xref: /openbmc/linux/net/tipc/socket.c (revision 12eb4683)
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 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
984 restart:
985 
986 	/* Look for a message in receive queue; wait if necessary */
987 	while (skb_queue_empty(&sk->sk_receive_queue)) {
988 		if (sock->state == SS_DISCONNECTING) {
989 			res = -ENOTCONN;
990 			goto exit;
991 		}
992 		if (timeout <= 0L) {
993 			res = timeout ? timeout : -EWOULDBLOCK;
994 			goto exit;
995 		}
996 		release_sock(sk);
997 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
998 							   tipc_rx_ready(sock),
999 							   timeout);
1000 		lock_sock(sk);
1001 	}
1002 
1003 	/* Look at first message in receive queue */
1004 	buf = skb_peek(&sk->sk_receive_queue);
1005 	msg = buf_msg(buf);
1006 	sz = msg_data_sz(msg);
1007 	err = msg_errcode(msg);
1008 
1009 	/* Discard an empty non-errored message & try again */
1010 	if ((!sz) && (!err)) {
1011 		advance_rx_queue(sk);
1012 		goto restart;
1013 	}
1014 
1015 	/* Capture sender's address (optional) */
1016 	set_orig_addr(m, msg);
1017 
1018 	/* Capture ancillary data (optional) */
1019 	res = anc_data_recv(m, msg, tport);
1020 	if (res)
1021 		goto exit;
1022 
1023 	/* Capture message data (if valid) & compute return value (always) */
1024 	if (!err) {
1025 		if (unlikely(buf_len < sz)) {
1026 			sz = buf_len;
1027 			m->msg_flags |= MSG_TRUNC;
1028 		}
1029 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1030 					      m->msg_iov, sz);
1031 		if (res)
1032 			goto exit;
1033 		res = sz;
1034 	} else {
1035 		if ((sock->state == SS_READY) ||
1036 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1037 			res = 0;
1038 		else
1039 			res = -ECONNRESET;
1040 	}
1041 
1042 	/* Consume received message (optional) */
1043 	if (likely(!(flags & MSG_PEEK))) {
1044 		if ((sock->state != SS_READY) &&
1045 		    (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1046 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1047 		advance_rx_queue(sk);
1048 	}
1049 exit:
1050 	release_sock(sk);
1051 	return res;
1052 }
1053 
1054 /**
1055  * recv_stream - receive stream-oriented data
1056  * @iocb: (unused)
1057  * @m: descriptor for message info
1058  * @buf_len: total size of user buffer area
1059  * @flags: receive flags
1060  *
1061  * Used for SOCK_STREAM messages only.  If not enough data is available
1062  * will optionally wait for more; never truncates data.
1063  *
1064  * Returns size of returned message data, errno otherwise
1065  */
1066 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1067 		       struct msghdr *m, size_t buf_len, int flags)
1068 {
1069 	struct sock *sk = sock->sk;
1070 	struct tipc_port *tport = tipc_sk_port(sk);
1071 	struct sk_buff *buf;
1072 	struct tipc_msg *msg;
1073 	long timeout;
1074 	unsigned int sz;
1075 	int sz_to_copy, target, needed;
1076 	int sz_copied = 0;
1077 	u32 err;
1078 	int res = 0;
1079 
1080 	/* Catch invalid receive attempts */
1081 	if (unlikely(!buf_len))
1082 		return -EINVAL;
1083 
1084 	lock_sock(sk);
1085 
1086 	if (unlikely((sock->state == SS_UNCONNECTED))) {
1087 		res = -ENOTCONN;
1088 		goto exit;
1089 	}
1090 
1091 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1092 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1093 
1094 restart:
1095 	/* Look for a message in receive queue; wait if necessary */
1096 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1097 		if (sock->state == SS_DISCONNECTING) {
1098 			res = -ENOTCONN;
1099 			goto exit;
1100 		}
1101 		if (timeout <= 0L) {
1102 			res = timeout ? timeout : -EWOULDBLOCK;
1103 			goto exit;
1104 		}
1105 		release_sock(sk);
1106 		timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1107 							   tipc_rx_ready(sock),
1108 							   timeout);
1109 		lock_sock(sk);
1110 	}
1111 
1112 	/* Look at first message in receive queue */
1113 	buf = skb_peek(&sk->sk_receive_queue);
1114 	msg = buf_msg(buf);
1115 	sz = msg_data_sz(msg);
1116 	err = msg_errcode(msg);
1117 
1118 	/* Discard an empty non-errored message & try again */
1119 	if ((!sz) && (!err)) {
1120 		advance_rx_queue(sk);
1121 		goto restart;
1122 	}
1123 
1124 	/* Optionally capture sender's address & ancillary data of first msg */
1125 	if (sz_copied == 0) {
1126 		set_orig_addr(m, msg);
1127 		res = anc_data_recv(m, msg, tport);
1128 		if (res)
1129 			goto exit;
1130 	}
1131 
1132 	/* Capture message data (if valid) & compute return value (always) */
1133 	if (!err) {
1134 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1135 
1136 		sz -= offset;
1137 		needed = (buf_len - sz_copied);
1138 		sz_to_copy = (sz <= needed) ? sz : needed;
1139 
1140 		res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1141 					      m->msg_iov, sz_to_copy);
1142 		if (res)
1143 			goto exit;
1144 
1145 		sz_copied += sz_to_copy;
1146 
1147 		if (sz_to_copy < sz) {
1148 			if (!(flags & MSG_PEEK))
1149 				TIPC_SKB_CB(buf)->handle =
1150 				(void *)(unsigned long)(offset + sz_to_copy);
1151 			goto exit;
1152 		}
1153 	} else {
1154 		if (sz_copied != 0)
1155 			goto exit; /* can't add error msg to valid data */
1156 
1157 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1158 			res = 0;
1159 		else
1160 			res = -ECONNRESET;
1161 	}
1162 
1163 	/* Consume received message (optional) */
1164 	if (likely(!(flags & MSG_PEEK))) {
1165 		if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1166 			tipc_acknowledge(tport->ref, tport->conn_unacked);
1167 		advance_rx_queue(sk);
1168 	}
1169 
1170 	/* Loop around if more data is required */
1171 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1172 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1173 	    (sz_copied < target)) &&	/* and more is ready or required */
1174 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1175 	    (!err))			/* and haven't reached a FIN */
1176 		goto restart;
1177 
1178 exit:
1179 	release_sock(sk);
1180 	return sz_copied ? sz_copied : res;
1181 }
1182 
1183 /**
1184  * tipc_write_space - wake up thread if port congestion is released
1185  * @sk: socket
1186  */
1187 static void tipc_write_space(struct sock *sk)
1188 {
1189 	struct socket_wq *wq;
1190 
1191 	rcu_read_lock();
1192 	wq = rcu_dereference(sk->sk_wq);
1193 	if (wq_has_sleeper(wq))
1194 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1195 						POLLWRNORM | POLLWRBAND);
1196 	rcu_read_unlock();
1197 }
1198 
1199 /**
1200  * tipc_data_ready - wake up threads to indicate messages have been received
1201  * @sk: socket
1202  * @len: the length of messages
1203  */
1204 static void tipc_data_ready(struct sock *sk, int len)
1205 {
1206 	struct socket_wq *wq;
1207 
1208 	rcu_read_lock();
1209 	wq = rcu_dereference(sk->sk_wq);
1210 	if (wq_has_sleeper(wq))
1211 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1212 						POLLRDNORM | POLLRDBAND);
1213 	rcu_read_unlock();
1214 }
1215 
1216 /**
1217  * filter_connect - Handle all incoming messages for a connection-based socket
1218  * @tsock: TIPC socket
1219  * @msg: message
1220  *
1221  * Returns TIPC error status code and socket error status code
1222  * once it encounters some errors
1223  */
1224 static u32 filter_connect(struct tipc_sock *tsock, struct sk_buff **buf)
1225 {
1226 	struct socket *sock = tsock->sk.sk_socket;
1227 	struct tipc_msg *msg = buf_msg(*buf);
1228 	struct sock *sk = &tsock->sk;
1229 	u32 retval = TIPC_ERR_NO_PORT;
1230 	int res;
1231 
1232 	if (msg_mcast(msg))
1233 		return retval;
1234 
1235 	switch ((int)sock->state) {
1236 	case SS_CONNECTED:
1237 		/* Accept only connection-based messages sent by peer */
1238 		if (msg_connected(msg) && tipc_port_peer_msg(tsock->p, msg)) {
1239 			if (unlikely(msg_errcode(msg))) {
1240 				sock->state = SS_DISCONNECTING;
1241 				__tipc_disconnect(tsock->p);
1242 			}
1243 			retval = TIPC_OK;
1244 		}
1245 		break;
1246 	case SS_CONNECTING:
1247 		/* Accept only ACK or NACK message */
1248 		if (unlikely(msg_errcode(msg))) {
1249 			sock->state = SS_DISCONNECTING;
1250 			sk->sk_err = ECONNREFUSED;
1251 			retval = TIPC_OK;
1252 			break;
1253 		}
1254 
1255 		if (unlikely(!msg_connected(msg)))
1256 			break;
1257 
1258 		res = auto_connect(sock, msg);
1259 		if (res) {
1260 			sock->state = SS_DISCONNECTING;
1261 			sk->sk_err = -res;
1262 			retval = TIPC_OK;
1263 			break;
1264 		}
1265 
1266 		/* If an incoming message is an 'ACK-', it should be
1267 		 * discarded here because it doesn't contain useful
1268 		 * data. In addition, we should try to wake up
1269 		 * connect() routine if sleeping.
1270 		 */
1271 		if (msg_data_sz(msg) == 0) {
1272 			kfree_skb(*buf);
1273 			*buf = NULL;
1274 			if (waitqueue_active(sk_sleep(sk)))
1275 				wake_up_interruptible(sk_sleep(sk));
1276 		}
1277 		retval = TIPC_OK;
1278 		break;
1279 	case SS_LISTENING:
1280 	case SS_UNCONNECTED:
1281 		/* Accept only SYN message */
1282 		if (!msg_connected(msg) && !(msg_errcode(msg)))
1283 			retval = TIPC_OK;
1284 		break;
1285 	case SS_DISCONNECTING:
1286 		break;
1287 	default:
1288 		pr_err("Unknown socket state %u\n", sock->state);
1289 	}
1290 	return retval;
1291 }
1292 
1293 /**
1294  * rcvbuf_limit - get proper overload limit of socket receive queue
1295  * @sk: socket
1296  * @buf: message
1297  *
1298  * For all connection oriented messages, irrespective of importance,
1299  * the default overload value (i.e. 67MB) is set as limit.
1300  *
1301  * For all connectionless messages, by default new queue limits are
1302  * as belows:
1303  *
1304  * TIPC_LOW_IMPORTANCE       (4 MB)
1305  * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1306  * TIPC_HIGH_IMPORTANCE      (16 MB)
1307  * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1308  *
1309  * Returns overload limit according to corresponding message importance
1310  */
1311 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1312 {
1313 	struct tipc_msg *msg = buf_msg(buf);
1314 	unsigned int limit;
1315 
1316 	if (msg_connected(msg))
1317 		limit = sysctl_tipc_rmem[2];
1318 	else
1319 		limit = sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1320 			msg_importance(msg);
1321 	return limit;
1322 }
1323 
1324 /**
1325  * filter_rcv - validate incoming message
1326  * @sk: socket
1327  * @buf: message
1328  *
1329  * Enqueues message on receive queue if acceptable; optionally handles
1330  * disconnect indication for a connected socket.
1331  *
1332  * Called with socket lock already taken; port lock may also be taken.
1333  *
1334  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1335  */
1336 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1337 {
1338 	struct socket *sock = sk->sk_socket;
1339 	struct tipc_msg *msg = buf_msg(buf);
1340 	unsigned int limit = rcvbuf_limit(sk, buf);
1341 	u32 res = TIPC_OK;
1342 
1343 	/* Reject message if it is wrong sort of message for socket */
1344 	if (msg_type(msg) > TIPC_DIRECT_MSG)
1345 		return TIPC_ERR_NO_PORT;
1346 
1347 	if (sock->state == SS_READY) {
1348 		if (msg_connected(msg))
1349 			return TIPC_ERR_NO_PORT;
1350 	} else {
1351 		res = filter_connect(tipc_sk(sk), &buf);
1352 		if (res != TIPC_OK || buf == NULL)
1353 			return res;
1354 	}
1355 
1356 	/* Reject message if there isn't room to queue it */
1357 	if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1358 		return TIPC_ERR_OVERLOAD;
1359 
1360 	/* Enqueue message */
1361 	TIPC_SKB_CB(buf)->handle = NULL;
1362 	__skb_queue_tail(&sk->sk_receive_queue, buf);
1363 	skb_set_owner_r(buf, sk);
1364 
1365 	sk->sk_data_ready(sk, 0);
1366 	return TIPC_OK;
1367 }
1368 
1369 /**
1370  * backlog_rcv - handle incoming message from backlog queue
1371  * @sk: socket
1372  * @buf: message
1373  *
1374  * Caller must hold socket lock, but not port lock.
1375  *
1376  * Returns 0
1377  */
1378 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1379 {
1380 	u32 res;
1381 
1382 	res = filter_rcv(sk, buf);
1383 	if (res)
1384 		tipc_reject_msg(buf, res);
1385 	return 0;
1386 }
1387 
1388 /**
1389  * dispatch - handle incoming message
1390  * @tport: TIPC port that received message
1391  * @buf: message
1392  *
1393  * Called with port lock already taken.
1394  *
1395  * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1396  */
1397 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1398 {
1399 	struct sock *sk = tport->sk;
1400 	u32 res;
1401 
1402 	/*
1403 	 * Process message if socket is unlocked; otherwise add to backlog queue
1404 	 *
1405 	 * This code is based on sk_receive_skb(), but must be distinct from it
1406 	 * since a TIPC-specific filter/reject mechanism is utilized
1407 	 */
1408 	bh_lock_sock(sk);
1409 	if (!sock_owned_by_user(sk)) {
1410 		res = filter_rcv(sk, buf);
1411 	} else {
1412 		if (sk_add_backlog(sk, buf, rcvbuf_limit(sk, buf)))
1413 			res = TIPC_ERR_OVERLOAD;
1414 		else
1415 			res = TIPC_OK;
1416 	}
1417 	bh_unlock_sock(sk);
1418 
1419 	return res;
1420 }
1421 
1422 /**
1423  * wakeupdispatch - wake up port after congestion
1424  * @tport: port to wakeup
1425  *
1426  * Called with port lock already taken.
1427  */
1428 static void wakeupdispatch(struct tipc_port *tport)
1429 {
1430 	struct sock *sk = tport->sk;
1431 
1432 	sk->sk_write_space(sk);
1433 }
1434 
1435 /**
1436  * connect - establish a connection to another TIPC port
1437  * @sock: socket structure
1438  * @dest: socket address for destination port
1439  * @destlen: size of socket address data structure
1440  * @flags: file-related flags associated with socket
1441  *
1442  * Returns 0 on success, errno otherwise
1443  */
1444 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1445 		   int flags)
1446 {
1447 	struct sock *sk = sock->sk;
1448 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1449 	struct msghdr m = {NULL,};
1450 	unsigned int timeout;
1451 	int res;
1452 
1453 	lock_sock(sk);
1454 
1455 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1456 	if (sock->state == SS_READY) {
1457 		res = -EOPNOTSUPP;
1458 		goto exit;
1459 	}
1460 
1461 	/*
1462 	 * Reject connection attempt using multicast address
1463 	 *
1464 	 * Note: send_msg() validates the rest of the address fields,
1465 	 *       so there's no need to do it here
1466 	 */
1467 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1468 		res = -EINVAL;
1469 		goto exit;
1470 	}
1471 
1472 	timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1473 
1474 	switch (sock->state) {
1475 	case SS_UNCONNECTED:
1476 		/* Send a 'SYN-' to destination */
1477 		m.msg_name = dest;
1478 		m.msg_namelen = destlen;
1479 
1480 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1481 		 * indicate send_msg() is never blocked.
1482 		 */
1483 		if (!timeout)
1484 			m.msg_flags = MSG_DONTWAIT;
1485 
1486 		res = send_msg(NULL, sock, &m, 0);
1487 		if ((res < 0) && (res != -EWOULDBLOCK))
1488 			goto exit;
1489 
1490 		/* Just entered SS_CONNECTING state; the only
1491 		 * difference is that return value in non-blocking
1492 		 * case is EINPROGRESS, rather than EALREADY.
1493 		 */
1494 		res = -EINPROGRESS;
1495 		break;
1496 	case SS_CONNECTING:
1497 		res = -EALREADY;
1498 		break;
1499 	case SS_CONNECTED:
1500 		res = -EISCONN;
1501 		break;
1502 	default:
1503 		res = -EINVAL;
1504 		goto exit;
1505 	}
1506 
1507 	if (sock->state == SS_CONNECTING) {
1508 		if (!timeout)
1509 			goto exit;
1510 
1511 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1512 		release_sock(sk);
1513 		res = wait_event_interruptible_timeout(*sk_sleep(sk),
1514 				sock->state != SS_CONNECTING,
1515 				timeout ? (long)msecs_to_jiffies(timeout)
1516 					: MAX_SCHEDULE_TIMEOUT);
1517 		lock_sock(sk);
1518 		if (res <= 0) {
1519 			if (res == 0)
1520 				res = -ETIMEDOUT;
1521 			else
1522 				; /* leave "res" unchanged */
1523 			goto exit;
1524 		}
1525 	}
1526 
1527 	if (unlikely(sock->state == SS_DISCONNECTING))
1528 		res = sock_error(sk);
1529 	else
1530 		res = 0;
1531 
1532 exit:
1533 	release_sock(sk);
1534 	return res;
1535 }
1536 
1537 /**
1538  * listen - allow socket to listen for incoming connections
1539  * @sock: socket structure
1540  * @len: (unused)
1541  *
1542  * Returns 0 on success, errno otherwise
1543  */
1544 static int listen(struct socket *sock, int len)
1545 {
1546 	struct sock *sk = sock->sk;
1547 	int res;
1548 
1549 	lock_sock(sk);
1550 
1551 	if (sock->state != SS_UNCONNECTED)
1552 		res = -EINVAL;
1553 	else {
1554 		sock->state = SS_LISTENING;
1555 		res = 0;
1556 	}
1557 
1558 	release_sock(sk);
1559 	return res;
1560 }
1561 
1562 /**
1563  * accept - wait for connection request
1564  * @sock: listening socket
1565  * @newsock: new socket that is to be connected
1566  * @flags: file-related flags associated with socket
1567  *
1568  * Returns 0 on success, errno otherwise
1569  */
1570 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1571 {
1572 	struct sock *new_sk, *sk = sock->sk;
1573 	struct sk_buff *buf;
1574 	struct tipc_sock *new_tsock;
1575 	struct tipc_port *new_tport;
1576 	struct tipc_msg *msg;
1577 	u32 new_ref;
1578 
1579 	int res;
1580 
1581 	lock_sock(sk);
1582 
1583 	if (sock->state != SS_LISTENING) {
1584 		res = -EINVAL;
1585 		goto exit;
1586 	}
1587 
1588 	while (skb_queue_empty(&sk->sk_receive_queue)) {
1589 		if (flags & O_NONBLOCK) {
1590 			res = -EWOULDBLOCK;
1591 			goto exit;
1592 		}
1593 		release_sock(sk);
1594 		res = wait_event_interruptible(*sk_sleep(sk),
1595 				(!skb_queue_empty(&sk->sk_receive_queue)));
1596 		lock_sock(sk);
1597 		if (res)
1598 			goto exit;
1599 	}
1600 
1601 	buf = skb_peek(&sk->sk_receive_queue);
1602 
1603 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1604 	if (res)
1605 		goto exit;
1606 
1607 	new_sk = new_sock->sk;
1608 	new_tsock = tipc_sk(new_sk);
1609 	new_tport = new_tsock->p;
1610 	new_ref = new_tport->ref;
1611 	msg = buf_msg(buf);
1612 
1613 	/* we lock on new_sk; but lockdep sees the lock on sk */
1614 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1615 
1616 	/*
1617 	 * Reject any stray messages received by new socket
1618 	 * before the socket lock was taken (very, very unlikely)
1619 	 */
1620 	reject_rx_queue(new_sk);
1621 
1622 	/* Connect new socket to it's peer */
1623 	new_tsock->peer_name.ref = msg_origport(msg);
1624 	new_tsock->peer_name.node = msg_orignode(msg);
1625 	tipc_connect(new_ref, &new_tsock->peer_name);
1626 	new_sock->state = SS_CONNECTED;
1627 
1628 	tipc_set_portimportance(new_ref, msg_importance(msg));
1629 	if (msg_named(msg)) {
1630 		new_tport->conn_type = msg_nametype(msg);
1631 		new_tport->conn_instance = msg_nameinst(msg);
1632 	}
1633 
1634 	/*
1635 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1636 	 * Respond to 'SYN+' by queuing it on new socket.
1637 	 */
1638 	if (!msg_data_sz(msg)) {
1639 		struct msghdr m = {NULL,};
1640 
1641 		advance_rx_queue(sk);
1642 		send_packet(NULL, new_sock, &m, 0);
1643 	} else {
1644 		__skb_dequeue(&sk->sk_receive_queue);
1645 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
1646 		skb_set_owner_r(buf, new_sk);
1647 	}
1648 	release_sock(new_sk);
1649 
1650 exit:
1651 	release_sock(sk);
1652 	return res;
1653 }
1654 
1655 /**
1656  * shutdown - shutdown socket connection
1657  * @sock: socket structure
1658  * @how: direction to close (must be SHUT_RDWR)
1659  *
1660  * Terminates connection (if necessary), then purges socket's receive queue.
1661  *
1662  * Returns 0 on success, errno otherwise
1663  */
1664 static int shutdown(struct socket *sock, int how)
1665 {
1666 	struct sock *sk = sock->sk;
1667 	struct tipc_port *tport = tipc_sk_port(sk);
1668 	struct sk_buff *buf;
1669 	int res;
1670 
1671 	if (how != SHUT_RDWR)
1672 		return -EINVAL;
1673 
1674 	lock_sock(sk);
1675 
1676 	switch (sock->state) {
1677 	case SS_CONNECTING:
1678 	case SS_CONNECTED:
1679 
1680 restart:
1681 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1682 		buf = __skb_dequeue(&sk->sk_receive_queue);
1683 		if (buf) {
1684 			if (TIPC_SKB_CB(buf)->handle != NULL) {
1685 				kfree_skb(buf);
1686 				goto restart;
1687 			}
1688 			tipc_disconnect(tport->ref);
1689 			tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1690 		} else {
1691 			tipc_shutdown(tport->ref);
1692 		}
1693 
1694 		sock->state = SS_DISCONNECTING;
1695 
1696 		/* fall through */
1697 
1698 	case SS_DISCONNECTING:
1699 
1700 		/* Discard any unreceived messages */
1701 		__skb_queue_purge(&sk->sk_receive_queue);
1702 
1703 		/* Wake up anyone sleeping in poll */
1704 		sk->sk_state_change(sk);
1705 		res = 0;
1706 		break;
1707 
1708 	default:
1709 		res = -ENOTCONN;
1710 	}
1711 
1712 	release_sock(sk);
1713 	return res;
1714 }
1715 
1716 /**
1717  * setsockopt - set socket option
1718  * @sock: socket structure
1719  * @lvl: option level
1720  * @opt: option identifier
1721  * @ov: pointer to new option value
1722  * @ol: length of option value
1723  *
1724  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1725  * (to ease compatibility).
1726  *
1727  * Returns 0 on success, errno otherwise
1728  */
1729 static int setsockopt(struct socket *sock, int lvl, int opt, char __user *ov,
1730 		      unsigned int ol)
1731 {
1732 	struct sock *sk = sock->sk;
1733 	struct tipc_port *tport = tipc_sk_port(sk);
1734 	u32 value;
1735 	int res;
1736 
1737 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1738 		return 0;
1739 	if (lvl != SOL_TIPC)
1740 		return -ENOPROTOOPT;
1741 	if (ol < sizeof(value))
1742 		return -EINVAL;
1743 	res = get_user(value, (u32 __user *)ov);
1744 	if (res)
1745 		return res;
1746 
1747 	lock_sock(sk);
1748 
1749 	switch (opt) {
1750 	case TIPC_IMPORTANCE:
1751 		res = tipc_set_portimportance(tport->ref, value);
1752 		break;
1753 	case TIPC_SRC_DROPPABLE:
1754 		if (sock->type != SOCK_STREAM)
1755 			res = tipc_set_portunreliable(tport->ref, value);
1756 		else
1757 			res = -ENOPROTOOPT;
1758 		break;
1759 	case TIPC_DEST_DROPPABLE:
1760 		res = tipc_set_portunreturnable(tport->ref, value);
1761 		break;
1762 	case TIPC_CONN_TIMEOUT:
1763 		tipc_sk(sk)->conn_timeout = value;
1764 		/* no need to set "res", since already 0 at this point */
1765 		break;
1766 	default:
1767 		res = -EINVAL;
1768 	}
1769 
1770 	release_sock(sk);
1771 
1772 	return res;
1773 }
1774 
1775 /**
1776  * getsockopt - get socket option
1777  * @sock: socket structure
1778  * @lvl: option level
1779  * @opt: option identifier
1780  * @ov: receptacle for option value
1781  * @ol: receptacle for length of option value
1782  *
1783  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1784  * (to ease compatibility).
1785  *
1786  * Returns 0 on success, errno otherwise
1787  */
1788 static int getsockopt(struct socket *sock, int lvl, int opt, char __user *ov,
1789 		      int __user *ol)
1790 {
1791 	struct sock *sk = sock->sk;
1792 	struct tipc_port *tport = tipc_sk_port(sk);
1793 	int len;
1794 	u32 value;
1795 	int res;
1796 
1797 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1798 		return put_user(0, ol);
1799 	if (lvl != SOL_TIPC)
1800 		return -ENOPROTOOPT;
1801 	res = get_user(len, ol);
1802 	if (res)
1803 		return res;
1804 
1805 	lock_sock(sk);
1806 
1807 	switch (opt) {
1808 	case TIPC_IMPORTANCE:
1809 		res = tipc_portimportance(tport->ref, &value);
1810 		break;
1811 	case TIPC_SRC_DROPPABLE:
1812 		res = tipc_portunreliable(tport->ref, &value);
1813 		break;
1814 	case TIPC_DEST_DROPPABLE:
1815 		res = tipc_portunreturnable(tport->ref, &value);
1816 		break;
1817 	case TIPC_CONN_TIMEOUT:
1818 		value = tipc_sk(sk)->conn_timeout;
1819 		/* no need to set "res", since already 0 at this point */
1820 		break;
1821 	case TIPC_NODE_RECVQ_DEPTH:
1822 		value = 0; /* was tipc_queue_size, now obsolete */
1823 		break;
1824 	case TIPC_SOCK_RECVQ_DEPTH:
1825 		value = skb_queue_len(&sk->sk_receive_queue);
1826 		break;
1827 	default:
1828 		res = -EINVAL;
1829 	}
1830 
1831 	release_sock(sk);
1832 
1833 	if (res)
1834 		return res;	/* "get" failed */
1835 
1836 	if (len < sizeof(value))
1837 		return -EINVAL;
1838 
1839 	if (copy_to_user(ov, &value, sizeof(value)))
1840 		return -EFAULT;
1841 
1842 	return put_user(sizeof(value), ol);
1843 }
1844 
1845 /* Protocol switches for the various types of TIPC sockets */
1846 
1847 static const struct proto_ops msg_ops = {
1848 	.owner		= THIS_MODULE,
1849 	.family		= AF_TIPC,
1850 	.release	= release,
1851 	.bind		= bind,
1852 	.connect	= connect,
1853 	.socketpair	= sock_no_socketpair,
1854 	.accept		= sock_no_accept,
1855 	.getname	= get_name,
1856 	.poll		= poll,
1857 	.ioctl		= sock_no_ioctl,
1858 	.listen		= sock_no_listen,
1859 	.shutdown	= shutdown,
1860 	.setsockopt	= setsockopt,
1861 	.getsockopt	= getsockopt,
1862 	.sendmsg	= send_msg,
1863 	.recvmsg	= recv_msg,
1864 	.mmap		= sock_no_mmap,
1865 	.sendpage	= sock_no_sendpage
1866 };
1867 
1868 static const struct proto_ops packet_ops = {
1869 	.owner		= THIS_MODULE,
1870 	.family		= AF_TIPC,
1871 	.release	= release,
1872 	.bind		= bind,
1873 	.connect	= connect,
1874 	.socketpair	= sock_no_socketpair,
1875 	.accept		= accept,
1876 	.getname	= get_name,
1877 	.poll		= poll,
1878 	.ioctl		= sock_no_ioctl,
1879 	.listen		= listen,
1880 	.shutdown	= shutdown,
1881 	.setsockopt	= setsockopt,
1882 	.getsockopt	= getsockopt,
1883 	.sendmsg	= send_packet,
1884 	.recvmsg	= recv_msg,
1885 	.mmap		= sock_no_mmap,
1886 	.sendpage	= sock_no_sendpage
1887 };
1888 
1889 static const struct proto_ops stream_ops = {
1890 	.owner		= THIS_MODULE,
1891 	.family		= AF_TIPC,
1892 	.release	= release,
1893 	.bind		= bind,
1894 	.connect	= connect,
1895 	.socketpair	= sock_no_socketpair,
1896 	.accept		= accept,
1897 	.getname	= get_name,
1898 	.poll		= poll,
1899 	.ioctl		= sock_no_ioctl,
1900 	.listen		= listen,
1901 	.shutdown	= shutdown,
1902 	.setsockopt	= setsockopt,
1903 	.getsockopt	= getsockopt,
1904 	.sendmsg	= send_stream,
1905 	.recvmsg	= recv_stream,
1906 	.mmap		= sock_no_mmap,
1907 	.sendpage	= sock_no_sendpage
1908 };
1909 
1910 static const struct net_proto_family tipc_family_ops = {
1911 	.owner		= THIS_MODULE,
1912 	.family		= AF_TIPC,
1913 	.create		= tipc_sk_create
1914 };
1915 
1916 static struct proto tipc_proto = {
1917 	.name		= "TIPC",
1918 	.owner		= THIS_MODULE,
1919 	.obj_size	= sizeof(struct tipc_sock),
1920 	.sysctl_rmem	= sysctl_tipc_rmem
1921 };
1922 
1923 static struct proto tipc_proto_kern = {
1924 	.name		= "TIPC",
1925 	.obj_size	= sizeof(struct tipc_sock),
1926 	.sysctl_rmem	= sysctl_tipc_rmem
1927 };
1928 
1929 /**
1930  * tipc_socket_init - initialize TIPC socket interface
1931  *
1932  * Returns 0 on success, errno otherwise
1933  */
1934 int tipc_socket_init(void)
1935 {
1936 	int res;
1937 
1938 	res = proto_register(&tipc_proto, 1);
1939 	if (res) {
1940 		pr_err("Failed to register TIPC protocol type\n");
1941 		goto out;
1942 	}
1943 
1944 	res = sock_register(&tipc_family_ops);
1945 	if (res) {
1946 		pr_err("Failed to register TIPC socket type\n");
1947 		proto_unregister(&tipc_proto);
1948 		goto out;
1949 	}
1950 
1951 	sockets_enabled = 1;
1952  out:
1953 	return res;
1954 }
1955 
1956 /**
1957  * tipc_socket_stop - stop TIPC socket interface
1958  */
1959 void tipc_socket_stop(void)
1960 {
1961 	if (!sockets_enabled)
1962 		return;
1963 
1964 	sockets_enabled = 0;
1965 	sock_unregister(tipc_family_ops.family);
1966 	proto_unregister(&tipc_proto);
1967 }
1968