xref: /openbmc/linux/net/tipc/socket.c (revision a03a8dbe20eff6d57aae3147577bf84b52aba4e6)
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 <linux/jhash.h>
39 #include "core.h"
40 #include "name_table.h"
41 #include "node.h"
42 #include "link.h"
43 #include "name_distr.h"
44 #include "socket.h"
45 
46 #define SS_LISTENING		-1	/* socket is listening */
47 #define SS_READY		-2	/* socket is connectionless */
48 
49 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
50 #define CONN_PROBING_INTERVAL	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
51 #define TIPC_FWD_MSG		1
52 #define TIPC_CONN_OK		0
53 #define TIPC_CONN_PROBING	1
54 #define TIPC_MAX_PORT		0xffffffff
55 #define TIPC_MIN_PORT		1
56 
57 /**
58  * struct tipc_sock - TIPC socket structure
59  * @sk: socket - interacts with 'port' and with user via the socket API
60  * @connected: non-zero if port is currently connected to a peer port
61  * @conn_type: TIPC type used when connection was established
62  * @conn_instance: TIPC instance used when connection was established
63  * @published: non-zero if port has one or more associated names
64  * @max_pkt: maximum packet size "hint" used when building messages sent by port
65  * @portid: unique port identity in TIPC socket hash table
66  * @phdr: preformatted message header used when sending messages
67  * @port_list: adjacent ports in TIPC's global list of ports
68  * @publications: list of publications for port
69  * @pub_count: total # of publications port has made during its lifetime
70  * @probing_state:
71  * @probing_intv:
72  * @conn_timeout: the time we can wait for an unresponded setup request
73  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
74  * @link_cong: non-zero if owner must sleep because of link congestion
75  * @sent_unacked: # messages sent by socket, and not yet acked by peer
76  * @rcv_unacked: # messages read by user, but not yet acked back to peer
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 rhash_head node;
100 	struct rcu_head rcu;
101 };
102 
103 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
104 static void tipc_data_ready(struct sock *sk);
105 static void tipc_write_space(struct sock *sk);
106 static int tipc_release(struct socket *sock);
107 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
108 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
109 static void tipc_sk_timeout(unsigned long data);
110 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
111 			   struct tipc_name_seq const *seq);
112 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
113 			    struct tipc_name_seq const *seq);
114 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
115 static int tipc_sk_insert(struct tipc_sock *tsk);
116 static void tipc_sk_remove(struct tipc_sock *tsk);
117 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
118 			      size_t dsz);
119 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
120 
121 static const struct proto_ops packet_ops;
122 static const struct proto_ops stream_ops;
123 static const struct proto_ops msg_ops;
124 
125 static struct proto tipc_proto;
126 static struct proto tipc_proto_kern;
127 
128 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
129 	[TIPC_NLA_SOCK_UNSPEC]		= { .type = NLA_UNSPEC },
130 	[TIPC_NLA_SOCK_ADDR]		= { .type = NLA_U32 },
131 	[TIPC_NLA_SOCK_REF]		= { .type = NLA_U32 },
132 	[TIPC_NLA_SOCK_CON]		= { .type = NLA_NESTED },
133 	[TIPC_NLA_SOCK_HAS_PUBL]	= { .type = NLA_FLAG }
134 };
135 
136 /*
137  * Revised TIPC socket locking policy:
138  *
139  * Most socket operations take the standard socket lock when they start
140  * and hold it until they finish (or until they need to sleep).  Acquiring
141  * this lock grants the owner exclusive access to the fields of the socket
142  * data structures, with the exception of the backlog queue.  A few socket
143  * operations can be done without taking the socket lock because they only
144  * read socket information that never changes during the life of the socket.
145  *
146  * Socket operations may acquire the lock for the associated TIPC port if they
147  * need to perform an operation on the port.  If any routine needs to acquire
148  * both the socket lock and the port lock it must take the socket lock first
149  * to avoid the risk of deadlock.
150  *
151  * The dispatcher handling incoming messages cannot grab the socket lock in
152  * the standard fashion, since invoked it runs at the BH level and cannot block.
153  * Instead, it checks to see if the socket lock is currently owned by someone,
154  * and either handles the message itself or adds it to the socket's backlog
155  * queue; in the latter case the queued message is processed once the process
156  * owning the socket lock releases it.
157  *
158  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
159  * the problem of a blocked socket operation preventing any other operations
160  * from occurring.  However, applications must be careful if they have
161  * multiple threads trying to send (or receive) on the same socket, as these
162  * operations might interfere with each other.  For example, doing a connect
163  * and a receive at the same time might allow the receive to consume the
164  * ACK message meant for the connect.  While additional work could be done
165  * to try and overcome this, it doesn't seem to be worthwhile at the present.
166  *
167  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
168  * that another operation that must be performed in a non-blocking manner is
169  * not delayed for very long because the lock has already been taken.
170  *
171  * NOTE: This code assumes that certain fields of a port/socket pair are
172  * constant over its lifetime; such fields can be examined without taking
173  * the socket lock and/or port lock, and do not need to be re-read even
174  * after resuming processing after waiting.  These fields include:
175  *   - socket type
176  *   - pointer to socket sk structure (aka tipc_sock structure)
177  *   - pointer to port structure
178  *   - port reference
179  */
180 
181 static u32 tsk_own_node(struct tipc_sock *tsk)
182 {
183 	return msg_prevnode(&tsk->phdr);
184 }
185 
186 static u32 tsk_peer_node(struct tipc_sock *tsk)
187 {
188 	return msg_destnode(&tsk->phdr);
189 }
190 
191 static u32 tsk_peer_port(struct tipc_sock *tsk)
192 {
193 	return msg_destport(&tsk->phdr);
194 }
195 
196 static  bool tsk_unreliable(struct tipc_sock *tsk)
197 {
198 	return msg_src_droppable(&tsk->phdr) != 0;
199 }
200 
201 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
202 {
203 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
204 }
205 
206 static bool tsk_unreturnable(struct tipc_sock *tsk)
207 {
208 	return msg_dest_droppable(&tsk->phdr) != 0;
209 }
210 
211 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
212 {
213 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
214 }
215 
216 static int tsk_importance(struct tipc_sock *tsk)
217 {
218 	return msg_importance(&tsk->phdr);
219 }
220 
221 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
222 {
223 	if (imp > TIPC_CRITICAL_IMPORTANCE)
224 		return -EINVAL;
225 	msg_set_importance(&tsk->phdr, (u32)imp);
226 	return 0;
227 }
228 
229 static struct tipc_sock *tipc_sk(const struct sock *sk)
230 {
231 	return container_of(sk, struct tipc_sock, sk);
232 }
233 
234 static int tsk_conn_cong(struct tipc_sock *tsk)
235 {
236 	return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
237 }
238 
239 /**
240  * tsk_advance_rx_queue - discard first buffer in socket receive queue
241  *
242  * Caller must hold socket lock
243  */
244 static void tsk_advance_rx_queue(struct sock *sk)
245 {
246 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
247 }
248 
249 /**
250  * tsk_rej_rx_queue - reject all buffers in socket receive queue
251  *
252  * Caller must hold socket lock
253  */
254 static void tsk_rej_rx_queue(struct sock *sk)
255 {
256 	struct sk_buff *skb;
257 	u32 dnode;
258 	u32 own_node = tsk_own_node(tipc_sk(sk));
259 
260 	while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
261 		if (tipc_msg_reverse(own_node, skb, &dnode, TIPC_ERR_NO_PORT))
262 			tipc_link_xmit_skb(sock_net(sk), skb, dnode, 0);
263 	}
264 }
265 
266 /* tsk_peer_msg - verify if message was sent by connected port's peer
267  *
268  * Handles cases where the node's network address has changed from
269  * the default of <0.0.0> to its configured setting.
270  */
271 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
272 {
273 	struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
274 	u32 peer_port = tsk_peer_port(tsk);
275 	u32 orig_node;
276 	u32 peer_node;
277 
278 	if (unlikely(!tsk->connected))
279 		return false;
280 
281 	if (unlikely(msg_origport(msg) != peer_port))
282 		return false;
283 
284 	orig_node = msg_orignode(msg);
285 	peer_node = tsk_peer_node(tsk);
286 
287 	if (likely(orig_node == peer_node))
288 		return true;
289 
290 	if (!orig_node && (peer_node == tn->own_addr))
291 		return true;
292 
293 	if (!peer_node && (orig_node == tn->own_addr))
294 		return true;
295 
296 	return false;
297 }
298 
299 /**
300  * tipc_sk_create - create a TIPC socket
301  * @net: network namespace (must be default network)
302  * @sock: pre-allocated socket structure
303  * @protocol: protocol indicator (must be 0)
304  * @kern: caused by kernel or by userspace?
305  *
306  * This routine creates additional data structures used by the TIPC socket,
307  * initializes them, and links them together.
308  *
309  * Returns 0 on success, errno otherwise
310  */
311 static int tipc_sk_create(struct net *net, struct socket *sock,
312 			  int protocol, int kern)
313 {
314 	struct tipc_net *tn;
315 	const struct proto_ops *ops;
316 	socket_state state;
317 	struct sock *sk;
318 	struct tipc_sock *tsk;
319 	struct tipc_msg *msg;
320 
321 	/* Validate arguments */
322 	if (unlikely(protocol != 0))
323 		return -EPROTONOSUPPORT;
324 
325 	switch (sock->type) {
326 	case SOCK_STREAM:
327 		ops = &stream_ops;
328 		state = SS_UNCONNECTED;
329 		break;
330 	case SOCK_SEQPACKET:
331 		ops = &packet_ops;
332 		state = SS_UNCONNECTED;
333 		break;
334 	case SOCK_DGRAM:
335 	case SOCK_RDM:
336 		ops = &msg_ops;
337 		state = SS_READY;
338 		break;
339 	default:
340 		return -EPROTOTYPE;
341 	}
342 
343 	/* Allocate socket's protocol area */
344 	if (!kern)
345 		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
346 	else
347 		sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
348 
349 	if (sk == NULL)
350 		return -ENOMEM;
351 
352 	tsk = tipc_sk(sk);
353 	tsk->max_pkt = MAX_PKT_DEFAULT;
354 	INIT_LIST_HEAD(&tsk->publications);
355 	msg = &tsk->phdr;
356 	tn = net_generic(sock_net(sk), tipc_net_id);
357 	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
358 		      NAMED_H_SIZE, 0);
359 
360 	/* Finish initializing socket data structures */
361 	sock->ops = ops;
362 	sock->state = state;
363 	sock_init_data(sock, sk);
364 	if (tipc_sk_insert(tsk)) {
365 		pr_warn("Socket create failed; port numbrer exhausted\n");
366 		return -EINVAL;
367 	}
368 	msg_set_origport(msg, tsk->portid);
369 	setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
370 	sk->sk_backlog_rcv = tipc_backlog_rcv;
371 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
372 	sk->sk_data_ready = tipc_data_ready;
373 	sk->sk_write_space = tipc_write_space;
374 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
375 	tsk->sent_unacked = 0;
376 	atomic_set(&tsk->dupl_rcvcnt, 0);
377 
378 	if (sock->state == SS_READY) {
379 		tsk_set_unreturnable(tsk, true);
380 		if (sock->type == SOCK_DGRAM)
381 			tsk_set_unreliable(tsk, true);
382 	}
383 	return 0;
384 }
385 
386 /**
387  * tipc_sock_create_local - create TIPC socket from inside TIPC module
388  * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
389  *
390  * We cannot use sock_creat_kern here because it bumps module user count.
391  * Since socket owner and creator is the same module we must make sure
392  * that module count remains zero for module local sockets, otherwise
393  * we cannot do rmmod.
394  *
395  * Returns 0 on success, errno otherwise
396  */
397 int tipc_sock_create_local(struct net *net, int type, struct socket **res)
398 {
399 	int rc;
400 
401 	rc = sock_create_lite(AF_TIPC, type, 0, res);
402 	if (rc < 0) {
403 		pr_err("Failed to create kernel socket\n");
404 		return rc;
405 	}
406 	tipc_sk_create(net, *res, 0, 1);
407 
408 	return 0;
409 }
410 
411 /**
412  * tipc_sock_release_local - release socket created by tipc_sock_create_local
413  * @sock: the socket to be released.
414  *
415  * Module reference count is not incremented when such sockets are created,
416  * so we must keep it from being decremented when they are released.
417  */
418 void tipc_sock_release_local(struct socket *sock)
419 {
420 	tipc_release(sock);
421 	sock->ops = NULL;
422 	sock_release(sock);
423 }
424 
425 /**
426  * tipc_sock_accept_local - accept a connection on a socket created
427  * with tipc_sock_create_local. Use this function to avoid that
428  * module reference count is inadvertently incremented.
429  *
430  * @sock:    the accepting socket
431  * @newsock: reference to the new socket to be created
432  * @flags:   socket flags
433  */
434 
435 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
436 			   int flags)
437 {
438 	struct sock *sk = sock->sk;
439 	int ret;
440 
441 	ret = sock_create_lite(sk->sk_family, sk->sk_type,
442 			       sk->sk_protocol, newsock);
443 	if (ret < 0)
444 		return ret;
445 
446 	ret = tipc_accept(sock, *newsock, flags);
447 	if (ret < 0) {
448 		sock_release(*newsock);
449 		return ret;
450 	}
451 	(*newsock)->ops = sock->ops;
452 	return ret;
453 }
454 
455 static void tipc_sk_callback(struct rcu_head *head)
456 {
457 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
458 
459 	sock_put(&tsk->sk);
460 }
461 
462 /**
463  * tipc_release - destroy a TIPC socket
464  * @sock: socket to destroy
465  *
466  * This routine cleans up any messages that are still queued on the socket.
467  * For DGRAM and RDM socket types, all queued messages are rejected.
468  * For SEQPACKET and STREAM socket types, the first message is rejected
469  * and any others are discarded.  (If the first message on a STREAM socket
470  * is partially-read, it is discarded and the next one is rejected instead.)
471  *
472  * NOTE: Rejected messages are not necessarily returned to the sender!  They
473  * are returned or discarded according to the "destination droppable" setting
474  * specified for the message by the sender.
475  *
476  * Returns 0 on success, errno otherwise
477  */
478 static int tipc_release(struct socket *sock)
479 {
480 	struct sock *sk = sock->sk;
481 	struct net *net;
482 	struct tipc_sock *tsk;
483 	struct sk_buff *skb;
484 	u32 dnode, probing_state;
485 
486 	/*
487 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
488 	 * releases a pre-allocated child socket that was never used)
489 	 */
490 	if (sk == NULL)
491 		return 0;
492 
493 	net = sock_net(sk);
494 	tsk = tipc_sk(sk);
495 	lock_sock(sk);
496 
497 	/*
498 	 * Reject all unreceived messages, except on an active connection
499 	 * (which disconnects locally & sends a 'FIN+' to peer)
500 	 */
501 	dnode = tsk_peer_node(tsk);
502 	while (sock->state != SS_DISCONNECTING) {
503 		skb = __skb_dequeue(&sk->sk_receive_queue);
504 		if (skb == NULL)
505 			break;
506 		if (TIPC_SKB_CB(skb)->handle != NULL)
507 			kfree_skb(skb);
508 		else {
509 			if ((sock->state == SS_CONNECTING) ||
510 			    (sock->state == SS_CONNECTED)) {
511 				sock->state = SS_DISCONNECTING;
512 				tsk->connected = 0;
513 				tipc_node_remove_conn(net, dnode, tsk->portid);
514 			}
515 			if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
516 					     TIPC_ERR_NO_PORT))
517 				tipc_link_xmit_skb(net, skb, dnode, 0);
518 		}
519 	}
520 
521 	tipc_sk_withdraw(tsk, 0, NULL);
522 	probing_state = tsk->probing_state;
523 	if (del_timer_sync(&sk->sk_timer) &&
524 	    probing_state != TIPC_CONN_PROBING)
525 		sock_put(sk);
526 	tipc_sk_remove(tsk);
527 	if (tsk->connected) {
528 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
529 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
530 				      tsk_own_node(tsk), tsk_peer_port(tsk),
531 				      tsk->portid, TIPC_ERR_NO_PORT);
532 		if (skb)
533 			tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
534 		tipc_node_remove_conn(net, dnode, tsk->portid);
535 	}
536 
537 	/* Discard any remaining (connection-based) messages in receive queue */
538 	__skb_queue_purge(&sk->sk_receive_queue);
539 
540 	/* Reject any messages that accumulated in backlog queue */
541 	sock->state = SS_DISCONNECTING;
542 	release_sock(sk);
543 
544 	call_rcu(&tsk->rcu, tipc_sk_callback);
545 	sock->sk = NULL;
546 
547 	return 0;
548 }
549 
550 /**
551  * tipc_bind - associate or disassocate TIPC name(s) with a socket
552  * @sock: socket structure
553  * @uaddr: socket address describing name(s) and desired operation
554  * @uaddr_len: size of socket address data structure
555  *
556  * Name and name sequence binding is indicated using a positive scope value;
557  * a negative scope value unbinds the specified name.  Specifying no name
558  * (i.e. a socket address length of 0) unbinds all names from the socket.
559  *
560  * Returns 0 on success, errno otherwise
561  *
562  * NOTE: This routine doesn't need to take the socket lock since it doesn't
563  *       access any non-constant socket information.
564  */
565 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
566 		     int uaddr_len)
567 {
568 	struct sock *sk = sock->sk;
569 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
570 	struct tipc_sock *tsk = tipc_sk(sk);
571 	int res = -EINVAL;
572 
573 	lock_sock(sk);
574 	if (unlikely(!uaddr_len)) {
575 		res = tipc_sk_withdraw(tsk, 0, NULL);
576 		goto exit;
577 	}
578 
579 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
580 		res = -EINVAL;
581 		goto exit;
582 	}
583 	if (addr->family != AF_TIPC) {
584 		res = -EAFNOSUPPORT;
585 		goto exit;
586 	}
587 
588 	if (addr->addrtype == TIPC_ADDR_NAME)
589 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
590 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
591 		res = -EAFNOSUPPORT;
592 		goto exit;
593 	}
594 
595 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
596 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
597 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
598 		res = -EACCES;
599 		goto exit;
600 	}
601 
602 	res = (addr->scope > 0) ?
603 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
604 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
605 exit:
606 	release_sock(sk);
607 	return res;
608 }
609 
610 /**
611  * tipc_getname - get port ID of socket or peer socket
612  * @sock: socket structure
613  * @uaddr: area for returned socket address
614  * @uaddr_len: area for returned length of socket address
615  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
616  *
617  * Returns 0 on success, errno otherwise
618  *
619  * NOTE: This routine doesn't need to take the socket lock since it only
620  *       accesses socket information that is unchanging (or which changes in
621  *       a completely predictable manner).
622  */
623 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
624 			int *uaddr_len, int peer)
625 {
626 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
627 	struct tipc_sock *tsk = tipc_sk(sock->sk);
628 	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
629 
630 	memset(addr, 0, sizeof(*addr));
631 	if (peer) {
632 		if ((sock->state != SS_CONNECTED) &&
633 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
634 			return -ENOTCONN;
635 		addr->addr.id.ref = tsk_peer_port(tsk);
636 		addr->addr.id.node = tsk_peer_node(tsk);
637 	} else {
638 		addr->addr.id.ref = tsk->portid;
639 		addr->addr.id.node = tn->own_addr;
640 	}
641 
642 	*uaddr_len = sizeof(*addr);
643 	addr->addrtype = TIPC_ADDR_ID;
644 	addr->family = AF_TIPC;
645 	addr->scope = 0;
646 	addr->addr.name.domain = 0;
647 
648 	return 0;
649 }
650 
651 /**
652  * tipc_poll - read and possibly block on pollmask
653  * @file: file structure associated with the socket
654  * @sock: socket for which to calculate the poll bits
655  * @wait: ???
656  *
657  * Returns pollmask value
658  *
659  * COMMENTARY:
660  * It appears that the usual socket locking mechanisms are not useful here
661  * since the pollmask info is potentially out-of-date the moment this routine
662  * exits.  TCP and other protocols seem to rely on higher level poll routines
663  * to handle any preventable race conditions, so TIPC will do the same ...
664  *
665  * TIPC sets the returned events as follows:
666  *
667  * socket state		flags set
668  * ------------		---------
669  * unconnected		no read flags
670  *			POLLOUT if port is not congested
671  *
672  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
673  *			no write flags
674  *
675  * connected		POLLIN/POLLRDNORM if data in rx queue
676  *			POLLOUT if port is not congested
677  *
678  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
679  *			no write flags
680  *
681  * listening		POLLIN if SYN in rx queue
682  *			no write flags
683  *
684  * ready		POLLIN/POLLRDNORM if data in rx queue
685  * [connectionless]	POLLOUT (since port cannot be congested)
686  *
687  * IMPORTANT: The fact that a read or write operation is indicated does NOT
688  * imply that the operation will succeed, merely that it should be performed
689  * and will not block.
690  */
691 static unsigned int tipc_poll(struct file *file, struct socket *sock,
692 			      poll_table *wait)
693 {
694 	struct sock *sk = sock->sk;
695 	struct tipc_sock *tsk = tipc_sk(sk);
696 	u32 mask = 0;
697 
698 	sock_poll_wait(file, sk_sleep(sk), wait);
699 
700 	switch ((int)sock->state) {
701 	case SS_UNCONNECTED:
702 		if (!tsk->link_cong)
703 			mask |= POLLOUT;
704 		break;
705 	case SS_READY:
706 	case SS_CONNECTED:
707 		if (!tsk->link_cong && !tsk_conn_cong(tsk))
708 			mask |= POLLOUT;
709 		/* fall thru' */
710 	case SS_CONNECTING:
711 	case SS_LISTENING:
712 		if (!skb_queue_empty(&sk->sk_receive_queue))
713 			mask |= (POLLIN | POLLRDNORM);
714 		break;
715 	case SS_DISCONNECTING:
716 		mask = (POLLIN | POLLRDNORM | POLLHUP);
717 		break;
718 	}
719 
720 	return mask;
721 }
722 
723 /**
724  * tipc_sendmcast - send multicast message
725  * @sock: socket structure
726  * @seq: destination address
727  * @msg: message to send
728  * @dsz: total length of message data
729  * @timeo: timeout to wait for wakeup
730  *
731  * Called from function tipc_sendmsg(), which has done all sanity checks
732  * Returns the number of bytes sent on success, or errno
733  */
734 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
735 			  struct msghdr *msg, size_t dsz, long timeo)
736 {
737 	struct sock *sk = sock->sk;
738 	struct tipc_sock *tsk = tipc_sk(sk);
739 	struct net *net = sock_net(sk);
740 	struct tipc_msg *mhdr = &tsk->phdr;
741 	struct sk_buff_head *pktchain = &sk->sk_write_queue;
742 	struct iov_iter save = msg->msg_iter;
743 	uint mtu;
744 	int rc;
745 
746 	msg_set_type(mhdr, TIPC_MCAST_MSG);
747 	msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
748 	msg_set_destport(mhdr, 0);
749 	msg_set_destnode(mhdr, 0);
750 	msg_set_nametype(mhdr, seq->type);
751 	msg_set_namelower(mhdr, seq->lower);
752 	msg_set_nameupper(mhdr, seq->upper);
753 	msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
754 
755 new_mtu:
756 	mtu = tipc_bclink_get_mtu();
757 	rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, pktchain);
758 	if (unlikely(rc < 0))
759 		return rc;
760 
761 	do {
762 		rc = tipc_bclink_xmit(net, pktchain);
763 		if (likely(rc >= 0)) {
764 			rc = dsz;
765 			break;
766 		}
767 		if (rc == -EMSGSIZE) {
768 			msg->msg_iter = save;
769 			goto new_mtu;
770 		}
771 		if (rc != -ELINKCONG)
772 			break;
773 		tipc_sk(sk)->link_cong = 1;
774 		rc = tipc_wait_for_sndmsg(sock, &timeo);
775 		if (rc)
776 			__skb_queue_purge(pktchain);
777 	} while (!rc);
778 	return rc;
779 }
780 
781 /**
782  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
783  * @arrvq: queue with arriving messages, to be cloned after destination lookup
784  * @inputq: queue with cloned messages, delivered to socket after dest lookup
785  *
786  * Multi-threaded: parallel calls with reference to same queues may occur
787  */
788 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
789 		       struct sk_buff_head *inputq)
790 {
791 	struct tipc_msg *msg;
792 	struct tipc_plist dports;
793 	u32 portid;
794 	u32 scope = TIPC_CLUSTER_SCOPE;
795 	struct sk_buff_head tmpq;
796 	uint hsz;
797 	struct sk_buff *skb, *_skb;
798 
799 	__skb_queue_head_init(&tmpq);
800 	tipc_plist_init(&dports);
801 
802 	skb = tipc_skb_peek(arrvq, &inputq->lock);
803 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
804 		msg = buf_msg(skb);
805 		hsz = skb_headroom(skb) + msg_hdr_sz(msg);
806 
807 		if (in_own_node(net, msg_orignode(msg)))
808 			scope = TIPC_NODE_SCOPE;
809 
810 		/* Create destination port list and message clones: */
811 		tipc_nametbl_mc_translate(net,
812 					  msg_nametype(msg), msg_namelower(msg),
813 					  msg_nameupper(msg), scope, &dports);
814 		portid = tipc_plist_pop(&dports);
815 		for (; portid; portid = tipc_plist_pop(&dports)) {
816 			_skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
817 			if (_skb) {
818 				msg_set_destport(buf_msg(_skb), portid);
819 				__skb_queue_tail(&tmpq, _skb);
820 				continue;
821 			}
822 			pr_warn("Failed to clone mcast rcv buffer\n");
823 		}
824 		/* Append to inputq if not already done by other thread */
825 		spin_lock_bh(&inputq->lock);
826 		if (skb_peek(arrvq) == skb) {
827 			skb_queue_splice_tail_init(&tmpq, inputq);
828 			kfree_skb(__skb_dequeue(arrvq));
829 		}
830 		spin_unlock_bh(&inputq->lock);
831 		__skb_queue_purge(&tmpq);
832 		kfree_skb(skb);
833 	}
834 	tipc_sk_rcv(net, inputq);
835 }
836 
837 /**
838  * tipc_sk_proto_rcv - receive a connection mng protocol message
839  * @tsk: receiving socket
840  * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
841  */
842 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff **skb)
843 {
844 	struct tipc_msg *msg = buf_msg(*skb);
845 	int conn_cong;
846 	u32 dnode;
847 	u32 own_node = tsk_own_node(tsk);
848 	/* Ignore if connection cannot be validated: */
849 	if (!tsk_peer_msg(tsk, msg))
850 		goto exit;
851 
852 	tsk->probing_state = TIPC_CONN_OK;
853 
854 	if (msg_type(msg) == CONN_ACK) {
855 		conn_cong = tsk_conn_cong(tsk);
856 		tsk->sent_unacked -= msg_msgcnt(msg);
857 		if (conn_cong)
858 			tsk->sk.sk_write_space(&tsk->sk);
859 	} else if (msg_type(msg) == CONN_PROBE) {
860 		if (tipc_msg_reverse(own_node, *skb, &dnode, TIPC_OK)) {
861 			msg_set_type(msg, CONN_PROBE_REPLY);
862 			return;
863 		}
864 	}
865 	/* Do nothing if msg_type() == CONN_PROBE_REPLY */
866 exit:
867 	kfree_skb(*skb);
868 	*skb = NULL;
869 }
870 
871 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
872 {
873 	struct sock *sk = sock->sk;
874 	struct tipc_sock *tsk = tipc_sk(sk);
875 	DEFINE_WAIT(wait);
876 	int done;
877 
878 	do {
879 		int err = sock_error(sk);
880 		if (err)
881 			return err;
882 		if (sock->state == SS_DISCONNECTING)
883 			return -EPIPE;
884 		if (!*timeo_p)
885 			return -EAGAIN;
886 		if (signal_pending(current))
887 			return sock_intr_errno(*timeo_p);
888 
889 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
890 		done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
891 		finish_wait(sk_sleep(sk), &wait);
892 	} while (!done);
893 	return 0;
894 }
895 
896 /**
897  * tipc_sendmsg - send message in connectionless manner
898  * @sock: socket structure
899  * @m: message to send
900  * @dsz: amount of user data to be sent
901  *
902  * Message must have an destination specified explicitly.
903  * Used for SOCK_RDM and SOCK_DGRAM messages,
904  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
905  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
906  *
907  * Returns the number of bytes sent on success, or errno otherwise
908  */
909 static int tipc_sendmsg(struct socket *sock,
910 			struct msghdr *m, size_t dsz)
911 {
912 	struct sock *sk = sock->sk;
913 	int ret;
914 
915 	lock_sock(sk);
916 	ret = __tipc_sendmsg(sock, m, dsz);
917 	release_sock(sk);
918 
919 	return ret;
920 }
921 
922 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
923 {
924 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
925 	struct sock *sk = sock->sk;
926 	struct tipc_sock *tsk = tipc_sk(sk);
927 	struct net *net = sock_net(sk);
928 	struct tipc_msg *mhdr = &tsk->phdr;
929 	u32 dnode, dport;
930 	struct sk_buff_head *pktchain = &sk->sk_write_queue;
931 	struct sk_buff *skb;
932 	struct tipc_name_seq *seq = &dest->addr.nameseq;
933 	struct iov_iter save;
934 	u32 mtu;
935 	long timeo;
936 	int rc;
937 
938 	if (unlikely(!dest))
939 		return -EDESTADDRREQ;
940 
941 	if (unlikely((m->msg_namelen < sizeof(*dest)) ||
942 		     (dest->family != AF_TIPC)))
943 		return -EINVAL;
944 
945 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
946 		return -EMSGSIZE;
947 
948 	if (unlikely(sock->state != SS_READY)) {
949 		if (sock->state == SS_LISTENING)
950 			return -EPIPE;
951 		if (sock->state != SS_UNCONNECTED)
952 			return -EISCONN;
953 		if (tsk->published)
954 			return -EOPNOTSUPP;
955 		if (dest->addrtype == TIPC_ADDR_NAME) {
956 			tsk->conn_type = dest->addr.name.name.type;
957 			tsk->conn_instance = dest->addr.name.name.instance;
958 		}
959 	}
960 
961 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
962 
963 	if (dest->addrtype == TIPC_ADDR_MCAST) {
964 		return tipc_sendmcast(sock, seq, m, dsz, timeo);
965 	} else if (dest->addrtype == TIPC_ADDR_NAME) {
966 		u32 type = dest->addr.name.name.type;
967 		u32 inst = dest->addr.name.name.instance;
968 		u32 domain = dest->addr.name.domain;
969 
970 		dnode = domain;
971 		msg_set_type(mhdr, TIPC_NAMED_MSG);
972 		msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
973 		msg_set_nametype(mhdr, type);
974 		msg_set_nameinst(mhdr, inst);
975 		msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
976 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
977 		msg_set_destnode(mhdr, dnode);
978 		msg_set_destport(mhdr, dport);
979 		if (unlikely(!dport && !dnode))
980 			return -EHOSTUNREACH;
981 	} else if (dest->addrtype == TIPC_ADDR_ID) {
982 		dnode = dest->addr.id.node;
983 		msg_set_type(mhdr, TIPC_DIRECT_MSG);
984 		msg_set_lookup_scope(mhdr, 0);
985 		msg_set_destnode(mhdr, dnode);
986 		msg_set_destport(mhdr, dest->addr.id.ref);
987 		msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
988 	}
989 
990 	save = m->msg_iter;
991 new_mtu:
992 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
993 	rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, pktchain);
994 	if (rc < 0)
995 		return rc;
996 
997 	do {
998 		skb = skb_peek(pktchain);
999 		TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
1000 		rc = tipc_link_xmit(net, pktchain, dnode, tsk->portid);
1001 		if (likely(rc >= 0)) {
1002 			if (sock->state != SS_READY)
1003 				sock->state = SS_CONNECTING;
1004 			rc = dsz;
1005 			break;
1006 		}
1007 		if (rc == -EMSGSIZE) {
1008 			m->msg_iter = save;
1009 			goto new_mtu;
1010 		}
1011 		if (rc != -ELINKCONG)
1012 			break;
1013 		tsk->link_cong = 1;
1014 		rc = tipc_wait_for_sndmsg(sock, &timeo);
1015 		if (rc)
1016 			__skb_queue_purge(pktchain);
1017 	} while (!rc);
1018 
1019 	return rc;
1020 }
1021 
1022 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
1023 {
1024 	struct sock *sk = sock->sk;
1025 	struct tipc_sock *tsk = tipc_sk(sk);
1026 	DEFINE_WAIT(wait);
1027 	int done;
1028 
1029 	do {
1030 		int err = sock_error(sk);
1031 		if (err)
1032 			return err;
1033 		if (sock->state == SS_DISCONNECTING)
1034 			return -EPIPE;
1035 		else if (sock->state != SS_CONNECTED)
1036 			return -ENOTCONN;
1037 		if (!*timeo_p)
1038 			return -EAGAIN;
1039 		if (signal_pending(current))
1040 			return sock_intr_errno(*timeo_p);
1041 
1042 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1043 		done = sk_wait_event(sk, timeo_p,
1044 				     (!tsk->link_cong &&
1045 				      !tsk_conn_cong(tsk)) ||
1046 				     !tsk->connected);
1047 		finish_wait(sk_sleep(sk), &wait);
1048 	} while (!done);
1049 	return 0;
1050 }
1051 
1052 /**
1053  * tipc_send_stream - send stream-oriented data
1054  * @sock: socket structure
1055  * @m: data to send
1056  * @dsz: total length of data to be transmitted
1057  *
1058  * Used for SOCK_STREAM data.
1059  *
1060  * Returns the number of bytes sent on success (or partial success),
1061  * or errno if no data sent
1062  */
1063 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1064 {
1065 	struct sock *sk = sock->sk;
1066 	int ret;
1067 
1068 	lock_sock(sk);
1069 	ret = __tipc_send_stream(sock, m, dsz);
1070 	release_sock(sk);
1071 
1072 	return ret;
1073 }
1074 
1075 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1076 {
1077 	struct sock *sk = sock->sk;
1078 	struct net *net = sock_net(sk);
1079 	struct tipc_sock *tsk = tipc_sk(sk);
1080 	struct tipc_msg *mhdr = &tsk->phdr;
1081 	struct sk_buff_head *pktchain = &sk->sk_write_queue;
1082 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1083 	u32 portid = tsk->portid;
1084 	int rc = -EINVAL;
1085 	long timeo;
1086 	u32 dnode;
1087 	uint mtu, send, sent = 0;
1088 	struct iov_iter save;
1089 
1090 	/* Handle implied connection establishment */
1091 	if (unlikely(dest)) {
1092 		rc = __tipc_sendmsg(sock, m, dsz);
1093 		if (dsz && (dsz == rc))
1094 			tsk->sent_unacked = 1;
1095 		return rc;
1096 	}
1097 	if (dsz > (uint)INT_MAX)
1098 		return -EMSGSIZE;
1099 
1100 	if (unlikely(sock->state != SS_CONNECTED)) {
1101 		if (sock->state == SS_DISCONNECTING)
1102 			return -EPIPE;
1103 		else
1104 			return -ENOTCONN;
1105 	}
1106 
1107 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1108 	dnode = tsk_peer_node(tsk);
1109 
1110 next:
1111 	save = m->msg_iter;
1112 	mtu = tsk->max_pkt;
1113 	send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1114 	rc = tipc_msg_build(mhdr, m, sent, send, mtu, pktchain);
1115 	if (unlikely(rc < 0))
1116 		return rc;
1117 	do {
1118 		if (likely(!tsk_conn_cong(tsk))) {
1119 			rc = tipc_link_xmit(net, pktchain, dnode, portid);
1120 			if (likely(!rc)) {
1121 				tsk->sent_unacked++;
1122 				sent += send;
1123 				if (sent == dsz)
1124 					break;
1125 				goto next;
1126 			}
1127 			if (rc == -EMSGSIZE) {
1128 				tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1129 								 portid);
1130 				m->msg_iter = save;
1131 				goto next;
1132 			}
1133 			if (rc != -ELINKCONG)
1134 				break;
1135 			tsk->link_cong = 1;
1136 		}
1137 		rc = tipc_wait_for_sndpkt(sock, &timeo);
1138 		if (rc)
1139 			__skb_queue_purge(pktchain);
1140 	} while (!rc);
1141 
1142 	return sent ? sent : rc;
1143 }
1144 
1145 /**
1146  * tipc_send_packet - send a connection-oriented message
1147  * @sock: socket structure
1148  * @m: message to send
1149  * @dsz: length of data to be transmitted
1150  *
1151  * Used for SOCK_SEQPACKET messages.
1152  *
1153  * Returns the number of bytes sent on success, or errno otherwise
1154  */
1155 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1156 {
1157 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1158 		return -EMSGSIZE;
1159 
1160 	return tipc_send_stream(sock, m, dsz);
1161 }
1162 
1163 /* tipc_sk_finish_conn - complete the setup of a connection
1164  */
1165 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1166 				u32 peer_node)
1167 {
1168 	struct sock *sk = &tsk->sk;
1169 	struct net *net = sock_net(sk);
1170 	struct tipc_msg *msg = &tsk->phdr;
1171 
1172 	msg_set_destnode(msg, peer_node);
1173 	msg_set_destport(msg, peer_port);
1174 	msg_set_type(msg, TIPC_CONN_MSG);
1175 	msg_set_lookup_scope(msg, 0);
1176 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1177 
1178 	tsk->probing_intv = CONN_PROBING_INTERVAL;
1179 	tsk->probing_state = TIPC_CONN_OK;
1180 	tsk->connected = 1;
1181 	sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1182 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1183 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1184 }
1185 
1186 /**
1187  * set_orig_addr - capture sender's address for received message
1188  * @m: descriptor for message info
1189  * @msg: received message header
1190  *
1191  * Note: Address is not captured if not requested by receiver.
1192  */
1193 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1194 {
1195 	DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1196 
1197 	if (addr) {
1198 		addr->family = AF_TIPC;
1199 		addr->addrtype = TIPC_ADDR_ID;
1200 		memset(&addr->addr, 0, sizeof(addr->addr));
1201 		addr->addr.id.ref = msg_origport(msg);
1202 		addr->addr.id.node = msg_orignode(msg);
1203 		addr->addr.name.domain = 0;	/* could leave uninitialized */
1204 		addr->scope = 0;		/* could leave uninitialized */
1205 		m->msg_namelen = sizeof(struct sockaddr_tipc);
1206 	}
1207 }
1208 
1209 /**
1210  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1211  * @m: descriptor for message info
1212  * @msg: received message header
1213  * @tsk: TIPC port associated with message
1214  *
1215  * Note: Ancillary data is not captured if not requested by receiver.
1216  *
1217  * Returns 0 if successful, otherwise errno
1218  */
1219 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1220 				 struct tipc_sock *tsk)
1221 {
1222 	u32 anc_data[3];
1223 	u32 err;
1224 	u32 dest_type;
1225 	int has_name;
1226 	int res;
1227 
1228 	if (likely(m->msg_controllen == 0))
1229 		return 0;
1230 
1231 	/* Optionally capture errored message object(s) */
1232 	err = msg ? msg_errcode(msg) : 0;
1233 	if (unlikely(err)) {
1234 		anc_data[0] = err;
1235 		anc_data[1] = msg_data_sz(msg);
1236 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1237 		if (res)
1238 			return res;
1239 		if (anc_data[1]) {
1240 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1241 				       msg_data(msg));
1242 			if (res)
1243 				return res;
1244 		}
1245 	}
1246 
1247 	/* Optionally capture message destination object */
1248 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1249 	switch (dest_type) {
1250 	case TIPC_NAMED_MSG:
1251 		has_name = 1;
1252 		anc_data[0] = msg_nametype(msg);
1253 		anc_data[1] = msg_namelower(msg);
1254 		anc_data[2] = msg_namelower(msg);
1255 		break;
1256 	case TIPC_MCAST_MSG:
1257 		has_name = 1;
1258 		anc_data[0] = msg_nametype(msg);
1259 		anc_data[1] = msg_namelower(msg);
1260 		anc_data[2] = msg_nameupper(msg);
1261 		break;
1262 	case TIPC_CONN_MSG:
1263 		has_name = (tsk->conn_type != 0);
1264 		anc_data[0] = tsk->conn_type;
1265 		anc_data[1] = tsk->conn_instance;
1266 		anc_data[2] = tsk->conn_instance;
1267 		break;
1268 	default:
1269 		has_name = 0;
1270 	}
1271 	if (has_name) {
1272 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1273 		if (res)
1274 			return res;
1275 	}
1276 
1277 	return 0;
1278 }
1279 
1280 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1281 {
1282 	struct net *net = sock_net(&tsk->sk);
1283 	struct sk_buff *skb = NULL;
1284 	struct tipc_msg *msg;
1285 	u32 peer_port = tsk_peer_port(tsk);
1286 	u32 dnode = tsk_peer_node(tsk);
1287 
1288 	if (!tsk->connected)
1289 		return;
1290 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1291 			      dnode, tsk_own_node(tsk), peer_port,
1292 			      tsk->portid, TIPC_OK);
1293 	if (!skb)
1294 		return;
1295 	msg = buf_msg(skb);
1296 	msg_set_msgcnt(msg, ack);
1297 	tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1298 }
1299 
1300 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1301 {
1302 	struct sock *sk = sock->sk;
1303 	DEFINE_WAIT(wait);
1304 	long timeo = *timeop;
1305 	int err;
1306 
1307 	for (;;) {
1308 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1309 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1310 			if (sock->state == SS_DISCONNECTING) {
1311 				err = -ENOTCONN;
1312 				break;
1313 			}
1314 			release_sock(sk);
1315 			timeo = schedule_timeout(timeo);
1316 			lock_sock(sk);
1317 		}
1318 		err = 0;
1319 		if (!skb_queue_empty(&sk->sk_receive_queue))
1320 			break;
1321 		err = -EAGAIN;
1322 		if (!timeo)
1323 			break;
1324 		err = sock_intr_errno(timeo);
1325 		if (signal_pending(current))
1326 			break;
1327 	}
1328 	finish_wait(sk_sleep(sk), &wait);
1329 	*timeop = timeo;
1330 	return err;
1331 }
1332 
1333 /**
1334  * tipc_recvmsg - receive packet-oriented message
1335  * @m: descriptor for message info
1336  * @buf_len: total size of user buffer area
1337  * @flags: receive flags
1338  *
1339  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1340  * If the complete message doesn't fit in user area, truncate it.
1341  *
1342  * Returns size of returned message data, errno otherwise
1343  */
1344 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1345 			int flags)
1346 {
1347 	struct sock *sk = sock->sk;
1348 	struct tipc_sock *tsk = tipc_sk(sk);
1349 	struct sk_buff *buf;
1350 	struct tipc_msg *msg;
1351 	long timeo;
1352 	unsigned int sz;
1353 	u32 err;
1354 	int res;
1355 
1356 	/* Catch invalid receive requests */
1357 	if (unlikely(!buf_len))
1358 		return -EINVAL;
1359 
1360 	lock_sock(sk);
1361 
1362 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1363 		res = -ENOTCONN;
1364 		goto exit;
1365 	}
1366 
1367 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1368 restart:
1369 
1370 	/* Look for a message in receive queue; wait if necessary */
1371 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1372 	if (res)
1373 		goto exit;
1374 
1375 	/* Look at first message in receive queue */
1376 	buf = skb_peek(&sk->sk_receive_queue);
1377 	msg = buf_msg(buf);
1378 	sz = msg_data_sz(msg);
1379 	err = msg_errcode(msg);
1380 
1381 	/* Discard an empty non-errored message & try again */
1382 	if ((!sz) && (!err)) {
1383 		tsk_advance_rx_queue(sk);
1384 		goto restart;
1385 	}
1386 
1387 	/* Capture sender's address (optional) */
1388 	set_orig_addr(m, msg);
1389 
1390 	/* Capture ancillary data (optional) */
1391 	res = tipc_sk_anc_data_recv(m, msg, tsk);
1392 	if (res)
1393 		goto exit;
1394 
1395 	/* Capture message data (if valid) & compute return value (always) */
1396 	if (!err) {
1397 		if (unlikely(buf_len < sz)) {
1398 			sz = buf_len;
1399 			m->msg_flags |= MSG_TRUNC;
1400 		}
1401 		res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1402 		if (res)
1403 			goto exit;
1404 		res = sz;
1405 	} else {
1406 		if ((sock->state == SS_READY) ||
1407 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1408 			res = 0;
1409 		else
1410 			res = -ECONNRESET;
1411 	}
1412 
1413 	/* Consume received message (optional) */
1414 	if (likely(!(flags & MSG_PEEK))) {
1415 		if ((sock->state != SS_READY) &&
1416 		    (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1417 			tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1418 			tsk->rcv_unacked = 0;
1419 		}
1420 		tsk_advance_rx_queue(sk);
1421 	}
1422 exit:
1423 	release_sock(sk);
1424 	return res;
1425 }
1426 
1427 /**
1428  * tipc_recv_stream - receive stream-oriented data
1429  * @m: descriptor for message info
1430  * @buf_len: total size of user buffer area
1431  * @flags: receive flags
1432  *
1433  * Used for SOCK_STREAM messages only.  If not enough data is available
1434  * will optionally wait for more; never truncates data.
1435  *
1436  * Returns size of returned message data, errno otherwise
1437  */
1438 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1439 			    size_t buf_len, int flags)
1440 {
1441 	struct sock *sk = sock->sk;
1442 	struct tipc_sock *tsk = tipc_sk(sk);
1443 	struct sk_buff *buf;
1444 	struct tipc_msg *msg;
1445 	long timeo;
1446 	unsigned int sz;
1447 	int sz_to_copy, target, needed;
1448 	int sz_copied = 0;
1449 	u32 err;
1450 	int res = 0;
1451 
1452 	/* Catch invalid receive attempts */
1453 	if (unlikely(!buf_len))
1454 		return -EINVAL;
1455 
1456 	lock_sock(sk);
1457 
1458 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1459 		res = -ENOTCONN;
1460 		goto exit;
1461 	}
1462 
1463 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1464 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1465 
1466 restart:
1467 	/* Look for a message in receive queue; wait if necessary */
1468 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1469 	if (res)
1470 		goto exit;
1471 
1472 	/* Look at first message in receive queue */
1473 	buf = skb_peek(&sk->sk_receive_queue);
1474 	msg = buf_msg(buf);
1475 	sz = msg_data_sz(msg);
1476 	err = msg_errcode(msg);
1477 
1478 	/* Discard an empty non-errored message & try again */
1479 	if ((!sz) && (!err)) {
1480 		tsk_advance_rx_queue(sk);
1481 		goto restart;
1482 	}
1483 
1484 	/* Optionally capture sender's address & ancillary data of first msg */
1485 	if (sz_copied == 0) {
1486 		set_orig_addr(m, msg);
1487 		res = tipc_sk_anc_data_recv(m, msg, tsk);
1488 		if (res)
1489 			goto exit;
1490 	}
1491 
1492 	/* Capture message data (if valid) & compute return value (always) */
1493 	if (!err) {
1494 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1495 
1496 		sz -= offset;
1497 		needed = (buf_len - sz_copied);
1498 		sz_to_copy = (sz <= needed) ? sz : needed;
1499 
1500 		res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1501 					    m, sz_to_copy);
1502 		if (res)
1503 			goto exit;
1504 
1505 		sz_copied += sz_to_copy;
1506 
1507 		if (sz_to_copy < sz) {
1508 			if (!(flags & MSG_PEEK))
1509 				TIPC_SKB_CB(buf)->handle =
1510 				(void *)(unsigned long)(offset + sz_to_copy);
1511 			goto exit;
1512 		}
1513 	} else {
1514 		if (sz_copied != 0)
1515 			goto exit; /* can't add error msg to valid data */
1516 
1517 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1518 			res = 0;
1519 		else
1520 			res = -ECONNRESET;
1521 	}
1522 
1523 	/* Consume received message (optional) */
1524 	if (likely(!(flags & MSG_PEEK))) {
1525 		if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1526 			tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1527 			tsk->rcv_unacked = 0;
1528 		}
1529 		tsk_advance_rx_queue(sk);
1530 	}
1531 
1532 	/* Loop around if more data is required */
1533 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1534 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1535 	    (sz_copied < target)) &&	/* and more is ready or required */
1536 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1537 	    (!err))			/* and haven't reached a FIN */
1538 		goto restart;
1539 
1540 exit:
1541 	release_sock(sk);
1542 	return sz_copied ? sz_copied : res;
1543 }
1544 
1545 /**
1546  * tipc_write_space - wake up thread if port congestion is released
1547  * @sk: socket
1548  */
1549 static void tipc_write_space(struct sock *sk)
1550 {
1551 	struct socket_wq *wq;
1552 
1553 	rcu_read_lock();
1554 	wq = rcu_dereference(sk->sk_wq);
1555 	if (wq_has_sleeper(wq))
1556 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1557 						POLLWRNORM | POLLWRBAND);
1558 	rcu_read_unlock();
1559 }
1560 
1561 /**
1562  * tipc_data_ready - wake up threads to indicate messages have been received
1563  * @sk: socket
1564  * @len: the length of messages
1565  */
1566 static void tipc_data_ready(struct sock *sk)
1567 {
1568 	struct socket_wq *wq;
1569 
1570 	rcu_read_lock();
1571 	wq = rcu_dereference(sk->sk_wq);
1572 	if (wq_has_sleeper(wq))
1573 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1574 						POLLRDNORM | POLLRDBAND);
1575 	rcu_read_unlock();
1576 }
1577 
1578 /**
1579  * filter_connect - Handle all incoming messages for a connection-based socket
1580  * @tsk: TIPC socket
1581  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1582  *
1583  * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1584  */
1585 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **skb)
1586 {
1587 	struct sock *sk = &tsk->sk;
1588 	struct net *net = sock_net(sk);
1589 	struct socket *sock = sk->sk_socket;
1590 	struct tipc_msg *msg = buf_msg(*skb);
1591 	int retval = -TIPC_ERR_NO_PORT;
1592 
1593 	if (msg_mcast(msg))
1594 		return retval;
1595 
1596 	switch ((int)sock->state) {
1597 	case SS_CONNECTED:
1598 		/* Accept only connection-based messages sent by peer */
1599 		if (tsk_peer_msg(tsk, msg)) {
1600 			if (unlikely(msg_errcode(msg))) {
1601 				sock->state = SS_DISCONNECTING;
1602 				tsk->connected = 0;
1603 				/* let timer expire on it's own */
1604 				tipc_node_remove_conn(net, tsk_peer_node(tsk),
1605 						      tsk->portid);
1606 			}
1607 			retval = TIPC_OK;
1608 		}
1609 		break;
1610 	case SS_CONNECTING:
1611 		/* Accept only ACK or NACK message */
1612 
1613 		if (unlikely(!msg_connected(msg)))
1614 			break;
1615 
1616 		if (unlikely(msg_errcode(msg))) {
1617 			sock->state = SS_DISCONNECTING;
1618 			sk->sk_err = ECONNREFUSED;
1619 			retval = TIPC_OK;
1620 			break;
1621 		}
1622 
1623 		if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1624 			sock->state = SS_DISCONNECTING;
1625 			sk->sk_err = EINVAL;
1626 			retval = TIPC_OK;
1627 			break;
1628 		}
1629 
1630 		tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1631 		msg_set_importance(&tsk->phdr, msg_importance(msg));
1632 		sock->state = SS_CONNECTED;
1633 
1634 		/* If an incoming message is an 'ACK-', it should be
1635 		 * discarded here because it doesn't contain useful
1636 		 * data. In addition, we should try to wake up
1637 		 * connect() routine if sleeping.
1638 		 */
1639 		if (msg_data_sz(msg) == 0) {
1640 			kfree_skb(*skb);
1641 			*skb = NULL;
1642 			if (waitqueue_active(sk_sleep(sk)))
1643 				wake_up_interruptible(sk_sleep(sk));
1644 		}
1645 		retval = TIPC_OK;
1646 		break;
1647 	case SS_LISTENING:
1648 	case SS_UNCONNECTED:
1649 		/* Accept only SYN message */
1650 		if (!msg_connected(msg) && !(msg_errcode(msg)))
1651 			retval = TIPC_OK;
1652 		break;
1653 	case SS_DISCONNECTING:
1654 		break;
1655 	default:
1656 		pr_err("Unknown socket state %u\n", sock->state);
1657 	}
1658 	return retval;
1659 }
1660 
1661 /**
1662  * rcvbuf_limit - get proper overload limit of socket receive queue
1663  * @sk: socket
1664  * @buf: message
1665  *
1666  * For all connection oriented messages, irrespective of importance,
1667  * the default overload value (i.e. 67MB) is set as limit.
1668  *
1669  * For all connectionless messages, by default new queue limits are
1670  * as belows:
1671  *
1672  * TIPC_LOW_IMPORTANCE       (4 MB)
1673  * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1674  * TIPC_HIGH_IMPORTANCE      (16 MB)
1675  * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1676  *
1677  * Returns overload limit according to corresponding message importance
1678  */
1679 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1680 {
1681 	struct tipc_msg *msg = buf_msg(buf);
1682 
1683 	if (msg_connected(msg))
1684 		return sysctl_tipc_rmem[2];
1685 
1686 	return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1687 		msg_importance(msg);
1688 }
1689 
1690 /**
1691  * filter_rcv - validate incoming message
1692  * @sk: socket
1693  * @skb: pointer to message. Set to NULL if buffer is consumed.
1694  *
1695  * Enqueues message on receive queue if acceptable; optionally handles
1696  * disconnect indication for a connected socket.
1697  *
1698  * Called with socket lock already taken
1699  *
1700  * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1701  */
1702 static int filter_rcv(struct sock *sk, struct sk_buff **skb)
1703 {
1704 	struct socket *sock = sk->sk_socket;
1705 	struct tipc_sock *tsk = tipc_sk(sk);
1706 	struct tipc_msg *msg = buf_msg(*skb);
1707 	unsigned int limit = rcvbuf_limit(sk, *skb);
1708 	int rc = TIPC_OK;
1709 
1710 	if (unlikely(msg_user(msg) == CONN_MANAGER)) {
1711 		tipc_sk_proto_rcv(tsk, skb);
1712 		return TIPC_OK;
1713 	}
1714 
1715 	if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1716 		kfree_skb(*skb);
1717 		tsk->link_cong = 0;
1718 		sk->sk_write_space(sk);
1719 		*skb = NULL;
1720 		return TIPC_OK;
1721 	}
1722 
1723 	/* Reject message if it is wrong sort of message for socket */
1724 	if (msg_type(msg) > TIPC_DIRECT_MSG)
1725 		return -TIPC_ERR_NO_PORT;
1726 
1727 	if (sock->state == SS_READY) {
1728 		if (msg_connected(msg))
1729 			return -TIPC_ERR_NO_PORT;
1730 	} else {
1731 		rc = filter_connect(tsk, skb);
1732 		if (rc != TIPC_OK || !*skb)
1733 			return rc;
1734 	}
1735 
1736 	/* Reject message if there isn't room to queue it */
1737 	if (sk_rmem_alloc_get(sk) + (*skb)->truesize >= limit)
1738 		return -TIPC_ERR_OVERLOAD;
1739 
1740 	/* Enqueue message */
1741 	TIPC_SKB_CB(*skb)->handle = NULL;
1742 	__skb_queue_tail(&sk->sk_receive_queue, *skb);
1743 	skb_set_owner_r(*skb, sk);
1744 
1745 	sk->sk_data_ready(sk);
1746 	*skb = NULL;
1747 	return TIPC_OK;
1748 }
1749 
1750 /**
1751  * tipc_backlog_rcv - handle incoming message from backlog queue
1752  * @sk: socket
1753  * @skb: message
1754  *
1755  * Caller must hold socket lock
1756  *
1757  * Returns 0
1758  */
1759 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1760 {
1761 	int err;
1762 	atomic_t *dcnt;
1763 	u32 dnode;
1764 	struct tipc_sock *tsk = tipc_sk(sk);
1765 	struct net *net = sock_net(sk);
1766 	uint truesize = skb->truesize;
1767 
1768 	err = filter_rcv(sk, &skb);
1769 	if (likely(!skb)) {
1770 		dcnt = &tsk->dupl_rcvcnt;
1771 		if (atomic_read(dcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1772 			atomic_add(truesize, dcnt);
1773 		return 0;
1774 	}
1775 	if (!err || tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode, -err))
1776 		tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
1777 	return 0;
1778 }
1779 
1780 /**
1781  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1782  *                   inputq and try adding them to socket or backlog queue
1783  * @inputq: list of incoming buffers with potentially different destinations
1784  * @sk: socket where the buffers should be enqueued
1785  * @dport: port number for the socket
1786  * @_skb: returned buffer to be forwarded or rejected, if applicable
1787  *
1788  * Caller must hold socket lock
1789  *
1790  * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1791  * or -TIPC_ERR_NO_PORT
1792  */
1793 static int tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1794 			   u32 dport, struct sk_buff **_skb)
1795 {
1796 	unsigned int lim;
1797 	atomic_t *dcnt;
1798 	int err;
1799 	struct sk_buff *skb;
1800 	unsigned long time_limit = jiffies + 2;
1801 
1802 	while (skb_queue_len(inputq)) {
1803 		if (unlikely(time_after_eq(jiffies, time_limit)))
1804 			return TIPC_OK;
1805 		skb = tipc_skb_dequeue(inputq, dport);
1806 		if (unlikely(!skb))
1807 			return TIPC_OK;
1808 		if (!sock_owned_by_user(sk)) {
1809 			err = filter_rcv(sk, &skb);
1810 			if (likely(!skb))
1811 				continue;
1812 			*_skb = skb;
1813 			return err;
1814 		}
1815 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1816 		if (sk->sk_backlog.len)
1817 			atomic_set(dcnt, 0);
1818 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1819 		if (likely(!sk_add_backlog(sk, skb, lim)))
1820 			continue;
1821 		*_skb = skb;
1822 		return -TIPC_ERR_OVERLOAD;
1823 	}
1824 	return TIPC_OK;
1825 }
1826 
1827 /**
1828  * tipc_sk_rcv - handle a chain of incoming buffers
1829  * @inputq: buffer list containing the buffers
1830  * Consumes all buffers in list until inputq is empty
1831  * Note: may be called in multiple threads referring to the same queue
1832  * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1833  * Only node local calls check the return value, sending single-buffer queues
1834  */
1835 int tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1836 {
1837 	u32 dnode, dport = 0;
1838 	int err = -TIPC_ERR_NO_PORT;
1839 	struct sk_buff *skb;
1840 	struct tipc_sock *tsk;
1841 	struct tipc_net *tn;
1842 	struct sock *sk;
1843 
1844 	while (skb_queue_len(inputq)) {
1845 		skb = NULL;
1846 		dport = tipc_skb_peek_port(inputq, dport);
1847 		tsk = tipc_sk_lookup(net, dport);
1848 		if (likely(tsk)) {
1849 			sk = &tsk->sk;
1850 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1851 				err = tipc_sk_enqueue(inputq, sk, dport, &skb);
1852 				spin_unlock_bh(&sk->sk_lock.slock);
1853 				dport = 0;
1854 			}
1855 			sock_put(sk);
1856 		} else {
1857 			skb = tipc_skb_dequeue(inputq, dport);
1858 		}
1859 		if (likely(!skb))
1860 			continue;
1861 		if (tipc_msg_lookup_dest(net, skb, &dnode, &err))
1862 			goto xmit;
1863 		if (!err) {
1864 			dnode = msg_destnode(buf_msg(skb));
1865 			goto xmit;
1866 		}
1867 		tn = net_generic(net, tipc_net_id);
1868 		if (!tipc_msg_reverse(tn->own_addr, skb, &dnode, -err))
1869 			continue;
1870 xmit:
1871 		tipc_link_xmit_skb(net, skb, dnode, dport);
1872 	}
1873 	return err ? -EHOSTUNREACH : 0;
1874 }
1875 
1876 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1877 {
1878 	struct sock *sk = sock->sk;
1879 	DEFINE_WAIT(wait);
1880 	int done;
1881 
1882 	do {
1883 		int err = sock_error(sk);
1884 		if (err)
1885 			return err;
1886 		if (!*timeo_p)
1887 			return -ETIMEDOUT;
1888 		if (signal_pending(current))
1889 			return sock_intr_errno(*timeo_p);
1890 
1891 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1892 		done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1893 		finish_wait(sk_sleep(sk), &wait);
1894 	} while (!done);
1895 	return 0;
1896 }
1897 
1898 /**
1899  * tipc_connect - establish a connection to another TIPC port
1900  * @sock: socket structure
1901  * @dest: socket address for destination port
1902  * @destlen: size of socket address data structure
1903  * @flags: file-related flags associated with socket
1904  *
1905  * Returns 0 on success, errno otherwise
1906  */
1907 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1908 			int destlen, int flags)
1909 {
1910 	struct sock *sk = sock->sk;
1911 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1912 	struct msghdr m = {NULL,};
1913 	long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1914 	socket_state previous;
1915 	int res;
1916 
1917 	lock_sock(sk);
1918 
1919 	/* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1920 	if (sock->state == SS_READY) {
1921 		res = -EOPNOTSUPP;
1922 		goto exit;
1923 	}
1924 
1925 	/*
1926 	 * Reject connection attempt using multicast address
1927 	 *
1928 	 * Note: send_msg() validates the rest of the address fields,
1929 	 *       so there's no need to do it here
1930 	 */
1931 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1932 		res = -EINVAL;
1933 		goto exit;
1934 	}
1935 
1936 	previous = sock->state;
1937 	switch (sock->state) {
1938 	case SS_UNCONNECTED:
1939 		/* Send a 'SYN-' to destination */
1940 		m.msg_name = dest;
1941 		m.msg_namelen = destlen;
1942 
1943 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1944 		 * indicate send_msg() is never blocked.
1945 		 */
1946 		if (!timeout)
1947 			m.msg_flags = MSG_DONTWAIT;
1948 
1949 		res = __tipc_sendmsg(sock, &m, 0);
1950 		if ((res < 0) && (res != -EWOULDBLOCK))
1951 			goto exit;
1952 
1953 		/* Just entered SS_CONNECTING state; the only
1954 		 * difference is that return value in non-blocking
1955 		 * case is EINPROGRESS, rather than EALREADY.
1956 		 */
1957 		res = -EINPROGRESS;
1958 	case SS_CONNECTING:
1959 		if (previous == SS_CONNECTING)
1960 			res = -EALREADY;
1961 		if (!timeout)
1962 			goto exit;
1963 		timeout = msecs_to_jiffies(timeout);
1964 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1965 		res = tipc_wait_for_connect(sock, &timeout);
1966 		break;
1967 	case SS_CONNECTED:
1968 		res = -EISCONN;
1969 		break;
1970 	default:
1971 		res = -EINVAL;
1972 		break;
1973 	}
1974 exit:
1975 	release_sock(sk);
1976 	return res;
1977 }
1978 
1979 /**
1980  * tipc_listen - allow socket to listen for incoming connections
1981  * @sock: socket structure
1982  * @len: (unused)
1983  *
1984  * Returns 0 on success, errno otherwise
1985  */
1986 static int tipc_listen(struct socket *sock, int len)
1987 {
1988 	struct sock *sk = sock->sk;
1989 	int res;
1990 
1991 	lock_sock(sk);
1992 
1993 	if (sock->state != SS_UNCONNECTED)
1994 		res = -EINVAL;
1995 	else {
1996 		sock->state = SS_LISTENING;
1997 		res = 0;
1998 	}
1999 
2000 	release_sock(sk);
2001 	return res;
2002 }
2003 
2004 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2005 {
2006 	struct sock *sk = sock->sk;
2007 	DEFINE_WAIT(wait);
2008 	int err;
2009 
2010 	/* True wake-one mechanism for incoming connections: only
2011 	 * one process gets woken up, not the 'whole herd'.
2012 	 * Since we do not 'race & poll' for established sockets
2013 	 * anymore, the common case will execute the loop only once.
2014 	*/
2015 	for (;;) {
2016 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2017 					  TASK_INTERRUPTIBLE);
2018 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2019 			release_sock(sk);
2020 			timeo = schedule_timeout(timeo);
2021 			lock_sock(sk);
2022 		}
2023 		err = 0;
2024 		if (!skb_queue_empty(&sk->sk_receive_queue))
2025 			break;
2026 		err = -EINVAL;
2027 		if (sock->state != SS_LISTENING)
2028 			break;
2029 		err = -EAGAIN;
2030 		if (!timeo)
2031 			break;
2032 		err = sock_intr_errno(timeo);
2033 		if (signal_pending(current))
2034 			break;
2035 	}
2036 	finish_wait(sk_sleep(sk), &wait);
2037 	return err;
2038 }
2039 
2040 /**
2041  * tipc_accept - wait for connection request
2042  * @sock: listening socket
2043  * @newsock: new socket that is to be connected
2044  * @flags: file-related flags associated with socket
2045  *
2046  * Returns 0 on success, errno otherwise
2047  */
2048 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2049 {
2050 	struct sock *new_sk, *sk = sock->sk;
2051 	struct sk_buff *buf;
2052 	struct tipc_sock *new_tsock;
2053 	struct tipc_msg *msg;
2054 	long timeo;
2055 	int res;
2056 
2057 	lock_sock(sk);
2058 
2059 	if (sock->state != SS_LISTENING) {
2060 		res = -EINVAL;
2061 		goto exit;
2062 	}
2063 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2064 	res = tipc_wait_for_accept(sock, timeo);
2065 	if (res)
2066 		goto exit;
2067 
2068 	buf = skb_peek(&sk->sk_receive_queue);
2069 
2070 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2071 	if (res)
2072 		goto exit;
2073 
2074 	new_sk = new_sock->sk;
2075 	new_tsock = tipc_sk(new_sk);
2076 	msg = buf_msg(buf);
2077 
2078 	/* we lock on new_sk; but lockdep sees the lock on sk */
2079 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2080 
2081 	/*
2082 	 * Reject any stray messages received by new socket
2083 	 * before the socket lock was taken (very, very unlikely)
2084 	 */
2085 	tsk_rej_rx_queue(new_sk);
2086 
2087 	/* Connect new socket to it's peer */
2088 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2089 	new_sock->state = SS_CONNECTED;
2090 
2091 	tsk_set_importance(new_tsock, msg_importance(msg));
2092 	if (msg_named(msg)) {
2093 		new_tsock->conn_type = msg_nametype(msg);
2094 		new_tsock->conn_instance = msg_nameinst(msg);
2095 	}
2096 
2097 	/*
2098 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2099 	 * Respond to 'SYN+' by queuing it on new socket.
2100 	 */
2101 	if (!msg_data_sz(msg)) {
2102 		struct msghdr m = {NULL,};
2103 
2104 		tsk_advance_rx_queue(sk);
2105 		__tipc_send_stream(new_sock, &m, 0);
2106 	} else {
2107 		__skb_dequeue(&sk->sk_receive_queue);
2108 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2109 		skb_set_owner_r(buf, new_sk);
2110 	}
2111 	release_sock(new_sk);
2112 exit:
2113 	release_sock(sk);
2114 	return res;
2115 }
2116 
2117 /**
2118  * tipc_shutdown - shutdown socket connection
2119  * @sock: socket structure
2120  * @how: direction to close (must be SHUT_RDWR)
2121  *
2122  * Terminates connection (if necessary), then purges socket's receive queue.
2123  *
2124  * Returns 0 on success, errno otherwise
2125  */
2126 static int tipc_shutdown(struct socket *sock, int how)
2127 {
2128 	struct sock *sk = sock->sk;
2129 	struct net *net = sock_net(sk);
2130 	struct tipc_sock *tsk = tipc_sk(sk);
2131 	struct sk_buff *skb;
2132 	u32 dnode;
2133 	int res;
2134 
2135 	if (how != SHUT_RDWR)
2136 		return -EINVAL;
2137 
2138 	lock_sock(sk);
2139 
2140 	switch (sock->state) {
2141 	case SS_CONNECTING:
2142 	case SS_CONNECTED:
2143 
2144 restart:
2145 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2146 		skb = __skb_dequeue(&sk->sk_receive_queue);
2147 		if (skb) {
2148 			if (TIPC_SKB_CB(skb)->handle != NULL) {
2149 				kfree_skb(skb);
2150 				goto restart;
2151 			}
2152 			if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
2153 					     TIPC_CONN_SHUTDOWN))
2154 				tipc_link_xmit_skb(net, skb, dnode,
2155 						   tsk->portid);
2156 			tipc_node_remove_conn(net, dnode, tsk->portid);
2157 		} else {
2158 			dnode = tsk_peer_node(tsk);
2159 
2160 			skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2161 					      TIPC_CONN_MSG, SHORT_H_SIZE,
2162 					      0, dnode, tsk_own_node(tsk),
2163 					      tsk_peer_port(tsk),
2164 					      tsk->portid, TIPC_CONN_SHUTDOWN);
2165 			tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2166 		}
2167 		tsk->connected = 0;
2168 		sock->state = SS_DISCONNECTING;
2169 		tipc_node_remove_conn(net, dnode, tsk->portid);
2170 		/* fall through */
2171 
2172 	case SS_DISCONNECTING:
2173 
2174 		/* Discard any unreceived messages */
2175 		__skb_queue_purge(&sk->sk_receive_queue);
2176 
2177 		/* Wake up anyone sleeping in poll */
2178 		sk->sk_state_change(sk);
2179 		res = 0;
2180 		break;
2181 
2182 	default:
2183 		res = -ENOTCONN;
2184 	}
2185 
2186 	release_sock(sk);
2187 	return res;
2188 }
2189 
2190 static void tipc_sk_timeout(unsigned long data)
2191 {
2192 	struct tipc_sock *tsk = (struct tipc_sock *)data;
2193 	struct sock *sk = &tsk->sk;
2194 	struct sk_buff *skb = NULL;
2195 	u32 peer_port, peer_node;
2196 	u32 own_node = tsk_own_node(tsk);
2197 
2198 	bh_lock_sock(sk);
2199 	if (!tsk->connected) {
2200 		bh_unlock_sock(sk);
2201 		goto exit;
2202 	}
2203 	peer_port = tsk_peer_port(tsk);
2204 	peer_node = tsk_peer_node(tsk);
2205 
2206 	if (tsk->probing_state == TIPC_CONN_PROBING) {
2207 		/* Previous probe not answered -> self abort */
2208 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2209 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0,
2210 				      own_node, peer_node, tsk->portid,
2211 				      peer_port, TIPC_ERR_NO_PORT);
2212 	} else {
2213 		skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2214 				      INT_H_SIZE, 0, peer_node, own_node,
2215 				      peer_port, tsk->portid, TIPC_OK);
2216 		tsk->probing_state = TIPC_CONN_PROBING;
2217 		sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2218 	}
2219 	bh_unlock_sock(sk);
2220 	if (skb)
2221 		tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2222 exit:
2223 	sock_put(sk);
2224 }
2225 
2226 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2227 			   struct tipc_name_seq const *seq)
2228 {
2229 	struct net *net = sock_net(&tsk->sk);
2230 	struct publication *publ;
2231 	u32 key;
2232 
2233 	if (tsk->connected)
2234 		return -EINVAL;
2235 	key = tsk->portid + tsk->pub_count + 1;
2236 	if (key == tsk->portid)
2237 		return -EADDRINUSE;
2238 
2239 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2240 				    scope, tsk->portid, key);
2241 	if (unlikely(!publ))
2242 		return -EINVAL;
2243 
2244 	list_add(&publ->pport_list, &tsk->publications);
2245 	tsk->pub_count++;
2246 	tsk->published = 1;
2247 	return 0;
2248 }
2249 
2250 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2251 			    struct tipc_name_seq const *seq)
2252 {
2253 	struct net *net = sock_net(&tsk->sk);
2254 	struct publication *publ;
2255 	struct publication *safe;
2256 	int rc = -EINVAL;
2257 
2258 	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2259 		if (seq) {
2260 			if (publ->scope != scope)
2261 				continue;
2262 			if (publ->type != seq->type)
2263 				continue;
2264 			if (publ->lower != seq->lower)
2265 				continue;
2266 			if (publ->upper != seq->upper)
2267 				break;
2268 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2269 					      publ->ref, publ->key);
2270 			rc = 0;
2271 			break;
2272 		}
2273 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2274 				      publ->ref, publ->key);
2275 		rc = 0;
2276 	}
2277 	if (list_empty(&tsk->publications))
2278 		tsk->published = 0;
2279 	return rc;
2280 }
2281 
2282 /* tipc_sk_reinit: set non-zero address in all existing sockets
2283  *                 when we go from standalone to network mode.
2284  */
2285 void tipc_sk_reinit(struct net *net)
2286 {
2287 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2288 	const struct bucket_table *tbl;
2289 	struct rhash_head *pos;
2290 	struct tipc_sock *tsk;
2291 	struct tipc_msg *msg;
2292 	int i;
2293 
2294 	rcu_read_lock();
2295 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2296 	for (i = 0; i < tbl->size; i++) {
2297 		rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2298 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2299 			msg = &tsk->phdr;
2300 			msg_set_prevnode(msg, tn->own_addr);
2301 			msg_set_orignode(msg, tn->own_addr);
2302 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2303 		}
2304 	}
2305 	rcu_read_unlock();
2306 }
2307 
2308 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2309 {
2310 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2311 	struct tipc_sock *tsk;
2312 
2313 	rcu_read_lock();
2314 	tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2315 	if (tsk)
2316 		sock_hold(&tsk->sk);
2317 	rcu_read_unlock();
2318 
2319 	return tsk;
2320 }
2321 
2322 static int tipc_sk_insert(struct tipc_sock *tsk)
2323 {
2324 	struct sock *sk = &tsk->sk;
2325 	struct net *net = sock_net(sk);
2326 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2327 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2328 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2329 
2330 	while (remaining--) {
2331 		portid++;
2332 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2333 			portid = TIPC_MIN_PORT;
2334 		tsk->portid = portid;
2335 		sock_hold(&tsk->sk);
2336 		if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2337 			return 0;
2338 		sock_put(&tsk->sk);
2339 	}
2340 
2341 	return -1;
2342 }
2343 
2344 static void tipc_sk_remove(struct tipc_sock *tsk)
2345 {
2346 	struct sock *sk = &tsk->sk;
2347 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2348 
2349 	if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2350 		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2351 		__sock_put(sk);
2352 	}
2353 }
2354 
2355 int tipc_sk_rht_init(struct net *net)
2356 {
2357 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2358 	struct rhashtable_params rht_params = {
2359 		.nelem_hint = 192,
2360 		.head_offset = offsetof(struct tipc_sock, node),
2361 		.key_offset = offsetof(struct tipc_sock, portid),
2362 		.key_len = sizeof(u32), /* portid */
2363 		.hashfn = jhash,
2364 		.max_shift = 20, /* 1M */
2365 		.min_shift = 8,  /* 256 */
2366 	};
2367 
2368 	return rhashtable_init(&tn->sk_rht, &rht_params);
2369 }
2370 
2371 void tipc_sk_rht_destroy(struct net *net)
2372 {
2373 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2374 
2375 	/* Wait for socket readers to complete */
2376 	synchronize_net();
2377 
2378 	rhashtable_destroy(&tn->sk_rht);
2379 }
2380 
2381 /**
2382  * tipc_setsockopt - set socket option
2383  * @sock: socket structure
2384  * @lvl: option level
2385  * @opt: option identifier
2386  * @ov: pointer to new option value
2387  * @ol: length of option value
2388  *
2389  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2390  * (to ease compatibility).
2391  *
2392  * Returns 0 on success, errno otherwise
2393  */
2394 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2395 			   char __user *ov, unsigned int ol)
2396 {
2397 	struct sock *sk = sock->sk;
2398 	struct tipc_sock *tsk = tipc_sk(sk);
2399 	u32 value;
2400 	int res;
2401 
2402 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2403 		return 0;
2404 	if (lvl != SOL_TIPC)
2405 		return -ENOPROTOOPT;
2406 	if (ol < sizeof(value))
2407 		return -EINVAL;
2408 	res = get_user(value, (u32 __user *)ov);
2409 	if (res)
2410 		return res;
2411 
2412 	lock_sock(sk);
2413 
2414 	switch (opt) {
2415 	case TIPC_IMPORTANCE:
2416 		res = tsk_set_importance(tsk, value);
2417 		break;
2418 	case TIPC_SRC_DROPPABLE:
2419 		if (sock->type != SOCK_STREAM)
2420 			tsk_set_unreliable(tsk, value);
2421 		else
2422 			res = -ENOPROTOOPT;
2423 		break;
2424 	case TIPC_DEST_DROPPABLE:
2425 		tsk_set_unreturnable(tsk, value);
2426 		break;
2427 	case TIPC_CONN_TIMEOUT:
2428 		tipc_sk(sk)->conn_timeout = value;
2429 		/* no need to set "res", since already 0 at this point */
2430 		break;
2431 	default:
2432 		res = -EINVAL;
2433 	}
2434 
2435 	release_sock(sk);
2436 
2437 	return res;
2438 }
2439 
2440 /**
2441  * tipc_getsockopt - get socket option
2442  * @sock: socket structure
2443  * @lvl: option level
2444  * @opt: option identifier
2445  * @ov: receptacle for option value
2446  * @ol: receptacle for length of option value
2447  *
2448  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2449  * (to ease compatibility).
2450  *
2451  * Returns 0 on success, errno otherwise
2452  */
2453 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2454 			   char __user *ov, int __user *ol)
2455 {
2456 	struct sock *sk = sock->sk;
2457 	struct tipc_sock *tsk = tipc_sk(sk);
2458 	int len;
2459 	u32 value;
2460 	int res;
2461 
2462 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2463 		return put_user(0, ol);
2464 	if (lvl != SOL_TIPC)
2465 		return -ENOPROTOOPT;
2466 	res = get_user(len, ol);
2467 	if (res)
2468 		return res;
2469 
2470 	lock_sock(sk);
2471 
2472 	switch (opt) {
2473 	case TIPC_IMPORTANCE:
2474 		value = tsk_importance(tsk);
2475 		break;
2476 	case TIPC_SRC_DROPPABLE:
2477 		value = tsk_unreliable(tsk);
2478 		break;
2479 	case TIPC_DEST_DROPPABLE:
2480 		value = tsk_unreturnable(tsk);
2481 		break;
2482 	case TIPC_CONN_TIMEOUT:
2483 		value = tsk->conn_timeout;
2484 		/* no need to set "res", since already 0 at this point */
2485 		break;
2486 	case TIPC_NODE_RECVQ_DEPTH:
2487 		value = 0; /* was tipc_queue_size, now obsolete */
2488 		break;
2489 	case TIPC_SOCK_RECVQ_DEPTH:
2490 		value = skb_queue_len(&sk->sk_receive_queue);
2491 		break;
2492 	default:
2493 		res = -EINVAL;
2494 	}
2495 
2496 	release_sock(sk);
2497 
2498 	if (res)
2499 		return res;	/* "get" failed */
2500 
2501 	if (len < sizeof(value))
2502 		return -EINVAL;
2503 
2504 	if (copy_to_user(ov, &value, sizeof(value)))
2505 		return -EFAULT;
2506 
2507 	return put_user(sizeof(value), ol);
2508 }
2509 
2510 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2511 {
2512 	struct sock *sk = sock->sk;
2513 	struct tipc_sioc_ln_req lnr;
2514 	void __user *argp = (void __user *)arg;
2515 
2516 	switch (cmd) {
2517 	case SIOCGETLINKNAME:
2518 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2519 			return -EFAULT;
2520 		if (!tipc_node_get_linkname(sock_net(sk),
2521 					    lnr.bearer_id & 0xffff, lnr.peer,
2522 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2523 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2524 				return -EFAULT;
2525 			return 0;
2526 		}
2527 		return -EADDRNOTAVAIL;
2528 	default:
2529 		return -ENOIOCTLCMD;
2530 	}
2531 }
2532 
2533 /* Protocol switches for the various types of TIPC sockets */
2534 
2535 static const struct proto_ops msg_ops = {
2536 	.owner		= THIS_MODULE,
2537 	.family		= AF_TIPC,
2538 	.release	= tipc_release,
2539 	.bind		= tipc_bind,
2540 	.connect	= tipc_connect,
2541 	.socketpair	= sock_no_socketpair,
2542 	.accept		= sock_no_accept,
2543 	.getname	= tipc_getname,
2544 	.poll		= tipc_poll,
2545 	.ioctl		= tipc_ioctl,
2546 	.listen		= sock_no_listen,
2547 	.shutdown	= tipc_shutdown,
2548 	.setsockopt	= tipc_setsockopt,
2549 	.getsockopt	= tipc_getsockopt,
2550 	.sendmsg	= tipc_sendmsg,
2551 	.recvmsg	= tipc_recvmsg,
2552 	.mmap		= sock_no_mmap,
2553 	.sendpage	= sock_no_sendpage
2554 };
2555 
2556 static const struct proto_ops packet_ops = {
2557 	.owner		= THIS_MODULE,
2558 	.family		= AF_TIPC,
2559 	.release	= tipc_release,
2560 	.bind		= tipc_bind,
2561 	.connect	= tipc_connect,
2562 	.socketpair	= sock_no_socketpair,
2563 	.accept		= tipc_accept,
2564 	.getname	= tipc_getname,
2565 	.poll		= tipc_poll,
2566 	.ioctl		= tipc_ioctl,
2567 	.listen		= tipc_listen,
2568 	.shutdown	= tipc_shutdown,
2569 	.setsockopt	= tipc_setsockopt,
2570 	.getsockopt	= tipc_getsockopt,
2571 	.sendmsg	= tipc_send_packet,
2572 	.recvmsg	= tipc_recvmsg,
2573 	.mmap		= sock_no_mmap,
2574 	.sendpage	= sock_no_sendpage
2575 };
2576 
2577 static const struct proto_ops stream_ops = {
2578 	.owner		= THIS_MODULE,
2579 	.family		= AF_TIPC,
2580 	.release	= tipc_release,
2581 	.bind		= tipc_bind,
2582 	.connect	= tipc_connect,
2583 	.socketpair	= sock_no_socketpair,
2584 	.accept		= tipc_accept,
2585 	.getname	= tipc_getname,
2586 	.poll		= tipc_poll,
2587 	.ioctl		= tipc_ioctl,
2588 	.listen		= tipc_listen,
2589 	.shutdown	= tipc_shutdown,
2590 	.setsockopt	= tipc_setsockopt,
2591 	.getsockopt	= tipc_getsockopt,
2592 	.sendmsg	= tipc_send_stream,
2593 	.recvmsg	= tipc_recv_stream,
2594 	.mmap		= sock_no_mmap,
2595 	.sendpage	= sock_no_sendpage
2596 };
2597 
2598 static const struct net_proto_family tipc_family_ops = {
2599 	.owner		= THIS_MODULE,
2600 	.family		= AF_TIPC,
2601 	.create		= tipc_sk_create
2602 };
2603 
2604 static struct proto tipc_proto = {
2605 	.name		= "TIPC",
2606 	.owner		= THIS_MODULE,
2607 	.obj_size	= sizeof(struct tipc_sock),
2608 	.sysctl_rmem	= sysctl_tipc_rmem
2609 };
2610 
2611 static struct proto tipc_proto_kern = {
2612 	.name		= "TIPC",
2613 	.obj_size	= sizeof(struct tipc_sock),
2614 	.sysctl_rmem	= sysctl_tipc_rmem
2615 };
2616 
2617 /**
2618  * tipc_socket_init - initialize TIPC socket interface
2619  *
2620  * Returns 0 on success, errno otherwise
2621  */
2622 int tipc_socket_init(void)
2623 {
2624 	int res;
2625 
2626 	res = proto_register(&tipc_proto, 1);
2627 	if (res) {
2628 		pr_err("Failed to register TIPC protocol type\n");
2629 		goto out;
2630 	}
2631 
2632 	res = sock_register(&tipc_family_ops);
2633 	if (res) {
2634 		pr_err("Failed to register TIPC socket type\n");
2635 		proto_unregister(&tipc_proto);
2636 		goto out;
2637 	}
2638  out:
2639 	return res;
2640 }
2641 
2642 /**
2643  * tipc_socket_stop - stop TIPC socket interface
2644  */
2645 void tipc_socket_stop(void)
2646 {
2647 	sock_unregister(tipc_family_ops.family);
2648 	proto_unregister(&tipc_proto);
2649 }
2650 
2651 /* Caller should hold socket lock for the passed tipc socket. */
2652 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2653 {
2654 	u32 peer_node;
2655 	u32 peer_port;
2656 	struct nlattr *nest;
2657 
2658 	peer_node = tsk_peer_node(tsk);
2659 	peer_port = tsk_peer_port(tsk);
2660 
2661 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2662 
2663 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2664 		goto msg_full;
2665 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2666 		goto msg_full;
2667 
2668 	if (tsk->conn_type != 0) {
2669 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2670 			goto msg_full;
2671 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2672 			goto msg_full;
2673 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2674 			goto msg_full;
2675 	}
2676 	nla_nest_end(skb, nest);
2677 
2678 	return 0;
2679 
2680 msg_full:
2681 	nla_nest_cancel(skb, nest);
2682 
2683 	return -EMSGSIZE;
2684 }
2685 
2686 /* Caller should hold socket lock for the passed tipc socket. */
2687 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2688 			    struct tipc_sock *tsk)
2689 {
2690 	int err;
2691 	void *hdr;
2692 	struct nlattr *attrs;
2693 	struct net *net = sock_net(skb->sk);
2694 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2695 
2696 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2697 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2698 	if (!hdr)
2699 		goto msg_cancel;
2700 
2701 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2702 	if (!attrs)
2703 		goto genlmsg_cancel;
2704 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2705 		goto attr_msg_cancel;
2706 	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2707 		goto attr_msg_cancel;
2708 
2709 	if (tsk->connected) {
2710 		err = __tipc_nl_add_sk_con(skb, tsk);
2711 		if (err)
2712 			goto attr_msg_cancel;
2713 	} else if (!list_empty(&tsk->publications)) {
2714 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2715 			goto attr_msg_cancel;
2716 	}
2717 	nla_nest_end(skb, attrs);
2718 	genlmsg_end(skb, hdr);
2719 
2720 	return 0;
2721 
2722 attr_msg_cancel:
2723 	nla_nest_cancel(skb, attrs);
2724 genlmsg_cancel:
2725 	genlmsg_cancel(skb, hdr);
2726 msg_cancel:
2727 	return -EMSGSIZE;
2728 }
2729 
2730 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2731 {
2732 	int err;
2733 	struct tipc_sock *tsk;
2734 	const struct bucket_table *tbl;
2735 	struct rhash_head *pos;
2736 	struct net *net = sock_net(skb->sk);
2737 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2738 	u32 tbl_id = cb->args[0];
2739 	u32 prev_portid = cb->args[1];
2740 
2741 	rcu_read_lock();
2742 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2743 	for (; tbl_id < tbl->size; tbl_id++) {
2744 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2745 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2746 			if (prev_portid && prev_portid != tsk->portid) {
2747 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2748 				continue;
2749 			}
2750 
2751 			err = __tipc_nl_add_sk(skb, cb, tsk);
2752 			if (err) {
2753 				prev_portid = tsk->portid;
2754 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2755 				goto out;
2756 			}
2757 			prev_portid = 0;
2758 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2759 		}
2760 	}
2761 out:
2762 	rcu_read_unlock();
2763 	cb->args[0] = tbl_id;
2764 	cb->args[1] = prev_portid;
2765 
2766 	return skb->len;
2767 }
2768 
2769 /* Caller should hold socket lock for the passed tipc socket. */
2770 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2771 				 struct netlink_callback *cb,
2772 				 struct publication *publ)
2773 {
2774 	void *hdr;
2775 	struct nlattr *attrs;
2776 
2777 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2778 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2779 	if (!hdr)
2780 		goto msg_cancel;
2781 
2782 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2783 	if (!attrs)
2784 		goto genlmsg_cancel;
2785 
2786 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2787 		goto attr_msg_cancel;
2788 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2789 		goto attr_msg_cancel;
2790 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2791 		goto attr_msg_cancel;
2792 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2793 		goto attr_msg_cancel;
2794 
2795 	nla_nest_end(skb, attrs);
2796 	genlmsg_end(skb, hdr);
2797 
2798 	return 0;
2799 
2800 attr_msg_cancel:
2801 	nla_nest_cancel(skb, attrs);
2802 genlmsg_cancel:
2803 	genlmsg_cancel(skb, hdr);
2804 msg_cancel:
2805 	return -EMSGSIZE;
2806 }
2807 
2808 /* Caller should hold socket lock for the passed tipc socket. */
2809 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2810 				  struct netlink_callback *cb,
2811 				  struct tipc_sock *tsk, u32 *last_publ)
2812 {
2813 	int err;
2814 	struct publication *p;
2815 
2816 	if (*last_publ) {
2817 		list_for_each_entry(p, &tsk->publications, pport_list) {
2818 			if (p->key == *last_publ)
2819 				break;
2820 		}
2821 		if (p->key != *last_publ) {
2822 			/* We never set seq or call nl_dump_check_consistent()
2823 			 * this means that setting prev_seq here will cause the
2824 			 * consistence check to fail in the netlink callback
2825 			 * handler. Resulting in the last NLMSG_DONE message
2826 			 * having the NLM_F_DUMP_INTR flag set.
2827 			 */
2828 			cb->prev_seq = 1;
2829 			*last_publ = 0;
2830 			return -EPIPE;
2831 		}
2832 	} else {
2833 		p = list_first_entry(&tsk->publications, struct publication,
2834 				     pport_list);
2835 	}
2836 
2837 	list_for_each_entry_from(p, &tsk->publications, pport_list) {
2838 		err = __tipc_nl_add_sk_publ(skb, cb, p);
2839 		if (err) {
2840 			*last_publ = p->key;
2841 			return err;
2842 		}
2843 	}
2844 	*last_publ = 0;
2845 
2846 	return 0;
2847 }
2848 
2849 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2850 {
2851 	int err;
2852 	u32 tsk_portid = cb->args[0];
2853 	u32 last_publ = cb->args[1];
2854 	u32 done = cb->args[2];
2855 	struct net *net = sock_net(skb->sk);
2856 	struct tipc_sock *tsk;
2857 
2858 	if (!tsk_portid) {
2859 		struct nlattr **attrs;
2860 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2861 
2862 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
2863 		if (err)
2864 			return err;
2865 
2866 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2867 				       attrs[TIPC_NLA_SOCK],
2868 				       tipc_nl_sock_policy);
2869 		if (err)
2870 			return err;
2871 
2872 		if (!sock[TIPC_NLA_SOCK_REF])
2873 			return -EINVAL;
2874 
2875 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2876 	}
2877 
2878 	if (done)
2879 		return 0;
2880 
2881 	tsk = tipc_sk_lookup(net, tsk_portid);
2882 	if (!tsk)
2883 		return -EINVAL;
2884 
2885 	lock_sock(&tsk->sk);
2886 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2887 	if (!err)
2888 		done = 1;
2889 	release_sock(&tsk->sk);
2890 	sock_put(&tsk->sk);
2891 
2892 	cb->args[0] = tsk_portid;
2893 	cb->args[1] = last_publ;
2894 	cb->args[2] = done;
2895 
2896 	return skb->len;
2897 }
2898