xref: /linux/net/tipc/socket.c (revision 498d319bb512992ef0784c278fa03679f2f5649d)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012 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 
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 CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
47 
48 struct tipc_sock {
49 	struct sock sk;
50 	struct tipc_port *p;
51 	struct tipc_portid peer_name;
52 	unsigned int conn_timeout;
53 };
54 
55 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
56 #define tipc_sk_port(sk) (tipc_sk(sk)->p)
57 
58 #define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
59 			(sock->state == SS_DISCONNECTING))
60 
61 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
62 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
63 static void wakeupdispatch(struct tipc_port *tport);
64 static void tipc_data_ready(struct sock *sk, int len);
65 static void tipc_write_space(struct sock *sk);
66 static int release(struct socket *sock);
67 static int accept(struct socket *sock, struct socket *new_sock, int flags);
68 
69 static const struct proto_ops packet_ops;
70 static const struct proto_ops stream_ops;
71 static const struct proto_ops msg_ops;
72 
73 static struct proto tipc_proto;
74 static struct proto tipc_proto_kern;
75 
76 static int sockets_enabled;
77 
78 /*
79  * Revised TIPC socket locking policy:
80  *
81  * Most socket operations take the standard socket lock when they start
82  * and hold it until they finish (or until they need to sleep).  Acquiring
83  * this lock grants the owner exclusive access to the fields of the socket
84  * data structures, with the exception of the backlog queue.  A few socket
85  * operations can be done without taking the socket lock because they only
86  * read socket information that never changes during the life of the socket.
87  *
88  * Socket operations may acquire the lock for the associated TIPC port if they
89  * need to perform an operation on the port.  If any routine needs to acquire
90  * both the socket lock and the port lock it must take the socket lock first
91  * to avoid the risk of deadlock.
92  *
93  * The dispatcher handling incoming messages cannot grab the socket lock in
94  * the standard fashion, since invoked it runs at the BH level and cannot block.
95  * Instead, it checks to see if the socket lock is currently owned by someone,
96  * and either handles the message itself or adds it to the socket's backlog
97  * queue; in the latter case the queued message is processed once the process
98  * owning the socket lock releases it.
99  *
100  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
101  * the problem of a blocked socket operation preventing any other operations
102  * from occurring.  However, applications must be careful if they have
103  * multiple threads trying to send (or receive) on the same socket, as these
104  * operations might interfere with each other.  For example, doing a connect
105  * and a receive at the same time might allow the receive to consume the
106  * ACK message meant for the connect.  While additional work could be done
107  * to try and overcome this, it doesn't seem to be worthwhile at the present.
108  *
109  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
110  * that another operation that must be performed in a non-blocking manner is
111  * not delayed for very long because the lock has already been taken.
112  *
113  * NOTE: This code assumes that certain fields of a port/socket pair are
114  * constant over its lifetime; such fields can be examined without taking
115  * the socket lock and/or port lock, and do not need to be re-read even
116  * after resuming processing after waiting.  These fields include:
117  *   - socket type
118  *   - pointer to socket sk structure (aka tipc_sock structure)
119  *   - pointer to port structure
120  *   - port reference
121  */
122 
123 /**
124  * advance_rx_queue - discard first buffer in socket receive queue
125  *
126  * Caller must hold socket lock
127  */
128 static void advance_rx_queue(struct sock *sk)
129 {
130 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
131 }
132 
133 /**
134  * reject_rx_queue - reject all buffers in socket receive queue
135  *
136  * Caller must hold socket lock
137  */
138 static void reject_rx_queue(struct sock *sk)
139 {
140 	struct sk_buff *buf;
141 
142 	while ((buf = __skb_dequeue(&sk->sk_receive_queue)))
143 		tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
144 }
145 
146 /**
147  * tipc_sk_create - create a TIPC socket
148  * @net: network namespace (must be default network)
149  * @sock: pre-allocated socket structure
150  * @protocol: protocol indicator (must be 0)
151  * @kern: caused by kernel or by userspace?
152  *
153  * This routine creates additional data structures used by the TIPC socket,
154  * initializes them, and links them together.
155  *
156  * Returns 0 on success, errno otherwise
157  */
158 static int tipc_sk_create(struct net *net, struct socket *sock, int protocol,
159 			  int kern)
160 {
161 	const struct proto_ops *ops;
162 	socket_state state;
163 	struct sock *sk;
164 	struct tipc_port *tp_ptr;
165 
166 	/* Validate arguments */
167 	if (unlikely(protocol != 0))
168 		return -EPROTONOSUPPORT;
169 
170 	switch (sock->type) {
171 	case SOCK_STREAM:
172 		ops = &stream_ops;
173 		state = SS_UNCONNECTED;
174 		break;
175 	case SOCK_SEQPACKET:
176 		ops = &packet_ops;
177 		state = SS_UNCONNECTED;
178 		break;
179 	case SOCK_DGRAM:
180 	case SOCK_RDM:
181 		ops = &msg_ops;
182 		state = SS_READY;
183 		break;
184 	default:
185 		return -EPROTOTYPE;
186 	}
187 
188 	/* Allocate socket's protocol area */
189 	if (!kern)
190 		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
191 	else
192 		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
193 
194 	if (sk == NULL)
195 		return -ENOMEM;
196 
197 	/* Allocate TIPC port for socket to use */
198 	tp_ptr = tipc_createport(sk, &dispatch, &wakeupdispatch,
199 				 TIPC_LOW_IMPORTANCE);
200 	if (unlikely(!tp_ptr)) {
201 		sk_free(sk);
202 		return -ENOMEM;
203 	}
204 
205 	/* Finish initializing socket data structures */
206 	sock->ops = ops;
207 	sock->state = state;
208 
209 	sock_init_data(sock, sk);
210 	sk->sk_backlog_rcv = backlog_rcv;
211 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
212 	sk->sk_data_ready = tipc_data_ready;
213 	sk->sk_write_space = tipc_write_space;
214 	tipc_sk(sk)->p = tp_ptr;
215 	tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT;
216 
217 	spin_unlock_bh(tp_ptr->lock);
218 
219 	if (sock->state == SS_READY) {
220 		tipc_set_portunreturnable(tp_ptr->ref, 1);
221 		if (sock->type == SOCK_DGRAM)
222 			tipc_set_portunreliable(tp_ptr->ref, 1);
223 	}
224 
225 	return 0;
226 }
227 
228 /**
229  * tipc_sock_create_local - create TIPC socket from inside TIPC module
230  * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
231  *
232  * We cannot use sock_creat_kern here because it bumps module user count.
233  * Since socket owner and creator is the same module we must make sure
234  * that module count remains zero for module local sockets, otherwise
235  * we cannot do rmmod.
236  *
237  * Returns 0 on success, errno otherwise
238  */
239 int tipc_sock_create_local(int type, struct socket **res)
240 {
241 	int rc;
242 	struct sock *sk;
243 
244 	rc = sock_create_lite(AF_TIPC, type, 0, res);
245 	if (rc < 0) {
246 		pr_err("Failed to create kernel socket\n");
247 		return rc;
248 	}
249 	tipc_sk_create(&init_net, *res, 0, 1);
250 
251 	sk = (*res)->sk;
252 
253 	return 0;
254 }
255 
256 /**
257  * tipc_sock_release_local - release socket created by tipc_sock_create_local
258  * @sock: the socket to be released.
259  *
260  * Module reference count is not incremented when such sockets are created,
261  * so we must keep it from being decremented when they are released.
262  */
263 void tipc_sock_release_local(struct socket *sock)
264 {
265 	release(sock);
266 	sock->ops = NULL;
267 	sock_release(sock);
268 }
269 
270 /**
271  * tipc_sock_accept_local - accept a connection on a socket created
272  * with tipc_sock_create_local. Use this function to avoid that
273  * module reference count is inadvertently incremented.
274  *
275  * @sock:    the accepting socket
276  * @newsock: reference to the new socket to be created
277  * @flags:   socket flags
278  */
279 
280 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
281 			   int flags)
282 {
283 	struct sock *sk = sock->sk;
284 	int ret;
285 
286 	ret = sock_create_lite(sk->sk_family, sk->sk_type,
287 			       sk->sk_protocol, newsock);
288 	if (ret < 0)
289 		return ret;
290 
291 	ret = accept(sock, *newsock, flags);
292 	if (ret < 0) {
293 		sock_release(*newsock);
294 		return ret;
295 	}
296 	(*newsock)->ops = sock->ops;
297 	return ret;
298 }
299 
300 /**
301  * release - destroy a TIPC socket
302  * @sock: socket to destroy
303  *
304  * This routine cleans up any messages that are still queued on the socket.
305  * For DGRAM and RDM socket types, all queued messages are rejected.
306  * For SEQPACKET and STREAM socket types, the first message is rejected
307  * and any others are discarded.  (If the first message on a STREAM socket
308  * is partially-read, it is discarded and the next one is rejected instead.)
309  *
310  * NOTE: Rejected messages are not necessarily returned to the sender!  They
311  * are returned or discarded according to the "destination droppable" setting
312  * specified for the message by the sender.
313  *
314  * Returns 0 on success, errno otherwise
315  */
316 static int release(struct socket *sock)
317 {
318 	struct sock *sk = sock->sk;
319 	struct tipc_port *tport;
320 	struct sk_buff *buf;
321 	int res;
322 
323 	/*
324 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
325 	 * releases a pre-allocated child socket that was never used)
326 	 */
327 	if (sk == NULL)
328 		return 0;
329 
330 	tport = tipc_sk_port(sk);
331 	lock_sock(sk);
332 
333 	/*
334 	 * Reject all unreceived messages, except on an active connection
335 	 * (which disconnects locally & sends a 'FIN+' to peer)
336 	 */
337 	while (sock->state != SS_DISCONNECTING) {
338 		buf = __skb_dequeue(&sk->sk_receive_queue);
339 		if (buf == NULL)
340 			break;
341 		if (TIPC_SKB_CB(buf)->handle != NULL)
342 			kfree_skb(buf);
343 		else {
344 			if ((sock->state == SS_CONNECTING) ||
345 			    (sock->state == SS_CONNECTED)) {
346 				sock->state = SS_DISCONNECTING;
347 				tipc_disconnect(tport->ref);
348 			}
349 			tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
350 		}
351 	}
352 
353 	/*
354 	 * Delete TIPC port; this ensures no more messages are queued
355 	 * (also disconnects an active connection & sends a 'FIN-' to peer)
356 	 */
357 	res = tipc_deleteport(tport->ref);
358 
359 	/* Discard any remaining (connection-based) messages in receive queue */
360 	__skb_queue_purge(&sk->sk_receive_queue);
361 
362 	/* Reject any messages that accumulated in backlog queue */
363 	sock->state = SS_DISCONNECTING;
364 	release_sock(sk);
365 
366 	sock_put(sk);
367 	sock->sk = NULL;
368 
369 	return res;
370 }
371 
372 /**
373  * bind - associate or disassocate TIPC name(s) with a socket
374  * @sock: socket structure
375  * @uaddr: socket address describing name(s) and desired operation
376  * @uaddr_len: size of socket address data structure
377  *
378  * Name and name sequence binding is indicated using a positive scope value;
379  * a negative scope value unbinds the specified name.  Specifying no name
380  * (i.e. a socket address length of 0) unbinds all names from the socket.
381  *
382  * Returns 0 on success, errno otherwise
383  *
384  * NOTE: This routine doesn't need to take the socket lock since it doesn't
385  *       access any non-constant socket information.
386  */
387 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
388 {
389 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
390 	u32 portref = tipc_sk_port(sock->sk)->ref;
391 
392 	if (unlikely(!uaddr_len))
393 		return tipc_withdraw(portref, 0, NULL);
394 
395 	if (uaddr_len < sizeof(struct sockaddr_tipc))
396 		return -EINVAL;
397 	if (addr->family != AF_TIPC)
398 		return -EAFNOSUPPORT;
399 
400 	if (addr->addrtype == TIPC_ADDR_NAME)
401 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
402 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
403 		return -EAFNOSUPPORT;
404 
405 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
406 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
407 	    (addr->addr.nameseq.type != TIPC_CFG_SRV))
408 		return -EACCES;
409 
410 	return (addr->scope > 0) ?
411 		tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
412 		tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
413 }
414 
415 /**
416  * get_name - get port ID of socket or peer socket
417  * @sock: socket structure
418  * @uaddr: area for returned socket address
419  * @uaddr_len: area for returned length of socket address
420  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
421  *
422  * Returns 0 on success, errno otherwise
423  *
424  * NOTE: This routine doesn't need to take the socket lock since it only
425  *       accesses socket information that is unchanging (or which changes in
426  *       a completely predictable manner).
427  */
428 static int get_name(struct socket *sock, struct sockaddr *uaddr,
429 		    int *uaddr_len, int peer)
430 {
431 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
432 	struct tipc_sock *tsock = tipc_sk(sock->sk);
433 
434 	memset(addr, 0, sizeof(*addr));
435 	if (peer) {
436 		if ((sock->state != SS_CONNECTED) &&
437 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
438 			return -ENOTCONN;
439 		addr->addr.id.ref = tsock->peer_name.ref;
440 		addr->addr.id.node = tsock->peer_name.node;
441 	} else {
442 		addr->addr.id.ref = tsock->p->ref;
443 		addr->addr.id.node = tipc_own_addr;
444 	}
445 
446 	*uaddr_len = sizeof(*addr);
447 	addr->addrtype = TIPC_ADDR_ID;
448 	addr->family = AF_TIPC;
449 	addr->scope = 0;
450 	addr->addr.name.domain = 0;
451 
452 	return 0;
453 }
454 
455 /**
456  * poll - read and possibly block on pollmask
457  * @file: file structure associated with the socket
458  * @sock: socket for which to calculate the poll bits
459  * @wait: ???
460  *
461  * Returns pollmask value
462  *
463  * COMMENTARY:
464  * It appears that the usual socket locking mechanisms are not useful here
465  * since the pollmask info is potentially out-of-date the moment this routine
466  * exits.  TCP and other protocols seem to rely on higher level poll routines
467  * to handle any preventable race conditions, so TIPC will do the same ...
468  *
469  * TIPC sets the returned events as follows:
470  *
471  * socket state		flags set
472  * ------------		---------
473  * unconnected		no read flags
474  *			POLLOUT if port is not congested
475  *
476  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
477  *			no write flags
478  *
479  * connected		POLLIN/POLLRDNORM if data in rx queue
480  *			POLLOUT if port is not congested
481  *
482  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
483  *			no write flags
484  *
485  * listening		POLLIN if SYN in rx queue
486  *			no write flags
487  *
488  * ready		POLLIN/POLLRDNORM if data in rx queue
489  * [connectionless]	POLLOUT (since port cannot be congested)
490  *
491  * IMPORTANT: The fact that a read or write operation is indicated does NOT
492  * imply that the operation will succeed, merely that it should be performed
493  * and will not block.
494  */
495 static unsigned int poll(struct file *file, struct socket *sock,
496 			 poll_table *wait)
497 {
498 	struct sock *sk = sock->sk;
499 	u32 mask = 0;
500 
501 	sock_poll_wait(file, sk_sleep(sk), wait);
502 
503 	switch ((int)sock->state) {
504 	case SS_UNCONNECTED:
505 		if (!tipc_sk_port(sk)->congested)
506 			mask |= POLLOUT;
507 		break;
508 	case SS_READY:
509 	case SS_CONNECTED:
510 		if (!tipc_sk_port(sk)->congested)
511 			mask |= POLLOUT;
512 		/* fall thru' */
513 	case SS_CONNECTING:
514 	case SS_LISTENING:
515 		if (!skb_queue_empty(&sk->sk_receive_queue))
516 			mask |= (POLLIN | POLLRDNORM);
517 		break;
518 	case SS_DISCONNECTING:
519 		mask = (POLLIN | POLLRDNORM | POLLHUP);
520 		break;
521 	}
522 
523 	return mask;
524 }
525 
526 /**
527  * dest_name_check - verify user is permitted to send to specified port name
528  * @dest: destination address
529  * @m: descriptor for message to be sent
530  *
531  * Prevents restricted configuration commands from being issued by
532  * unauthorized users.
533  *
534  * Returns 0 if permission is granted, otherwise errno
535  */
536 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
537 {
538 	struct tipc_cfg_msg_hdr hdr;
539 
540 	if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
541 		return 0;
542 	if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
543 		return 0;
544 	if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
545 		return -EACCES;
546 
547 	if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
548 		return -EMSGSIZE;
549 	if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
550 		return -EFAULT;
551 	if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
552 		return -EACCES;
553 
554 	return 0;
555 }
556 
557 /**
558  * send_msg - send message in connectionless manner
559  * @iocb: if NULL, indicates that socket lock is already held
560  * @sock: socket structure
561  * @m: message to send
562  * @total_len: length of message
563  *
564  * Message must have an destination specified explicitly.
565  * Used for SOCK_RDM and SOCK_DGRAM messages,
566  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
567  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
568  *
569  * Returns the number of bytes sent on success, or errno otherwise
570  */
571 static int send_msg(struct kiocb *iocb, struct socket *sock,
572 		    struct msghdr *m, size_t total_len)
573 {
574 	struct sock *sk = sock->sk;
575 	struct tipc_port *tport = tipc_sk_port(sk);
576 	struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
577 	int needs_conn;
578 	long timeout_val;
579 	int res = -EINVAL;
580 
581 	if (unlikely(!dest))
582 		return -EDESTADDRREQ;
583 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
584 		     (dest->family != AF_TIPC)))
585 		return -EINVAL;
586 	if (total_len > TIPC_MAX_USER_MSG_SIZE)
587 		return -EMSGSIZE;
588 
589 	if (iocb)
590 		lock_sock(sk);
591 
592 	needs_conn = (sock->state != SS_READY);
593 	if (unlikely(needs_conn)) {
594 		if (sock->state == SS_LISTENING) {
595 			res = -EPIPE;
596 			goto exit;
597 		}
598 		if (sock->state != SS_UNCONNECTED) {
599 			res = -EISCONN;
600 			goto exit;
601 		}
602 		if (tport->published) {
603 			res = -EOPNOTSUPP;
604 			goto exit;
605 		}
606 		if (dest->addrtype == TIPC_ADDR_NAME) {
607 			tport->conn_type = dest->addr.name.name.type;
608 			tport->conn_instance = dest->addr.name.name.instance;
609 		}
610 
611 		/* Abort any pending connection attempts (very unlikely) */
612 		reject_rx_queue(sk);
613 	}
614 
615 	timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
616 
617 	do {
618 		if (dest->addrtype == TIPC_ADDR_NAME) {
619 			res = dest_name_check(dest, m);
620 			if (res)
621 				break;
622 			res = tipc_send2name(tport->ref,
623 					     &dest->addr.name.name,
624 					     dest->addr.name.domain,
625 					     m->msg_iov,
626 					     total_len);
627 		} else if (dest->addrtype == TIPC_ADDR_ID) {
628 			res = tipc_send2port(tport->ref,
629 					     &dest->addr.id,
630 					     m->msg_iov,
631 					     total_len);
632 		} else if (dest->addrtype == TIPC_ADDR_MCAST) {
633 			if (needs_conn) {
634 				res = -EOPNOTSUPP;
635 				break;
636 			}
637 			res = dest_name_check(dest, m);
638 			if (res)
639 				break;
640 			res = tipc_multicast(tport->ref,
641 					     &dest->addr.nameseq,
642 					     m->msg_iov,
643 					     total_len);
644 		}
645 		if (likely(res != -ELINKCONG)) {
646 			if (needs_conn && (res >= 0))
647 				sock->state = SS_CONNECTING;
648 			break;
649 		}
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, timeout_val);
657 		lock_sock(sk);
658 	} while (1);
659 
660 exit:
661 	if (iocb)
662 		release_sock(sk);
663 	return res;
664 }
665 
666 /**
667  * send_packet - send a connection-oriented message
668  * @iocb: if NULL, indicates that socket lock is already held
669  * @sock: socket structure
670  * @m: message to send
671  * @total_len: length of message
672  *
673  * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
674  *
675  * Returns the number of bytes sent on success, or errno otherwise
676  */
677 static int send_packet(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 sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
683 	long timeout_val;
684 	int res;
685 
686 	/* Handle implied connection establishment */
687 	if (unlikely(dest))
688 		return send_msg(iocb, sock, m, total_len);
689 
690 	if (total_len > TIPC_MAX_USER_MSG_SIZE)
691 		return -EMSGSIZE;
692 
693 	if (iocb)
694 		lock_sock(sk);
695 
696 	timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
697 
698 	do {
699 		if (unlikely(sock->state != SS_CONNECTED)) {
700 			if (sock->state == SS_DISCONNECTING)
701 				res = -EPIPE;
702 			else
703 				res = -ENOTCONN;
704 			break;
705 		}
706 
707 		res = tipc_send(tport->ref, m->msg_iov, total_len);
708 		if (likely(res != -ELINKCONG))
709 			break;
710 		if (timeout_val <= 0L) {
711 			res = timeout_val ? timeout_val : -EWOULDBLOCK;
712 			break;
713 		}
714 		release_sock(sk);
715 		timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
716 			(!tport->congested || !tport->connected), timeout_val);
717 		lock_sock(sk);
718 	} while (1);
719 
720 	if (iocb)
721 		release_sock(sk);
722 	return res;
723 }
724 
725 /**
726  * send_stream - send stream-oriented data
727  * @iocb: (unused)
728  * @sock: socket structure
729  * @m: data to send
730  * @total_len: total length of data to be sent
731  *
732  * Used for SOCK_STREAM data.
733  *
734  * Returns the number of bytes sent on success (or partial success),
735  * or errno if no data sent
736  */
737 static int send_stream(struct kiocb *iocb, struct socket *sock,
738 		       struct msghdr *m, size_t total_len)
739 {
740 	struct sock *sk = sock->sk;
741 	struct tipc_port *tport = tipc_sk_port(sk);
742 	struct msghdr my_msg;
743 	struct iovec my_iov;
744 	struct iovec *curr_iov;
745 	int curr_iovlen;
746 	char __user *curr_start;
747 	u32 hdr_size;
748 	int curr_left;
749 	int bytes_to_send;
750 	int bytes_sent;
751 	int res;
752 
753 	lock_sock(sk);
754 
755 	/* Handle special cases where there is no connection */
756 	if (unlikely(sock->state != SS_CONNECTED)) {
757 		if (sock->state == SS_UNCONNECTED) {
758 			res = send_packet(NULL, sock, m, total_len);
759 			goto exit;
760 		} else if (sock->state == SS_DISCONNECTING) {
761 			res = -EPIPE;
762 			goto exit;
763 		} else {
764 			res = -ENOTCONN;
765 			goto exit;
766 		}
767 	}
768 
769 	if (unlikely(m->msg_name)) {
770 		res = -EISCONN;
771 		goto exit;
772 	}
773 
774 	if (total_len > (unsigned int)INT_MAX) {
775 		res = -EMSGSIZE;
776 		goto exit;
777 	}
778 
779 	/*
780 	 * Send each iovec entry using one or more messages
781 	 *
782 	 * Note: This algorithm is good for the most likely case
783 	 * (i.e. one large iovec entry), but could be improved to pass sets
784 	 * of small iovec entries into send_packet().
785 	 */
786 	curr_iov = m->msg_iov;
787 	curr_iovlen = m->msg_iovlen;
788 	my_msg.msg_iov = &my_iov;
789 	my_msg.msg_iovlen = 1;
790 	my_msg.msg_flags = m->msg_flags;
791 	my_msg.msg_name = NULL;
792 	bytes_sent = 0;
793 
794 	hdr_size = msg_hdr_sz(&tport->phdr);
795 
796 	while (curr_iovlen--) {
797 		curr_start = curr_iov->iov_base;
798 		curr_left = curr_iov->iov_len;
799 
800 		while (curr_left) {
801 			bytes_to_send = tport->max_pkt - hdr_size;
802 			if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
803 				bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
804 			if (curr_left < bytes_to_send)
805 				bytes_to_send = curr_left;
806 			my_iov.iov_base = curr_start;
807 			my_iov.iov_len = bytes_to_send;
808 			res = send_packet(NULL, sock, &my_msg, bytes_to_send);
809 			if (res < 0) {
810 				if (bytes_sent)
811 					res = bytes_sent;
812 				goto exit;
813 			}
814 			curr_left -= bytes_to_send;
815 			curr_start += bytes_to_send;
816 			bytes_sent += bytes_to_send;
817 		}
818 
819 		curr_iov++;
820 	}
821 	res = bytes_sent;
822 exit:
823 	release_sock(sk);
824 	return res;
825 }
826 
827 /**
828  * auto_connect - complete connection setup to a remote port
829  * @sock: socket structure
830  * @msg: peer's response message
831  *
832  * Returns 0 on success, errno otherwise
833  */
834 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
835 {
836 	struct tipc_sock *tsock = tipc_sk(sock->sk);
837 	struct tipc_port *p_ptr;
838 
839 	tsock->peer_name.ref = msg_origport(msg);
840 	tsock->peer_name.node = msg_orignode(msg);
841 	p_ptr = tipc_port_deref(tsock->p->ref);
842 	if (!p_ptr)
843 		return -EINVAL;
844 
845 	__tipc_connect(tsock->p->ref, p_ptr, &tsock->peer_name);
846 
847 	if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)
848 		return -EINVAL;
849 	msg_set_importance(&p_ptr->phdr, (u32)msg_importance(msg));
850 	sock->state = SS_CONNECTED;
851 	return 0;
852 }
853 
854 /**
855  * set_orig_addr - capture sender's address for received message
856  * @m: descriptor for message info
857  * @msg: received message header
858  *
859  * Note: Address is not captured if not requested by receiver.
860  */
861 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
862 {
863 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
864 
865 	if (addr) {
866 		addr->family = AF_TIPC;
867 		addr->addrtype = TIPC_ADDR_ID;
868 		memset(&addr->addr, 0, sizeof(addr->addr));
869 		addr->addr.id.ref = msg_origport(msg);
870 		addr->addr.id.node = msg_orignode(msg);
871 		addr->addr.name.domain = 0;	/* could leave uninitialized */
872 		addr->scope = 0;		/* could leave uninitialized */
873 		m->msg_namelen = sizeof(struct sockaddr_tipc);
874 	}
875 }
876 
877 /**
878  * anc_data_recv - optionally capture ancillary data for received message
879  * @m: descriptor for message info
880  * @msg: received message header
881  * @tport: TIPC port associated with message
882  *
883  * Note: Ancillary data is not captured if not requested by receiver.
884  *
885  * Returns 0 if successful, otherwise errno
886  */
887 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
888 			 struct tipc_port *tport)
889 {
890 	u32 anc_data[3];
891 	u32 err;
892 	u32 dest_type;
893 	int has_name;
894 	int res;
895 
896 	if (likely(m->msg_controllen == 0))
897 		return 0;
898 
899 	/* Optionally capture errored message object(s) */
900 	err = msg ? msg_errcode(msg) : 0;
901 	if (unlikely(err)) {
902 		anc_data[0] = err;
903 		anc_data[1] = msg_data_sz(msg);
904 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
905 		if (res)
906 			return res;
907 		if (anc_data[1]) {
908 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
909 				       msg_data(msg));
910 			if (res)
911 				return res;
912 		}
913 	}
914 
915 	/* Optionally capture message destination object */
916 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
917 	switch (dest_type) {
918 	case TIPC_NAMED_MSG:
919 		has_name = 1;
920 		anc_data[0] = msg_nametype(msg);
921 		anc_data[1] = msg_namelower(msg);
922 		anc_data[2] = msg_namelower(msg);
923 		break;
924 	case TIPC_MCAST_MSG:
925 		has_name = 1;
926 		anc_data[0] = msg_nametype(msg);
927 		anc_data[1] = msg_namelower(msg);
928 		anc_data[2] = msg_nameupper(msg);
929 		break;
930 	case TIPC_CONN_MSG:
931 		has_name = (tport->conn_type != 0);
932 		anc_data[0] = tport->conn_type;
933 		anc_data[1] = tport->conn_instance;
934 		anc_data[2] = tport->conn_instance;
935 		break;
936 	default:
937 		has_name = 0;
938 	}
939 	if (has_name) {
940 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
941 		if (res)
942 			return res;
943 	}
944 
945 	return 0;
946 }
947 
948 /**
949  * recv_msg - receive packet-oriented message
950  * @iocb: (unused)
951  * @m: descriptor for message info
952  * @buf_len: total size of user buffer area
953  * @flags: receive flags
954  *
955  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
956  * If the complete message doesn't fit in user area, truncate it.
957  *
958  * Returns size of returned message data, errno otherwise
959  */
960 static int recv_msg(struct kiocb *iocb, struct socket *sock,
961 		    struct msghdr *m, size_t buf_len, int flags)
962 {
963 	struct sock *sk = sock->sk;
964 	struct tipc_port *tport = tipc_sk_port(sk);
965 	struct sk_buff *buf;
966 	struct tipc_msg *msg;
967 	long timeout;
968 	unsigned int sz;
969 	u32 err;
970 	int res;
971 
972 	/* Catch invalid receive requests */
973 	if (unlikely(!buf_len))
974 		return -EINVAL;
975 
976 	lock_sock(sk);
977 
978 	if (unlikely(sock->state == SS_UNCONNECTED)) {
979 		res = -ENOTCONN;
980 		goto exit;
981 	}
982 
983 	/* will be updated in set_orig_addr() if needed */
984 	m->msg_namelen = 0;
985 
986 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
987 restart:
988 
989 	/* Look for a message in receive queue; wait if necessary */
990 	while (skb_queue_empty(&sk->sk_receive_queue)) {
991 		if (sock->state == SS_DISCONNECTING) {
992 			res = -ENOTCONN;
993 			goto exit;
994 		}
995 		if (timeout <= 0L) {
996 			res = timeout ? timeout : -EWOULDBLOCK;
997 			goto exit;
998 		}
999 		release_sock(sk);
1000 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1001 							   tipc_rx_ready(sock),
1002 							   timeout);
1003 		lock_sock(sk);
1004 	}
1005 
1006 	/* Look at first message in receive queue */
1007 	buf = skb_peek(&sk->sk_receive_queue);
1008 	msg = buf_msg(buf);
1009 	sz = msg_data_sz(msg);
1010 	err = msg_errcode(msg);
1011 
1012 	/* Discard an empty non-errored message & try again */
1013 	if ((!sz) && (!err)) {
1014 		advance_rx_queue(sk);
1015 		goto restart;
1016 	}
1017 
1018 	/* Capture sender's address (optional) */
1019 	set_orig_addr(m, msg);
1020 
1021 	/* Capture ancillary data (optional) */
1022 	res = anc_data_recv(m, msg, tport);
1023 	if (res)
1024 		goto exit;
1025 
1026 	/* Capture message data (if valid) & compute return value (always) */
1027 	if (!err) {
1028 		if (unlikely(buf_len < sz)) {
1029 			sz = buf_len;
1030 			m->msg_flags |= MSG_TRUNC;
1031 		}
1032 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1033 					      m->msg_iov, sz);
1034 		if (res)
1035 			goto exit;
1036 		res = sz;
1037 	} else {
1038 		if ((sock->state == SS_READY) ||
1039 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1040 			res = 0;
1041 		else
1042 			res = -ECONNRESET;
1043 	}
1044 
1045 	/* Consume received message (optional) */
1046 	if (likely(!(flags & MSG_PEEK))) {
1047 		if ((sock->state != SS_READY) &&
1048 		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1049 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1050 		advance_rx_queue(sk);
1051 	}
1052 exit:
1053 	release_sock(sk);
1054 	return res;
1055 }
1056 
1057 /**
1058  * recv_stream - receive stream-oriented data
1059  * @iocb: (unused)
1060  * @m: descriptor for message info
1061  * @buf_len: total size of user buffer area
1062  * @flags: receive flags
1063  *
1064  * Used for SOCK_STREAM messages only.  If not enough data is available
1065  * will optionally wait for more; never truncates data.
1066  *
1067  * Returns size of returned message data, errno otherwise
1068  */
1069 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1070 		       struct msghdr *m, size_t buf_len, int flags)
1071 {
1072 	struct sock *sk = sock->sk;
1073 	struct tipc_port *tport = tipc_sk_port(sk);
1074 	struct sk_buff *buf;
1075 	struct tipc_msg *msg;
1076 	long timeout;
1077 	unsigned int sz;
1078 	int sz_to_copy, target, needed;
1079 	int sz_copied = 0;
1080 	u32 err;
1081 	int res = 0;
1082 
1083 	/* Catch invalid receive attempts */
1084 	if (unlikely(!buf_len))
1085 		return -EINVAL;
1086 
1087 	lock_sock(sk);
1088 
1089 	if (unlikely((sock->state == SS_UNCONNECTED))) {
1090 		res = -ENOTCONN;
1091 		goto exit;
1092 	}
1093 
1094 	/* will be updated in set_orig_addr() if needed */
1095 	m->msg_namelen = 0;
1096 
1097 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1098 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1099 
1100 restart:
1101 	/* Look for a message in receive queue; wait if necessary */
1102 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1103 		if (sock->state == SS_DISCONNECTING) {
1104 			res = -ENOTCONN;
1105 			goto exit;
1106 		}
1107 		if (timeout <= 0L) {
1108 			res = timeout ? timeout : -EWOULDBLOCK;
1109 			goto exit;
1110 		}
1111 		release_sock(sk);
1112 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1113 							   tipc_rx_ready(sock),
1114 							   timeout);
1115 		lock_sock(sk);
1116 	}
1117 
1118 	/* Look at first message in receive queue */
1119 	buf = skb_peek(&sk->sk_receive_queue);
1120 	msg = buf_msg(buf);
1121 	sz = msg_data_sz(msg);
1122 	err = msg_errcode(msg);
1123 
1124 	/* Discard an empty non-errored message & try again */
1125 	if ((!sz) && (!err)) {
1126 		advance_rx_queue(sk);
1127 		goto restart;
1128 	}
1129 
1130 	/* Optionally capture sender's address & ancillary data of first msg */
1131 	if (sz_copied == 0) {
1132 		set_orig_addr(m, msg);
1133 		res = anc_data_recv(m, msg, tport);
1134 		if (res)
1135 			goto exit;
1136 	}
1137 
1138 	/* Capture message data (if valid) & compute return value (always) */
1139 	if (!err) {
1140 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1141 
1142 		sz -= offset;
1143 		needed = (buf_len - sz_copied);
1144 		sz_to_copy = (sz <= needed) ? sz : needed;
1145 
1146 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1147 					      m->msg_iov, sz_to_copy);
1148 		if (res)
1149 			goto exit;
1150 
1151 		sz_copied += sz_to_copy;
1152 
1153 		if (sz_to_copy < sz) {
1154 			if (!(flags & MSG_PEEK))
1155 				TIPC_SKB_CB(buf)->handle =
1156 				(void *)(unsigned long)(offset + sz_to_copy);
1157 			goto exit;
1158 		}
1159 	} else {
1160 		if (sz_copied != 0)
1161 			goto exit; /* can't add error msg to valid data */
1162 
1163 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1164 			res = 0;
1165 		else
1166 			res = -ECONNRESET;
1167 	}
1168 
1169 	/* Consume received message (optional) */
1170 	if (likely(!(flags & MSG_PEEK))) {
1171 		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1172 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1173 		advance_rx_queue(sk);
1174 	}
1175 
1176 	/* Loop around if more data is required */
1177 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1178 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1179 	    (sz_copied < target)) &&	/* and more is ready or required */
1180 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1181 	    (!err))			/* and haven't reached a FIN */
1182 		goto restart;
1183 
1184 exit:
1185 	release_sock(sk);
1186 	return sz_copied ? sz_copied : res;
1187 }
1188 
1189 /**
1190  * tipc_write_space - wake up thread if port congestion is released
1191  * @sk: socket
1192  */
1193 static void tipc_write_space(struct sock *sk)
1194 {
1195 	struct socket_wq *wq;
1196 
1197 	rcu_read_lock();
1198 	wq = rcu_dereference(sk->sk_wq);
1199 	if (wq_has_sleeper(wq))
1200 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1201 						POLLWRNORM | POLLWRBAND);
1202 	rcu_read_unlock();
1203 }
1204 
1205 /**
1206  * tipc_data_ready - wake up threads to indicate messages have been received
1207  * @sk: socket
1208  * @len: the length of messages
1209  */
1210 static void tipc_data_ready(struct sock *sk, int len)
1211 {
1212 	struct socket_wq *wq;
1213 
1214 	rcu_read_lock();
1215 	wq = rcu_dereference(sk->sk_wq);
1216 	if (wq_has_sleeper(wq))
1217 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1218 						POLLRDNORM | POLLRDBAND);
1219 	rcu_read_unlock();
1220 }
1221 
1222 /**
1223  * filter_connect - Handle all incoming messages for a connection-based socket
1224  * @tsock: TIPC socket
1225  * @msg: message
1226  *
1227  * Returns TIPC error status code and socket error status code
1228  * once it encounters some errors
1229  */
1230 static u32 filter_connect(struct tipc_sock *tsock, struct sk_buff **buf)
1231 {
1232 	struct socket *sock = tsock->sk.sk_socket;
1233 	struct tipc_msg *msg = buf_msg(*buf);
1234 	struct sock *sk = &tsock->sk;
1235 	u32 retval = TIPC_ERR_NO_PORT;
1236 	int res;
1237 
1238 	if (msg_mcast(msg))
1239 		return retval;
1240 
1241 	switch ((int)sock->state) {
1242 	case SS_CONNECTED:
1243 		/* Accept only connection-based messages sent by peer */
1244 		if (msg_connected(msg) && tipc_port_peer_msg(tsock->p, msg)) {
1245 			if (unlikely(msg_errcode(msg))) {
1246 				sock->state = SS_DISCONNECTING;
1247 				__tipc_disconnect(tsock->p);
1248 			}
1249 			retval = TIPC_OK;
1250 		}
1251 		break;
1252 	case SS_CONNECTING:
1253 		/* Accept only ACK or NACK message */
1254 		if (unlikely(msg_errcode(msg))) {
1255 			sock->state = SS_DISCONNECTING;
1256 			sk->sk_err = ECONNREFUSED;
1257 			retval = TIPC_OK;
1258 			break;
1259 		}
1260 
1261 		if (unlikely(!msg_connected(msg)))
1262 			break;
1263 
1264 		res = auto_connect(sock, msg);
1265 		if (res) {
1266 			sock->state = SS_DISCONNECTING;
1267 			sk->sk_err = -res;
1268 			retval = TIPC_OK;
1269 			break;
1270 		}
1271 
1272 		/* If an incoming message is an 'ACK-', it should be
1273 		 * discarded here because it doesn't contain useful
1274 		 * data. In addition, we should try to wake up
1275 		 * connect() routine if sleeping.
1276 		 */
1277 		if (msg_data_sz(msg) == 0) {
1278 			kfree_skb(*buf);
1279 			*buf = NULL;
1280 			if (waitqueue_active(sk_sleep(sk)))
1281 				wake_up_interruptible(sk_sleep(sk));
1282 		}
1283 		retval = TIPC_OK;
1284 		break;
1285 	case SS_LISTENING:
1286 	case SS_UNCONNECTED:
1287 		/* Accept only SYN message */
1288 		if (!msg_connected(msg) && !(msg_errcode(msg)))
1289 			retval = TIPC_OK;
1290 		break;
1291 	case SS_DISCONNECTING:
1292 		break;
1293 	default:
1294 		pr_err("Unknown socket state %u\n", sock->state);
1295 	}
1296 	return retval;
1297 }
1298 
1299 /**
1300  * rcvbuf_limit - get proper overload limit of socket receive queue
1301  * @sk: socket
1302  * @buf: message
1303  *
1304  * For all connection oriented messages, irrespective of importance,
1305  * the default overload value (i.e. 67MB) is set as limit.
1306  *
1307  * For all connectionless messages, by default new queue limits are
1308  * as belows:
1309  *
1310  * TIPC_LOW_IMPORTANCE       (4 MB)
1311  * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1312  * TIPC_HIGH_IMPORTANCE      (16 MB)
1313  * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1314  *
1315  * Returns overload limit according to corresponding message importance
1316  */
1317 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1318 {
1319 	struct tipc_msg *msg = buf_msg(buf);
1320 	unsigned int limit;
1321 
1322 	if (msg_connected(msg))
1323 		limit = sysctl_tipc_rmem[2];
1324 	else
1325 		limit = sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1326 			msg_importance(msg);
1327 	return limit;
1328 }
1329 
1330 /**
1331  * filter_rcv - validate incoming message
1332  * @sk: socket
1333  * @buf: message
1334  *
1335  * Enqueues message on receive queue if acceptable; optionally handles
1336  * disconnect indication for a connected socket.
1337  *
1338  * Called with socket lock already taken; port lock may also be taken.
1339  *
1340  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1341  */
1342 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1343 {
1344 	struct socket *sock = sk->sk_socket;
1345 	struct tipc_msg *msg = buf_msg(buf);
1346 	unsigned int limit = rcvbuf_limit(sk, buf);
1347 	u32 res = TIPC_OK;
1348 
1349 	/* Reject message if it is wrong sort of message for socket */
1350 	if (msg_type(msg) > TIPC_DIRECT_MSG)
1351 		return TIPC_ERR_NO_PORT;
1352 
1353 	if (sock->state == SS_READY) {
1354 		if (msg_connected(msg))
1355 			return TIPC_ERR_NO_PORT;
1356 	} else {
1357 		res = filter_connect(tipc_sk(sk), &buf);
1358 		if (res != TIPC_OK || buf == NULL)
1359 			return res;
1360 	}
1361 
1362 	/* Reject message if there isn't room to queue it */
1363 	if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1364 		return TIPC_ERR_OVERLOAD;
1365 
1366 	/* Enqueue message */
1367 	TIPC_SKB_CB(buf)->handle = NULL;
1368 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1369 	skb_set_owner_r(buf, sk);
1370 
1371 	sk->sk_data_ready(sk, 0);
1372 	return TIPC_OK;
1373 }
1374 
1375 /**
1376  * backlog_rcv - handle incoming message from backlog queue
1377  * @sk: socket
1378  * @buf: message
1379  *
1380  * Caller must hold socket lock, but not port lock.
1381  *
1382  * Returns 0
1383  */
1384 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1385 {
1386 	u32 res;
1387 
1388 	res = filter_rcv(sk, buf);
1389 	if (res)
1390 		tipc_reject_msg(buf, res);
1391 	return 0;
1392 }
1393 
1394 /**
1395  * dispatch - handle incoming message
1396  * @tport: TIPC port that received message
1397  * @buf: message
1398  *
1399  * Called with port lock already taken.
1400  *
1401  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1402  */
1403 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1404 {
1405 	struct sock *sk = tport->sk;
1406 	u32 res;
1407 
1408 	/*
1409 	 * Process message if socket is unlocked; otherwise add to backlog queue
1410 	 *
1411 	 * This code is based on sk_receive_skb(), but must be distinct from it
1412 	 * since a TIPC-specific filter/reject mechanism is utilized
1413 	 */
1414 	bh_lock_sock(sk);
1415 	if (!sock_owned_by_user(sk)) {
1416 		res = filter_rcv(sk, buf);
1417 	} else {
1418 		if (sk_add_backlog(sk, buf, rcvbuf_limit(sk, buf)))
1419 			res = TIPC_ERR_OVERLOAD;
1420 		else
1421 			res = TIPC_OK;
1422 	}
1423 	bh_unlock_sock(sk);
1424 
1425 	return res;
1426 }
1427 
1428 /**
1429  * wakeupdispatch - wake up port after congestion
1430  * @tport: port to wakeup
1431  *
1432  * Called with port lock already taken.
1433  */
1434 static void wakeupdispatch(struct tipc_port *tport)
1435 {
1436 	struct sock *sk = tport->sk;
1437 
1438 	sk->sk_write_space(sk);
1439 }
1440 
1441 /**
1442  * connect - establish a connection to another TIPC port
1443  * @sock: socket structure
1444  * @dest: socket address for destination port
1445  * @destlen: size of socket address data structure
1446  * @flags: file-related flags associated with socket
1447  *
1448  * Returns 0 on success, errno otherwise
1449  */
1450 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1451 		   int flags)
1452 {
1453 	struct sock *sk = sock->sk;
1454 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1455 	struct msghdr m = {NULL,};
1456 	unsigned int timeout;
1457 	int res;
1458 
1459 	lock_sock(sk);
1460 
1461 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1462 	if (sock->state == SS_READY) {
1463 		res = -EOPNOTSUPP;
1464 		goto exit;
1465 	}
1466 
1467 	/*
1468 	 * Reject connection attempt using multicast address
1469 	 *
1470 	 * Note: send_msg() validates the rest of the address fields,
1471 	 *       so there's no need to do it here
1472 	 */
1473 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1474 		res = -EINVAL;
1475 		goto exit;
1476 	}
1477 
1478 	timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1479 
1480 	switch (sock->state) {
1481 	case SS_UNCONNECTED:
1482 		/* Send a 'SYN-' to destination */
1483 		m.msg_name = dest;
1484 		m.msg_namelen = destlen;
1485 
1486 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1487 		 * indicate send_msg() is never blocked.
1488 		 */
1489 		if (!timeout)
1490 			m.msg_flags = MSG_DONTWAIT;
1491 
1492 		res = send_msg(NULL, sock, &m, 0);
1493 		if ((res < 0) && (res != -EWOULDBLOCK))
1494 			goto exit;
1495 
1496 		/* Just entered SS_CONNECTING state; the only
1497 		 * difference is that return value in non-blocking
1498 		 * case is EINPROGRESS, rather than EALREADY.
1499 		 */
1500 		res = -EINPROGRESS;
1501 		break;
1502 	case SS_CONNECTING:
1503 		res = -EALREADY;
1504 		break;
1505 	case SS_CONNECTED:
1506 		res = -EISCONN;
1507 		break;
1508 	default:
1509 		res = -EINVAL;
1510 		goto exit;
1511 	}
1512 
1513 	if (sock->state == SS_CONNECTING) {
1514 		if (!timeout)
1515 			goto exit;
1516 
1517 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1518 		release_sock(sk);
1519 		res = wait_event_interruptible_timeout(*sk_sleep(sk),
1520 				sock->state != SS_CONNECTING,
1521 				timeout ? (long)msecs_to_jiffies(timeout)
1522 					: MAX_SCHEDULE_TIMEOUT);
1523 		lock_sock(sk);
1524 		if (res <= 0) {
1525 			if (res == 0)
1526 				res = -ETIMEDOUT;
1527 			else
1528 				; /* leave "res" unchanged */
1529 			goto exit;
1530 		}
1531 	}
1532 
1533 	if (unlikely(sock->state == SS_DISCONNECTING))
1534 		res = sock_error(sk);
1535 	else
1536 		res = 0;
1537 
1538 exit:
1539 	release_sock(sk);
1540 	return res;
1541 }
1542 
1543 /**
1544  * listen - allow socket to listen for incoming connections
1545  * @sock: socket structure
1546  * @len: (unused)
1547  *
1548  * Returns 0 on success, errno otherwise
1549  */
1550 static int listen(struct socket *sock, int len)
1551 {
1552 	struct sock *sk = sock->sk;
1553 	int res;
1554 
1555 	lock_sock(sk);
1556 
1557 	if (sock->state != SS_UNCONNECTED)
1558 		res = -EINVAL;
1559 	else {
1560 		sock->state = SS_LISTENING;
1561 		res = 0;
1562 	}
1563 
1564 	release_sock(sk);
1565 	return res;
1566 }
1567 
1568 /**
1569  * accept - wait for connection request
1570  * @sock: listening socket
1571  * @newsock: new socket that is to be connected
1572  * @flags: file-related flags associated with socket
1573  *
1574  * Returns 0 on success, errno otherwise
1575  */
1576 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1577 {
1578 	struct sock *new_sk, *sk = sock->sk;
1579 	struct sk_buff *buf;
1580 	struct tipc_sock *new_tsock;
1581 	struct tipc_port *new_tport;
1582 	struct tipc_msg *msg;
1583 	u32 new_ref;
1584 
1585 	int res;
1586 
1587 	lock_sock(sk);
1588 
1589 	if (sock->state != SS_LISTENING) {
1590 		res = -EINVAL;
1591 		goto exit;
1592 	}
1593 
1594 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1595 		if (flags & O_NONBLOCK) {
1596 			res = -EWOULDBLOCK;
1597 			goto exit;
1598 		}
1599 		release_sock(sk);
1600 		res = wait_event_interruptible(*sk_sleep(sk),
1601 				(!skb_queue_empty(&sk->sk_receive_queue)));
1602 		lock_sock(sk);
1603 		if (res)
1604 			goto exit;
1605 	}
1606 
1607 	buf = skb_peek(&sk->sk_receive_queue);
1608 
1609 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1610 	if (res)
1611 		goto exit;
1612 
1613 	new_sk = new_sock->sk;
1614 	new_tsock = tipc_sk(new_sk);
1615 	new_tport = new_tsock->p;
1616 	new_ref = new_tport->ref;
1617 	msg = buf_msg(buf);
1618 
1619 	/* we lock on new_sk; but lockdep sees the lock on sk */
1620 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1621 
1622 	/*
1623 	 * Reject any stray messages received by new socket
1624 	 * before the socket lock was taken (very, very unlikely)
1625 	 */
1626 	reject_rx_queue(new_sk);
1627 
1628 	/* Connect new socket to it's peer */
1629 	new_tsock->peer_name.ref = msg_origport(msg);
1630 	new_tsock->peer_name.node = msg_orignode(msg);
1631 	tipc_connect(new_ref, &new_tsock->peer_name);
1632 	new_sock->state = SS_CONNECTED;
1633 
1634 	tipc_set_portimportance(new_ref, msg_importance(msg));
1635 	if (msg_named(msg)) {
1636 		new_tport->conn_type = msg_nametype(msg);
1637 		new_tport->conn_instance = msg_nameinst(msg);
1638 	}
1639 
1640 	/*
1641 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1642 	 * Respond to 'SYN+' by queuing it on new socket.
1643 	 */
1644 	if (!msg_data_sz(msg)) {
1645 		struct msghdr m = {NULL,};
1646 
1647 		advance_rx_queue(sk);
1648 		send_packet(NULL, new_sock, &m, 0);
1649 	} else {
1650 		__skb_dequeue(&sk->sk_receive_queue);
1651 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
1652 		skb_set_owner_r(buf, new_sk);
1653 	}
1654 	release_sock(new_sk);
1655 
1656 exit:
1657 	release_sock(sk);
1658 	return res;
1659 }
1660 
1661 /**
1662  * shutdown - shutdown socket connection
1663  * @sock: socket structure
1664  * @how: direction to close (must be SHUT_RDWR)
1665  *
1666  * Terminates connection (if necessary), then purges socket's receive queue.
1667  *
1668  * Returns 0 on success, errno otherwise
1669  */
1670 static int shutdown(struct socket *sock, int how)
1671 {
1672 	struct sock *sk = sock->sk;
1673 	struct tipc_port *tport = tipc_sk_port(sk);
1674 	struct sk_buff *buf;
1675 	int res;
1676 
1677 	if (how != SHUT_RDWR)
1678 		return -EINVAL;
1679 
1680 	lock_sock(sk);
1681 
1682 	switch (sock->state) {
1683 	case SS_CONNECTING:
1684 	case SS_CONNECTED:
1685 
1686 restart:
1687 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1688 		buf = __skb_dequeue(&sk->sk_receive_queue);
1689 		if (buf) {
1690 			if (TIPC_SKB_CB(buf)->handle != NULL) {
1691 				kfree_skb(buf);
1692 				goto restart;
1693 			}
1694 			tipc_disconnect(tport->ref);
1695 			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1696 		} else {
1697 			tipc_shutdown(tport->ref);
1698 		}
1699 
1700 		sock->state = SS_DISCONNECTING;
1701 
1702 		/* fall through */
1703 
1704 	case SS_DISCONNECTING:
1705 
1706 		/* Discard any unreceived messages */
1707 		__skb_queue_purge(&sk->sk_receive_queue);
1708 
1709 		/* Wake up anyone sleeping in poll */
1710 		sk->sk_state_change(sk);
1711 		res = 0;
1712 		break;
1713 
1714 	default:
1715 		res = -ENOTCONN;
1716 	}
1717 
1718 	release_sock(sk);
1719 	return res;
1720 }
1721 
1722 /**
1723  * setsockopt - set socket option
1724  * @sock: socket structure
1725  * @lvl: option level
1726  * @opt: option identifier
1727  * @ov: pointer to new option value
1728  * @ol: length of option value
1729  *
1730  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1731  * (to ease compatibility).
1732  *
1733  * Returns 0 on success, errno otherwise
1734  */
1735 static int setsockopt(struct socket *sock, int lvl, int opt, char __user *ov,
1736 		      unsigned int ol)
1737 {
1738 	struct sock *sk = sock->sk;
1739 	struct tipc_port *tport = tipc_sk_port(sk);
1740 	u32 value;
1741 	int res;
1742 
1743 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1744 		return 0;
1745 	if (lvl != SOL_TIPC)
1746 		return -ENOPROTOOPT;
1747 	if (ol < sizeof(value))
1748 		return -EINVAL;
1749 	res = get_user(value, (u32 __user *)ov);
1750 	if (res)
1751 		return res;
1752 
1753 	lock_sock(sk);
1754 
1755 	switch (opt) {
1756 	case TIPC_IMPORTANCE:
1757 		res = tipc_set_portimportance(tport->ref, value);
1758 		break;
1759 	case TIPC_SRC_DROPPABLE:
1760 		if (sock->type != SOCK_STREAM)
1761 			res = tipc_set_portunreliable(tport->ref, value);
1762 		else
1763 			res = -ENOPROTOOPT;
1764 		break;
1765 	case TIPC_DEST_DROPPABLE:
1766 		res = tipc_set_portunreturnable(tport->ref, value);
1767 		break;
1768 	case TIPC_CONN_TIMEOUT:
1769 		tipc_sk(sk)->conn_timeout = value;
1770 		/* no need to set "res", since already 0 at this point */
1771 		break;
1772 	default:
1773 		res = -EINVAL;
1774 	}
1775 
1776 	release_sock(sk);
1777 
1778 	return res;
1779 }
1780 
1781 /**
1782  * getsockopt - get socket option
1783  * @sock: socket structure
1784  * @lvl: option level
1785  * @opt: option identifier
1786  * @ov: receptacle for option value
1787  * @ol: receptacle for length of option value
1788  *
1789  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1790  * (to ease compatibility).
1791  *
1792  * Returns 0 on success, errno otherwise
1793  */
1794 static int getsockopt(struct socket *sock, int lvl, int opt, char __user *ov,
1795 		      int __user *ol)
1796 {
1797 	struct sock *sk = sock->sk;
1798 	struct tipc_port *tport = tipc_sk_port(sk);
1799 	int len;
1800 	u32 value;
1801 	int res;
1802 
1803 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1804 		return put_user(0, ol);
1805 	if (lvl != SOL_TIPC)
1806 		return -ENOPROTOOPT;
1807 	res = get_user(len, ol);
1808 	if (res)
1809 		return res;
1810 
1811 	lock_sock(sk);
1812 
1813 	switch (opt) {
1814 	case TIPC_IMPORTANCE:
1815 		res = tipc_portimportance(tport->ref, &value);
1816 		break;
1817 	case TIPC_SRC_DROPPABLE:
1818 		res = tipc_portunreliable(tport->ref, &value);
1819 		break;
1820 	case TIPC_DEST_DROPPABLE:
1821 		res = tipc_portunreturnable(tport->ref, &value);
1822 		break;
1823 	case TIPC_CONN_TIMEOUT:
1824 		value = tipc_sk(sk)->conn_timeout;
1825 		/* no need to set "res", since already 0 at this point */
1826 		break;
1827 	case TIPC_NODE_RECVQ_DEPTH:
1828 		value = 0; /* was tipc_queue_size, now obsolete */
1829 		break;
1830 	case TIPC_SOCK_RECVQ_DEPTH:
1831 		value = skb_queue_len(&sk->sk_receive_queue);
1832 		break;
1833 	default:
1834 		res = -EINVAL;
1835 	}
1836 
1837 	release_sock(sk);
1838 
1839 	if (res)
1840 		return res;	/* "get" failed */
1841 
1842 	if (len < sizeof(value))
1843 		return -EINVAL;
1844 
1845 	if (copy_to_user(ov, &value, sizeof(value)))
1846 		return -EFAULT;
1847 
1848 	return put_user(sizeof(value), ol);
1849 }
1850 
1851 /* Protocol switches for the various types of TIPC sockets */
1852 
1853 static const struct proto_ops msg_ops = {
1854 	.owner		= THIS_MODULE,
1855 	.family		= AF_TIPC,
1856 	.release	= release,
1857 	.bind		= bind,
1858 	.connect	= connect,
1859 	.socketpair	= sock_no_socketpair,
1860 	.accept		= sock_no_accept,
1861 	.getname	= get_name,
1862 	.poll		= poll,
1863 	.ioctl		= sock_no_ioctl,
1864 	.listen		= sock_no_listen,
1865 	.shutdown	= shutdown,
1866 	.setsockopt	= setsockopt,
1867 	.getsockopt	= getsockopt,
1868 	.sendmsg	= send_msg,
1869 	.recvmsg	= recv_msg,
1870 	.mmap		= sock_no_mmap,
1871 	.sendpage	= sock_no_sendpage
1872 };
1873 
1874 static const struct proto_ops packet_ops = {
1875 	.owner		= THIS_MODULE,
1876 	.family		= AF_TIPC,
1877 	.release	= release,
1878 	.bind		= bind,
1879 	.connect	= connect,
1880 	.socketpair	= sock_no_socketpair,
1881 	.accept		= accept,
1882 	.getname	= get_name,
1883 	.poll		= poll,
1884 	.ioctl		= sock_no_ioctl,
1885 	.listen		= listen,
1886 	.shutdown	= shutdown,
1887 	.setsockopt	= setsockopt,
1888 	.getsockopt	= getsockopt,
1889 	.sendmsg	= send_packet,
1890 	.recvmsg	= recv_msg,
1891 	.mmap		= sock_no_mmap,
1892 	.sendpage	= sock_no_sendpage
1893 };
1894 
1895 static const struct proto_ops stream_ops = {
1896 	.owner		= THIS_MODULE,
1897 	.family		= AF_TIPC,
1898 	.release	= release,
1899 	.bind		= bind,
1900 	.connect	= connect,
1901 	.socketpair	= sock_no_socketpair,
1902 	.accept		= accept,
1903 	.getname	= get_name,
1904 	.poll		= poll,
1905 	.ioctl		= sock_no_ioctl,
1906 	.listen		= listen,
1907 	.shutdown	= shutdown,
1908 	.setsockopt	= setsockopt,
1909 	.getsockopt	= getsockopt,
1910 	.sendmsg	= send_stream,
1911 	.recvmsg	= recv_stream,
1912 	.mmap		= sock_no_mmap,
1913 	.sendpage	= sock_no_sendpage
1914 };
1915 
1916 static const struct net_proto_family tipc_family_ops = {
1917 	.owner		= THIS_MODULE,
1918 	.family		= AF_TIPC,
1919 	.create		= tipc_sk_create
1920 };
1921 
1922 static struct proto tipc_proto = {
1923 	.name		= "TIPC",
1924 	.owner		= THIS_MODULE,
1925 	.obj_size	= sizeof(struct tipc_sock),
1926 	.sysctl_rmem	= sysctl_tipc_rmem
1927 };
1928 
1929 static struct proto tipc_proto_kern = {
1930 	.name		= "TIPC",
1931 	.obj_size	= sizeof(struct tipc_sock),
1932 	.sysctl_rmem	= sysctl_tipc_rmem
1933 };
1934 
1935 /**
1936  * tipc_socket_init - initialize TIPC socket interface
1937  *
1938  * Returns 0 on success, errno otherwise
1939  */
1940 int tipc_socket_init(void)
1941 {
1942 	int res;
1943 
1944 	res = proto_register(&tipc_proto, 1);
1945 	if (res) {
1946 		pr_err("Failed to register TIPC protocol type\n");
1947 		goto out;
1948 	}
1949 
1950 	res = sock_register(&tipc_family_ops);
1951 	if (res) {
1952 		pr_err("Failed to register TIPC socket type\n");
1953 		proto_unregister(&tipc_proto);
1954 		goto out;
1955 	}
1956 
1957 	sockets_enabled = 1;
1958  out:
1959 	return res;
1960 }
1961 
1962 /**
1963  * tipc_socket_stop - stop TIPC socket interface
1964  */
1965 void tipc_socket_stop(void)
1966 {
1967 	if (!sockets_enabled)
1968 		return;
1969 
1970 	sockets_enabled = 0;
1971 	sock_unregister(tipc_family_ops.family);
1972 	proto_unregister(&tipc_proto);
1973 }
1974