true);
216 		if (sock->type == SOCK_DGRAM)
217 			tipc_port_set_unreliable(port, true);
218 	}
219 	return 0;
220 }
221 
222 /**
223  * tipc_sock_create_local - create TIPC socket from inside TIPC module
224  * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
225  *
226  * We cannot use sock_creat_kern here because it bumps module user count.
227  * Since socket owner and creator is the same module we must make sure
228  * that module count remains zero for module local sockets, otherwise
229  * we cannot do rmmod.
230  *
231  * Returns 0 on success, errno otherwise
232  */
233 int tipc_sock_create_local(int type, struct socket **res)
234 {
235 	int rc;
236 
237 	rc = sock_create_lite(AF_TIPC, type, 0, res);
238 	if (rc < 0) {
239 		pr_err("Failed to create kernel socket\n");
240 		return rc;
241 	}
242 	tipc_sk_create(&init_net, *res, 0, 1);
243 
244 	return 0;
245 }
246 
247 /**
248  * tipc_sock_release_local - release socket created by tipc_sock_create_local
249  * @sock: the socket to be released.
250  *
251  * Module reference count is not incremented when such sockets are created,
252  * so we must keep it from being decremented when they are released.
253  */
254 void tipc_sock_release_local(struct socket *sock)
255 {
256 	tipc_release(sock);
257 	sock->ops = NULL;
258 	sock_release(sock);
259 }
260 
261 /**
262  * tipc_sock_accept_local - accept a connection on a socket created
263  * with tipc_sock_create_local. Use this function to avoid that
264  * module reference count is inadvertently incremented.
265  *
266  * @sock:    the accepting socket
267  * @newsock: reference to the new socket to be created
268  * @flags:   socket flags
269  */
270 
271 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
272 			   int flags)
273 {
274 	struct sock *sk = sock->sk;
275 	int ret;
276 
277 	ret = sock_create_lite(sk->sk_family, sk->sk_type,
278 			       sk->sk_protocol, newsock);
279 	if (ret < 0)
280 		return ret;
281 
282 	ret = tipc_accept(sock, *newsock, flags);
283 	if (ret < 0) {
284 		sock_release(*newsock);
285 		return ret;
286 	}
287 	(*newsock)->ops = sock->ops;
288 	return ret;
289 }
290 
291 /**
292  * tipc_release - destroy a TIPC socket
293  * @sock: socket to destroy
294  *
295  * This routine cleans up any messages that are still queued on the socket.
296  * For DGRAM and RDM socket types, all queued messages are rejected.
297  * For SEQPACKET and STREAM socket types, the first message is rejected
298  * and any others are discarded.  (If the first message on a STREAM socket
299  * is partially-read, it is discarded and the next one is rejected instead.)
300  *
301  * NOTE: Rejected messages are not necessarily returned to the sender!  They
302  * are returned or discarded according to the "destination droppable" setting
303  * specified for the message by the sender.
304  *
305  * Returns 0 on success, errno otherwise
306  */
307 static int tipc_release(struct socket *sock)
308 {
309 	struct sock *sk = sock->sk;
310 	struct tipc_sock *tsk;
311 	struct tipc_port *port;
312 	struct sk_buff *buf;
313 	u32 dnode;
314 
315 	/*
316 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
317 	 * releases a pre-allocated child socket that was never used)
318 	 */
319 	if (sk == NULL)
320 		return 0;
321 
322 	tsk = tipc_sk(sk);
323 	port = &tsk->port;
324 	lock_sock(sk);
325 
326 	/*
327 	 * Reject all unreceived messages, except on an active connection
328 	 * (which disconnects locally & sends a 'FIN+' to peer)
329 	 */
330 	while (sock->state != SS_DISCONNECTING) {
331 		buf = __skb_dequeue(&sk->sk_receive_queue);
332 		if (buf == NULL)
333 			break;
334 		if (TIPC_SKB_CB(buf)->handle != NULL)
335 			kfree_skb(buf);
336 		else {
337 			if ((sock->state == SS_CONNECTING) ||
338 			    (sock->state == SS_CONNECTED)) {
339 				sock->state = SS_DISCONNECTING;
340 				tipc_port_disconnect(port->ref);
341 			}
342 			if (tipc_msg_reverse(buf, &dnode, TIPC_ERR_NO_PORT))
343 				tipc_link_xmit(buf, dnode, 0);
344 		}
345 	}
346 
347 	/* Destroy TIPC port; also disconnects an active connection and
348 	 * sends a 'FIN-' to peer.
349 	 */
350 	tipc_port_destroy(port);
351 
352 	/* Discard any remaining (connection-based) messages in receive queue */
353 	__skb_queue_purge(&sk->sk_receive_queue);
354 
355 	/* Reject any messages that accumulated in backlog queue */
356 	sock->state = SS_DISCONNECTING;
357 	release_sock(sk);
358 
359 	sock_put(sk);
360 	sock->sk = NULL;
361 
362 	return 0;
363 }
364 
365 /**
366  * tipc_bind - associate or disassocate TIPC name(s) with a socket
367  * @sock: socket structure
368  * @uaddr: socket address describing name(s) and desired operation
369  * @uaddr_len: size of socket address data structure
370  *
371  * Name and name sequence binding is indicated using a positive scope value;
372  * a negative scope value unbinds the specified name.  Specifying no name
373  * (i.e. a socket address length of 0) unbinds all names from the socket.
374  *
375  * Returns 0 on success, errno otherwise
376  *
377  * NOTE: This routine doesn't need to take the socket lock since it doesn't
378  *       access any non-constant socket information.
379  */
380 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
381 		     int uaddr_len)
382 {
383 	struct sock *sk = sock->sk;
384 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
385 	struct tipc_sock *tsk = tipc_sk(sk);
386 	int res = -EINVAL;
387 
388 	lock_sock(sk);
389 	if (unlikely(!uaddr_len)) {
390 		res = tipc_withdraw(&tsk->port, 0, NULL);
391 		goto exit;
392 	}
393 
394 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
395 		res = -EINVAL;
396 		goto exit;
397 	}
398 	if (addr->family != AF_TIPC) {
399 		res = -EAFNOSUPPORT;
400 		goto exit;
401 	}
402 
403 	if (addr->addrtype == TIPC_ADDR_NAME)
404 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
405 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
406 		res = -EAFNOSUPPORT;
407 		goto exit;
408 	}
409 
410 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
411 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
412 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
413 		res = -EACCES;
414 		goto exit;
415 	}
416 
417 	res = (addr->scope > 0) ?
418 		tipc_publish(&tsk->port, addr->scope, &addr->addr.nameseq) :
419 		tipc_withdraw(&tsk->port, -addr->scope, &addr->addr.nameseq);
420 exit:
421 	release_sock(sk);
422 	return res;
423 }
424 
425 /**
426  * tipc_getname - get port ID of socket or peer socket
427  * @sock: socket structure
428  * @uaddr: area for returned socket address
429  * @uaddr_len: area for returned length of socket address
430  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
431  *
432  * Returns 0 on success, errno otherwise
433  *
434  * NOTE: This routine doesn't need to take the socket lock since it only
435  *       accesses socket information that is unchanging (or which changes in
436  *       a completely predictable manner).
437  */
438 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
439 			int *uaddr_len, int peer)
440 {
441 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
442 	struct tipc_sock *tsk = tipc_sk(sock->sk);
443 
444 	memset(addr, 0, sizeof(*addr));
445 	if (peer) {
446 		if ((sock->state != SS_CONNECTED) &&
447 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
448 			return -ENOTCONN;
449 		addr->addr.id.ref = tipc_port_peerport(&tsk->port);
450 		addr->addr.id.node = tipc_port_peernode(&tsk->port);
451 	} else {
452 		addr->addr.id.ref = tsk->port.ref;
453 		addr->addr.id.node = tipc_own_addr;
454 	}
455 
456 	*uaddr_len = sizeof(*addr);
457 	addr->addrtype = TIPC_ADDR_ID;
458 	addr->family = AF_TIPC;
459 	addr->scope = 0;
460 	addr->addr.name.domain = 0;
461 
462 	return 0;
463 }
464 
465 /**
466  * tipc_poll - read and possibly block on pollmask
467  * @file: file structure associated with the socket
468  * @sock: socket for which to calculate the poll bits
469  * @wait: ???
470  *
471  * Returns pollmask value
472  *
473  * COMMENTARY:
474  * It appears that the usual socket locking mechanisms are not useful here
475  * since the pollmask info is potentially out-of-date the moment this routine
476  * exits.  TCP and other protocols seem to rely on higher level poll routines
477  * to handle any preventable race conditions, so TIPC will do the same ...
478  *
479  * TIPC sets the returned events as follows:
480  *
481  * socket state		flags set
482  * ------------		---------
483  * unconnected		no read flags
484  *			POLLOUT if port is not congested
485  *
486  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
487  *			no write flags
488  *
489  * connected		POLLIN/POLLRDNORM if data in rx queue
490  *			POLLOUT if port is not congested
491  *
492  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
493  *			no write flags
494  *
495  * listening		POLLIN if SYN in rx queue
496  *			no write flags
497  *
498  * ready		POLLIN/POLLRDNORM if data in rx queue
499  * [connectionless]	POLLOUT (since port cannot be congested)
500  *
501  * IMPORTANT: The fact that a read or write operation is indicated does NOT
502  * imply that the operation will succeed, merely that it should be performed
503  * and will not block.
504  */
505 static unsigned int tipc_poll(struct file *file, struct socket *sock,
506 			      poll_table *wait)
507 {
508 	struct sock *sk = sock->sk;
509 	struct tipc_sock *tsk = tipc_sk(sk);
510 	u32 mask = 0;
511 
512 	sock_poll_wait(file, sk_sleep(sk), wait);
513 
514 	switch ((int)sock->state) {
515 	case SS_UNCONNECTED:
516 		if (!tsk->link_cong)
517 			mask |= POLLOUT;
518 		break;
519 	case SS_READY:
520 	case SS_CONNECTED:
521 		if (!tsk->link_cong && !tipc_sk_conn_cong(tsk))
522 			mask |= POLLOUT;
523 		/* fall thru' */
524 	case SS_CONNECTING:
525 	case SS_LISTENING:
526 		if (!skb_queue_empty(&sk->sk_receive_queue))
527 			mask |= (POLLIN | POLLRDNORM);
528 		break;
529 	case SS_DISCONNECTING:
530 		mask = (POLLIN | POLLRDNORM | POLLHUP);
531 		break;
532 	}
533 
534 	return mask;
535 }
536 
537 /**
538  * tipc_sendmcast - send multicast message
539  * @sock: socket structure
540  * @seq: destination address
541  * @iov: message data to send
542  * @dsz: total length of message data
543  * @timeo: timeout to wait for wakeup
544  *
545  * Called from function tipc_sendmsg(), which has done all sanity checks
546  * Returns the number of bytes sent on success, or errno
547  */
548 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
549 			  struct iovec *iov, size_t dsz, long timeo)
550 {
551 	struct sock *sk = sock->sk;
552 	struct tipc_msg *mhdr = &tipc_sk(sk)->port.phdr;
553 	struct sk_buff *buf;
554 	uint mtu;
555 	int rc;
556 
557 	msg_set_type(mhdr, TIPC_MCAST_MSG);
558 	msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
559 	msg_set_destport(mhdr, 0);
560 	msg_set_destnode(mhdr, 0);
561 	msg_set_nametype(mhdr, seq->type);
562 	msg_set_namelower(mhdr, seq->lower);
563 	msg_set_nameupper(mhdr, seq->upper);
564 	msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
565 
566 new_mtu:
567 	mtu = tipc_bclink_get_mtu();
568 	rc = tipc_msg_build(mhdr, iov, 0, dsz, mtu, &buf);
569 	if (unlikely(rc < 0))
570 		return rc;
571 
572 	do {
573 		rc = tipc_bclink_xmit(buf);
574 		if (likely(rc >= 0)) {
575 			rc = dsz;
576 			break;
577 		}
578 		if (rc == -EMSGSIZE)
579 			goto new_mtu;
580 		if (rc != -ELINKCONG)
581 			break;
582 		rc = tipc_wait_for_sndmsg(sock, &timeo);
583 		if (rc)
584 			kfree_skb_list(buf);
585 	} while (!rc);
586 	return rc;
587 }
588 
589 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
590  */
591 void tipc_sk_mcast_rcv(struct sk_buff *buf)
592 {
593 	struct tipc_msg *msg = buf_msg(buf);
594 	struct tipc_port_list dports = {0, NULL, };
595 	struct tipc_port_list *item;
596 	struct sk_buff *b;
597 	uint i, last, dst = 0;
598 	u32 scope = TIPC_CLUSTER_SCOPE;
599 
600 	if (in_own_node(msg_orignode(msg)))
601 		scope = TIPC_NODE_SCOPE;
602 
603 	/* Create destination port list: */
604 	tipc_nametbl_mc_translate(msg_nametype(msg),
605 				  msg_namelower(msg),
606 				  msg_nameupper(msg),
607 				  scope,
608 				  &dports);
609 	last = dports.count;
610 	if (!last) {
611 		kfree_skb(buf);
612 		return;
613 	}
614 
615 	for (item = &dports; item; item = item->next) {
616 		for (i = 0; i < PLSIZE && ++dst <= last; i++) {
617 			b = (dst != last) ? skb_clone(buf, GFP_ATOMIC) : buf;
618 			if (!b) {
619 				pr_warn("Failed do clone mcast rcv buffer\n");
620 				continue;
621 			}
622 			msg_set_destport(msg, item->ports[i]);
623 			tipc_sk_rcv(b);
624 		}
625 	}
626 	tipc_port_list_free(&dports);
627 }
628 
629 /**
630  * tipc_sk_proto_rcv - receive a connection mng protocol message
631  * @tsk: receiving socket
632  * @dnode: node to send response message to, if any
633  * @buf: buffer containing protocol message
634  * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
635  * (CONN_PROBE_REPLY) message should be forwarded.
636  */
637 static int tipc_sk_proto_rcv(struct tipc_sock *tsk, u32 *dnode,
638 			     struct sk_buff *buf)
639 {
640 	struct tipc_msg *msg = buf_msg(buf);
641 	struct tipc_port *port = &tsk->port;
642 	int conn_cong;
643 
644 	/* Ignore if connection cannot be validated: */
645 	if (!port->connected || !tipc_port_peer_msg(port, msg))
646 		goto exit;
647 
648 	port->probing_state = TIPC_CONN_OK;
649 
650 	if (msg_type(msg) == CONN_ACK) {
651 		conn_cong = tipc_sk_conn_cong(tsk);
652 		tsk->sent_unacked -= msg_msgcnt(msg);
653 		if (conn_cong)
654 			tipc_sock_wakeup(tsk);
655 	} else if (msg_type(msg) == CONN_PROBE) {
656 		if (!tipc_msg_reverse(buf, dnode, TIPC_OK))
657 			return TIPC_OK;
658 		msg_set_type(msg, CONN_PROBE_REPLY);
659 		return TIPC_FWD_MSG;
660 	}
661 	/* Do nothing if msg_type() == CONN_PROBE_REPLY */
662 exit:
663 	kfree_skb(buf);
664 	return TIPC_OK;
665 }
666 
667 /**
668  * dest_name_check - verify user is permitted to send to specified port name
669  * @dest: destination address
670  * @m: descriptor for message to be sent
671  *
672  * Prevents restricted configuration commands from being issued by
673  * unauthorized users.
674  *
675  * Returns 0 if permission is granted, otherwise errno
676  */
677 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
678 {
679 	struct tipc_cfg_msg_hdr hdr;
680 
681 	if (unlikely(dest->addrtype == TIPC_ADDR_ID))
682 		return 0;
683 	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
684 		return 0;
685 	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
686 		return 0;
687 	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
688 		return -EACCES;
689 
690 	if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
691 		return -EMSGSIZE;
692 	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
693 		return -EFAULT;
694 	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
695 		return -EACCES;
696 
697 	return 0;
698 }
699 
700 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
701 {
702 	struct sock *sk = sock->sk;
703 	struct tipc_sock *tsk = tipc_sk(sk);
704 	DEFINE_WAIT(wait);
705 	int done;
706 
707 	do {
708 		int err = sock_error(sk);
709 		if (err)
710 			return err;
711 		if (sock->state == SS_DISCONNECTING)
712 			return -EPIPE;
713 		if (!*timeo_p)
714 			return -EAGAIN;
715 		if (signal_pending(current))
716 			return sock_intr_errno(*timeo_p);
717 
718 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
719 		done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
720 		finish_wait(sk_sleep(sk), &wait);
721 	} while (!done);
722 	return 0;
723 }
724 
725 /**
726  * tipc_sendmsg - send message in connectionless manner
727  * @iocb: if NULL, indicates that socket lock is already held
728  * @sock: socket structure
729  * @m: message to send
730  * @dsz: amount of user data to be sent
731  *
732  * Message must have an destination specified explicitly.
733  * Used for SOCK_RDM and SOCK_DGRAM messages,
734  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
735  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
736  *
737  * Returns the number of bytes sent on success, or errno otherwise
738  */
739 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
740 			struct msghdr *m, size_t dsz)
741 {
742 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
743 	struct sock *sk = sock->sk;
744 	struct tipc_sock *tsk = tipc_sk(sk);
745 	struct tipc_port *port = &tsk->port;
746 	struct tipc_msg *mhdr = &port->phdr;
747 	struct iovec *iov = m->msg_iov;
748 	u32 dnode, dport;
749 	struct sk_buff *buf;
750 	struct tipc_name_seq *seq = &dest->addr.nameseq;
751 	u32 mtu;
752 	long timeo;
753 	int rc = -EINVAL;
754 
755 	if (unlikely(!dest))
756 		return -EDESTADDRREQ;
757 
758 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
759 		     (dest->family != AF_TIPC)))
760 		return -EINVAL;
761 
762 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
763 		return -EMSGSIZE;
764 
765 	if (iocb)
766 		lock_sock(sk);
767 
768 	if (unlikely(sock->state != SS_READY)) {
769 		if (sock->state == SS_LISTENING) {
770 			rc = -EPIPE;
771 			goto exit;
772 		}
773 		if (sock->state != SS_UNCONNECTED) {
774 			rc = -EISCONN;
775 			goto exit;
776 		}
777 		if (tsk->port.published) {
778 			rc = -EOPNOTSUPP;
779 			goto exit;
780 		}
781 		if (dest->addrtype == TIPC_ADDR_NAME) {
782 			tsk->port.conn_type = dest->addr.name.name.type;
783 			tsk->port.conn_instance = dest->addr.name.name.instance;
784 		}
785 	}
786 	rc = dest_name_check(dest, m);
787 	if (rc)
788 		goto exit;
789 
790 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
791 
792 	if (dest->addrtype == TIPC_ADDR_MCAST) {
793 		rc = tipc_sendmcast(sock, seq, iov, dsz, timeo);
794 		goto exit;
795 	} else if (dest->addrtype == TIPC_ADDR_NAME) {
796 		u32 type = dest->addr.name.name.type;
797 		u32 inst = dest->addr.name.name.instance;
798 		u32 domain = dest->addr.name.domain;
799 
800 		dnode = domain;
801 		msg_set_type(mhdr, TIPC_NAMED_MSG);
802 		msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
803 		msg_set_nametype(mhdr, type);
804 		msg_set_nameinst(mhdr, inst);
805 		msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
806 		dport = tipc_nametbl_translate(type, inst, &dnode);
807 		msg_set_destnode(mhdr, dnode);
808 		msg_set_destport(mhdr, dport);
809 		if (unlikely(!dport && !dnode)) {
810 			rc = -EHOSTUNREACH;
811 			goto exit;
812 		}
813 	} else if (dest->addrtype == TIPC_ADDR_ID) {
814 		dnode = dest->addr.id.node;
815 		msg_set_type(mhdr, TIPC_DIRECT_MSG);
816 		msg_set_lookup_scope(mhdr, 0);
817 		msg_set_destnode(mhdr, dnode);
818 		msg_set_destport(mhdr, dest->addr.id.ref);
819 		msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
820 	}
821 
822 new_mtu:
823 	mtu = tipc_node_get_mtu(dnode, tsk->port.ref);
824 	rc = tipc_msg_build(mhdr, iov, 0, dsz, mtu, &buf);
825 	if (rc < 0)
826 		goto exit;
827 
828 	do {
829 		rc = tipc_link_xmit(buf, dnode, tsk->port.ref);
830 		if (likely(rc >= 0)) {
831 			if (sock->state != SS_READY)
832 				sock->state = SS_CONNECTING;
833 			rc = dsz;
834 			break;
835 		}
836 		if (rc == -EMSGSIZE)
837 			goto new_mtu;
838 
839 		if (rc != -ELINKCONG)
840 			break;
841 
842 		rc = tipc_wait_for_sndmsg(sock, &timeo);
843 		if (rc)
844 			kfree_skb_list(buf);
845 	} while (!rc);
846 exit:
847 	if (iocb)
848 		release_sock(sk);
849 
850 	return rc;
851 }
852 
853 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
854 {
855 	struct sock *sk = sock->sk;
856 	struct tipc_sock *tsk = tipc_sk(sk);
857 	DEFINE_WAIT(wait);
858 	int done;
859 
860 	do {
861 		int err = sock_error(sk);
862 		if (err)
863 			return err;
864 		if (sock->state == SS_DISCONNECTING)
865 			return -EPIPE;
866 		else if (sock->state != SS_CONNECTED)
867 			return -ENOTCONN;
868 		if (!*timeo_p)
869 			return -EAGAIN;
870 		if (signal_pending(current))
871 			return sock_intr_errno(*timeo_p);
872 
873 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
874 		done = sk_wait_event(sk, timeo_p,
875 				     (!tsk->link_cong &&
876 				      !tipc_sk_conn_cong(tsk)) ||
877 				     !tsk->port.connected);
878 		finish_wait(sk_sleep(sk), &wait);
879 	} while (!done);
880 	return 0;
881 }
882 
883 /**
884  * tipc_send_stream - send stream-oriented data
885  * @iocb: (unused)
886  * @sock: socket structure
887  * @m: data to send
888  * @dsz: total length of data to be transmitted
889  *
890  * Used for SOCK_STREAM data.
891  *
892  * Returns the number of bytes sent on success (or partial success),
893  * or errno if no data sent
894  */
895 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
896 			    struct msghdr *m, size_t dsz)
897 {
898 	struct sock *sk = sock->sk;
899 	struct tipc_sock *tsk = tipc_sk(sk);
900 	struct tipc_port *port = &tsk->port;
901 	struct tipc_msg *mhdr = &port->phdr;
902 	struct sk_buff *buf;
903 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
904 	u32 ref = port->ref;
905 	int rc = -EINVAL;
906 	long timeo;
907 	u32 dnode;
908 	uint mtu, send, sent = 0;
909 
910 	/* Handle implied connection establishment */
911 	if (unlikely(dest)) {
912 		rc = tipc_sendmsg(iocb, sock, m, dsz);
913 		if (dsz && (dsz == rc))
914 			tsk->sent_unacked = 1;
915 		return rc;
916 	}
917 	if (dsz > (uint)INT_MAX)
918 		return -EMSGSIZE;
919 
920 	if (iocb)
921 		lock_sock(sk);
922 
923 	if (unlikely(sock->state != SS_CONNECTED)) {
924 		if (sock->state == SS_DISCONNECTING)
925 			rc = -EPIPE;
926 		else
927 			rc = -ENOTCONN;
928 		goto exit;
929 	}
930 
931 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
932 	dnode = tipc_port_peernode(port);
933 
934 next:
935 	mtu = port->max_pkt;
936 	send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
937 	rc = tipc_msg_build(mhdr, m->msg_iov, sent, send, mtu, &buf);
938 	if (unlikely(rc < 0))
939 		goto exit;
940 	do {
941 		if (likely(!tipc_sk_conn_cong(tsk))) {
942 			rc = tipc_link_xmit(buf, dnode, ref);
943 			if (likely(!rc)) {
944 				tsk->sent_unacked++;
945 				sent += send;
946 				if (sent == dsz)
947 					break;
948 				goto next;
949 			}
950 			if (rc == -EMSGSIZE) {
951 				port->max_pkt = tipc_node_get_mtu(dnode, ref);
952 				goto next;
953 			}
954 			if (rc != -ELINKCONG)
955 				break;
956 		}
957 		rc = tipc_wait_for_sndpkt(sock, &timeo);
958 		if (rc)
959 			kfree_skb_list(buf);
960 	} while (!rc);
961 exit:
962 	if (iocb)
963 		release_sock(sk);
964 	return sent ? sent : rc;
965 }
966 
967 /**
968  * tipc_send_packet - send a connection-oriented message
969  * @iocb: if NULL, indicates that socket lock is already held
970  * @sock: socket structure
971  * @m: message to send
972  * @dsz: length of data to be transmitted
973  *
974  * Used for SOCK_SEQPACKET messages.
975  *
976  * Returns the number of bytes sent on success, or errno otherwise
977  */
978 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
979 			    struct msghdr *m, size_t dsz)
980 {
981 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
982 		return -EMSGSIZE;
983 
984 	return tipc_send_stream(iocb, sock, m, dsz);
985 }
986 
987 /**
988  * auto_connect - complete connection setup to a remote port
989  * @tsk: tipc socket structure
990  * @msg: peer's response message
991  *
992  * Returns 0 on success, errno otherwise
993  */
994 static int auto_connect(struct tipc_sock *tsk, struct tipc_msg *msg)
995 {
996 	struct tipc_port *port = &tsk->port;
997 	struct socket *sock = tsk->sk.sk_socket;
998 	struct tipc_portid peer;
999 
1000 	peer.ref = msg_origport(msg);
1001 	peer.node = msg_orignode(msg);
1002 
1003 	__tipc_port_connect(port->ref, port, &peer);
1004 
1005 	if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)
1006 		return -EINVAL;
1007 	msg_set_importance(&port->phdr, (u32)msg_importance(msg));
1008 	sock->state = SS_CONNECTED;
1009 	return 0;
1010 }
1011 
1012 /**
1013  * set_orig_addr - capture sender's address for received message
1014  * @m: descriptor for message info
1015  * @msg: received message header
1016  *
1017  * Note: Address is not captured if not requested by receiver.
1018  */
1019 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1020 {
1021 	DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1022 
1023 	if (addr) {
1024 		addr->family = AF_TIPC;
1025 		addr->addrtype = TIPC_ADDR_ID;
1026 		memset(&addr->addr, 0, sizeof(addr->addr));
1027 		addr->addr.id.ref = msg_origport(msg);
1028 		addr->addr.id.node = msg_orignode(msg);
1029 		addr->addr.name.domain = 0;	/* could leave uninitialized */
1030 		addr->scope = 0;		/* could leave uninitialized */
1031 		m->msg_namelen = sizeof(struct sockaddr_tipc);
1032 	}
1033 }
1034 
1035 /**
1036  * anc_data_recv - optionally capture ancillary data for received message
1037  * @m: descriptor for message info
1038  * @msg: received message header
1039  * @tport: TIPC port associated with message
1040  *
1041  * Note: Ancillary data is not captured if not requested by receiver.
1042  *
1043  * Returns 0 if successful, otherwise errno
1044  */
1045 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1046 			 struct tipc_port *tport)
1047 {
1048 	u32 anc_data[3];
1049 	u32 err;
1050 	u32 dest_type;
1051 	int has_name;
1052 	int res;
1053 
1054 	if (likely(m->msg_controllen == 0))
1055 		return 0;
1056 
1057 	/* Optionally capture errored message object(s) */
1058 	err = msg ? msg_errcode(msg) : 0;
1059 	if (unlikely(err)) {
1060 		anc_data[0] = err;
1061 		anc_data[1] = msg_data_sz(msg);
1062 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1063 		if (res)
1064 			return res;
1065 		if (anc_data[1]) {
1066 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1067 				       msg_data(msg));
1068 			if (res)
1069 				return res;
1070 		}
1071 	}
1072 
1073 	/* Optionally capture message destination object */
1074 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1075 	switch (dest_type) {
1076 	case TIPC_NAMED_MSG:
1077 		has_name = 1;
1078 		anc_data[0] = msg_nametype(msg);
1079 		anc_data[1] = msg_namelower(msg);
1080 		anc_data[2] = msg_namelower(msg);
1081 		break;
1082 	case TIPC_MCAST_MSG:
1083 		has_name = 1;
1084 		anc_data[0] = msg_nametype(msg);
1085 		anc_data[1] = msg_namelower(msg);
1086 		anc_data[2] = msg_nameupper(msg);
1087 		break;
1088 	case TIPC_CONN_MSG:
1089 		has_name = (tport->conn_type != 0);
1090 		anc_data[0] = tport->conn_type;
1091 		anc_data[1] = tport->conn_instance;
1092 		anc_data[2] = tport->conn_instance;
1093 		break;
1094 	default:
1095 		has_name = 0;
1096 	}
1097 	if (has_name) {
1098 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1099 		if (res)
1100 			return res;
1101 	}
1102 
1103 	return 0;
1104 }
1105 
1106 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1107 {
1108 	struct sock *sk = sock->sk;
1109 	DEFINE_WAIT(wait);
1110 	long timeo = *timeop;
1111 	int err;
1112 
1113 	for (;;) {
1114 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1115 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1116 			if (sock->state == SS_DISCONNECTING) {
1117 				err = -ENOTCONN;
1118 				break;
1119 			}
1120 			release_sock(sk);
1121 			timeo = schedule_timeout(timeo);
1122 			lock_sock(sk);
1123 		}
1124 		err = 0;
1125 		if (!skb_queue_empty(&sk->sk_receive_queue))
1126 			break;
1127 		err = sock_intr_errno(timeo);
1128 		if (signal_pending(current))
1129 			break;
1130 		err = -EAGAIN;
1131 		if (!timeo)
1132 			break;
1133 	}
1134 	finish_wait(sk_sleep(sk), &wait);
1135 	*timeop = timeo;
1136 	return err;
1137 }
1138 
1139 /**
1140  * tipc_recvmsg - receive packet-oriented message
1141  * @iocb: (unused)
1142  * @m: descriptor for message info
1143  * @buf_len: total size of user buffer area
1144  * @flags: receive flags
1145  *
1146  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1147  * If the complete message doesn't fit in user area, truncate it.
1148  *
1149  * Returns size of returned message data, errno otherwise
1150  */
1151 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1152 			struct msghdr *m, size_t buf_len, int flags)
1153 {
1154 	struct sock *sk = sock->sk;
1155 	struct tipc_sock *tsk = tipc_sk(sk);
1156 	struct tipc_port *port = &tsk->port;
1157 	struct sk_buff *buf;
1158 	struct tipc_msg *msg;
1159 	long timeo;
1160 	unsigned int sz;
1161 	u32 err;
1162 	int res;
1163 
1164 	/* Catch invalid receive requests */
1165 	if (unlikely(!buf_len))
1166 		return -EINVAL;
1167 
1168 	lock_sock(sk);
1169 
1170 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1171 		res = -ENOTCONN;
1172 		goto exit;
1173 	}
1174 
1175 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1176 restart:
1177 
1178 	/* Look for a message in receive queue; wait if necessary */
1179 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1180 	if (res)
1181 		goto exit;
1182 
1183 	/* Look at first message in receive queue */
1184 	buf = skb_peek(&sk->sk_receive_queue);
1185 	msg = buf_msg(buf);
1186 	sz = msg_data_sz(msg);
1187 	err = msg_errcode(msg);
1188 
1189 	/* Discard an empty non-errored message & try again */
1190 	if ((!sz) && (!err)) {
1191 		advance_rx_queue(sk);
1192 		goto restart;
1193 	}
1194 
1195 	/* Capture sender's address (optional) */
1196 	set_orig_addr(m, msg);
1197 
1198 	/* Capture ancillary data (optional) */
1199 	res = anc_data_recv(m, msg, port);
1200 	if (res)
1201 		goto exit;
1202 
1203 	/* Capture message data (if valid) & compute return value (always) */
1204 	if (!err) {
1205 		if (unlikely(buf_len < sz)) {
1206 			sz = buf_len;
1207 			m->msg_flags |= MSG_TRUNC;
1208 		}
1209 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1210 					      m->msg_iov, sz);
1211 		if (res)
1212 			goto exit;
1213 		res = sz;
1214 	} else {
1215 		if ((sock->state == SS_READY) ||
1216 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1217 			res = 0;
1218 		else
1219 			res = -ECONNRESET;
1220 	}
1221 
1222 	/* Consume received message (optional) */
1223 	if (likely(!(flags & MSG_PEEK))) {
1224 		if ((sock->state != SS_READY) &&
1225 		    (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1226 			tipc_acknowledge(port->ref, tsk->rcv_unacked);
1227 			tsk->rcv_unacked = 0;
1228 		}
1229 		advance_rx_queue(sk);
1230 	}
1231 exit:
1232 	release_sock(sk);
1233 	return res;
1234 }
1235 
1236 /**
1237  * tipc_recv_stream - receive stream-oriented data
1238  * @iocb: (unused)
1239  * @m: descriptor for message info
1240  * @buf_len: total size of user buffer area
1241  * @flags: receive flags
1242  *
1243  * Used for SOCK_STREAM messages only.  If not enough data is available
1244  * will optionally wait for more; never truncates data.
1245  *
1246  * Returns size of returned message data, errno otherwise
1247  */
1248 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1249 			    struct msghdr *m, size_t buf_len, int flags)
1250 {
1251 	struct sock *sk = sock->sk;
1252 	struct tipc_sock *tsk = tipc_sk(sk);
1253 	struct tipc_port *port = &tsk->port;
1254 	struct sk_buff *buf;
1255 	struct tipc_msg *msg;
1256 	long timeo;
1257 	unsigned int sz;
1258 	int sz_to_copy, target, needed;
1259 	int sz_copied = 0;
1260 	u32 err;
1261 	int res = 0;
1262 
1263 	/* Catch invalid receive attempts */
1264 	if (unlikely(!buf_len))
1265 		return -EINVAL;
1266 
1267 	lock_sock(sk);
1268 
1269 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1270 		res = -ENOTCONN;
1271 		goto exit;
1272 	}
1273 
1274 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1275 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1276 
1277 restart:
1278 	/* Look for a message in receive queue; wait if necessary */
1279 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1280 	if (res)
1281 		goto exit;
1282 
1283 	/* Look at first message in receive queue */
1284 	buf = skb_peek(&sk->sk_receive_queue);
1285 	msg = buf_msg(buf);
1286 	sz = msg_data_sz(msg);
1287 	err = msg_errcode(msg);
1288 
1289 	/* Discard an empty non-errored message & try again */
1290 	if ((!sz) && (!err)) {
1291 		advance_rx_queue(sk);
1292 		goto restart;
1293 	}
1294 
1295 	/* Optionally capture sender's address & ancillary data of first msg */
1296 	if (sz_copied == 0) {
1297 		set_orig_addr(m, msg);
1298 		res = anc_data_recv(m, msg, port);
1299 		if (res)
1300 			goto exit;
1301 	}
1302 
1303 	/* Capture message data (if valid) & compute return value (always) */
1304 	if (!err) {
1305 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1306 
1307 		sz -= offset;
1308 		needed = (buf_len - sz_copied);
1309 		sz_to_copy = (sz <= needed) ? sz : needed;
1310 
1311 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1312 					      m->msg_iov, sz_to_copy);
1313 		if (res)
1314 			goto exit;
1315 
1316 		sz_copied += sz_to_copy;
1317 
1318 		if (sz_to_copy < sz) {
1319 			if (!(flags & MSG_PEEK))
1320 				TIPC_SKB_CB(buf)->handle =
1321 				(void *)(unsigned long)(offset + sz_to_copy);
1322 			goto exit;
1323 		}
1324 	} else {
1325 		if (sz_copied != 0)
1326 			goto exit; /* can't add error msg to valid data */
1327 
1328 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1329 			res = 0;
1330 		else
1331 			res = -ECONNRESET;
1332 	}
1333 
1334 	/* Consume received message (optional) */
1335 	if (likely(!(flags & MSG_PEEK))) {
1336 		if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1337 			tipc_acknowledge(port->ref, tsk->rcv_unacked);
1338 			tsk->rcv_unacked = 0;
1339 		}
1340 		advance_rx_queue(sk);
1341 	}
1342 
1343 	/* Loop around if more data is required */
1344 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1345 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1346 	    (sz_copied < target)) &&	/* and more is ready or required */
1347 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1348 	    (!err))			/* and haven't reached a FIN */
1349 		goto restart;
1350 
1351 exit:
1352 	release_sock(sk);
1353 	return sz_copied ? sz_copied : res;
1354 }
1355 
1356 /**
1357  * tipc_write_space - wake up thread if port congestion is released
1358  * @sk: socket
1359  */
1360 static void tipc_write_space(struct sock *sk)
1361 {
1362 	struct socket_wq *wq;
1363 
1364 	rcu_read_lock();
1365 	wq = rcu_dereference(sk->sk_wq);
1366 	if (wq_has_sleeper(wq))
1367 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1368 						POLLWRNORM | POLLWRBAND);
1369 	rcu_read_unlock();
1370 }
1371 
1372 /**
1373  * tipc_data_ready - wake up threads to indicate messages have been received
1374  * @sk: socket
1375  * @len: the length of messages
1376  */
1377 static void tipc_data_ready(struct sock *sk)
1378 {
1379 	struct socket_wq *wq;
1380 
1381 	rcu_read_lock();
1382 	wq = rcu_dereference(sk->sk_wq);
1383 	if (wq_has_sleeper(wq))
1384 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1385 						POLLRDNORM | POLLRDBAND);
1386 	rcu_read_unlock();
1387 }
1388 
1389 /**
1390  * filter_connect - Handle all incoming messages for a connection-based socket
1391  * @tsk: TIPC socket
1392  * @msg: message
1393  *
1394  * Returns 0 (TIPC_OK) if everyting ok, -TIPC_ERR_NO_PORT otherwise
1395  */
1396 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1397 {
1398 	struct sock *sk = &tsk->sk;
1399 	struct tipc_port *port = &tsk->port;
1400 	struct socket *sock = sk->sk_socket;
1401 	struct tipc_msg *msg = buf_msg(*buf);
1402 
1403 	int retval = -TIPC_ERR_NO_PORT;
1404 	int res;
1405 
1406 	if (msg_mcast(msg))
1407 		return retval;
1408 
1409 	switch ((int)sock->state) {
1410 	case SS_CONNECTED:
1411 		/* Accept only connection-based messages sent by peer */
1412 		if (msg_connected(msg) && tipc_port_peer_msg(port, msg)) {
1413 			if (unlikely(msg_errcode(msg))) {
1414 				sock->state = SS_DISCONNECTING;
1415 				__tipc_port_disconnect(port);
1416 			}
1417 			retval = TIPC_OK;
1418 		}
1419 		break;
1420 	case SS_CONNECTING:
1421 		/* Accept only ACK or NACK message */
1422 		if (unlikely(msg_errcode(msg))) {
1423 			sock->state = SS_DISCONNECTING;
1424 			sk->sk_err = ECONNREFUSED;
1425 			retval = TIPC_OK;
1426 			break;
1427 		}
1428 
1429 		if (unlikely(!msg_connected(msg)))
1430 			break;
1431 
1432 		res = auto_connect(tsk, msg);
1433 		if (res) {
1434 			sock->state = SS_DISCONNECTING;
1435 			sk->sk_err = -res;
1436 			retval = TIPC_OK;
1437 			break;
1438 		}
1439 
1440 		/* If an incoming message is an 'ACK-', it should be
1441 		 * discarded here because it doesn't contain useful
1442 		 * data. In addition, we should try to wake up
1443 		 * connect() routine if sleeping.
1444 		 */
1445 		if (msg_data_sz(msg) == 0) {
1446 			kfree_skb(*buf);
1447 			*buf = NULL;
1448 			if (waitqueue_active(sk_sleep(sk)))
1449 				wake_up_interruptible(sk_sleep(sk));
1450 		}
1451 		retval = TIPC_OK;
1452 		break;
1453 	case SS_LISTENING:
1454 	case SS_UNCONNECTED:
1455 		/* Accept only SYN message */
1456 		if (!msg_connected(msg) && !(msg_errcode(msg)))
1457 			retval = TIPC_OK;
1458 		break;
1459 	case SS_DISCONNECTING:
1460 		break;
1461 	default:
1462 		pr_err("Unknown socket state %u\n", sock->state);
1463 	}
1464 	return retval;
1465 }
1466 
1467 /**
1468  * rcvbuf_limit - get proper overload limit of socket receive queue
1469  * @sk: socket
1470  * @buf: message
1471  *
1472  * For all connection oriented messages, irrespective of importance,
1473  * the default overload value (i.e. 67MB) is set as limit.
1474  *
1475  * For all connectionless messages, by default new queue limits are
1476  * as belows:
1477  *
1478  * TIPC_LOW_IMPORTANCE       (4 MB)
1479  * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1480  * TIPC_HIGH_IMPORTANCE      (16 MB)
1481  * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1482  *
1483  * Returns overload limit according to corresponding message importance
1484  */
1485 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1486 {
1487 	struct tipc_msg *msg = buf_msg(buf);
1488 
1489 	if (msg_connected(msg))
1490 		return sysctl_tipc_rmem[2];
1491 
1492 	return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1493 		msg_importance(msg);
1494 }
1495 
1496 /**
1497  * filter_rcv - validate incoming message
1498  * @sk: socket
1499  * @buf: message
1500  *
1501  * Enqueues message on receive queue if acceptable; optionally handles
1502  * disconnect indication for a connected socket.
1503  *
1504  * Called with socket lock already taken; port lock may also be taken.
1505  *
1506  * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1507  * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1508  */
1509 static int filter_rcv(struct sock *sk, struct sk_buff *buf)
1510 {
1511 	struct socket *sock = sk->sk_socket;
1512 	struct tipc_sock *tsk = tipc_sk(sk);
1513 	struct tipc_msg *msg = buf_msg(buf);
1514 	unsigned int limit = rcvbuf_limit(sk, buf);
1515 	u32 onode;
1516 	int rc = TIPC_OK;
1517 
1518 	if (unlikely(msg_user(msg) == CONN_MANAGER))
1519 		return tipc_sk_proto_rcv(tsk, &onode, buf);
1520 
1521 	/* Reject message if it is wrong sort of message for socket */
1522 	if (msg_type(msg) > TIPC_DIRECT_MSG)
1523 		return -TIPC_ERR_NO_PORT;
1524 
1525 	if (sock->state == SS_READY) {
1526 		if (msg_connected(msg))
1527 			return -TIPC_ERR_NO_PORT;
1528 	} else {
1529 		rc = filter_connect(tsk, &buf);
1530 		if (rc != TIPC_OK || buf == NULL)
1531 			return rc;
1532 	}
1533 
1534 	/* Reject message if there isn't room to queue it */
1535 	if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1536 		return -TIPC_ERR_OVERLOAD;
1537 
1538 	/* Enqueue message */
1539 	TIPC_SKB_CB(buf)->handle = NULL;
1540 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1541 	skb_set_owner_r(buf, sk);
1542 
1543 	sk->sk_data_ready(sk);
1544 	return TIPC_OK;
1545 }
1546 
1547 /**
1548  * tipc_backlog_rcv - handle incoming message from backlog queue
1549  * @sk: socket
1550  * @buf: message
1551  *
1552  * Caller must hold socket lock, but not port lock.
1553  *
1554  * Returns 0
1555  */
1556 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *buf)
1557 {
1558 	int rc;
1559 	u32 onode;
1560 	struct tipc_sock *tsk = tipc_sk(sk);
1561 	uint truesize = buf->truesize;
1562 
1563 	rc = filter_rcv(sk, buf);
1564 
1565 	if (likely(!rc)) {
1566 		if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1567 			atomic_add(truesize, &tsk->dupl_rcvcnt);
1568 		return 0;
1569 	}
1570 
1571 	if ((rc < 0) && !tipc_msg_reverse(buf, &onode, -rc))
1572 		return 0;
1573 
1574 	tipc_link_xmit(buf, onode, 0);
1575 
1576 	return 0;
1577 }
1578 
1579 /**
1580  * tipc_sk_rcv - handle incoming message
1581  * @buf: buffer containing arriving message
1582  * Consumes buffer
1583  * Returns 0 if success, or errno: -EHOSTUNREACH
1584  */
1585 int tipc_sk_rcv(struct sk_buff *buf)
1586 {
1587 	struct tipc_sock *tsk;
1588 	struct tipc_port *port;
1589 	struct sock *sk;
1590 	u32 dport = msg_destport(buf_msg(buf));
1591 	int rc = TIPC_OK;
1592 	uint limit;
1593 	u32 dnode;
1594 
1595 	/* Validate destination and message */
1596 	port = tipc_port_lock(dport);
1597 	if (unlikely(!port)) {
1598 		rc = tipc_msg_eval(buf, &dnode);
1599 		goto exit;
1600 	}
1601 
1602 	tsk = tipc_port_to_sock(port);
1603 	sk = &tsk->sk;
1604 
1605 	/* Queue message */
1606 	bh_lock_sock(sk);
1607 
1608 	if (!sock_owned_by_user(sk)) {
1609 		rc = filter_rcv(sk, buf);
1610 	} else {
1611 		if (sk->sk_backlog.len == 0)
1612 			atomic_set(&tsk->dupl_rcvcnt, 0);
1613 		limit = rcvbuf_limit(sk, buf) + atomic_read(&tsk->dupl_rcvcnt);
1614 		if (sk_add_backlog(sk, buf, limit))
1615 			rc = -TIPC_ERR_OVERLOAD;
1616 	}
1617 	bh_unlock_sock(sk);
1618 	tipc_port_unlock(port);
1619 
1620 	if (likely(!rc))
1621 		return 0;
1622 exit:
1623 	if ((rc < 0) && !tipc_msg_reverse(buf, &dnode, -rc))
1624 		return -EHOSTUNREACH;
1625 
1626 	tipc_link_xmit(buf, dnode, 0);
1627 	return (rc < 0) ? -EHOSTUNREACH : 0;
1628 }
1629 
1630 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1631 {
1632 	struct sock *sk = sock->sk;
1633 	DEFINE_WAIT(wait);
1634 	int done;
1635 
1636 	do {
1637 		int err = sock_error(sk);
1638 		if (err)
1639 			return err;
1640 		if (!*timeo_p)
1641 			return -ETIMEDOUT;
1642 		if (signal_pending(current))
1643 			return sock_intr_errno(*timeo_p);
1644 
1645 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1646 		done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1647 		finish_wait(sk_sleep(sk), &wait);
1648 	} while (!done);
1649 	return 0;
1650 }
1651 
1652 /**
1653  * tipc_connect - establish a connection to another TIPC port
1654  * @sock: socket structure
1655  * @dest: socket address for destination port
1656  * @destlen: size of socket address data structure
1657  * @flags: file-related flags associated with socket
1658  *
1659  * Returns 0 on success, errno otherwise
1660  */
1661 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1662 			int destlen, int flags)
1663 {
1664 	struct sock *sk = sock->sk;
1665 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1666 	struct msghdr m = {NULL,};
1667 	long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1668 	socket_state previous;
1669 	int res;
1670 
1671 	lock_sock(sk);
1672 
1673 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1674 	if (sock->state == SS_READY) {
1675 		res = -EOPNOTSUPP;
1676 		goto exit;
1677 	}
1678 
1679 	/*
1680 	 * Reject connection attempt using multicast address
1681 	 *
1682 	 * Note: send_msg() validates the rest of the address fields,
1683 	 *       so there's no need to do it here
1684 	 */
1685 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1686 		res = -EINVAL;
1687 		goto exit;
1688 	}
1689 
1690 	previous = sock->state;
1691 	switch (sock->state) {
1692 	case SS_UNCONNECTED:
1693 		/* Send a 'SYN-' to destination */
1694 		m.msg_name = dest;
1695 		m.msg_namelen = destlen;
1696 
1697 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1698 		 * indicate send_msg() is never blocked.
1699 		 */
1700 		if (!timeout)
1701 			m.msg_flags = MSG_DONTWAIT;
1702 
1703 		res = tipc_sendmsg(NULL, sock, &m, 0);
1704 		if ((res < 0) && (res != -EWOULDBLOCK))
1705 			goto exit;
1706 
1707 		/* Just entered SS_CONNECTING state; the only
1708 		 * difference is that return value in non-blocking
1709 		 * case is EINPROGRESS, rather than EALREADY.
1710 		 */
1711 		res = -EINPROGRESS;
1712 	case SS_CONNECTING:
1713 		if (previous == SS_CONNECTING)
1714 			res = -EALREADY;
1715 		if (!timeout)
1716 			goto exit;
1717 		timeout = msecs_to_jiffies(timeout);
1718 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1719 		res = tipc_wait_for_connect(sock, &timeout);
1720 		break;
1721 	case SS_CONNECTED:
1722 		res = -EISCONN;
1723 		break;
1724 	default:
1725 		res = -EINVAL;
1726 		break;
1727 	}
1728 exit:
1729 	release_sock(sk);
1730 	return res;
1731 }
1732 
1733 /**
1734  * tipc_listen - allow socket to listen for incoming connections
1735  * @sock: socket structure
1736  * @len: (unused)
1737  *
1738  * Returns 0 on success, errno otherwise
1739  */
1740 static int tipc_listen(struct socket *sock, int len)
1741 {
1742 	struct sock *sk = sock->sk;
1743 	int res;
1744 
1745 	lock_sock(sk);
1746 
1747 	if (sock->state != SS_UNCONNECTED)
1748 		res = -EINVAL;
1749 	else {
1750 		sock->state = SS_LISTENING;
1751 		res = 0;
1752 	}
1753 
1754 	release_sock(sk);
1755 	return res;
1756 }
1757 
1758 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1759 {
1760 	struct sock *sk = sock->sk;
1761 	DEFINE_WAIT(wait);
1762 	int err;
1763 
1764 	/* True wake-one mechanism for incoming connections: only
1765 	 * one process gets woken up, not the 'whole herd'.
1766 	 * Since we do not 'race & poll' for established sockets
1767 	 * anymore, the common case will execute the loop only once.
1768 	*/
1769 	for (;;) {
1770 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1771 					  TASK_INTERRUPTIBLE);
1772 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1773 			release_sock(sk);
1774 			timeo = schedule_timeout(timeo);
1775 			lock_sock(sk);
1776 		}
1777 		err = 0;
1778 		if (!skb_queue_empty(&sk->sk_receive_queue))
1779 			break;
1780 		err = -EINVAL;
1781 		if (sock->state != SS_LISTENING)
1782 			break;
1783 		err = sock_intr_errno(timeo);
1784 		if (signal_pending(current))
1785 			break;
1786 		err = -EAGAIN;
1787 		if (!timeo)
1788 			break;
1789 	}
1790 	finish_wait(sk_sleep(sk), &wait);
1791 	return err;
1792 }
1793 
1794 /**
1795  * tipc_accept - wait for connection request
1796  * @sock: listening socket
1797  * @newsock: new socket that is to be connected
1798  * @flags: file-related flags associated with socket
1799  *
1800  * Returns 0 on success, errno otherwise
1801  */
1802 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1803 {
1804 	struct sock *new_sk, *sk = sock->sk;
1805 	struct sk_buff *buf;
1806 	struct tipc_port *new_port;
1807 	struct tipc_msg *msg;
1808 	struct tipc_portid peer;
1809 	u32 new_ref;
1810 	long timeo;
1811 	int res;
1812 
1813 	lock_sock(sk);
1814 
1815 	if (sock->state != SS_LISTENING) {
1816 		res = -EINVAL;
1817 		goto exit;
1818 	}
1819 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1820 	res = tipc_wait_for_accept(sock, timeo);
1821 	if (res)
1822 		goto exit;
1823 
1824 	buf = skb_peek(&sk->sk_receive_queue);
1825 
1826 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1827 	if (res)
1828 		goto exit;
1829 
1830 	new_sk = new_sock->sk;
1831 	new_port = &tipc_sk(new_sk)->port;
1832 	new_ref = new_port->ref;
1833 	msg = buf_msg(buf);
1834 
1835 	/* we lock on new_sk; but lockdep sees the lock on sk */
1836 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1837 
1838 	/*
1839 	 * Reject any stray messages received by new socket
1840 	 * before the socket lock was taken (very, very unlikely)
1841 	 */
1842 	reject_rx_queue(new_sk);
1843 
1844 	/* Connect new socket to it's peer */
1845 	peer.ref = msg_origport(msg);
1846 	peer.node = msg_orignode(msg);
1847 	tipc_port_connect(new_ref, &peer);
1848 	new_sock->state = SS_CONNECTED;
1849 
1850 	tipc_port_set_importance(new_port, msg_importance(msg));
1851 	if (msg_named(msg)) {
1852 		new_port->conn_type = msg_nametype(msg);
1853 		new_port->conn_instance = msg_nameinst(msg);
1854 	}
1855 
1856 	/*
1857 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1858 	 * Respond to 'SYN+' by queuing it on new socket.
1859 	 */
1860 	if (!msg_data_sz(msg)) {
1861 		struct msghdr m = {NULL,};
1862 
1863 		advance_rx_queue(sk);
1864 		tipc_send_packet(NULL, new_sock, &m, 0);
1865 	} else {
1866 		__skb_dequeue(&sk->sk_receive_queue);
1867 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
1868 		skb_set_owner_r(buf, new_sk);
1869 	}
1870 	release_sock(new_sk);
1871 exit:
1872 	release_sock(sk);
1873 	return res;
1874 }
1875 
1876 /**
1877  * tipc_shutdown - shutdown socket connection
1878  * @sock: socket structure
1879  * @how: direction to close (must be SHUT_RDWR)
1880  *
1881  * Terminates connection (if necessary), then purges socket's receive queue.
1882  *
1883  * Returns 0 on success, errno otherwise
1884  */
1885 static int tipc_shutdown(struct socket *sock, int how)
1886 {
1887 	struct sock *sk = sock->sk;
1888 	struct tipc_sock *tsk = tipc_sk(sk);
1889 	struct tipc_port *port = &tsk->port;
1890 	struct sk_buff *buf;
1891 	u32 peer;
1892 	int res;
1893 
1894 	if (how != SHUT_RDWR)
1895 		return -EINVAL;
1896 
1897 	lock_sock(sk);
1898 
1899 	switch (sock->state) {
1900 	case SS_CONNECTING:
1901 	case SS_CONNECTED:
1902 
1903 restart:
1904 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1905 		buf = __skb_dequeue(&sk->sk_receive_queue);
1906 		if (buf) {
1907 			if (TIPC_SKB_CB(buf)->handle != NULL) {
1908 				kfree_skb(buf);
1909 				goto restart;
1910 			}
1911 			tipc_port_disconnect(port->ref);
1912 			if (tipc_msg_reverse(buf, &peer, TIPC_CONN_SHUTDOWN))
1913 				tipc_link_xmit(buf, peer, 0);
1914 		} else {
1915 			tipc_port_shutdown(port->ref);
1916 		}
1917 
1918 		sock->state = SS_DISCONNECTING;
1919 
1920 		/* fall through */
1921 
1922 	case SS_DISCONNECTING:
1923 
1924 		/* Discard any unreceived messages */
1925 		__skb_queue_purge(&sk->sk_receive_queue);
1926 
1927 		/* Wake up anyone sleeping in poll */
1928 		sk->sk_state_change(sk);
1929 		res = 0;
1930 		break;
1931 
1932 	default:
1933 		res = -ENOTCONN;
1934 	}
1935 
1936 	release_sock(sk);
1937 	return res;
1938 }
1939 
1940 /**
1941  * tipc_setsockopt - set socket option
1942  * @sock: socket structure
1943  * @lvl: option level
1944  * @opt: option identifier
1945  * @ov: pointer to new option value
1946  * @ol: length of option value
1947  *
1948  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1949  * (to ease compatibility).
1950  *
1951  * Returns 0 on success, errno otherwise
1952  */
1953 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
1954 			   char __user *ov, unsigned int ol)
1955 {
1956 	struct sock *sk = sock->sk;
1957 	struct tipc_sock *tsk = tipc_sk(sk);
1958 	struct tipc_port *port = &tsk->port;
1959 	u32 value;
1960 	int res;
1961 
1962 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1963 		return 0;
1964 	if (lvl != SOL_TIPC)
1965 		return -ENOPROTOOPT;
1966 	if (ol < sizeof(value))
1967 		return -EINVAL;
1968 	res = get_user(value, (u32 __user *)ov);
1969 	if (res)
1970 		return res;
1971 
1972 	lock_sock(sk);
1973 
1974 	switch (opt) {
1975 	case TIPC_IMPORTANCE:
1976 		tipc_port_set_importance(port, value);
1977 		break;
1978 	case TIPC_SRC_DROPPABLE:
1979 		if (sock->type != SOCK_STREAM)
1980 			tipc_port_set_unreliable(port, value);
1981 		else
1982 			res = -ENOPROTOOPT;
1983 		break;
1984 	case TIPC_DEST_DROPPABLE:
1985 		tipc_port_set_unreturnable(port, value);
1986 		break;
1987 	case TIPC_CONN_TIMEOUT:
1988 		tipc_sk(sk)->conn_timeout = value;
1989 		/* no need to set "res", since already 0 at this point */
1990 		break;
1991 	default:
1992 		res = -EINVAL;
1993 	}
1994 
1995 	release_sock(sk);
1996 
1997 	return res;
1998 }
1999 
2000 /**
2001  * tipc_getsockopt - get socket option
2002  * @sock: socket structure
2003  * @lvl: option level
2004  * @opt: option identifier
2005  * @ov: receptacle for option value
2006  * @ol: receptacle for length of option value
2007  *
2008  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2009  * (to ease compatibility).
2010  *
2011  * Returns 0 on success, errno otherwise
2012  */
2013 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2014 			   char __user *ov, int __user *ol)
2015 {
2016 	struct sock *sk = sock->sk;
2017 	struct tipc_sock *tsk = tipc_sk(sk);
2018 	struct tipc_port *port = &tsk->port;
2019 	int len;
2020 	u32 value;
2021 	int res;
2022 
2023 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2024 		return put_user(0, ol);
2025 	if (lvl != SOL_TIPC)
2026 		return -ENOPROTOOPT;
2027 	res = get_user(len, ol);
2028 	if (res)
2029 		return res;
2030 
2031 	lock_sock(sk);
2032 
2033 	switch (opt) {
2034 	case TIPC_IMPORTANCE:
2035 		value = tipc_port_importance(port);
2036 		break;
2037 	case TIPC_SRC_DROPPABLE:
2038 		value = tipc_port_unreliable(port);
2039 		break;
2040 	case TIPC_DEST_DROPPABLE:
2041 		value = tipc_port_unreturnable(port);
2042 		break;
2043 	case TIPC_CONN_TIMEOUT:
2044 		value = tipc_sk(sk)->conn_timeout;
2045 		/* no need to set "res", since already 0 at this point */
2046 		break;
2047 	case TIPC_NODE_RECVQ_DEPTH:
2048 		value = 0; /* was tipc_queue_size, now obsolete */
2049 		break;
2050 	case TIPC_SOCK_RECVQ_DEPTH:
2051 		value = skb_queue_len(&sk->sk_receive_queue);
2052 		break;
2053 	default:
2054 		res = -EINVAL;
2055 	}
2056 
2057 	release_sock(sk);
2058 
2059 	if (res)
2060 		return res;	/* "get" failed */
2061 
2062 	if (len < sizeof(value))
2063 		return -EINVAL;
2064 
2065 	if (copy_to_user(ov, &value, sizeof(value)))
2066 		return -EFAULT;
2067 
2068 	return put_user(sizeof(value), ol);
2069 }
2070 
2071 static int tipc_ioctl(struct socket *sk, unsigned int cmd, unsigned long arg)
2072 {
2073 	struct tipc_sioc_ln_req lnr;
2074 	void __user *argp = (void __user *)arg;
2075 
2076 	switch (cmd) {
2077 	case SIOCGETLINKNAME:
2078 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2079 			return -EFAULT;
2080 		if (!tipc_node_get_linkname(lnr.bearer_id, lnr.peer,
2081 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2082 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2083 				return -EFAULT;
2084 			return 0;
2085 		}
2086 		return -EADDRNOTAVAIL;
2087 	default:
2088 		return -ENOIOCTLCMD;
2089 	}
2090 }
2091 
2092 /* Protocol switches for the various types of TIPC sockets */
2093 
2094 static const struct proto_ops msg_ops = {
2095 	.owner		= THIS_MODULE,
2096 	.family		= AF_TIPC,
2097 	.release	= tipc_release,
2098 	.bind		= tipc_bind,
2099 	.connect	= tipc_connect,
2100 	.socketpair	= sock_no_socketpair,
2101 	.accept		= sock_no_accept,
2102 	.getname	= tipc_getname,
2103 	.poll		= tipc_poll,
2104 	.ioctl		= tipc_ioctl,
2105 	.listen		= sock_no_listen,
2106 	.shutdown	= tipc_shutdown,
2107 	.setsockopt	= tipc_setsockopt,
2108 	.getsockopt	= tipc_getsockopt,
2109 	.sendmsg	= tipc_sendmsg,
2110 	.recvmsg	= tipc_recvmsg,
2111 	.mmap		= sock_no_mmap,
2112 	.sendpage	= sock_no_sendpage
2113 };
2114 
2115 static const struct proto_ops packet_ops = {
2116 	.owner		= THIS_MODULE,
2117 	.family		= AF_TIPC,
2118 	.release	= tipc_release,
2119 	.bind		= tipc_bind,
2120 	.connect	= tipc_connect,
2121 	.socketpair	= sock_no_socketpair,
2122 	.accept		= tipc_accept,
2123 	.getname	= tipc_getname,
2124 	.poll		= tipc_poll,
2125 	.ioctl		= tipc_ioctl,
2126 	.listen		= tipc_listen,
2127 	.shutdown	= tipc_shutdown,
2128 	.setsockopt	= tipc_setsockopt,
2129 	.getsockopt	= tipc_getsockopt,
2130 	.sendmsg	= tipc_send_packet,
2131 	.recvmsg	= tipc_recvmsg,
2132 	.mmap		= sock_no_mmap,
2133 	.sendpage	= sock_no_sendpage
2134 };
2135 
2136 static const struct proto_ops stream_ops = {
2137 	.owner		= THIS_MODULE,
2138 	.family		= AF_TIPC,
2139 	.release	= tipc_release,
2140 	.bind		= tipc_bind,
2141 	.connect	= tipc_connect,
2142 	.socketpair	= sock_no_socketpair,
2143 	.accept		= tipc_accept,
2144 	.getname	= tipc_getname,
2145 	.poll		= tipc_poll,
2146 	.ioctl		= tipc_ioctl,
2147 	.listen		= tipc_listen,
2148 	.shutdown	= tipc_shutdown,
2149 	.setsockopt	= tipc_setsockopt,
2150 	.getsockopt	= tipc_getsockopt,
2151 	.sendmsg	= tipc_send_stream,
2152 	.recvmsg	= tipc_recv_stream,
2153 	.mmap		= sock_no_mmap,
2154 	.sendpage	= sock_no_sendpage
2155 };
2156 
2157 static const struct net_proto_family tipc_family_ops = {
2158 	.owner		= THIS_MODULE,
2159 	.family		= AF_TIPC,
2160 	.create		= tipc_sk_create
2161 };
2162 
2163 static struct proto tipc_proto = {
2164 	.name		= "TIPC",
2165 	.owner		= THIS_MODULE,
2166 	.obj_size	= sizeof(struct tipc_sock),
2167 	.sysctl_rmem	= sysctl_tipc_rmem
2168 };
2169 
2170 static struct proto tipc_proto_kern = {
2171 	.name		= "TIPC",
2172 	.obj_size	= sizeof(struct tipc_sock),
2173 	.sysctl_rmem	= sysctl_tipc_rmem
2174 };
2175 
2176 /**
2177  * tipc_socket_init - initialize TIPC socket interface
2178  *
2179  * Returns 0 on success, errno otherwise
2180  */
2181 int tipc_socket_init(void)
2182 {
2183 	int res;
2184 
2185 	res = proto_register(&tipc_proto, 1);
2186 	if (res) {
2187 		pr_err("Failed to register TIPC protocol type\n");
2188 		goto out;
2189 	}
2190 
2191 	res = sock_register(&tipc_family_ops);
2192 	if (res) {
2193 		pr_err("Failed to register TIPC socket type\n");
2194 		proto_unregister(&tipc_proto);
2195 		goto out;
2196 	}
2197  out:
2198 	return res;
2199 }
2200 
2201 /**
2202  * tipc_socket_stop - stop TIPC socket interface
2203  */
2204 void tipc_socket_stop(void)
2205 {
2206 	sock_unregister(tipc_family_ops.family);
2207 	proto_unregister(&tipc_proto);
2208 }
2209