xref: /linux/net/tipc/socket.c (revision 3ce095c16263630dde46d6051854073edaacf3d7)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2015, 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 <linux/rhashtable.h>
38 #include "core.h"
39 #include "name_table.h"
40 #include "node.h"
41 #include "link.h"
42 #include "name_distr.h"
43 #include "socket.h"
44 
45 #define SS_LISTENING		-1	/* socket is listening */
46 #define SS_READY		-2	/* socket is connectionless */
47 
48 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
49 #define CONN_PROBING_INTERVAL	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
50 #define TIPC_FWD_MSG		1
51 #define TIPC_CONN_OK		0
52 #define TIPC_CONN_PROBING	1
53 #define TIPC_MAX_PORT		0xffffffff
54 #define TIPC_MIN_PORT		1
55 
56 /**
57  * struct tipc_sock - TIPC socket structure
58  * @sk: socket - interacts with 'port' and with user via the socket API
59  * @connected: non-zero if port is currently connected to a peer port
60  * @conn_type: TIPC type used when connection was established
61  * @conn_instance: TIPC instance used when connection was established
62  * @published: non-zero if port has one or more associated names
63  * @max_pkt: maximum packet size "hint" used when building messages sent by port
64  * @portid: unique port identity in TIPC socket hash table
65  * @phdr: preformatted message header used when sending messages
66  * @port_list: adjacent ports in TIPC's global list of ports
67  * @publications: list of publications for port
68  * @pub_count: total # of publications port has made during its lifetime
69  * @probing_state:
70  * @probing_intv:
71  * @conn_timeout: the time we can wait for an unresponded setup request
72  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
73  * @link_cong: non-zero if owner must sleep because of link congestion
74  * @sent_unacked: # messages sent by socket, and not yet acked by peer
75  * @rcv_unacked: # messages read by user, but not yet acked back to peer
76  * @remote: 'connected' peer for dgram/rdm
77  * @node: hash table node
78  * @rcu: rcu struct for tipc_sock
79  */
80 struct tipc_sock {
81 	struct sock sk;
82 	int connected;
83 	u32 conn_type;
84 	u32 conn_instance;
85 	int published;
86 	u32 max_pkt;
87 	u32 portid;
88 	struct tipc_msg phdr;
89 	struct list_head sock_list;
90 	struct list_head publications;
91 	u32 pub_count;
92 	u32 probing_state;
93 	unsigned long probing_intv;
94 	uint conn_timeout;
95 	atomic_t dupl_rcvcnt;
96 	bool link_cong;
97 	uint sent_unacked;
98 	uint rcv_unacked;
99 	struct sockaddr_tipc remote;
100 	struct rhash_head node;
101 	struct rcu_head rcu;
102 };
103 
104 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
105 static void tipc_data_ready(struct sock *sk);
106 static void tipc_write_space(struct sock *sk);
107 static int tipc_release(struct socket *sock);
108 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
109 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
110 static void tipc_sk_timeout(unsigned long data);
111 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
112 			   struct tipc_name_seq const *seq);
113 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
114 			    struct tipc_name_seq const *seq);
115 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
116 static int tipc_sk_insert(struct tipc_sock *tsk);
117 static void tipc_sk_remove(struct tipc_sock *tsk);
118 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
119 			      size_t dsz);
120 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
121 
122 static const struct proto_ops packet_ops;
123 static const struct proto_ops stream_ops;
124 static const struct proto_ops msg_ops;
125 static struct proto tipc_proto;
126 
127 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
128 	[TIPC_NLA_SOCK_UNSPEC]		= { .type = NLA_UNSPEC },
129 	[TIPC_NLA_SOCK_ADDR]		= { .type = NLA_U32 },
130 	[TIPC_NLA_SOCK_REF]		= { .type = NLA_U32 },
131 	[TIPC_NLA_SOCK_CON]		= { .type = NLA_NESTED },
132 	[TIPC_NLA_SOCK_HAS_PUBL]	= { .type = NLA_FLAG }
133 };
134 
135 static const struct rhashtable_params tsk_rht_params;
136 
137 /*
138  * Revised TIPC socket locking policy:
139  *
140  * Most socket operations take the standard socket lock when they start
141  * and hold it until they finish (or until they need to sleep).  Acquiring
142  * this lock grants the owner exclusive access to the fields of the socket
143  * data structures, with the exception of the backlog queue.  A few socket
144  * operations can be done without taking the socket lock because they only
145  * read socket information that never changes during the life of the socket.
146  *
147  * Socket operations may acquire the lock for the associated TIPC port if they
148  * need to perform an operation on the port.  If any routine needs to acquire
149  * both the socket lock and the port lock it must take the socket lock first
150  * to avoid the risk of deadlock.
151  *
152  * The dispatcher handling incoming messages cannot grab the socket lock in
153  * the standard fashion, since invoked it runs at the BH level and cannot block.
154  * Instead, it checks to see if the socket lock is currently owned by someone,
155  * and either handles the message itself or adds it to the socket's backlog
156  * queue; in the latter case the queued message is processed once the process
157  * owning the socket lock releases it.
158  *
159  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
160  * the problem of a blocked socket operation preventing any other operations
161  * from occurring.  However, applications must be careful if they have
162  * multiple threads trying to send (or receive) on the same socket, as these
163  * operations might interfere with each other.  For example, doing a connect
164  * and a receive at the same time might allow the receive to consume the
165  * ACK message meant for the connect.  While additional work could be done
166  * to try and overcome this, it doesn't seem to be worthwhile at the present.
167  *
168  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
169  * that another operation that must be performed in a non-blocking manner is
170  * not delayed for very long because the lock has already been taken.
171  *
172  * NOTE: This code assumes that certain fields of a port/socket pair are
173  * constant over its lifetime; such fields can be examined without taking
174  * the socket lock and/or port lock, and do not need to be re-read even
175  * after resuming processing after waiting.  These fields include:
176  *   - socket type
177  *   - pointer to socket sk structure (aka tipc_sock structure)
178  *   - pointer to port structure
179  *   - port reference
180  */
181 
182 static u32 tsk_own_node(struct tipc_sock *tsk)
183 {
184 	return msg_prevnode(&tsk->phdr);
185 }
186 
187 static u32 tsk_peer_node(struct tipc_sock *tsk)
188 {
189 	return msg_destnode(&tsk->phdr);
190 }
191 
192 static u32 tsk_peer_port(struct tipc_sock *tsk)
193 {
194 	return msg_destport(&tsk->phdr);
195 }
196 
197 static  bool tsk_unreliable(struct tipc_sock *tsk)
198 {
199 	return msg_src_droppable(&tsk->phdr) != 0;
200 }
201 
202 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
203 {
204 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
205 }
206 
207 static bool tsk_unreturnable(struct tipc_sock *tsk)
208 {
209 	return msg_dest_droppable(&tsk->phdr) != 0;
210 }
211 
212 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
213 {
214 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
215 }
216 
217 static int tsk_importance(struct tipc_sock *tsk)
218 {
219 	return msg_importance(&tsk->phdr);
220 }
221 
222 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
223 {
224 	if (imp > TIPC_CRITICAL_IMPORTANCE)
225 		return -EINVAL;
226 	msg_set_importance(&tsk->phdr, (u32)imp);
227 	return 0;
228 }
229 
230 static struct tipc_sock *tipc_sk(const struct sock *sk)
231 {
232 	return container_of(sk, struct tipc_sock, sk);
233 }
234 
235 static int tsk_conn_cong(struct tipc_sock *tsk)
236 {
237 	return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
238 }
239 
240 /**
241  * tsk_advance_rx_queue - discard first buffer in socket receive queue
242  *
243  * Caller must hold socket lock
244  */
245 static void tsk_advance_rx_queue(struct sock *sk)
246 {
247 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
248 }
249 
250 /**
251  * tsk_rej_rx_queue - reject all buffers in socket receive queue
252  *
253  * Caller must hold socket lock
254  */
255 static void tsk_rej_rx_queue(struct sock *sk)
256 {
257 	struct sk_buff *skb;
258 	u32 dnode;
259 	u32 own_node = tsk_own_node(tipc_sk(sk));
260 
261 	while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
262 		if (tipc_msg_reverse(own_node, skb, &dnode, TIPC_ERR_NO_PORT))
263 			tipc_link_xmit_skb(sock_net(sk), skb, dnode, 0);
264 	}
265 }
266 
267 /* tsk_peer_msg - verify if message was sent by connected port's peer
268  *
269  * Handles cases where the node's network address has changed from
270  * the default of <0.0.0> to its configured setting.
271  */
272 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
273 {
274 	struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
275 	u32 peer_port = tsk_peer_port(tsk);
276 	u32 orig_node;
277 	u32 peer_node;
278 
279 	if (unlikely(!tsk->connected))
280 		return false;
281 
282 	if (unlikely(msg_origport(msg) != peer_port))
283 		return false;
284 
285 	orig_node = msg_orignode(msg);
286 	peer_node = tsk_peer_node(tsk);
287 
288 	if (likely(orig_node == peer_node))
289 		return true;
290 
291 	if (!orig_node && (peer_node == tn->own_addr))
292 		return true;
293 
294 	if (!peer_node && (orig_node == tn->own_addr))
295 		return true;
296 
297 	return false;
298 }
299 
300 /**
301  * tipc_sk_create - create a TIPC socket
302  * @net: network namespace (must be default network)
303  * @sock: pre-allocated socket structure
304  * @protocol: protocol indicator (must be 0)
305  * @kern: caused by kernel or by userspace?
306  *
307  * This routine creates additional data structures used by the TIPC socket,
308  * initializes them, and links them together.
309  *
310  * Returns 0 on success, errno otherwise
311  */
312 static int tipc_sk_create(struct net *net, struct socket *sock,
313 			  int protocol, int kern)
314 {
315 	struct tipc_net *tn;
316 	const struct proto_ops *ops;
317 	socket_state state;
318 	struct sock *sk;
319 	struct tipc_sock *tsk;
320 	struct tipc_msg *msg;
321 
322 	/* Validate arguments */
323 	if (unlikely(protocol != 0))
324 		return -EPROTONOSUPPORT;
325 
326 	switch (sock->type) {
327 	case SOCK_STREAM:
328 		ops = &stream_ops;
329 		state = SS_UNCONNECTED;
330 		break;
331 	case SOCK_SEQPACKET:
332 		ops = &packet_ops;
333 		state = SS_UNCONNECTED;
334 		break;
335 	case SOCK_DGRAM:
336 	case SOCK_RDM:
337 		ops = &msg_ops;
338 		state = SS_READY;
339 		break;
340 	default:
341 		return -EPROTOTYPE;
342 	}
343 
344 	/* Allocate socket's protocol area */
345 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
346 	if (sk == NULL)
347 		return -ENOMEM;
348 
349 	tsk = tipc_sk(sk);
350 	tsk->max_pkt = MAX_PKT_DEFAULT;
351 	INIT_LIST_HEAD(&tsk->publications);
352 	msg = &tsk->phdr;
353 	tn = net_generic(sock_net(sk), tipc_net_id);
354 	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
355 		      NAMED_H_SIZE, 0);
356 
357 	/* Finish initializing socket data structures */
358 	sock->ops = ops;
359 	sock->state = state;
360 	sock_init_data(sock, sk);
361 	if (tipc_sk_insert(tsk)) {
362 		pr_warn("Socket create failed; port numbrer exhausted\n");
363 		return -EINVAL;
364 	}
365 	msg_set_origport(msg, tsk->portid);
366 	setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
367 	sk->sk_backlog_rcv = tipc_backlog_rcv;
368 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
369 	sk->sk_data_ready = tipc_data_ready;
370 	sk->sk_write_space = tipc_write_space;
371 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
372 	tsk->sent_unacked = 0;
373 	atomic_set(&tsk->dupl_rcvcnt, 0);
374 
375 	if (sock->state == SS_READY) {
376 		tsk_set_unreturnable(tsk, true);
377 		if (sock->type == SOCK_DGRAM)
378 			tsk_set_unreliable(tsk, true);
379 	}
380 	return 0;
381 }
382 
383 static void tipc_sk_callback(struct rcu_head *head)
384 {
385 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
386 
387 	sock_put(&tsk->sk);
388 }
389 
390 /**
391  * tipc_release - destroy a TIPC socket
392  * @sock: socket to destroy
393  *
394  * This routine cleans up any messages that are still queued on the socket.
395  * For DGRAM and RDM socket types, all queued messages are rejected.
396  * For SEQPACKET and STREAM socket types, the first message is rejected
397  * and any others are discarded.  (If the first message on a STREAM socket
398  * is partially-read, it is discarded and the next one is rejected instead.)
399  *
400  * NOTE: Rejected messages are not necessarily returned to the sender!  They
401  * are returned or discarded according to the "destination droppable" setting
402  * specified for the message by the sender.
403  *
404  * Returns 0 on success, errno otherwise
405  */
406 static int tipc_release(struct socket *sock)
407 {
408 	struct sock *sk = sock->sk;
409 	struct net *net;
410 	struct tipc_sock *tsk;
411 	struct sk_buff *skb;
412 	u32 dnode, probing_state;
413 
414 	/*
415 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
416 	 * releases a pre-allocated child socket that was never used)
417 	 */
418 	if (sk == NULL)
419 		return 0;
420 
421 	net = sock_net(sk);
422 	tsk = tipc_sk(sk);
423 	lock_sock(sk);
424 
425 	/*
426 	 * Reject all unreceived messages, except on an active connection
427 	 * (which disconnects locally & sends a 'FIN+' to peer)
428 	 */
429 	dnode = tsk_peer_node(tsk);
430 	while (sock->state != SS_DISCONNECTING) {
431 		skb = __skb_dequeue(&sk->sk_receive_queue);
432 		if (skb == NULL)
433 			break;
434 		if (TIPC_SKB_CB(skb)->handle != NULL)
435 			kfree_skb(skb);
436 		else {
437 			if ((sock->state == SS_CONNECTING) ||
438 			    (sock->state == SS_CONNECTED)) {
439 				sock->state = SS_DISCONNECTING;
440 				tsk->connected = 0;
441 				tipc_node_remove_conn(net, dnode, tsk->portid);
442 			}
443 			if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
444 					     TIPC_ERR_NO_PORT))
445 				tipc_link_xmit_skb(net, skb, dnode, 0);
446 		}
447 	}
448 
449 	tipc_sk_withdraw(tsk, 0, NULL);
450 	probing_state = tsk->probing_state;
451 	if (del_timer_sync(&sk->sk_timer) &&
452 	    probing_state != TIPC_CONN_PROBING)
453 		sock_put(sk);
454 	tipc_sk_remove(tsk);
455 	if (tsk->connected) {
456 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
457 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
458 				      tsk_own_node(tsk), tsk_peer_port(tsk),
459 				      tsk->portid, TIPC_ERR_NO_PORT);
460 		if (skb)
461 			tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
462 		tipc_node_remove_conn(net, dnode, tsk->portid);
463 	}
464 
465 	/* Discard any remaining (connection-based) messages in receive queue */
466 	__skb_queue_purge(&sk->sk_receive_queue);
467 
468 	/* Reject any messages that accumulated in backlog queue */
469 	sock->state = SS_DISCONNECTING;
470 	release_sock(sk);
471 
472 	call_rcu(&tsk->rcu, tipc_sk_callback);
473 	sock->sk = NULL;
474 
475 	return 0;
476 }
477 
478 /**
479  * tipc_bind - associate or disassocate TIPC name(s) with a socket
480  * @sock: socket structure
481  * @uaddr: socket address describing name(s) and desired operation
482  * @uaddr_len: size of socket address data structure
483  *
484  * Name and name sequence binding is indicated using a positive scope value;
485  * a negative scope value unbinds the specified name.  Specifying no name
486  * (i.e. a socket address length of 0) unbinds all names from the socket.
487  *
488  * Returns 0 on success, errno otherwise
489  *
490  * NOTE: This routine doesn't need to take the socket lock since it doesn't
491  *       access any non-constant socket information.
492  */
493 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
494 		     int uaddr_len)
495 {
496 	struct sock *sk = sock->sk;
497 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
498 	struct tipc_sock *tsk = tipc_sk(sk);
499 	int res = -EINVAL;
500 
501 	lock_sock(sk);
502 	if (unlikely(!uaddr_len)) {
503 		res = tipc_sk_withdraw(tsk, 0, NULL);
504 		goto exit;
505 	}
506 
507 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
508 		res = -EINVAL;
509 		goto exit;
510 	}
511 	if (addr->family != AF_TIPC) {
512 		res = -EAFNOSUPPORT;
513 		goto exit;
514 	}
515 
516 	if (addr->addrtype == TIPC_ADDR_NAME)
517 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
518 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
519 		res = -EAFNOSUPPORT;
520 		goto exit;
521 	}
522 
523 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
524 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
525 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
526 		res = -EACCES;
527 		goto exit;
528 	}
529 
530 	res = (addr->scope > 0) ?
531 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
532 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
533 exit:
534 	release_sock(sk);
535 	return res;
536 }
537 
538 /**
539  * tipc_getname - get port ID of socket or peer socket
540  * @sock: socket structure
541  * @uaddr: area for returned socket address
542  * @uaddr_len: area for returned length of socket address
543  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
544  *
545  * Returns 0 on success, errno otherwise
546  *
547  * NOTE: This routine doesn't need to take the socket lock since it only
548  *       accesses socket information that is unchanging (or which changes in
549  *       a completely predictable manner).
550  */
551 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
552 			int *uaddr_len, int peer)
553 {
554 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
555 	struct tipc_sock *tsk = tipc_sk(sock->sk);
556 	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
557 
558 	memset(addr, 0, sizeof(*addr));
559 	if (peer) {
560 		if ((sock->state != SS_CONNECTED) &&
561 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
562 			return -ENOTCONN;
563 		addr->addr.id.ref = tsk_peer_port(tsk);
564 		addr->addr.id.node = tsk_peer_node(tsk);
565 	} else {
566 		addr->addr.id.ref = tsk->portid;
567 		addr->addr.id.node = tn->own_addr;
568 	}
569 
570 	*uaddr_len = sizeof(*addr);
571 	addr->addrtype = TIPC_ADDR_ID;
572 	addr->family = AF_TIPC;
573 	addr->scope = 0;
574 	addr->addr.name.domain = 0;
575 
576 	return 0;
577 }
578 
579 /**
580  * tipc_poll - read and possibly block on pollmask
581  * @file: file structure associated with the socket
582  * @sock: socket for which to calculate the poll bits
583  * @wait: ???
584  *
585  * Returns pollmask value
586  *
587  * COMMENTARY:
588  * It appears that the usual socket locking mechanisms are not useful here
589  * since the pollmask info is potentially out-of-date the moment this routine
590  * exits.  TCP and other protocols seem to rely on higher level poll routines
591  * to handle any preventable race conditions, so TIPC will do the same ...
592  *
593  * TIPC sets the returned events as follows:
594  *
595  * socket state		flags set
596  * ------------		---------
597  * unconnected		no read flags
598  *			POLLOUT if port is not congested
599  *
600  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
601  *			no write flags
602  *
603  * connected		POLLIN/POLLRDNORM if data in rx queue
604  *			POLLOUT if port is not congested
605  *
606  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
607  *			no write flags
608  *
609  * listening		POLLIN if SYN in rx queue
610  *			no write flags
611  *
612  * ready		POLLIN/POLLRDNORM if data in rx queue
613  * [connectionless]	POLLOUT (since port cannot be congested)
614  *
615  * IMPORTANT: The fact that a read or write operation is indicated does NOT
616  * imply that the operation will succeed, merely that it should be performed
617  * and will not block.
618  */
619 static unsigned int tipc_poll(struct file *file, struct socket *sock,
620 			      poll_table *wait)
621 {
622 	struct sock *sk = sock->sk;
623 	struct tipc_sock *tsk = tipc_sk(sk);
624 	u32 mask = 0;
625 
626 	sock_poll_wait(file, sk_sleep(sk), wait);
627 
628 	switch ((int)sock->state) {
629 	case SS_UNCONNECTED:
630 		if (!tsk->link_cong)
631 			mask |= POLLOUT;
632 		break;
633 	case SS_READY:
634 	case SS_CONNECTED:
635 		if (!tsk->link_cong && !tsk_conn_cong(tsk))
636 			mask |= POLLOUT;
637 		/* fall thru' */
638 	case SS_CONNECTING:
639 	case SS_LISTENING:
640 		if (!skb_queue_empty(&sk->sk_receive_queue))
641 			mask |= (POLLIN | POLLRDNORM);
642 		break;
643 	case SS_DISCONNECTING:
644 		mask = (POLLIN | POLLRDNORM | POLLHUP);
645 		break;
646 	}
647 
648 	return mask;
649 }
650 
651 /**
652  * tipc_sendmcast - send multicast message
653  * @sock: socket structure
654  * @seq: destination address
655  * @msg: message to send
656  * @dsz: total length of message data
657  * @timeo: timeout to wait for wakeup
658  *
659  * Called from function tipc_sendmsg(), which has done all sanity checks
660  * Returns the number of bytes sent on success, or errno
661  */
662 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
663 			  struct msghdr *msg, size_t dsz, long timeo)
664 {
665 	struct sock *sk = sock->sk;
666 	struct tipc_sock *tsk = tipc_sk(sk);
667 	struct net *net = sock_net(sk);
668 	struct tipc_msg *mhdr = &tsk->phdr;
669 	struct sk_buff_head *pktchain = &sk->sk_write_queue;
670 	struct iov_iter save = msg->msg_iter;
671 	uint mtu;
672 	int rc;
673 
674 	msg_set_type(mhdr, TIPC_MCAST_MSG);
675 	msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
676 	msg_set_destport(mhdr, 0);
677 	msg_set_destnode(mhdr, 0);
678 	msg_set_nametype(mhdr, seq->type);
679 	msg_set_namelower(mhdr, seq->lower);
680 	msg_set_nameupper(mhdr, seq->upper);
681 	msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
682 
683 new_mtu:
684 	mtu = tipc_bclink_get_mtu();
685 	rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, pktchain);
686 	if (unlikely(rc < 0))
687 		return rc;
688 
689 	do {
690 		rc = tipc_bclink_xmit(net, pktchain);
691 		if (likely(rc >= 0)) {
692 			rc = dsz;
693 			break;
694 		}
695 		if (rc == -EMSGSIZE) {
696 			msg->msg_iter = save;
697 			goto new_mtu;
698 		}
699 		if (rc != -ELINKCONG)
700 			break;
701 		tipc_sk(sk)->link_cong = 1;
702 		rc = tipc_wait_for_sndmsg(sock, &timeo);
703 		if (rc)
704 			__skb_queue_purge(pktchain);
705 	} while (!rc);
706 	return rc;
707 }
708 
709 /**
710  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
711  * @arrvq: queue with arriving messages, to be cloned after destination lookup
712  * @inputq: queue with cloned messages, delivered to socket after dest lookup
713  *
714  * Multi-threaded: parallel calls with reference to same queues may occur
715  */
716 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
717 		       struct sk_buff_head *inputq)
718 {
719 	struct tipc_msg *msg;
720 	struct tipc_plist dports;
721 	u32 portid;
722 	u32 scope = TIPC_CLUSTER_SCOPE;
723 	struct sk_buff_head tmpq;
724 	uint hsz;
725 	struct sk_buff *skb, *_skb;
726 
727 	__skb_queue_head_init(&tmpq);
728 	tipc_plist_init(&dports);
729 
730 	skb = tipc_skb_peek(arrvq, &inputq->lock);
731 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
732 		msg = buf_msg(skb);
733 		hsz = skb_headroom(skb) + msg_hdr_sz(msg);
734 
735 		if (in_own_node(net, msg_orignode(msg)))
736 			scope = TIPC_NODE_SCOPE;
737 
738 		/* Create destination port list and message clones: */
739 		tipc_nametbl_mc_translate(net,
740 					  msg_nametype(msg), msg_namelower(msg),
741 					  msg_nameupper(msg), scope, &dports);
742 		portid = tipc_plist_pop(&dports);
743 		for (; portid; portid = tipc_plist_pop(&dports)) {
744 			_skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
745 			if (_skb) {
746 				msg_set_destport(buf_msg(_skb), portid);
747 				__skb_queue_tail(&tmpq, _skb);
748 				continue;
749 			}
750 			pr_warn("Failed to clone mcast rcv buffer\n");
751 		}
752 		/* Append to inputq if not already done by other thread */
753 		spin_lock_bh(&inputq->lock);
754 		if (skb_peek(arrvq) == skb) {
755 			skb_queue_splice_tail_init(&tmpq, inputq);
756 			kfree_skb(__skb_dequeue(arrvq));
757 		}
758 		spin_unlock_bh(&inputq->lock);
759 		__skb_queue_purge(&tmpq);
760 		kfree_skb(skb);
761 	}
762 	tipc_sk_rcv(net, inputq);
763 }
764 
765 /**
766  * tipc_sk_proto_rcv - receive a connection mng protocol message
767  * @tsk: receiving socket
768  * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
769  */
770 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff **skb)
771 {
772 	struct tipc_msg *msg = buf_msg(*skb);
773 	int conn_cong;
774 	u32 dnode;
775 	u32 own_node = tsk_own_node(tsk);
776 	/* Ignore if connection cannot be validated: */
777 	if (!tsk_peer_msg(tsk, msg))
778 		goto exit;
779 
780 	tsk->probing_state = TIPC_CONN_OK;
781 
782 	if (msg_type(msg) == CONN_ACK) {
783 		conn_cong = tsk_conn_cong(tsk);
784 		tsk->sent_unacked -= msg_msgcnt(msg);
785 		if (conn_cong)
786 			tsk->sk.sk_write_space(&tsk->sk);
787 	} else if (msg_type(msg) == CONN_PROBE) {
788 		if (tipc_msg_reverse(own_node, *skb, &dnode, TIPC_OK)) {
789 			msg_set_type(msg, CONN_PROBE_REPLY);
790 			return;
791 		}
792 	}
793 	/* Do nothing if msg_type() == CONN_PROBE_REPLY */
794 exit:
795 	kfree_skb(*skb);
796 	*skb = NULL;
797 }
798 
799 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
800 {
801 	struct sock *sk = sock->sk;
802 	struct tipc_sock *tsk = tipc_sk(sk);
803 	DEFINE_WAIT(wait);
804 	int done;
805 
806 	do {
807 		int err = sock_error(sk);
808 		if (err)
809 			return err;
810 		if (sock->state == SS_DISCONNECTING)
811 			return -EPIPE;
812 		if (!*timeo_p)
813 			return -EAGAIN;
814 		if (signal_pending(current))
815 			return sock_intr_errno(*timeo_p);
816 
817 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
818 		done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
819 		finish_wait(sk_sleep(sk), &wait);
820 	} while (!done);
821 	return 0;
822 }
823 
824 /**
825  * tipc_sendmsg - send message in connectionless manner
826  * @sock: socket structure
827  * @m: message to send
828  * @dsz: amount of user data to be sent
829  *
830  * Message must have an destination specified explicitly.
831  * Used for SOCK_RDM and SOCK_DGRAM messages,
832  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
833  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
834  *
835  * Returns the number of bytes sent on success, or errno otherwise
836  */
837 static int tipc_sendmsg(struct socket *sock,
838 			struct msghdr *m, size_t dsz)
839 {
840 	struct sock *sk = sock->sk;
841 	int ret;
842 
843 	lock_sock(sk);
844 	ret = __tipc_sendmsg(sock, m, dsz);
845 	release_sock(sk);
846 
847 	return ret;
848 }
849 
850 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
851 {
852 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
853 	struct sock *sk = sock->sk;
854 	struct tipc_sock *tsk = tipc_sk(sk);
855 	struct net *net = sock_net(sk);
856 	struct tipc_msg *mhdr = &tsk->phdr;
857 	u32 dnode, dport;
858 	struct sk_buff_head *pktchain = &sk->sk_write_queue;
859 	struct sk_buff *skb;
860 	struct tipc_name_seq *seq;
861 	struct iov_iter save;
862 	u32 mtu;
863 	long timeo;
864 	int rc;
865 
866 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
867 		return -EMSGSIZE;
868 	if (unlikely(!dest)) {
869 		if (tsk->connected && sock->state == SS_READY)
870 			dest = &tsk->remote;
871 		else
872 			return -EDESTADDRREQ;
873 	} else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
874 		   dest->family != AF_TIPC) {
875 		return -EINVAL;
876 	}
877 	if (unlikely(sock->state != SS_READY)) {
878 		if (sock->state == SS_LISTENING)
879 			return -EPIPE;
880 		if (sock->state != SS_UNCONNECTED)
881 			return -EISCONN;
882 		if (tsk->published)
883 			return -EOPNOTSUPP;
884 		if (dest->addrtype == TIPC_ADDR_NAME) {
885 			tsk->conn_type = dest->addr.name.name.type;
886 			tsk->conn_instance = dest->addr.name.name.instance;
887 		}
888 	}
889 	seq = &dest->addr.nameseq;
890 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
891 
892 	if (dest->addrtype == TIPC_ADDR_MCAST) {
893 		return tipc_sendmcast(sock, seq, m, dsz, timeo);
894 	} else if (dest->addrtype == TIPC_ADDR_NAME) {
895 		u32 type = dest->addr.name.name.type;
896 		u32 inst = dest->addr.name.name.instance;
897 		u32 domain = dest->addr.name.domain;
898 
899 		dnode = domain;
900 		msg_set_type(mhdr, TIPC_NAMED_MSG);
901 		msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
902 		msg_set_nametype(mhdr, type);
903 		msg_set_nameinst(mhdr, inst);
904 		msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
905 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
906 		msg_set_destnode(mhdr, dnode);
907 		msg_set_destport(mhdr, dport);
908 		if (unlikely(!dport && !dnode))
909 			return -EHOSTUNREACH;
910 	} else if (dest->addrtype == TIPC_ADDR_ID) {
911 		dnode = dest->addr.id.node;
912 		msg_set_type(mhdr, TIPC_DIRECT_MSG);
913 		msg_set_lookup_scope(mhdr, 0);
914 		msg_set_destnode(mhdr, dnode);
915 		msg_set_destport(mhdr, dest->addr.id.ref);
916 		msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
917 	}
918 
919 	save = m->msg_iter;
920 new_mtu:
921 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
922 	rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, pktchain);
923 	if (rc < 0)
924 		return rc;
925 
926 	do {
927 		skb = skb_peek(pktchain);
928 		TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
929 		rc = tipc_link_xmit(net, pktchain, dnode, tsk->portid);
930 		if (likely(rc >= 0)) {
931 			if (sock->state != SS_READY)
932 				sock->state = SS_CONNECTING;
933 			rc = dsz;
934 			break;
935 		}
936 		if (rc == -EMSGSIZE) {
937 			m->msg_iter = save;
938 			goto new_mtu;
939 		}
940 		if (rc != -ELINKCONG)
941 			break;
942 		tsk->link_cong = 1;
943 		rc = tipc_wait_for_sndmsg(sock, &timeo);
944 		if (rc)
945 			__skb_queue_purge(pktchain);
946 	} while (!rc);
947 
948 	return rc;
949 }
950 
951 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
952 {
953 	struct sock *sk = sock->sk;
954 	struct tipc_sock *tsk = tipc_sk(sk);
955 	DEFINE_WAIT(wait);
956 	int done;
957 
958 	do {
959 		int err = sock_error(sk);
960 		if (err)
961 			return err;
962 		if (sock->state == SS_DISCONNECTING)
963 			return -EPIPE;
964 		else if (sock->state != SS_CONNECTED)
965 			return -ENOTCONN;
966 		if (!*timeo_p)
967 			return -EAGAIN;
968 		if (signal_pending(current))
969 			return sock_intr_errno(*timeo_p);
970 
971 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
972 		done = sk_wait_event(sk, timeo_p,
973 				     (!tsk->link_cong &&
974 				      !tsk_conn_cong(tsk)) ||
975 				     !tsk->connected);
976 		finish_wait(sk_sleep(sk), &wait);
977 	} while (!done);
978 	return 0;
979 }
980 
981 /**
982  * tipc_send_stream - send stream-oriented data
983  * @sock: socket structure
984  * @m: data to send
985  * @dsz: total length of data to be transmitted
986  *
987  * Used for SOCK_STREAM data.
988  *
989  * Returns the number of bytes sent on success (or partial success),
990  * or errno if no data sent
991  */
992 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
993 {
994 	struct sock *sk = sock->sk;
995 	int ret;
996 
997 	lock_sock(sk);
998 	ret = __tipc_send_stream(sock, m, dsz);
999 	release_sock(sk);
1000 
1001 	return ret;
1002 }
1003 
1004 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1005 {
1006 	struct sock *sk = sock->sk;
1007 	struct net *net = sock_net(sk);
1008 	struct tipc_sock *tsk = tipc_sk(sk);
1009 	struct tipc_msg *mhdr = &tsk->phdr;
1010 	struct sk_buff_head *pktchain = &sk->sk_write_queue;
1011 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1012 	u32 portid = tsk->portid;
1013 	int rc = -EINVAL;
1014 	long timeo;
1015 	u32 dnode;
1016 	uint mtu, send, sent = 0;
1017 	struct iov_iter save;
1018 
1019 	/* Handle implied connection establishment */
1020 	if (unlikely(dest)) {
1021 		rc = __tipc_sendmsg(sock, m, dsz);
1022 		if (dsz && (dsz == rc))
1023 			tsk->sent_unacked = 1;
1024 		return rc;
1025 	}
1026 	if (dsz > (uint)INT_MAX)
1027 		return -EMSGSIZE;
1028 
1029 	if (unlikely(sock->state != SS_CONNECTED)) {
1030 		if (sock->state == SS_DISCONNECTING)
1031 			return -EPIPE;
1032 		else
1033 			return -ENOTCONN;
1034 	}
1035 
1036 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1037 	dnode = tsk_peer_node(tsk);
1038 
1039 next:
1040 	save = m->msg_iter;
1041 	mtu = tsk->max_pkt;
1042 	send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1043 	rc = tipc_msg_build(mhdr, m, sent, send, mtu, pktchain);
1044 	if (unlikely(rc < 0))
1045 		return rc;
1046 	do {
1047 		if (likely(!tsk_conn_cong(tsk))) {
1048 			rc = tipc_link_xmit(net, pktchain, dnode, portid);
1049 			if (likely(!rc)) {
1050 				tsk->sent_unacked++;
1051 				sent += send;
1052 				if (sent == dsz)
1053 					break;
1054 				goto next;
1055 			}
1056 			if (rc == -EMSGSIZE) {
1057 				tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1058 								 portid);
1059 				m->msg_iter = save;
1060 				goto next;
1061 			}
1062 			if (rc != -ELINKCONG)
1063 				break;
1064 			tsk->link_cong = 1;
1065 		}
1066 		rc = tipc_wait_for_sndpkt(sock, &timeo);
1067 		if (rc)
1068 			__skb_queue_purge(pktchain);
1069 	} while (!rc);
1070 
1071 	return sent ? sent : rc;
1072 }
1073 
1074 /**
1075  * tipc_send_packet - send a connection-oriented message
1076  * @sock: socket structure
1077  * @m: message to send
1078  * @dsz: length of data to be transmitted
1079  *
1080  * Used for SOCK_SEQPACKET messages.
1081  *
1082  * Returns the number of bytes sent on success, or errno otherwise
1083  */
1084 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1085 {
1086 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1087 		return -EMSGSIZE;
1088 
1089 	return tipc_send_stream(sock, m, dsz);
1090 }
1091 
1092 /* tipc_sk_finish_conn - complete the setup of a connection
1093  */
1094 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1095 				u32 peer_node)
1096 {
1097 	struct sock *sk = &tsk->sk;
1098 	struct net *net = sock_net(sk);
1099 	struct tipc_msg *msg = &tsk->phdr;
1100 
1101 	msg_set_destnode(msg, peer_node);
1102 	msg_set_destport(msg, peer_port);
1103 	msg_set_type(msg, TIPC_CONN_MSG);
1104 	msg_set_lookup_scope(msg, 0);
1105 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1106 
1107 	tsk->probing_intv = CONN_PROBING_INTERVAL;
1108 	tsk->probing_state = TIPC_CONN_OK;
1109 	tsk->connected = 1;
1110 	sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1111 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1112 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1113 }
1114 
1115 /**
1116  * set_orig_addr - capture sender's address for received message
1117  * @m: descriptor for message info
1118  * @msg: received message header
1119  *
1120  * Note: Address is not captured if not requested by receiver.
1121  */
1122 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1123 {
1124 	DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1125 
1126 	if (addr) {
1127 		addr->family = AF_TIPC;
1128 		addr->addrtype = TIPC_ADDR_ID;
1129 		memset(&addr->addr, 0, sizeof(addr->addr));
1130 		addr->addr.id.ref = msg_origport(msg);
1131 		addr->addr.id.node = msg_orignode(msg);
1132 		addr->addr.name.domain = 0;	/* could leave uninitialized */
1133 		addr->scope = 0;		/* could leave uninitialized */
1134 		m->msg_namelen = sizeof(struct sockaddr_tipc);
1135 	}
1136 }
1137 
1138 /**
1139  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1140  * @m: descriptor for message info
1141  * @msg: received message header
1142  * @tsk: TIPC port associated with message
1143  *
1144  * Note: Ancillary data is not captured if not requested by receiver.
1145  *
1146  * Returns 0 if successful, otherwise errno
1147  */
1148 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1149 				 struct tipc_sock *tsk)
1150 {
1151 	u32 anc_data[3];
1152 	u32 err;
1153 	u32 dest_type;
1154 	int has_name;
1155 	int res;
1156 
1157 	if (likely(m->msg_controllen == 0))
1158 		return 0;
1159 
1160 	/* Optionally capture errored message object(s) */
1161 	err = msg ? msg_errcode(msg) : 0;
1162 	if (unlikely(err)) {
1163 		anc_data[0] = err;
1164 		anc_data[1] = msg_data_sz(msg);
1165 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1166 		if (res)
1167 			return res;
1168 		if (anc_data[1]) {
1169 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1170 				       msg_data(msg));
1171 			if (res)
1172 				return res;
1173 		}
1174 	}
1175 
1176 	/* Optionally capture message destination object */
1177 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1178 	switch (dest_type) {
1179 	case TIPC_NAMED_MSG:
1180 		has_name = 1;
1181 		anc_data[0] = msg_nametype(msg);
1182 		anc_data[1] = msg_namelower(msg);
1183 		anc_data[2] = msg_namelower(msg);
1184 		break;
1185 	case TIPC_MCAST_MSG:
1186 		has_name = 1;
1187 		anc_data[0] = msg_nametype(msg);
1188 		anc_data[1] = msg_namelower(msg);
1189 		anc_data[2] = msg_nameupper(msg);
1190 		break;
1191 	case TIPC_CONN_MSG:
1192 		has_name = (tsk->conn_type != 0);
1193 		anc_data[0] = tsk->conn_type;
1194 		anc_data[1] = tsk->conn_instance;
1195 		anc_data[2] = tsk->conn_instance;
1196 		break;
1197 	default:
1198 		has_name = 0;
1199 	}
1200 	if (has_name) {
1201 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1202 		if (res)
1203 			return res;
1204 	}
1205 
1206 	return 0;
1207 }
1208 
1209 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1210 {
1211 	struct net *net = sock_net(&tsk->sk);
1212 	struct sk_buff *skb = NULL;
1213 	struct tipc_msg *msg;
1214 	u32 peer_port = tsk_peer_port(tsk);
1215 	u32 dnode = tsk_peer_node(tsk);
1216 
1217 	if (!tsk->connected)
1218 		return;
1219 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1220 			      dnode, tsk_own_node(tsk), peer_port,
1221 			      tsk->portid, TIPC_OK);
1222 	if (!skb)
1223 		return;
1224 	msg = buf_msg(skb);
1225 	msg_set_msgcnt(msg, ack);
1226 	tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1227 }
1228 
1229 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1230 {
1231 	struct sock *sk = sock->sk;
1232 	DEFINE_WAIT(wait);
1233 	long timeo = *timeop;
1234 	int err;
1235 
1236 	for (;;) {
1237 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1238 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1239 			if (sock->state == SS_DISCONNECTING) {
1240 				err = -ENOTCONN;
1241 				break;
1242 			}
1243 			release_sock(sk);
1244 			timeo = schedule_timeout(timeo);
1245 			lock_sock(sk);
1246 		}
1247 		err = 0;
1248 		if (!skb_queue_empty(&sk->sk_receive_queue))
1249 			break;
1250 		err = -EAGAIN;
1251 		if (!timeo)
1252 			break;
1253 		err = sock_intr_errno(timeo);
1254 		if (signal_pending(current))
1255 			break;
1256 	}
1257 	finish_wait(sk_sleep(sk), &wait);
1258 	*timeop = timeo;
1259 	return err;
1260 }
1261 
1262 /**
1263  * tipc_recvmsg - receive packet-oriented message
1264  * @m: descriptor for message info
1265  * @buf_len: total size of user buffer area
1266  * @flags: receive flags
1267  *
1268  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1269  * If the complete message doesn't fit in user area, truncate it.
1270  *
1271  * Returns size of returned message data, errno otherwise
1272  */
1273 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1274 			int flags)
1275 {
1276 	struct sock *sk = sock->sk;
1277 	struct tipc_sock *tsk = tipc_sk(sk);
1278 	struct sk_buff *buf;
1279 	struct tipc_msg *msg;
1280 	long timeo;
1281 	unsigned int sz;
1282 	u32 err;
1283 	int res;
1284 
1285 	/* Catch invalid receive requests */
1286 	if (unlikely(!buf_len))
1287 		return -EINVAL;
1288 
1289 	lock_sock(sk);
1290 
1291 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1292 		res = -ENOTCONN;
1293 		goto exit;
1294 	}
1295 
1296 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1297 restart:
1298 
1299 	/* Look for a message in receive queue; wait if necessary */
1300 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1301 	if (res)
1302 		goto exit;
1303 
1304 	/* Look at first message in receive queue */
1305 	buf = skb_peek(&sk->sk_receive_queue);
1306 	msg = buf_msg(buf);
1307 	sz = msg_data_sz(msg);
1308 	err = msg_errcode(msg);
1309 
1310 	/* Discard an empty non-errored message & try again */
1311 	if ((!sz) && (!err)) {
1312 		tsk_advance_rx_queue(sk);
1313 		goto restart;
1314 	}
1315 
1316 	/* Capture sender's address (optional) */
1317 	set_orig_addr(m, msg);
1318 
1319 	/* Capture ancillary data (optional) */
1320 	res = tipc_sk_anc_data_recv(m, msg, tsk);
1321 	if (res)
1322 		goto exit;
1323 
1324 	/* Capture message data (if valid) & compute return value (always) */
1325 	if (!err) {
1326 		if (unlikely(buf_len < sz)) {
1327 			sz = buf_len;
1328 			m->msg_flags |= MSG_TRUNC;
1329 		}
1330 		res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1331 		if (res)
1332 			goto exit;
1333 		res = sz;
1334 	} else {
1335 		if ((sock->state == SS_READY) ||
1336 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1337 			res = 0;
1338 		else
1339 			res = -ECONNRESET;
1340 	}
1341 
1342 	/* Consume received message (optional) */
1343 	if (likely(!(flags & MSG_PEEK))) {
1344 		if ((sock->state != SS_READY) &&
1345 		    (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1346 			tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1347 			tsk->rcv_unacked = 0;
1348 		}
1349 		tsk_advance_rx_queue(sk);
1350 	}
1351 exit:
1352 	release_sock(sk);
1353 	return res;
1354 }
1355 
1356 /**
1357  * tipc_recv_stream - receive stream-oriented data
1358  * @m: descriptor for message info
1359  * @buf_len: total size of user buffer area
1360  * @flags: receive flags
1361  *
1362  * Used for SOCK_STREAM messages only.  If not enough data is available
1363  * will optionally wait for more; never truncates data.
1364  *
1365  * Returns size of returned message data, errno otherwise
1366  */
1367 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1368 			    size_t buf_len, int flags)
1369 {
1370 	struct sock *sk = sock->sk;
1371 	struct tipc_sock *tsk = tipc_sk(sk);
1372 	struct sk_buff *buf;
1373 	struct tipc_msg *msg;
1374 	long timeo;
1375 	unsigned int sz;
1376 	int sz_to_copy, target, needed;
1377 	int sz_copied = 0;
1378 	u32 err;
1379 	int res = 0;
1380 
1381 	/* Catch invalid receive attempts */
1382 	if (unlikely(!buf_len))
1383 		return -EINVAL;
1384 
1385 	lock_sock(sk);
1386 
1387 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1388 		res = -ENOTCONN;
1389 		goto exit;
1390 	}
1391 
1392 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1393 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1394 
1395 restart:
1396 	/* Look for a message in receive queue; wait if necessary */
1397 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1398 	if (res)
1399 		goto exit;
1400 
1401 	/* Look at first message in receive queue */
1402 	buf = skb_peek(&sk->sk_receive_queue);
1403 	msg = buf_msg(buf);
1404 	sz = msg_data_sz(msg);
1405 	err = msg_errcode(msg);
1406 
1407 	/* Discard an empty non-errored message & try again */
1408 	if ((!sz) && (!err)) {
1409 		tsk_advance_rx_queue(sk);
1410 		goto restart;
1411 	}
1412 
1413 	/* Optionally capture sender's address & ancillary data of first msg */
1414 	if (sz_copied == 0) {
1415 		set_orig_addr(m, msg);
1416 		res = tipc_sk_anc_data_recv(m, msg, tsk);
1417 		if (res)
1418 			goto exit;
1419 	}
1420 
1421 	/* Capture message data (if valid) & compute return value (always) */
1422 	if (!err) {
1423 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1424 
1425 		sz -= offset;
1426 		needed = (buf_len - sz_copied);
1427 		sz_to_copy = (sz <= needed) ? sz : needed;
1428 
1429 		res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1430 					    m, sz_to_copy);
1431 		if (res)
1432 			goto exit;
1433 
1434 		sz_copied += sz_to_copy;
1435 
1436 		if (sz_to_copy < sz) {
1437 			if (!(flags & MSG_PEEK))
1438 				TIPC_SKB_CB(buf)->handle =
1439 				(void *)(unsigned long)(offset + sz_to_copy);
1440 			goto exit;
1441 		}
1442 	} else {
1443 		if (sz_copied != 0)
1444 			goto exit; /* can't add error msg to valid data */
1445 
1446 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1447 			res = 0;
1448 		else
1449 			res = -ECONNRESET;
1450 	}
1451 
1452 	/* Consume received message (optional) */
1453 	if (likely(!(flags & MSG_PEEK))) {
1454 		if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1455 			tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1456 			tsk->rcv_unacked = 0;
1457 		}
1458 		tsk_advance_rx_queue(sk);
1459 	}
1460 
1461 	/* Loop around if more data is required */
1462 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1463 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1464 	    (sz_copied < target)) &&	/* and more is ready or required */
1465 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1466 	    (!err))			/* and haven't reached a FIN */
1467 		goto restart;
1468 
1469 exit:
1470 	release_sock(sk);
1471 	return sz_copied ? sz_copied : res;
1472 }
1473 
1474 /**
1475  * tipc_write_space - wake up thread if port congestion is released
1476  * @sk: socket
1477  */
1478 static void tipc_write_space(struct sock *sk)
1479 {
1480 	struct socket_wq *wq;
1481 
1482 	rcu_read_lock();
1483 	wq = rcu_dereference(sk->sk_wq);
1484 	if (wq_has_sleeper(wq))
1485 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1486 						POLLWRNORM | POLLWRBAND);
1487 	rcu_read_unlock();
1488 }
1489 
1490 /**
1491  * tipc_data_ready - wake up threads to indicate messages have been received
1492  * @sk: socket
1493  * @len: the length of messages
1494  */
1495 static void tipc_data_ready(struct sock *sk)
1496 {
1497 	struct socket_wq *wq;
1498 
1499 	rcu_read_lock();
1500 	wq = rcu_dereference(sk->sk_wq);
1501 	if (wq_has_sleeper(wq))
1502 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1503 						POLLRDNORM | POLLRDBAND);
1504 	rcu_read_unlock();
1505 }
1506 
1507 /**
1508  * filter_connect - Handle all incoming messages for a connection-based socket
1509  * @tsk: TIPC socket
1510  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1511  *
1512  * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1513  */
1514 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **skb)
1515 {
1516 	struct sock *sk = &tsk->sk;
1517 	struct net *net = sock_net(sk);
1518 	struct socket *sock = sk->sk_socket;
1519 	struct tipc_msg *msg = buf_msg(*skb);
1520 	int retval = -TIPC_ERR_NO_PORT;
1521 
1522 	if (msg_mcast(msg))
1523 		return retval;
1524 
1525 	switch ((int)sock->state) {
1526 	case SS_CONNECTED:
1527 		/* Accept only connection-based messages sent by peer */
1528 		if (tsk_peer_msg(tsk, msg)) {
1529 			if (unlikely(msg_errcode(msg))) {
1530 				sock->state = SS_DISCONNECTING;
1531 				tsk->connected = 0;
1532 				/* let timer expire on it's own */
1533 				tipc_node_remove_conn(net, tsk_peer_node(tsk),
1534 						      tsk->portid);
1535 			}
1536 			retval = TIPC_OK;
1537 		}
1538 		break;
1539 	case SS_CONNECTING:
1540 		/* Accept only ACK or NACK message */
1541 
1542 		if (unlikely(!msg_connected(msg)))
1543 			break;
1544 
1545 		if (unlikely(msg_errcode(msg))) {
1546 			sock->state = SS_DISCONNECTING;
1547 			sk->sk_err = ECONNREFUSED;
1548 			retval = TIPC_OK;
1549 			break;
1550 		}
1551 
1552 		if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1553 			sock->state = SS_DISCONNECTING;
1554 			sk->sk_err = EINVAL;
1555 			retval = TIPC_OK;
1556 			break;
1557 		}
1558 
1559 		tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1560 		msg_set_importance(&tsk->phdr, msg_importance(msg));
1561 		sock->state = SS_CONNECTED;
1562 
1563 		/* If an incoming message is an 'ACK-', it should be
1564 		 * discarded here because it doesn't contain useful
1565 		 * data. In addition, we should try to wake up
1566 		 * connect() routine if sleeping.
1567 		 */
1568 		if (msg_data_sz(msg) == 0) {
1569 			kfree_skb(*skb);
1570 			*skb = NULL;
1571 			if (waitqueue_active(sk_sleep(sk)))
1572 				wake_up_interruptible(sk_sleep(sk));
1573 		}
1574 		retval = TIPC_OK;
1575 		break;
1576 	case SS_LISTENING:
1577 	case SS_UNCONNECTED:
1578 		/* Accept only SYN message */
1579 		if (!msg_connected(msg) && !(msg_errcode(msg)))
1580 			retval = TIPC_OK;
1581 		break;
1582 	case SS_DISCONNECTING:
1583 		break;
1584 	default:
1585 		pr_err("Unknown socket state %u\n", sock->state);
1586 	}
1587 	return retval;
1588 }
1589 
1590 /**
1591  * rcvbuf_limit - get proper overload limit of socket receive queue
1592  * @sk: socket
1593  * @buf: message
1594  *
1595  * For all connection oriented messages, irrespective of importance,
1596  * the default overload value (i.e. 67MB) is set as limit.
1597  *
1598  * For all connectionless messages, by default new queue limits are
1599  * as belows:
1600  *
1601  * TIPC_LOW_IMPORTANCE       (4 MB)
1602  * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1603  * TIPC_HIGH_IMPORTANCE      (16 MB)
1604  * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1605  *
1606  * Returns overload limit according to corresponding message importance
1607  */
1608 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1609 {
1610 	struct tipc_msg *msg = buf_msg(buf);
1611 
1612 	if (msg_connected(msg))
1613 		return sysctl_tipc_rmem[2];
1614 
1615 	return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1616 		msg_importance(msg);
1617 }
1618 
1619 /**
1620  * filter_rcv - validate incoming message
1621  * @sk: socket
1622  * @skb: pointer to message. Set to NULL if buffer is consumed.
1623  *
1624  * Enqueues message on receive queue if acceptable; optionally handles
1625  * disconnect indication for a connected socket.
1626  *
1627  * Called with socket lock already taken
1628  *
1629  * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1630  */
1631 static int filter_rcv(struct sock *sk, struct sk_buff **skb)
1632 {
1633 	struct socket *sock = sk->sk_socket;
1634 	struct tipc_sock *tsk = tipc_sk(sk);
1635 	struct tipc_msg *msg = buf_msg(*skb);
1636 	unsigned int limit = rcvbuf_limit(sk, *skb);
1637 	int rc = TIPC_OK;
1638 
1639 	if (unlikely(msg_user(msg) == CONN_MANAGER)) {
1640 		tipc_sk_proto_rcv(tsk, skb);
1641 		return TIPC_OK;
1642 	}
1643 
1644 	if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1645 		kfree_skb(*skb);
1646 		tsk->link_cong = 0;
1647 		sk->sk_write_space(sk);
1648 		*skb = NULL;
1649 		return TIPC_OK;
1650 	}
1651 
1652 	/* Reject message if it is wrong sort of message for socket */
1653 	if (msg_type(msg) > TIPC_DIRECT_MSG)
1654 		return -TIPC_ERR_NO_PORT;
1655 
1656 	if (sock->state == SS_READY) {
1657 		if (msg_connected(msg))
1658 			return -TIPC_ERR_NO_PORT;
1659 	} else {
1660 		rc = filter_connect(tsk, skb);
1661 		if (rc != TIPC_OK || !*skb)
1662 			return rc;
1663 	}
1664 
1665 	/* Reject message if there isn't room to queue it */
1666 	if (sk_rmem_alloc_get(sk) + (*skb)->truesize >= limit)
1667 		return -TIPC_ERR_OVERLOAD;
1668 
1669 	/* Enqueue message */
1670 	TIPC_SKB_CB(*skb)->handle = NULL;
1671 	__skb_queue_tail(&sk->sk_receive_queue, *skb);
1672 	skb_set_owner_r(*skb, sk);
1673 
1674 	sk->sk_data_ready(sk);
1675 	*skb = NULL;
1676 	return TIPC_OK;
1677 }
1678 
1679 /**
1680  * tipc_backlog_rcv - handle incoming message from backlog queue
1681  * @sk: socket
1682  * @skb: message
1683  *
1684  * Caller must hold socket lock
1685  *
1686  * Returns 0
1687  */
1688 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1689 {
1690 	int err;
1691 	atomic_t *dcnt;
1692 	u32 dnode;
1693 	struct tipc_sock *tsk = tipc_sk(sk);
1694 	struct net *net = sock_net(sk);
1695 	uint truesize = skb->truesize;
1696 
1697 	err = filter_rcv(sk, &skb);
1698 	if (likely(!skb)) {
1699 		dcnt = &tsk->dupl_rcvcnt;
1700 		if (atomic_read(dcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1701 			atomic_add(truesize, dcnt);
1702 		return 0;
1703 	}
1704 	if (!err || tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode, -err))
1705 		tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
1706 	return 0;
1707 }
1708 
1709 /**
1710  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1711  *                   inputq and try adding them to socket or backlog queue
1712  * @inputq: list of incoming buffers with potentially different destinations
1713  * @sk: socket where the buffers should be enqueued
1714  * @dport: port number for the socket
1715  * @_skb: returned buffer to be forwarded or rejected, if applicable
1716  *
1717  * Caller must hold socket lock
1718  *
1719  * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1720  * or -TIPC_ERR_NO_PORT
1721  */
1722 static int tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1723 			   u32 dport, struct sk_buff **_skb)
1724 {
1725 	unsigned int lim;
1726 	atomic_t *dcnt;
1727 	int err;
1728 	struct sk_buff *skb;
1729 	unsigned long time_limit = jiffies + 2;
1730 
1731 	while (skb_queue_len(inputq)) {
1732 		if (unlikely(time_after_eq(jiffies, time_limit)))
1733 			return TIPC_OK;
1734 		skb = tipc_skb_dequeue(inputq, dport);
1735 		if (unlikely(!skb))
1736 			return TIPC_OK;
1737 		if (!sock_owned_by_user(sk)) {
1738 			err = filter_rcv(sk, &skb);
1739 			if (likely(!skb))
1740 				continue;
1741 			*_skb = skb;
1742 			return err;
1743 		}
1744 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1745 		if (sk->sk_backlog.len)
1746 			atomic_set(dcnt, 0);
1747 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1748 		if (likely(!sk_add_backlog(sk, skb, lim)))
1749 			continue;
1750 		*_skb = skb;
1751 		return -TIPC_ERR_OVERLOAD;
1752 	}
1753 	return TIPC_OK;
1754 }
1755 
1756 /**
1757  * tipc_sk_rcv - handle a chain of incoming buffers
1758  * @inputq: buffer list containing the buffers
1759  * Consumes all buffers in list until inputq is empty
1760  * Note: may be called in multiple threads referring to the same queue
1761  * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1762  * Only node local calls check the return value, sending single-buffer queues
1763  */
1764 int tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1765 {
1766 	u32 dnode, dport = 0;
1767 	int err;
1768 	struct sk_buff *skb;
1769 	struct tipc_sock *tsk;
1770 	struct tipc_net *tn;
1771 	struct sock *sk;
1772 
1773 	while (skb_queue_len(inputq)) {
1774 		err = -TIPC_ERR_NO_PORT;
1775 		skb = NULL;
1776 		dport = tipc_skb_peek_port(inputq, dport);
1777 		tsk = tipc_sk_lookup(net, dport);
1778 		if (likely(tsk)) {
1779 			sk = &tsk->sk;
1780 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1781 				err = tipc_sk_enqueue(inputq, sk, dport, &skb);
1782 				spin_unlock_bh(&sk->sk_lock.slock);
1783 				dport = 0;
1784 			}
1785 			sock_put(sk);
1786 		} else {
1787 			skb = tipc_skb_dequeue(inputq, dport);
1788 		}
1789 		if (likely(!skb))
1790 			continue;
1791 		if (tipc_msg_lookup_dest(net, skb, &dnode, &err))
1792 			goto xmit;
1793 		if (!err) {
1794 			dnode = msg_destnode(buf_msg(skb));
1795 			goto xmit;
1796 		}
1797 		tn = net_generic(net, tipc_net_id);
1798 		if (!tipc_msg_reverse(tn->own_addr, skb, &dnode, -err))
1799 			continue;
1800 xmit:
1801 		tipc_link_xmit_skb(net, skb, dnode, dport);
1802 	}
1803 	return err ? -EHOSTUNREACH : 0;
1804 }
1805 
1806 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1807 {
1808 	struct sock *sk = sock->sk;
1809 	DEFINE_WAIT(wait);
1810 	int done;
1811 
1812 	do {
1813 		int err = sock_error(sk);
1814 		if (err)
1815 			return err;
1816 		if (!*timeo_p)
1817 			return -ETIMEDOUT;
1818 		if (signal_pending(current))
1819 			return sock_intr_errno(*timeo_p);
1820 
1821 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1822 		done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1823 		finish_wait(sk_sleep(sk), &wait);
1824 	} while (!done);
1825 	return 0;
1826 }
1827 
1828 /**
1829  * tipc_connect - establish a connection to another TIPC port
1830  * @sock: socket structure
1831  * @dest: socket address for destination port
1832  * @destlen: size of socket address data structure
1833  * @flags: file-related flags associated with socket
1834  *
1835  * Returns 0 on success, errno otherwise
1836  */
1837 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1838 			int destlen, int flags)
1839 {
1840 	struct sock *sk = sock->sk;
1841 	struct tipc_sock *tsk = tipc_sk(sk);
1842 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1843 	struct msghdr m = {NULL,};
1844 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1845 	socket_state previous;
1846 	int res = 0;
1847 
1848 	lock_sock(sk);
1849 
1850 	/* DGRAM/RDM connect(), just save the destaddr */
1851 	if (sock->state == SS_READY) {
1852 		if (dst->family == AF_UNSPEC) {
1853 			memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1854 			tsk->connected = 0;
1855 		} else if (destlen != sizeof(struct sockaddr_tipc)) {
1856 			res = -EINVAL;
1857 		} else {
1858 			memcpy(&tsk->remote, dest, destlen);
1859 			tsk->connected = 1;
1860 		}
1861 		goto exit;
1862 	}
1863 
1864 	/*
1865 	 * Reject connection attempt using multicast address
1866 	 *
1867 	 * Note: send_msg() validates the rest of the address fields,
1868 	 *       so there's no need to do it here
1869 	 */
1870 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1871 		res = -EINVAL;
1872 		goto exit;
1873 	}
1874 
1875 	previous = sock->state;
1876 	switch (sock->state) {
1877 	case SS_UNCONNECTED:
1878 		/* Send a 'SYN-' to destination */
1879 		m.msg_name = dest;
1880 		m.msg_namelen = destlen;
1881 
1882 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1883 		 * indicate send_msg() is never blocked.
1884 		 */
1885 		if (!timeout)
1886 			m.msg_flags = MSG_DONTWAIT;
1887 
1888 		res = __tipc_sendmsg(sock, &m, 0);
1889 		if ((res < 0) && (res != -EWOULDBLOCK))
1890 			goto exit;
1891 
1892 		/* Just entered SS_CONNECTING state; the only
1893 		 * difference is that return value in non-blocking
1894 		 * case is EINPROGRESS, rather than EALREADY.
1895 		 */
1896 		res = -EINPROGRESS;
1897 	case SS_CONNECTING:
1898 		if (previous == SS_CONNECTING)
1899 			res = -EALREADY;
1900 		if (!timeout)
1901 			goto exit;
1902 		timeout = msecs_to_jiffies(timeout);
1903 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1904 		res = tipc_wait_for_connect(sock, &timeout);
1905 		break;
1906 	case SS_CONNECTED:
1907 		res = -EISCONN;
1908 		break;
1909 	default:
1910 		res = -EINVAL;
1911 		break;
1912 	}
1913 exit:
1914 	release_sock(sk);
1915 	return res;
1916 }
1917 
1918 /**
1919  * tipc_listen - allow socket to listen for incoming connections
1920  * @sock: socket structure
1921  * @len: (unused)
1922  *
1923  * Returns 0 on success, errno otherwise
1924  */
1925 static int tipc_listen(struct socket *sock, int len)
1926 {
1927 	struct sock *sk = sock->sk;
1928 	int res;
1929 
1930 	lock_sock(sk);
1931 
1932 	if (sock->state != SS_UNCONNECTED)
1933 		res = -EINVAL;
1934 	else {
1935 		sock->state = SS_LISTENING;
1936 		res = 0;
1937 	}
1938 
1939 	release_sock(sk);
1940 	return res;
1941 }
1942 
1943 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1944 {
1945 	struct sock *sk = sock->sk;
1946 	DEFINE_WAIT(wait);
1947 	int err;
1948 
1949 	/* True wake-one mechanism for incoming connections: only
1950 	 * one process gets woken up, not the 'whole herd'.
1951 	 * Since we do not 'race & poll' for established sockets
1952 	 * anymore, the common case will execute the loop only once.
1953 	*/
1954 	for (;;) {
1955 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1956 					  TASK_INTERRUPTIBLE);
1957 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1958 			release_sock(sk);
1959 			timeo = schedule_timeout(timeo);
1960 			lock_sock(sk);
1961 		}
1962 		err = 0;
1963 		if (!skb_queue_empty(&sk->sk_receive_queue))
1964 			break;
1965 		err = -EINVAL;
1966 		if (sock->state != SS_LISTENING)
1967 			break;
1968 		err = -EAGAIN;
1969 		if (!timeo)
1970 			break;
1971 		err = sock_intr_errno(timeo);
1972 		if (signal_pending(current))
1973 			break;
1974 	}
1975 	finish_wait(sk_sleep(sk), &wait);
1976 	return err;
1977 }
1978 
1979 /**
1980  * tipc_accept - wait for connection request
1981  * @sock: listening socket
1982  * @newsock: new socket that is to be connected
1983  * @flags: file-related flags associated with socket
1984  *
1985  * Returns 0 on success, errno otherwise
1986  */
1987 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1988 {
1989 	struct sock *new_sk, *sk = sock->sk;
1990 	struct sk_buff *buf;
1991 	struct tipc_sock *new_tsock;
1992 	struct tipc_msg *msg;
1993 	long timeo;
1994 	int res;
1995 
1996 	lock_sock(sk);
1997 
1998 	if (sock->state != SS_LISTENING) {
1999 		res = -EINVAL;
2000 		goto exit;
2001 	}
2002 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2003 	res = tipc_wait_for_accept(sock, timeo);
2004 	if (res)
2005 		goto exit;
2006 
2007 	buf = skb_peek(&sk->sk_receive_queue);
2008 
2009 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2010 	if (res)
2011 		goto exit;
2012 
2013 	new_sk = new_sock->sk;
2014 	new_tsock = tipc_sk(new_sk);
2015 	msg = buf_msg(buf);
2016 
2017 	/* we lock on new_sk; but lockdep sees the lock on sk */
2018 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2019 
2020 	/*
2021 	 * Reject any stray messages received by new socket
2022 	 * before the socket lock was taken (very, very unlikely)
2023 	 */
2024 	tsk_rej_rx_queue(new_sk);
2025 
2026 	/* Connect new socket to it's peer */
2027 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2028 	new_sock->state = SS_CONNECTED;
2029 
2030 	tsk_set_importance(new_tsock, msg_importance(msg));
2031 	if (msg_named(msg)) {
2032 		new_tsock->conn_type = msg_nametype(msg);
2033 		new_tsock->conn_instance = msg_nameinst(msg);
2034 	}
2035 
2036 	/*
2037 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2038 	 * Respond to 'SYN+' by queuing it on new socket.
2039 	 */
2040 	if (!msg_data_sz(msg)) {
2041 		struct msghdr m = {NULL,};
2042 
2043 		tsk_advance_rx_queue(sk);
2044 		__tipc_send_stream(new_sock, &m, 0);
2045 	} else {
2046 		__skb_dequeue(&sk->sk_receive_queue);
2047 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2048 		skb_set_owner_r(buf, new_sk);
2049 	}
2050 	release_sock(new_sk);
2051 exit:
2052 	release_sock(sk);
2053 	return res;
2054 }
2055 
2056 /**
2057  * tipc_shutdown - shutdown socket connection
2058  * @sock: socket structure
2059  * @how: direction to close (must be SHUT_RDWR)
2060  *
2061  * Terminates connection (if necessary), then purges socket's receive queue.
2062  *
2063  * Returns 0 on success, errno otherwise
2064  */
2065 static int tipc_shutdown(struct socket *sock, int how)
2066 {
2067 	struct sock *sk = sock->sk;
2068 	struct net *net = sock_net(sk);
2069 	struct tipc_sock *tsk = tipc_sk(sk);
2070 	struct sk_buff *skb;
2071 	u32 dnode;
2072 	int res;
2073 
2074 	if (how != SHUT_RDWR)
2075 		return -EINVAL;
2076 
2077 	lock_sock(sk);
2078 
2079 	switch (sock->state) {
2080 	case SS_CONNECTING:
2081 	case SS_CONNECTED:
2082 
2083 restart:
2084 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2085 		skb = __skb_dequeue(&sk->sk_receive_queue);
2086 		if (skb) {
2087 			if (TIPC_SKB_CB(skb)->handle != NULL) {
2088 				kfree_skb(skb);
2089 				goto restart;
2090 			}
2091 			if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
2092 					     TIPC_CONN_SHUTDOWN))
2093 				tipc_link_xmit_skb(net, skb, dnode,
2094 						   tsk->portid);
2095 		} else {
2096 			dnode = tsk_peer_node(tsk);
2097 
2098 			skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2099 					      TIPC_CONN_MSG, SHORT_H_SIZE,
2100 					      0, dnode, tsk_own_node(tsk),
2101 					      tsk_peer_port(tsk),
2102 					      tsk->portid, TIPC_CONN_SHUTDOWN);
2103 			tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2104 		}
2105 		tsk->connected = 0;
2106 		sock->state = SS_DISCONNECTING;
2107 		tipc_node_remove_conn(net, dnode, tsk->portid);
2108 		/* fall through */
2109 
2110 	case SS_DISCONNECTING:
2111 
2112 		/* Discard any unreceived messages */
2113 		__skb_queue_purge(&sk->sk_receive_queue);
2114 
2115 		/* Wake up anyone sleeping in poll */
2116 		sk->sk_state_change(sk);
2117 		res = 0;
2118 		break;
2119 
2120 	default:
2121 		res = -ENOTCONN;
2122 	}
2123 
2124 	release_sock(sk);
2125 	return res;
2126 }
2127 
2128 static void tipc_sk_timeout(unsigned long data)
2129 {
2130 	struct tipc_sock *tsk = (struct tipc_sock *)data;
2131 	struct sock *sk = &tsk->sk;
2132 	struct sk_buff *skb = NULL;
2133 	u32 peer_port, peer_node;
2134 	u32 own_node = tsk_own_node(tsk);
2135 
2136 	bh_lock_sock(sk);
2137 	if (!tsk->connected) {
2138 		bh_unlock_sock(sk);
2139 		goto exit;
2140 	}
2141 	peer_port = tsk_peer_port(tsk);
2142 	peer_node = tsk_peer_node(tsk);
2143 
2144 	if (tsk->probing_state == TIPC_CONN_PROBING) {
2145 		if (!sock_owned_by_user(sk)) {
2146 			sk->sk_socket->state = SS_DISCONNECTING;
2147 			tsk->connected = 0;
2148 			tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2149 					      tsk_peer_port(tsk));
2150 			sk->sk_state_change(sk);
2151 		} else {
2152 			/* Try again later */
2153 			sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2154 		}
2155 
2156 	} else {
2157 		skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2158 				      INT_H_SIZE, 0, peer_node, own_node,
2159 				      peer_port, tsk->portid, TIPC_OK);
2160 		tsk->probing_state = TIPC_CONN_PROBING;
2161 		sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2162 	}
2163 	bh_unlock_sock(sk);
2164 	if (skb)
2165 		tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2166 exit:
2167 	sock_put(sk);
2168 }
2169 
2170 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2171 			   struct tipc_name_seq const *seq)
2172 {
2173 	struct net *net = sock_net(&tsk->sk);
2174 	struct publication *publ;
2175 	u32 key;
2176 
2177 	if (tsk->connected)
2178 		return -EINVAL;
2179 	key = tsk->portid + tsk->pub_count + 1;
2180 	if (key == tsk->portid)
2181 		return -EADDRINUSE;
2182 
2183 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2184 				    scope, tsk->portid, key);
2185 	if (unlikely(!publ))
2186 		return -EINVAL;
2187 
2188 	list_add(&publ->pport_list, &tsk->publications);
2189 	tsk->pub_count++;
2190 	tsk->published = 1;
2191 	return 0;
2192 }
2193 
2194 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2195 			    struct tipc_name_seq const *seq)
2196 {
2197 	struct net *net = sock_net(&tsk->sk);
2198 	struct publication *publ;
2199 	struct publication *safe;
2200 	int rc = -EINVAL;
2201 
2202 	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2203 		if (seq) {
2204 			if (publ->scope != scope)
2205 				continue;
2206 			if (publ->type != seq->type)
2207 				continue;
2208 			if (publ->lower != seq->lower)
2209 				continue;
2210 			if (publ->upper != seq->upper)
2211 				break;
2212 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2213 					      publ->ref, publ->key);
2214 			rc = 0;
2215 			break;
2216 		}
2217 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2218 				      publ->ref, publ->key);
2219 		rc = 0;
2220 	}
2221 	if (list_empty(&tsk->publications))
2222 		tsk->published = 0;
2223 	return rc;
2224 }
2225 
2226 /* tipc_sk_reinit: set non-zero address in all existing sockets
2227  *                 when we go from standalone to network mode.
2228  */
2229 void tipc_sk_reinit(struct net *net)
2230 {
2231 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2232 	const struct bucket_table *tbl;
2233 	struct rhash_head *pos;
2234 	struct tipc_sock *tsk;
2235 	struct tipc_msg *msg;
2236 	int i;
2237 
2238 	rcu_read_lock();
2239 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2240 	for (i = 0; i < tbl->size; i++) {
2241 		rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2242 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2243 			msg = &tsk->phdr;
2244 			msg_set_prevnode(msg, tn->own_addr);
2245 			msg_set_orignode(msg, tn->own_addr);
2246 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2247 		}
2248 	}
2249 	rcu_read_unlock();
2250 }
2251 
2252 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2253 {
2254 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2255 	struct tipc_sock *tsk;
2256 
2257 	rcu_read_lock();
2258 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2259 	if (tsk)
2260 		sock_hold(&tsk->sk);
2261 	rcu_read_unlock();
2262 
2263 	return tsk;
2264 }
2265 
2266 static int tipc_sk_insert(struct tipc_sock *tsk)
2267 {
2268 	struct sock *sk = &tsk->sk;
2269 	struct net *net = sock_net(sk);
2270 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2271 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2272 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2273 
2274 	while (remaining--) {
2275 		portid++;
2276 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2277 			portid = TIPC_MIN_PORT;
2278 		tsk->portid = portid;
2279 		sock_hold(&tsk->sk);
2280 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2281 						   tsk_rht_params))
2282 			return 0;
2283 		sock_put(&tsk->sk);
2284 	}
2285 
2286 	return -1;
2287 }
2288 
2289 static void tipc_sk_remove(struct tipc_sock *tsk)
2290 {
2291 	struct sock *sk = &tsk->sk;
2292 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2293 
2294 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2295 		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2296 		__sock_put(sk);
2297 	}
2298 }
2299 
2300 static const struct rhashtable_params tsk_rht_params = {
2301 	.nelem_hint = 192,
2302 	.head_offset = offsetof(struct tipc_sock, node),
2303 	.key_offset = offsetof(struct tipc_sock, portid),
2304 	.key_len = sizeof(u32), /* portid */
2305 	.max_size = 1048576,
2306 	.min_size = 256,
2307 	.automatic_shrinking = true,
2308 };
2309 
2310 int tipc_sk_rht_init(struct net *net)
2311 {
2312 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2313 
2314 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2315 }
2316 
2317 void tipc_sk_rht_destroy(struct net *net)
2318 {
2319 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2320 
2321 	/* Wait for socket readers to complete */
2322 	synchronize_net();
2323 
2324 	rhashtable_destroy(&tn->sk_rht);
2325 }
2326 
2327 /**
2328  * tipc_setsockopt - set socket option
2329  * @sock: socket structure
2330  * @lvl: option level
2331  * @opt: option identifier
2332  * @ov: pointer to new option value
2333  * @ol: length of option value
2334  *
2335  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2336  * (to ease compatibility).
2337  *
2338  * Returns 0 on success, errno otherwise
2339  */
2340 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2341 			   char __user *ov, unsigned int ol)
2342 {
2343 	struct sock *sk = sock->sk;
2344 	struct tipc_sock *tsk = tipc_sk(sk);
2345 	u32 value;
2346 	int res;
2347 
2348 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2349 		return 0;
2350 	if (lvl != SOL_TIPC)
2351 		return -ENOPROTOOPT;
2352 	if (ol < sizeof(value))
2353 		return -EINVAL;
2354 	res = get_user(value, (u32 __user *)ov);
2355 	if (res)
2356 		return res;
2357 
2358 	lock_sock(sk);
2359 
2360 	switch (opt) {
2361 	case TIPC_IMPORTANCE:
2362 		res = tsk_set_importance(tsk, value);
2363 		break;
2364 	case TIPC_SRC_DROPPABLE:
2365 		if (sock->type != SOCK_STREAM)
2366 			tsk_set_unreliable(tsk, value);
2367 		else
2368 			res = -ENOPROTOOPT;
2369 		break;
2370 	case TIPC_DEST_DROPPABLE:
2371 		tsk_set_unreturnable(tsk, value);
2372 		break;
2373 	case TIPC_CONN_TIMEOUT:
2374 		tipc_sk(sk)->conn_timeout = value;
2375 		/* no need to set "res", since already 0 at this point */
2376 		break;
2377 	default:
2378 		res = -EINVAL;
2379 	}
2380 
2381 	release_sock(sk);
2382 
2383 	return res;
2384 }
2385 
2386 /**
2387  * tipc_getsockopt - get socket option
2388  * @sock: socket structure
2389  * @lvl: option level
2390  * @opt: option identifier
2391  * @ov: receptacle for option value
2392  * @ol: receptacle for length of option value
2393  *
2394  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2395  * (to ease compatibility).
2396  *
2397  * Returns 0 on success, errno otherwise
2398  */
2399 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2400 			   char __user *ov, int __user *ol)
2401 {
2402 	struct sock *sk = sock->sk;
2403 	struct tipc_sock *tsk = tipc_sk(sk);
2404 	int len;
2405 	u32 value;
2406 	int res;
2407 
2408 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2409 		return put_user(0, ol);
2410 	if (lvl != SOL_TIPC)
2411 		return -ENOPROTOOPT;
2412 	res = get_user(len, ol);
2413 	if (res)
2414 		return res;
2415 
2416 	lock_sock(sk);
2417 
2418 	switch (opt) {
2419 	case TIPC_IMPORTANCE:
2420 		value = tsk_importance(tsk);
2421 		break;
2422 	case TIPC_SRC_DROPPABLE:
2423 		value = tsk_unreliable(tsk);
2424 		break;
2425 	case TIPC_DEST_DROPPABLE:
2426 		value = tsk_unreturnable(tsk);
2427 		break;
2428 	case TIPC_CONN_TIMEOUT:
2429 		value = tsk->conn_timeout;
2430 		/* no need to set "res", since already 0 at this point */
2431 		break;
2432 	case TIPC_NODE_RECVQ_DEPTH:
2433 		value = 0; /* was tipc_queue_size, now obsolete */
2434 		break;
2435 	case TIPC_SOCK_RECVQ_DEPTH:
2436 		value = skb_queue_len(&sk->sk_receive_queue);
2437 		break;
2438 	default:
2439 		res = -EINVAL;
2440 	}
2441 
2442 	release_sock(sk);
2443 
2444 	if (res)
2445 		return res;	/* "get" failed */
2446 
2447 	if (len < sizeof(value))
2448 		return -EINVAL;
2449 
2450 	if (copy_to_user(ov, &value, sizeof(value)))
2451 		return -EFAULT;
2452 
2453 	return put_user(sizeof(value), ol);
2454 }
2455 
2456 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2457 {
2458 	struct sock *sk = sock->sk;
2459 	struct tipc_sioc_ln_req lnr;
2460 	void __user *argp = (void __user *)arg;
2461 
2462 	switch (cmd) {
2463 	case SIOCGETLINKNAME:
2464 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2465 			return -EFAULT;
2466 		if (!tipc_node_get_linkname(sock_net(sk),
2467 					    lnr.bearer_id & 0xffff, lnr.peer,
2468 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2469 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2470 				return -EFAULT;
2471 			return 0;
2472 		}
2473 		return -EADDRNOTAVAIL;
2474 	default:
2475 		return -ENOIOCTLCMD;
2476 	}
2477 }
2478 
2479 /* Protocol switches for the various types of TIPC sockets */
2480 
2481 static const struct proto_ops msg_ops = {
2482 	.owner		= THIS_MODULE,
2483 	.family		= AF_TIPC,
2484 	.release	= tipc_release,
2485 	.bind		= tipc_bind,
2486 	.connect	= tipc_connect,
2487 	.socketpair	= sock_no_socketpair,
2488 	.accept		= sock_no_accept,
2489 	.getname	= tipc_getname,
2490 	.poll		= tipc_poll,
2491 	.ioctl		= tipc_ioctl,
2492 	.listen		= sock_no_listen,
2493 	.shutdown	= tipc_shutdown,
2494 	.setsockopt	= tipc_setsockopt,
2495 	.getsockopt	= tipc_getsockopt,
2496 	.sendmsg	= tipc_sendmsg,
2497 	.recvmsg	= tipc_recvmsg,
2498 	.mmap		= sock_no_mmap,
2499 	.sendpage	= sock_no_sendpage
2500 };
2501 
2502 static const struct proto_ops packet_ops = {
2503 	.owner		= THIS_MODULE,
2504 	.family		= AF_TIPC,
2505 	.release	= tipc_release,
2506 	.bind		= tipc_bind,
2507 	.connect	= tipc_connect,
2508 	.socketpair	= sock_no_socketpair,
2509 	.accept		= tipc_accept,
2510 	.getname	= tipc_getname,
2511 	.poll		= tipc_poll,
2512 	.ioctl		= tipc_ioctl,
2513 	.listen		= tipc_listen,
2514 	.shutdown	= tipc_shutdown,
2515 	.setsockopt	= tipc_setsockopt,
2516 	.getsockopt	= tipc_getsockopt,
2517 	.sendmsg	= tipc_send_packet,
2518 	.recvmsg	= tipc_recvmsg,
2519 	.mmap		= sock_no_mmap,
2520 	.sendpage	= sock_no_sendpage
2521 };
2522 
2523 static const struct proto_ops stream_ops = {
2524 	.owner		= THIS_MODULE,
2525 	.family		= AF_TIPC,
2526 	.release	= tipc_release,
2527 	.bind		= tipc_bind,
2528 	.connect	= tipc_connect,
2529 	.socketpair	= sock_no_socketpair,
2530 	.accept		= tipc_accept,
2531 	.getname	= tipc_getname,
2532 	.poll		= tipc_poll,
2533 	.ioctl		= tipc_ioctl,
2534 	.listen		= tipc_listen,
2535 	.shutdown	= tipc_shutdown,
2536 	.setsockopt	= tipc_setsockopt,
2537 	.getsockopt	= tipc_getsockopt,
2538 	.sendmsg	= tipc_send_stream,
2539 	.recvmsg	= tipc_recv_stream,
2540 	.mmap		= sock_no_mmap,
2541 	.sendpage	= sock_no_sendpage
2542 };
2543 
2544 static const struct net_proto_family tipc_family_ops = {
2545 	.owner		= THIS_MODULE,
2546 	.family		= AF_TIPC,
2547 	.create		= tipc_sk_create
2548 };
2549 
2550 static struct proto tipc_proto = {
2551 	.name		= "TIPC",
2552 	.owner		= THIS_MODULE,
2553 	.obj_size	= sizeof(struct tipc_sock),
2554 	.sysctl_rmem	= sysctl_tipc_rmem
2555 };
2556 
2557 /**
2558  * tipc_socket_init - initialize TIPC socket interface
2559  *
2560  * Returns 0 on success, errno otherwise
2561  */
2562 int tipc_socket_init(void)
2563 {
2564 	int res;
2565 
2566 	res = proto_register(&tipc_proto, 1);
2567 	if (res) {
2568 		pr_err("Failed to register TIPC protocol type\n");
2569 		goto out;
2570 	}
2571 
2572 	res = sock_register(&tipc_family_ops);
2573 	if (res) {
2574 		pr_err("Failed to register TIPC socket type\n");
2575 		proto_unregister(&tipc_proto);
2576 		goto out;
2577 	}
2578  out:
2579 	return res;
2580 }
2581 
2582 /**
2583  * tipc_socket_stop - stop TIPC socket interface
2584  */
2585 void tipc_socket_stop(void)
2586 {
2587 	sock_unregister(tipc_family_ops.family);
2588 	proto_unregister(&tipc_proto);
2589 }
2590 
2591 /* Caller should hold socket lock for the passed tipc socket. */
2592 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2593 {
2594 	u32 peer_node;
2595 	u32 peer_port;
2596 	struct nlattr *nest;
2597 
2598 	peer_node = tsk_peer_node(tsk);
2599 	peer_port = tsk_peer_port(tsk);
2600 
2601 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2602 
2603 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2604 		goto msg_full;
2605 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2606 		goto msg_full;
2607 
2608 	if (tsk->conn_type != 0) {
2609 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2610 			goto msg_full;
2611 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2612 			goto msg_full;
2613 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2614 			goto msg_full;
2615 	}
2616 	nla_nest_end(skb, nest);
2617 
2618 	return 0;
2619 
2620 msg_full:
2621 	nla_nest_cancel(skb, nest);
2622 
2623 	return -EMSGSIZE;
2624 }
2625 
2626 /* Caller should hold socket lock for the passed tipc socket. */
2627 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2628 			    struct tipc_sock *tsk)
2629 {
2630 	int err;
2631 	void *hdr;
2632 	struct nlattr *attrs;
2633 	struct net *net = sock_net(skb->sk);
2634 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2635 
2636 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2637 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2638 	if (!hdr)
2639 		goto msg_cancel;
2640 
2641 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2642 	if (!attrs)
2643 		goto genlmsg_cancel;
2644 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2645 		goto attr_msg_cancel;
2646 	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2647 		goto attr_msg_cancel;
2648 
2649 	if (tsk->connected) {
2650 		err = __tipc_nl_add_sk_con(skb, tsk);
2651 		if (err)
2652 			goto attr_msg_cancel;
2653 	} else if (!list_empty(&tsk->publications)) {
2654 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2655 			goto attr_msg_cancel;
2656 	}
2657 	nla_nest_end(skb, attrs);
2658 	genlmsg_end(skb, hdr);
2659 
2660 	return 0;
2661 
2662 attr_msg_cancel:
2663 	nla_nest_cancel(skb, attrs);
2664 genlmsg_cancel:
2665 	genlmsg_cancel(skb, hdr);
2666 msg_cancel:
2667 	return -EMSGSIZE;
2668 }
2669 
2670 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2671 {
2672 	int err;
2673 	struct tipc_sock *tsk;
2674 	const struct bucket_table *tbl;
2675 	struct rhash_head *pos;
2676 	struct net *net = sock_net(skb->sk);
2677 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2678 	u32 tbl_id = cb->args[0];
2679 	u32 prev_portid = cb->args[1];
2680 
2681 	rcu_read_lock();
2682 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2683 	for (; tbl_id < tbl->size; tbl_id++) {
2684 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2685 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2686 			if (prev_portid && prev_portid != tsk->portid) {
2687 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2688 				continue;
2689 			}
2690 
2691 			err = __tipc_nl_add_sk(skb, cb, tsk);
2692 			if (err) {
2693 				prev_portid = tsk->portid;
2694 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2695 				goto out;
2696 			}
2697 			prev_portid = 0;
2698 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2699 		}
2700 	}
2701 out:
2702 	rcu_read_unlock();
2703 	cb->args[0] = tbl_id;
2704 	cb->args[1] = prev_portid;
2705 
2706 	return skb->len;
2707 }
2708 
2709 /* Caller should hold socket lock for the passed tipc socket. */
2710 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2711 				 struct netlink_callback *cb,
2712 				 struct publication *publ)
2713 {
2714 	void *hdr;
2715 	struct nlattr *attrs;
2716 
2717 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2718 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2719 	if (!hdr)
2720 		goto msg_cancel;
2721 
2722 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2723 	if (!attrs)
2724 		goto genlmsg_cancel;
2725 
2726 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2727 		goto attr_msg_cancel;
2728 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2729 		goto attr_msg_cancel;
2730 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2731 		goto attr_msg_cancel;
2732 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2733 		goto attr_msg_cancel;
2734 
2735 	nla_nest_end(skb, attrs);
2736 	genlmsg_end(skb, hdr);
2737 
2738 	return 0;
2739 
2740 attr_msg_cancel:
2741 	nla_nest_cancel(skb, attrs);
2742 genlmsg_cancel:
2743 	genlmsg_cancel(skb, hdr);
2744 msg_cancel:
2745 	return -EMSGSIZE;
2746 }
2747 
2748 /* Caller should hold socket lock for the passed tipc socket. */
2749 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2750 				  struct netlink_callback *cb,
2751 				  struct tipc_sock *tsk, u32 *last_publ)
2752 {
2753 	int err;
2754 	struct publication *p;
2755 
2756 	if (*last_publ) {
2757 		list_for_each_entry(p, &tsk->publications, pport_list) {
2758 			if (p->key == *last_publ)
2759 				break;
2760 		}
2761 		if (p->key != *last_publ) {
2762 			/* We never set seq or call nl_dump_check_consistent()
2763 			 * this means that setting prev_seq here will cause the
2764 			 * consistence check to fail in the netlink callback
2765 			 * handler. Resulting in the last NLMSG_DONE message
2766 			 * having the NLM_F_DUMP_INTR flag set.
2767 			 */
2768 			cb->prev_seq = 1;
2769 			*last_publ = 0;
2770 			return -EPIPE;
2771 		}
2772 	} else {
2773 		p = list_first_entry(&tsk->publications, struct publication,
2774 				     pport_list);
2775 	}
2776 
2777 	list_for_each_entry_from(p, &tsk->publications, pport_list) {
2778 		err = __tipc_nl_add_sk_publ(skb, cb, p);
2779 		if (err) {
2780 			*last_publ = p->key;
2781 			return err;
2782 		}
2783 	}
2784 	*last_publ = 0;
2785 
2786 	return 0;
2787 }
2788 
2789 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2790 {
2791 	int err;
2792 	u32 tsk_portid = cb->args[0];
2793 	u32 last_publ = cb->args[1];
2794 	u32 done = cb->args[2];
2795 	struct net *net = sock_net(skb->sk);
2796 	struct tipc_sock *tsk;
2797 
2798 	if (!tsk_portid) {
2799 		struct nlattr **attrs;
2800 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2801 
2802 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
2803 		if (err)
2804 			return err;
2805 
2806 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2807 				       attrs[TIPC_NLA_SOCK],
2808 				       tipc_nl_sock_policy);
2809 		if (err)
2810 			return err;
2811 
2812 		if (!sock[TIPC_NLA_SOCK_REF])
2813 			return -EINVAL;
2814 
2815 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2816 	}
2817 
2818 	if (done)
2819 		return 0;
2820 
2821 	tsk = tipc_sk_lookup(net, tsk_portid);
2822 	if (!tsk)
2823 		return -EINVAL;
2824 
2825 	lock_sock(&tsk->sk);
2826 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2827 	if (!err)
2828 		done = 1;
2829 	release_sock(&tsk->sk);
2830 	sock_put(&tsk->sk);
2831 
2832 	cb->args[0] = tsk_portid;
2833 	cb->args[1] = last_publ;
2834 	cb->args[2] = done;
2835 
2836 	return skb->len;
2837 }
2838