xref: /linux/net/tipc/socket.c (revision 83a37b3292f4aca799b355179ad6fbdd78a08e10)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2017, 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/sched/signal.h>
39 
40 #include "core.h"
41 #include "name_table.h"
42 #include "node.h"
43 #include "link.h"
44 #include "name_distr.h"
45 #include "socket.h"
46 #include "bcast.h"
47 #include "netlink.h"
48 #include "group.h"
49 
50 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
51 #define CONN_PROBING_INTERVAL	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
52 #define TIPC_FWD_MSG		1
53 #define TIPC_MAX_PORT		0xffffffff
54 #define TIPC_MIN_PORT		1
55 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
56 
57 enum {
58 	TIPC_LISTEN = TCP_LISTEN,
59 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
60 	TIPC_OPEN = TCP_CLOSE,
61 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
62 	TIPC_CONNECTING = TCP_SYN_SENT,
63 };
64 
65 struct sockaddr_pair {
66 	struct sockaddr_tipc sock;
67 	struct sockaddr_tipc member;
68 };
69 
70 /**
71  * struct tipc_sock - TIPC socket structure
72  * @sk: socket - interacts with 'port' and with user via the socket API
73  * @conn_type: TIPC type used when connection was established
74  * @conn_instance: TIPC instance used when connection was established
75  * @published: non-zero if port has one or more associated names
76  * @max_pkt: maximum packet size "hint" used when building messages sent by port
77  * @portid: unique port identity in TIPC socket hash table
78  * @phdr: preformatted message header used when sending messages
79  * #cong_links: list of congested links
80  * @publications: list of publications for port
81  * @blocking_link: address of the congested link we are currently sleeping on
82  * @pub_count: total # of publications port has made during its lifetime
83  * @probing_state:
84  * @conn_timeout: the time we can wait for an unresponded setup request
85  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86  * @cong_link_cnt: number of congested links
87  * @snt_unacked: # messages sent by socket, and not yet acked by peer
88  * @rcv_unacked: # messages read by user, but not yet acked back to peer
89  * @peer: 'connected' peer for dgram/rdm
90  * @node: hash table node
91  * @mc_method: cookie for use between socket and broadcast layer
92  * @rcu: rcu struct for tipc_sock
93  */
94 struct tipc_sock {
95 	struct sock sk;
96 	u32 conn_type;
97 	u32 conn_instance;
98 	int published;
99 	u32 max_pkt;
100 	u32 portid;
101 	struct tipc_msg phdr;
102 	struct list_head cong_links;
103 	struct list_head publications;
104 	u32 pub_count;
105 	uint conn_timeout;
106 	atomic_t dupl_rcvcnt;
107 	bool probe_unacked;
108 	u16 cong_link_cnt;
109 	u16 snt_unacked;
110 	u16 snd_win;
111 	u16 peer_caps;
112 	u16 rcv_unacked;
113 	u16 rcv_win;
114 	struct sockaddr_tipc peer;
115 	struct rhash_head node;
116 	struct tipc_mc_method mc_method;
117 	struct rcu_head rcu;
118 	struct tipc_group *group;
119 };
120 
121 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
122 static void tipc_data_ready(struct sock *sk);
123 static void tipc_write_space(struct sock *sk);
124 static void tipc_sock_destruct(struct sock *sk);
125 static int tipc_release(struct socket *sock);
126 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
127 		       bool kern);
128 static void tipc_sk_timeout(unsigned long data);
129 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
130 			   struct tipc_name_seq const *seq);
131 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
132 			    struct tipc_name_seq const *seq);
133 static int tipc_sk_leave(struct tipc_sock *tsk);
134 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
135 static int tipc_sk_insert(struct tipc_sock *tsk);
136 static void tipc_sk_remove(struct tipc_sock *tsk);
137 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
138 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
139 
140 static const struct proto_ops packet_ops;
141 static const struct proto_ops stream_ops;
142 static const struct proto_ops msg_ops;
143 static struct proto tipc_proto;
144 static const struct rhashtable_params tsk_rht_params;
145 
146 static u32 tsk_own_node(struct tipc_sock *tsk)
147 {
148 	return msg_prevnode(&tsk->phdr);
149 }
150 
151 static u32 tsk_peer_node(struct tipc_sock *tsk)
152 {
153 	return msg_destnode(&tsk->phdr);
154 }
155 
156 static u32 tsk_peer_port(struct tipc_sock *tsk)
157 {
158 	return msg_destport(&tsk->phdr);
159 }
160 
161 static  bool tsk_unreliable(struct tipc_sock *tsk)
162 {
163 	return msg_src_droppable(&tsk->phdr) != 0;
164 }
165 
166 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
167 {
168 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
169 }
170 
171 static bool tsk_unreturnable(struct tipc_sock *tsk)
172 {
173 	return msg_dest_droppable(&tsk->phdr) != 0;
174 }
175 
176 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
177 {
178 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
179 }
180 
181 static int tsk_importance(struct tipc_sock *tsk)
182 {
183 	return msg_importance(&tsk->phdr);
184 }
185 
186 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
187 {
188 	if (imp > TIPC_CRITICAL_IMPORTANCE)
189 		return -EINVAL;
190 	msg_set_importance(&tsk->phdr, (u32)imp);
191 	return 0;
192 }
193 
194 static struct tipc_sock *tipc_sk(const struct sock *sk)
195 {
196 	return container_of(sk, struct tipc_sock, sk);
197 }
198 
199 static bool tsk_conn_cong(struct tipc_sock *tsk)
200 {
201 	return tsk->snt_unacked > tsk->snd_win;
202 }
203 
204 static u16 tsk_blocks(int len)
205 {
206 	return ((len / FLOWCTL_BLK_SZ) + 1);
207 }
208 
209 /* tsk_blocks(): translate a buffer size in bytes to number of
210  * advertisable blocks, taking into account the ratio truesize(len)/len
211  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
212  */
213 static u16 tsk_adv_blocks(int len)
214 {
215 	return len / FLOWCTL_BLK_SZ / 4;
216 }
217 
218 /* tsk_inc(): increment counter for sent or received data
219  * - If block based flow control is not supported by peer we
220  *   fall back to message based ditto, incrementing the counter
221  */
222 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
223 {
224 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
225 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
226 	return 1;
227 }
228 
229 /**
230  * tsk_advance_rx_queue - discard first buffer in socket receive queue
231  *
232  * Caller must hold socket lock
233  */
234 static void tsk_advance_rx_queue(struct sock *sk)
235 {
236 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
237 }
238 
239 /* tipc_sk_respond() : send response message back to sender
240  */
241 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
242 {
243 	u32 selector;
244 	u32 dnode;
245 	u32 onode = tipc_own_addr(sock_net(sk));
246 
247 	if (!tipc_msg_reverse(onode, &skb, err))
248 		return;
249 
250 	dnode = msg_destnode(buf_msg(skb));
251 	selector = msg_origport(buf_msg(skb));
252 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
253 }
254 
255 /**
256  * tsk_rej_rx_queue - reject all buffers in socket receive queue
257  *
258  * Caller must hold socket lock
259  */
260 static void tsk_rej_rx_queue(struct sock *sk)
261 {
262 	struct sk_buff *skb;
263 
264 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
265 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
266 }
267 
268 static bool tipc_sk_connected(struct sock *sk)
269 {
270 	return sk->sk_state == TIPC_ESTABLISHED;
271 }
272 
273 /* tipc_sk_type_connectionless - check if the socket is datagram socket
274  * @sk: socket
275  *
276  * Returns true if connection less, false otherwise
277  */
278 static bool tipc_sk_type_connectionless(struct sock *sk)
279 {
280 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
281 }
282 
283 /* tsk_peer_msg - verify if message was sent by connected port's peer
284  *
285  * Handles cases where the node's network address has changed from
286  * the default of <0.0.0> to its configured setting.
287  */
288 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
289 {
290 	struct sock *sk = &tsk->sk;
291 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
292 	u32 peer_port = tsk_peer_port(tsk);
293 	u32 orig_node;
294 	u32 peer_node;
295 
296 	if (unlikely(!tipc_sk_connected(sk)))
297 		return false;
298 
299 	if (unlikely(msg_origport(msg) != peer_port))
300 		return false;
301 
302 	orig_node = msg_orignode(msg);
303 	peer_node = tsk_peer_node(tsk);
304 
305 	if (likely(orig_node == peer_node))
306 		return true;
307 
308 	if (!orig_node && (peer_node == tn->own_addr))
309 		return true;
310 
311 	if (!peer_node && (orig_node == tn->own_addr))
312 		return true;
313 
314 	return false;
315 }
316 
317 /* tipc_set_sk_state - set the sk_state of the socket
318  * @sk: socket
319  *
320  * Caller must hold socket lock
321  *
322  * Returns 0 on success, errno otherwise
323  */
324 static int tipc_set_sk_state(struct sock *sk, int state)
325 {
326 	int oldsk_state = sk->sk_state;
327 	int res = -EINVAL;
328 
329 	switch (state) {
330 	case TIPC_OPEN:
331 		res = 0;
332 		break;
333 	case TIPC_LISTEN:
334 	case TIPC_CONNECTING:
335 		if (oldsk_state == TIPC_OPEN)
336 			res = 0;
337 		break;
338 	case TIPC_ESTABLISHED:
339 		if (oldsk_state == TIPC_CONNECTING ||
340 		    oldsk_state == TIPC_OPEN)
341 			res = 0;
342 		break;
343 	case TIPC_DISCONNECTING:
344 		if (oldsk_state == TIPC_CONNECTING ||
345 		    oldsk_state == TIPC_ESTABLISHED)
346 			res = 0;
347 		break;
348 	}
349 
350 	if (!res)
351 		sk->sk_state = state;
352 
353 	return res;
354 }
355 
356 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
357 {
358 	struct sock *sk = sock->sk;
359 	int err = sock_error(sk);
360 	int typ = sock->type;
361 
362 	if (err)
363 		return err;
364 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
365 		if (sk->sk_state == TIPC_DISCONNECTING)
366 			return -EPIPE;
367 		else if (!tipc_sk_connected(sk))
368 			return -ENOTCONN;
369 	}
370 	if (!*timeout)
371 		return -EAGAIN;
372 	if (signal_pending(current))
373 		return sock_intr_errno(*timeout);
374 
375 	return 0;
376 }
377 
378 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
379 ({                                                                             \
380 	struct sock *sk_;						       \
381 	int rc_;							       \
382 									       \
383 	while ((rc_ = !(condition_))) {					       \
384 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
385 		sk_ = (sock_)->sk;					       \
386 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
387 		if (rc_)						       \
388 			break;						       \
389 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
390 		release_sock(sk_);					       \
391 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
392 		sched_annotate_sleep();				               \
393 		lock_sock(sk_);						       \
394 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
395 	}								       \
396 	rc_;								       \
397 })
398 
399 /**
400  * tipc_sk_create - create a TIPC socket
401  * @net: network namespace (must be default network)
402  * @sock: pre-allocated socket structure
403  * @protocol: protocol indicator (must be 0)
404  * @kern: caused by kernel or by userspace?
405  *
406  * This routine creates additional data structures used by the TIPC socket,
407  * initializes them, and links them together.
408  *
409  * Returns 0 on success, errno otherwise
410  */
411 static int tipc_sk_create(struct net *net, struct socket *sock,
412 			  int protocol, int kern)
413 {
414 	struct tipc_net *tn;
415 	const struct proto_ops *ops;
416 	struct sock *sk;
417 	struct tipc_sock *tsk;
418 	struct tipc_msg *msg;
419 
420 	/* Validate arguments */
421 	if (unlikely(protocol != 0))
422 		return -EPROTONOSUPPORT;
423 
424 	switch (sock->type) {
425 	case SOCK_STREAM:
426 		ops = &stream_ops;
427 		break;
428 	case SOCK_SEQPACKET:
429 		ops = &packet_ops;
430 		break;
431 	case SOCK_DGRAM:
432 	case SOCK_RDM:
433 		ops = &msg_ops;
434 		break;
435 	default:
436 		return -EPROTOTYPE;
437 	}
438 
439 	/* Allocate socket's protocol area */
440 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
441 	if (sk == NULL)
442 		return -ENOMEM;
443 
444 	tsk = tipc_sk(sk);
445 	tsk->max_pkt = MAX_PKT_DEFAULT;
446 	INIT_LIST_HEAD(&tsk->publications);
447 	INIT_LIST_HEAD(&tsk->cong_links);
448 	msg = &tsk->phdr;
449 	tn = net_generic(sock_net(sk), tipc_net_id);
450 
451 	/* Finish initializing socket data structures */
452 	sock->ops = ops;
453 	sock_init_data(sock, sk);
454 	tipc_set_sk_state(sk, TIPC_OPEN);
455 	if (tipc_sk_insert(tsk)) {
456 		pr_warn("Socket create failed; port number exhausted\n");
457 		return -EINVAL;
458 	}
459 
460 	/* Ensure tsk is visible before we read own_addr. */
461 	smp_mb();
462 
463 	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
464 		      NAMED_H_SIZE, 0);
465 
466 	msg_set_origport(msg, tsk->portid);
467 	setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
468 	sk->sk_shutdown = 0;
469 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
470 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
471 	sk->sk_data_ready = tipc_data_ready;
472 	sk->sk_write_space = tipc_write_space;
473 	sk->sk_destruct = tipc_sock_destruct;
474 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
475 	atomic_set(&tsk->dupl_rcvcnt, 0);
476 
477 	/* Start out with safe limits until we receive an advertised window */
478 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
479 	tsk->rcv_win = tsk->snd_win;
480 
481 	if (tipc_sk_type_connectionless(sk)) {
482 		tsk_set_unreturnable(tsk, true);
483 		if (sock->type == SOCK_DGRAM)
484 			tsk_set_unreliable(tsk, true);
485 	}
486 
487 	return 0;
488 }
489 
490 static void tipc_sk_callback(struct rcu_head *head)
491 {
492 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
493 
494 	sock_put(&tsk->sk);
495 }
496 
497 /* Caller should hold socket lock for the socket. */
498 static void __tipc_shutdown(struct socket *sock, int error)
499 {
500 	struct sock *sk = sock->sk;
501 	struct tipc_sock *tsk = tipc_sk(sk);
502 	struct net *net = sock_net(sk);
503 	long timeout = CONN_TIMEOUT_DEFAULT;
504 	u32 dnode = tsk_peer_node(tsk);
505 	struct sk_buff *skb;
506 
507 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
508 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
509 					    !tsk_conn_cong(tsk)));
510 
511 	/* Reject all unreceived messages, except on an active connection
512 	 * (which disconnects locally & sends a 'FIN+' to peer).
513 	 */
514 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
515 		if (TIPC_SKB_CB(skb)->bytes_read) {
516 			kfree_skb(skb);
517 			continue;
518 		}
519 		if (!tipc_sk_type_connectionless(sk) &&
520 		    sk->sk_state != TIPC_DISCONNECTING) {
521 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
522 			tipc_node_remove_conn(net, dnode, tsk->portid);
523 		}
524 		tipc_sk_respond(sk, skb, error);
525 	}
526 
527 	if (tipc_sk_type_connectionless(sk))
528 		return;
529 
530 	if (sk->sk_state != TIPC_DISCONNECTING) {
531 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
532 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
533 				      tsk_own_node(tsk), tsk_peer_port(tsk),
534 				      tsk->portid, error);
535 		if (skb)
536 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
537 		tipc_node_remove_conn(net, dnode, tsk->portid);
538 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
539 	}
540 }
541 
542 /**
543  * tipc_release - destroy a TIPC socket
544  * @sock: socket to destroy
545  *
546  * This routine cleans up any messages that are still queued on the socket.
547  * For DGRAM and RDM socket types, all queued messages are rejected.
548  * For SEQPACKET and STREAM socket types, the first message is rejected
549  * and any others are discarded.  (If the first message on a STREAM socket
550  * is partially-read, it is discarded and the next one is rejected instead.)
551  *
552  * NOTE: Rejected messages are not necessarily returned to the sender!  They
553  * are returned or discarded according to the "destination droppable" setting
554  * specified for the message by the sender.
555  *
556  * Returns 0 on success, errno otherwise
557  */
558 static int tipc_release(struct socket *sock)
559 {
560 	struct sock *sk = sock->sk;
561 	struct tipc_sock *tsk;
562 
563 	/*
564 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
565 	 * releases a pre-allocated child socket that was never used)
566 	 */
567 	if (sk == NULL)
568 		return 0;
569 
570 	tsk = tipc_sk(sk);
571 	lock_sock(sk);
572 
573 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
574 	sk->sk_shutdown = SHUTDOWN_MASK;
575 	tipc_sk_leave(tsk);
576 	tipc_sk_withdraw(tsk, 0, NULL);
577 	sk_stop_timer(sk, &sk->sk_timer);
578 	tipc_sk_remove(tsk);
579 
580 	/* Reject any messages that accumulated in backlog queue */
581 	release_sock(sk);
582 	tipc_dest_list_purge(&tsk->cong_links);
583 	tsk->cong_link_cnt = 0;
584 	call_rcu(&tsk->rcu, tipc_sk_callback);
585 	sock->sk = NULL;
586 
587 	return 0;
588 }
589 
590 /**
591  * tipc_bind - associate or disassocate TIPC name(s) with a socket
592  * @sock: socket structure
593  * @uaddr: socket address describing name(s) and desired operation
594  * @uaddr_len: size of socket address data structure
595  *
596  * Name and name sequence binding is indicated using a positive scope value;
597  * a negative scope value unbinds the specified name.  Specifying no name
598  * (i.e. a socket address length of 0) unbinds all names from the socket.
599  *
600  * Returns 0 on success, errno otherwise
601  *
602  * NOTE: This routine doesn't need to take the socket lock since it doesn't
603  *       access any non-constant socket information.
604  */
605 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
606 		     int uaddr_len)
607 {
608 	struct sock *sk = sock->sk;
609 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
610 	struct tipc_sock *tsk = tipc_sk(sk);
611 	int res = -EINVAL;
612 
613 	lock_sock(sk);
614 	if (unlikely(!uaddr_len)) {
615 		res = tipc_sk_withdraw(tsk, 0, NULL);
616 		goto exit;
617 	}
618 	if (tsk->group) {
619 		res = -EACCES;
620 		goto exit;
621 	}
622 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
623 		res = -EINVAL;
624 		goto exit;
625 	}
626 	if (addr->family != AF_TIPC) {
627 		res = -EAFNOSUPPORT;
628 		goto exit;
629 	}
630 
631 	if (addr->addrtype == TIPC_ADDR_NAME)
632 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
633 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
634 		res = -EAFNOSUPPORT;
635 		goto exit;
636 	}
637 
638 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
639 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
640 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
641 		res = -EACCES;
642 		goto exit;
643 	}
644 
645 	res = (addr->scope > 0) ?
646 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
647 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
648 exit:
649 	release_sock(sk);
650 	return res;
651 }
652 
653 /**
654  * tipc_getname - get port ID of socket or peer socket
655  * @sock: socket structure
656  * @uaddr: area for returned socket address
657  * @uaddr_len: area for returned length of socket address
658  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
659  *
660  * Returns 0 on success, errno otherwise
661  *
662  * NOTE: This routine doesn't need to take the socket lock since it only
663  *       accesses socket information that is unchanging (or which changes in
664  *       a completely predictable manner).
665  */
666 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
667 			int *uaddr_len, int peer)
668 {
669 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
670 	struct sock *sk = sock->sk;
671 	struct tipc_sock *tsk = tipc_sk(sk);
672 	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
673 
674 	memset(addr, 0, sizeof(*addr));
675 	if (peer) {
676 		if ((!tipc_sk_connected(sk)) &&
677 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
678 			return -ENOTCONN;
679 		addr->addr.id.ref = tsk_peer_port(tsk);
680 		addr->addr.id.node = tsk_peer_node(tsk);
681 	} else {
682 		addr->addr.id.ref = tsk->portid;
683 		addr->addr.id.node = tn->own_addr;
684 	}
685 
686 	*uaddr_len = sizeof(*addr);
687 	addr->addrtype = TIPC_ADDR_ID;
688 	addr->family = AF_TIPC;
689 	addr->scope = 0;
690 	addr->addr.name.domain = 0;
691 
692 	return 0;
693 }
694 
695 /**
696  * tipc_poll - read and possibly block on pollmask
697  * @file: file structure associated with the socket
698  * @sock: socket for which to calculate the poll bits
699  * @wait: ???
700  *
701  * Returns pollmask value
702  *
703  * COMMENTARY:
704  * It appears that the usual socket locking mechanisms are not useful here
705  * since the pollmask info is potentially out-of-date the moment this routine
706  * exits.  TCP and other protocols seem to rely on higher level poll routines
707  * to handle any preventable race conditions, so TIPC will do the same ...
708  *
709  * IMPORTANT: The fact that a read or write operation is indicated does NOT
710  * imply that the operation will succeed, merely that it should be performed
711  * and will not block.
712  */
713 static unsigned int tipc_poll(struct file *file, struct socket *sock,
714 			      poll_table *wait)
715 {
716 	struct sock *sk = sock->sk;
717 	struct sk_buff *skb = skb_peek(&sk->sk_receive_queue);
718 	struct tipc_sock *tsk = tipc_sk(sk);
719 	struct tipc_group *grp = tsk->group;
720 	u32 revents = 0;
721 
722 	sock_poll_wait(file, sk_sleep(sk), wait);
723 
724 	if (sk->sk_shutdown & RCV_SHUTDOWN)
725 		revents |= POLLRDHUP | POLLIN | POLLRDNORM;
726 	if (sk->sk_shutdown == SHUTDOWN_MASK)
727 		revents |= POLLHUP;
728 
729 	switch (sk->sk_state) {
730 	case TIPC_ESTABLISHED:
731 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
732 			revents |= POLLOUT;
733 		/* fall thru' */
734 	case TIPC_LISTEN:
735 	case TIPC_CONNECTING:
736 		if (skb)
737 			revents |= POLLIN | POLLRDNORM;
738 		break;
739 	case TIPC_OPEN:
740 		if (!grp || tipc_group_size(grp))
741 			if (!tsk->cong_link_cnt)
742 				revents |= POLLOUT;
743 		if (!tipc_sk_type_connectionless(sk))
744 			break;
745 		if (!skb)
746 			break;
747 		revents |= POLLIN | POLLRDNORM;
748 		break;
749 	case TIPC_DISCONNECTING:
750 		revents = POLLIN | POLLRDNORM | POLLHUP;
751 		break;
752 	}
753 	return revents;
754 }
755 
756 /**
757  * tipc_sendmcast - send multicast message
758  * @sock: socket structure
759  * @seq: destination address
760  * @msg: message to send
761  * @dlen: length of data to send
762  * @timeout: timeout to wait for wakeup
763  *
764  * Called from function tipc_sendmsg(), which has done all sanity checks
765  * Returns the number of bytes sent on success, or errno
766  */
767 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
768 			  struct msghdr *msg, size_t dlen, long timeout)
769 {
770 	struct sock *sk = sock->sk;
771 	struct tipc_sock *tsk = tipc_sk(sk);
772 	struct tipc_msg *hdr = &tsk->phdr;
773 	struct net *net = sock_net(sk);
774 	int mtu = tipc_bcast_get_mtu(net);
775 	struct tipc_mc_method *method = &tsk->mc_method;
776 	u32 domain = addr_domain(net, TIPC_CLUSTER_SCOPE);
777 	struct sk_buff_head pkts;
778 	struct tipc_nlist dsts;
779 	int rc;
780 
781 	if (tsk->group)
782 		return -EACCES;
783 
784 	/* Block or return if any destination link is congested */
785 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
786 	if (unlikely(rc))
787 		return rc;
788 
789 	/* Lookup destination nodes */
790 	tipc_nlist_init(&dsts, tipc_own_addr(net));
791 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
792 				      seq->upper, domain, &dsts);
793 	if (!dsts.local && !dsts.remote)
794 		return -EHOSTUNREACH;
795 
796 	/* Build message header */
797 	msg_set_type(hdr, TIPC_MCAST_MSG);
798 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
799 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
800 	msg_set_destport(hdr, 0);
801 	msg_set_destnode(hdr, 0);
802 	msg_set_nametype(hdr, seq->type);
803 	msg_set_namelower(hdr, seq->lower);
804 	msg_set_nameupper(hdr, seq->upper);
805 
806 	/* Build message as chain of buffers */
807 	skb_queue_head_init(&pkts);
808 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
809 
810 	/* Send message if build was successful */
811 	if (unlikely(rc == dlen))
812 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
813 				     &tsk->cong_link_cnt);
814 
815 	tipc_nlist_purge(&dsts);
816 
817 	return rc ? rc : dlen;
818 }
819 
820 /**
821  * tipc_send_group_msg - send a message to a member in the group
822  * @net: network namespace
823  * @m: message to send
824  * @mb: group member
825  * @dnode: destination node
826  * @dport: destination port
827  * @dlen: total length of message data
828  */
829 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
830 			       struct msghdr *m, struct tipc_member *mb,
831 			       u32 dnode, u32 dport, int dlen)
832 {
833 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
834 	struct tipc_mc_method *method = &tsk->mc_method;
835 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
836 	struct tipc_msg *hdr = &tsk->phdr;
837 	struct sk_buff_head pkts;
838 	int mtu, rc;
839 
840 	/* Complete message header */
841 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
842 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
843 	msg_set_destport(hdr, dport);
844 	msg_set_destnode(hdr, dnode);
845 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
846 
847 	/* Build message as chain of buffers */
848 	skb_queue_head_init(&pkts);
849 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
850 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
851 	if (unlikely(rc != dlen))
852 		return rc;
853 
854 	/* Send message */
855 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
856 	if (unlikely(rc == -ELINKCONG)) {
857 		tipc_dest_push(&tsk->cong_links, dnode, 0);
858 		tsk->cong_link_cnt++;
859 	}
860 
861 	/* Update send window */
862 	tipc_group_update_member(mb, blks);
863 
864 	/* A broadcast sent within next EXPIRE period must follow same path */
865 	method->rcast = true;
866 	method->mandatory = true;
867 	return dlen;
868 }
869 
870 /**
871  * tipc_send_group_unicast - send message to a member in the group
872  * @sock: socket structure
873  * @m: message to send
874  * @dlen: total length of message data
875  * @timeout: timeout to wait for wakeup
876  *
877  * Called from function tipc_sendmsg(), which has done all sanity checks
878  * Returns the number of bytes sent on success, or errno
879  */
880 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
881 				   int dlen, long timeout)
882 {
883 	struct sock *sk = sock->sk;
884 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
885 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
886 	struct tipc_sock *tsk = tipc_sk(sk);
887 	struct tipc_group *grp = tsk->group;
888 	struct net *net = sock_net(sk);
889 	struct tipc_member *mb = NULL;
890 	u32 node, port;
891 	int rc;
892 
893 	node = dest->addr.id.node;
894 	port = dest->addr.id.ref;
895 	if (!port && !node)
896 		return -EHOSTUNREACH;
897 
898 	/* Block or return if destination link or member is congested */
899 	rc = tipc_wait_for_cond(sock, &timeout,
900 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
901 				!tipc_group_cong(grp, node, port, blks, &mb));
902 	if (unlikely(rc))
903 		return rc;
904 
905 	if (unlikely(!mb))
906 		return -EHOSTUNREACH;
907 
908 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
909 
910 	return rc ? rc : dlen;
911 }
912 
913 /**
914  * tipc_send_group_anycast - send message to any member with given identity
915  * @sock: socket structure
916  * @m: message to send
917  * @dlen: total length of message data
918  * @timeout: timeout to wait for wakeup
919  *
920  * Called from function tipc_sendmsg(), which has done all sanity checks
921  * Returns the number of bytes sent on success, or errno
922  */
923 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
924 				   int dlen, long timeout)
925 {
926 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
927 	struct sock *sk = sock->sk;
928 	struct tipc_sock *tsk = tipc_sk(sk);
929 	struct list_head *cong_links = &tsk->cong_links;
930 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
931 	struct tipc_group *grp = tsk->group;
932 	struct tipc_member *first = NULL;
933 	struct tipc_member *mbr = NULL;
934 	struct net *net = sock_net(sk);
935 	u32 node, port, exclude;
936 	u32 type, inst, domain;
937 	struct list_head dsts;
938 	int lookups = 0;
939 	int dstcnt, rc;
940 	bool cong;
941 
942 	INIT_LIST_HEAD(&dsts);
943 
944 	type = dest->addr.name.name.type;
945 	inst = dest->addr.name.name.instance;
946 	domain = addr_domain(net, dest->scope);
947 	exclude = tipc_group_exclude(grp);
948 
949 	while (++lookups < 4) {
950 		first = NULL;
951 
952 		/* Look for a non-congested destination member, if any */
953 		while (1) {
954 			if (!tipc_nametbl_lookup(net, type, inst, domain, &dsts,
955 						 &dstcnt, exclude, false))
956 				return -EHOSTUNREACH;
957 			tipc_dest_pop(&dsts, &node, &port);
958 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
959 			if (!cong)
960 				break;
961 			if (mbr == first)
962 				break;
963 			if (!first)
964 				first = mbr;
965 		}
966 
967 		/* Start over if destination was not in member list */
968 		if (unlikely(!mbr))
969 			continue;
970 
971 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
972 			break;
973 
974 		/* Block or return if destination link or member is congested */
975 		rc = tipc_wait_for_cond(sock, &timeout,
976 					!tipc_dest_find(cong_links, node, 0) &&
977 					!tipc_group_cong(grp, node, port,
978 							 blks, &mbr));
979 		if (unlikely(rc))
980 			return rc;
981 
982 		/* Send, unless destination disappeared while waiting */
983 		if (likely(mbr))
984 			break;
985 	}
986 
987 	if (unlikely(lookups >= 4))
988 		return -EHOSTUNREACH;
989 
990 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
991 
992 	return rc ? rc : dlen;
993 }
994 
995 /**
996  * tipc_send_group_bcast - send message to all members in communication group
997  * @sk: socket structure
998  * @m: message to send
999  * @dlen: total length of message data
1000  * @timeout: timeout to wait for wakeup
1001  *
1002  * Called from function tipc_sendmsg(), which has done all sanity checks
1003  * Returns the number of bytes sent on success, or errno
1004  */
1005 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1006 				 int dlen, long timeout)
1007 {
1008 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1009 	struct sock *sk = sock->sk;
1010 	struct net *net = sock_net(sk);
1011 	struct tipc_sock *tsk = tipc_sk(sk);
1012 	struct tipc_group *grp = tsk->group;
1013 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1014 	struct tipc_mc_method *method = &tsk->mc_method;
1015 	bool ack = method->mandatory && method->rcast;
1016 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1017 	struct tipc_msg *hdr = &tsk->phdr;
1018 	int mtu = tipc_bcast_get_mtu(net);
1019 	struct sk_buff_head pkts;
1020 	int rc = -EHOSTUNREACH;
1021 
1022 	if (!dsts->local && !dsts->remote)
1023 		return -EHOSTUNREACH;
1024 
1025 	/* Block or return if any destination link or member is congested */
1026 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1027 				!tipc_group_bc_cong(grp, blks));
1028 	if (unlikely(rc))
1029 		return rc;
1030 
1031 	/* Complete message header */
1032 	if (dest) {
1033 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1034 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1035 	} else {
1036 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1037 		msg_set_nameinst(hdr, 0);
1038 	}
1039 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1040 	msg_set_destport(hdr, 0);
1041 	msg_set_destnode(hdr, 0);
1042 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1043 
1044 	/* Avoid getting stuck with repeated forced replicasts */
1045 	msg_set_grp_bc_ack_req(hdr, ack);
1046 
1047 	/* Build message as chain of buffers */
1048 	skb_queue_head_init(&pkts);
1049 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1050 	if (unlikely(rc != dlen))
1051 		return rc;
1052 
1053 	/* Send message */
1054 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1055 	if (unlikely(rc))
1056 		return rc;
1057 
1058 	/* Update broadcast sequence number and send windows */
1059 	tipc_group_update_bc_members(tsk->group, blks, ack);
1060 
1061 	/* Broadcast link is now free to choose method for next broadcast */
1062 	method->mandatory = false;
1063 	method->expires = jiffies;
1064 
1065 	return dlen;
1066 }
1067 
1068 /**
1069  * tipc_send_group_mcast - send message to all members with given identity
1070  * @sock: socket structure
1071  * @m: message to send
1072  * @dlen: total length of message data
1073  * @timeout: timeout to wait for wakeup
1074  *
1075  * Called from function tipc_sendmsg(), which has done all sanity checks
1076  * Returns the number of bytes sent on success, or errno
1077  */
1078 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1079 				 int dlen, long timeout)
1080 {
1081 	struct sock *sk = sock->sk;
1082 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1083 	struct tipc_name_seq *seq = &dest->addr.nameseq;
1084 	struct tipc_sock *tsk = tipc_sk(sk);
1085 	struct tipc_group *grp = tsk->group;
1086 	struct net *net = sock_net(sk);
1087 	u32 domain, exclude, dstcnt;
1088 	struct list_head dsts;
1089 
1090 	INIT_LIST_HEAD(&dsts);
1091 
1092 	if (seq->lower != seq->upper)
1093 		return -ENOTSUPP;
1094 
1095 	domain = addr_domain(net, dest->scope);
1096 	exclude = tipc_group_exclude(grp);
1097 	if (!tipc_nametbl_lookup(net, seq->type, seq->lower, domain,
1098 				 &dsts, &dstcnt, exclude, true))
1099 		return -EHOSTUNREACH;
1100 
1101 	if (dstcnt == 1) {
1102 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1103 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1104 	}
1105 
1106 	tipc_dest_list_purge(&dsts);
1107 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1108 }
1109 
1110 /**
1111  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1112  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1113  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1114  *
1115  * Multi-threaded: parallel calls with reference to same queues may occur
1116  */
1117 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1118 		       struct sk_buff_head *inputq)
1119 {
1120 	u32 scope = TIPC_CLUSTER_SCOPE;
1121 	u32 self = tipc_own_addr(net);
1122 	struct sk_buff *skb, *_skb;
1123 	u32 lower = 0, upper = ~0;
1124 	struct sk_buff_head tmpq;
1125 	u32 portid, oport, onode;
1126 	struct list_head dports;
1127 	struct tipc_msg *msg;
1128 	int user, mtyp, hsz;
1129 
1130 	__skb_queue_head_init(&tmpq);
1131 	INIT_LIST_HEAD(&dports);
1132 
1133 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1134 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1135 		msg = buf_msg(skb);
1136 		user = msg_user(msg);
1137 		mtyp = msg_type(msg);
1138 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1139 			spin_lock_bh(&inputq->lock);
1140 			if (skb_peek(arrvq) == skb) {
1141 				__skb_dequeue(arrvq);
1142 				__skb_queue_tail(inputq, skb);
1143 			}
1144 			refcount_dec(&skb->users);
1145 			spin_unlock_bh(&inputq->lock);
1146 			continue;
1147 		}
1148 		hsz = skb_headroom(skb) + msg_hdr_sz(msg);
1149 		oport = msg_origport(msg);
1150 		onode = msg_orignode(msg);
1151 		if (onode == self)
1152 			scope = TIPC_NODE_SCOPE;
1153 
1154 		/* Create destination port list and message clones: */
1155 		if (!msg_in_group(msg)) {
1156 			lower = msg_namelower(msg);
1157 			upper = msg_nameupper(msg);
1158 		}
1159 		tipc_nametbl_mc_translate(net, msg_nametype(msg), lower, upper,
1160 					  scope, &dports);
1161 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1162 			_skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
1163 			if (_skb) {
1164 				msg_set_destport(buf_msg(_skb), portid);
1165 				__skb_queue_tail(&tmpq, _skb);
1166 				continue;
1167 			}
1168 			pr_warn("Failed to clone mcast rcv buffer\n");
1169 		}
1170 		/* Append to inputq if not already done by other thread */
1171 		spin_lock_bh(&inputq->lock);
1172 		if (skb_peek(arrvq) == skb) {
1173 			skb_queue_splice_tail_init(&tmpq, inputq);
1174 			kfree_skb(__skb_dequeue(arrvq));
1175 		}
1176 		spin_unlock_bh(&inputq->lock);
1177 		__skb_queue_purge(&tmpq);
1178 		kfree_skb(skb);
1179 	}
1180 	tipc_sk_rcv(net, inputq);
1181 }
1182 
1183 /**
1184  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1185  * @tsk: receiving socket
1186  * @skb: pointer to message buffer.
1187  */
1188 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1189 				   struct sk_buff_head *xmitq)
1190 {
1191 	struct tipc_msg *hdr = buf_msg(skb);
1192 	u32 onode = tsk_own_node(tsk);
1193 	struct sock *sk = &tsk->sk;
1194 	int mtyp = msg_type(hdr);
1195 	bool conn_cong;
1196 
1197 	/* Ignore if connection cannot be validated: */
1198 	if (!tsk_peer_msg(tsk, hdr))
1199 		goto exit;
1200 
1201 	if (unlikely(msg_errcode(hdr))) {
1202 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1203 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1204 				      tsk_peer_port(tsk));
1205 		sk->sk_state_change(sk);
1206 		goto exit;
1207 	}
1208 
1209 	tsk->probe_unacked = false;
1210 
1211 	if (mtyp == CONN_PROBE) {
1212 		msg_set_type(hdr, CONN_PROBE_REPLY);
1213 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1214 			__skb_queue_tail(xmitq, skb);
1215 		return;
1216 	} else if (mtyp == CONN_ACK) {
1217 		conn_cong = tsk_conn_cong(tsk);
1218 		tsk->snt_unacked -= msg_conn_ack(hdr);
1219 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1220 			tsk->snd_win = msg_adv_win(hdr);
1221 		if (conn_cong)
1222 			sk->sk_write_space(sk);
1223 	} else if (mtyp != CONN_PROBE_REPLY) {
1224 		pr_warn("Received unknown CONN_PROTO msg\n");
1225 	}
1226 exit:
1227 	kfree_skb(skb);
1228 }
1229 
1230 /**
1231  * tipc_sendmsg - send message in connectionless manner
1232  * @sock: socket structure
1233  * @m: message to send
1234  * @dsz: amount of user data to be sent
1235  *
1236  * Message must have an destination specified explicitly.
1237  * Used for SOCK_RDM and SOCK_DGRAM messages,
1238  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1239  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1240  *
1241  * Returns the number of bytes sent on success, or errno otherwise
1242  */
1243 static int tipc_sendmsg(struct socket *sock,
1244 			struct msghdr *m, size_t dsz)
1245 {
1246 	struct sock *sk = sock->sk;
1247 	int ret;
1248 
1249 	lock_sock(sk);
1250 	ret = __tipc_sendmsg(sock, m, dsz);
1251 	release_sock(sk);
1252 
1253 	return ret;
1254 }
1255 
1256 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1257 {
1258 	struct sock *sk = sock->sk;
1259 	struct net *net = sock_net(sk);
1260 	struct tipc_sock *tsk = tipc_sk(sk);
1261 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1262 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1263 	struct list_head *clinks = &tsk->cong_links;
1264 	bool syn = !tipc_sk_type_connectionless(sk);
1265 	struct tipc_group *grp = tsk->group;
1266 	struct tipc_msg *hdr = &tsk->phdr;
1267 	struct tipc_name_seq *seq;
1268 	struct sk_buff_head pkts;
1269 	u32 type, inst, domain;
1270 	u32 dnode, dport;
1271 	int mtu, rc;
1272 
1273 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1274 		return -EMSGSIZE;
1275 
1276 	if (likely(dest)) {
1277 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1278 			return -EINVAL;
1279 		if (unlikely(dest->family != AF_TIPC))
1280 			return -EINVAL;
1281 	}
1282 
1283 	if (grp) {
1284 		if (!dest)
1285 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1286 		if (dest->addrtype == TIPC_ADDR_NAME)
1287 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1288 		if (dest->addrtype == TIPC_ADDR_ID)
1289 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1290 		if (dest->addrtype == TIPC_ADDR_MCAST)
1291 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1292 		return -EINVAL;
1293 	}
1294 
1295 	if (unlikely(!dest)) {
1296 		dest = &tsk->peer;
1297 		if (!syn || dest->family != AF_TIPC)
1298 			return -EDESTADDRREQ;
1299 	}
1300 
1301 	if (unlikely(syn)) {
1302 		if (sk->sk_state == TIPC_LISTEN)
1303 			return -EPIPE;
1304 		if (sk->sk_state != TIPC_OPEN)
1305 			return -EISCONN;
1306 		if (tsk->published)
1307 			return -EOPNOTSUPP;
1308 		if (dest->addrtype == TIPC_ADDR_NAME) {
1309 			tsk->conn_type = dest->addr.name.name.type;
1310 			tsk->conn_instance = dest->addr.name.name.instance;
1311 		}
1312 	}
1313 
1314 	seq = &dest->addr.nameseq;
1315 	if (dest->addrtype == TIPC_ADDR_MCAST)
1316 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1317 
1318 	if (dest->addrtype == TIPC_ADDR_NAME) {
1319 		type = dest->addr.name.name.type;
1320 		inst = dest->addr.name.name.instance;
1321 		domain = dest->addr.name.domain;
1322 		dnode = domain;
1323 		msg_set_type(hdr, TIPC_NAMED_MSG);
1324 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1325 		msg_set_nametype(hdr, type);
1326 		msg_set_nameinst(hdr, inst);
1327 		msg_set_lookup_scope(hdr, tipc_addr_scope(domain));
1328 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1329 		msg_set_destnode(hdr, dnode);
1330 		msg_set_destport(hdr, dport);
1331 		if (unlikely(!dport && !dnode))
1332 			return -EHOSTUNREACH;
1333 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1334 		dnode = dest->addr.id.node;
1335 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1336 		msg_set_lookup_scope(hdr, 0);
1337 		msg_set_destnode(hdr, dnode);
1338 		msg_set_destport(hdr, dest->addr.id.ref);
1339 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1340 	}
1341 
1342 	/* Block or return if destination link is congested */
1343 	rc = tipc_wait_for_cond(sock, &timeout,
1344 				!tipc_dest_find(clinks, dnode, 0));
1345 	if (unlikely(rc))
1346 		return rc;
1347 
1348 	skb_queue_head_init(&pkts);
1349 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1350 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1351 	if (unlikely(rc != dlen))
1352 		return rc;
1353 
1354 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1355 	if (unlikely(rc == -ELINKCONG)) {
1356 		tipc_dest_push(clinks, dnode, 0);
1357 		tsk->cong_link_cnt++;
1358 		rc = 0;
1359 	}
1360 
1361 	if (unlikely(syn && !rc))
1362 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1363 
1364 	return rc ? rc : dlen;
1365 }
1366 
1367 /**
1368  * tipc_sendstream - send stream-oriented data
1369  * @sock: socket structure
1370  * @m: data to send
1371  * @dsz: total length of data to be transmitted
1372  *
1373  * Used for SOCK_STREAM data.
1374  *
1375  * Returns the number of bytes sent on success (or partial success),
1376  * or errno if no data sent
1377  */
1378 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1379 {
1380 	struct sock *sk = sock->sk;
1381 	int ret;
1382 
1383 	lock_sock(sk);
1384 	ret = __tipc_sendstream(sock, m, dsz);
1385 	release_sock(sk);
1386 
1387 	return ret;
1388 }
1389 
1390 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1391 {
1392 	struct sock *sk = sock->sk;
1393 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1394 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1395 	struct tipc_sock *tsk = tipc_sk(sk);
1396 	struct tipc_msg *hdr = &tsk->phdr;
1397 	struct net *net = sock_net(sk);
1398 	struct sk_buff_head pkts;
1399 	u32 dnode = tsk_peer_node(tsk);
1400 	int send, sent = 0;
1401 	int rc = 0;
1402 
1403 	skb_queue_head_init(&pkts);
1404 
1405 	if (unlikely(dlen > INT_MAX))
1406 		return -EMSGSIZE;
1407 
1408 	/* Handle implicit connection setup */
1409 	if (unlikely(dest)) {
1410 		rc = __tipc_sendmsg(sock, m, dlen);
1411 		if (dlen && (dlen == rc))
1412 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1413 		return rc;
1414 	}
1415 
1416 	do {
1417 		rc = tipc_wait_for_cond(sock, &timeout,
1418 					(!tsk->cong_link_cnt &&
1419 					 !tsk_conn_cong(tsk) &&
1420 					 tipc_sk_connected(sk)));
1421 		if (unlikely(rc))
1422 			break;
1423 
1424 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1425 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1426 		if (unlikely(rc != send))
1427 			break;
1428 
1429 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1430 		if (unlikely(rc == -ELINKCONG)) {
1431 			tsk->cong_link_cnt = 1;
1432 			rc = 0;
1433 		}
1434 		if (likely(!rc)) {
1435 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1436 			sent += send;
1437 		}
1438 	} while (sent < dlen && !rc);
1439 
1440 	return sent ? sent : rc;
1441 }
1442 
1443 /**
1444  * tipc_send_packet - send a connection-oriented message
1445  * @sock: socket structure
1446  * @m: message to send
1447  * @dsz: length of data to be transmitted
1448  *
1449  * Used for SOCK_SEQPACKET messages.
1450  *
1451  * Returns the number of bytes sent on success, or errno otherwise
1452  */
1453 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1454 {
1455 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1456 		return -EMSGSIZE;
1457 
1458 	return tipc_sendstream(sock, m, dsz);
1459 }
1460 
1461 /* tipc_sk_finish_conn - complete the setup of a connection
1462  */
1463 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1464 				u32 peer_node)
1465 {
1466 	struct sock *sk = &tsk->sk;
1467 	struct net *net = sock_net(sk);
1468 	struct tipc_msg *msg = &tsk->phdr;
1469 
1470 	msg_set_destnode(msg, peer_node);
1471 	msg_set_destport(msg, peer_port);
1472 	msg_set_type(msg, TIPC_CONN_MSG);
1473 	msg_set_lookup_scope(msg, 0);
1474 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1475 
1476 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
1477 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1478 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1479 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1480 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1481 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1482 		return;
1483 
1484 	/* Fall back to message based flow control */
1485 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1486 	tsk->snd_win = FLOWCTL_MSG_WIN;
1487 }
1488 
1489 /**
1490  * tipc_sk_set_orig_addr - capture sender's address for received message
1491  * @m: descriptor for message info
1492  * @hdr: received message header
1493  *
1494  * Note: Address is not captured if not requested by receiver.
1495  */
1496 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1497 {
1498 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1499 	struct tipc_msg *hdr = buf_msg(skb);
1500 
1501 	if (!srcaddr)
1502 		return;
1503 
1504 	srcaddr->sock.family = AF_TIPC;
1505 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1506 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1507 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1508 	srcaddr->sock.addr.name.domain = 0;
1509 	srcaddr->sock.scope = 0;
1510 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1511 
1512 	if (!msg_in_group(hdr))
1513 		return;
1514 
1515 	/* Group message users may also want to know sending member's id */
1516 	srcaddr->member.family = AF_TIPC;
1517 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1518 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1519 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1520 	srcaddr->member.addr.name.domain = 0;
1521 	m->msg_namelen = sizeof(*srcaddr);
1522 }
1523 
1524 /**
1525  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1526  * @m: descriptor for message info
1527  * @msg: received message header
1528  * @tsk: TIPC port associated with message
1529  *
1530  * Note: Ancillary data is not captured if not requested by receiver.
1531  *
1532  * Returns 0 if successful, otherwise errno
1533  */
1534 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1535 				 struct tipc_sock *tsk)
1536 {
1537 	u32 anc_data[3];
1538 	u32 err;
1539 	u32 dest_type;
1540 	int has_name;
1541 	int res;
1542 
1543 	if (likely(m->msg_controllen == 0))
1544 		return 0;
1545 
1546 	/* Optionally capture errored message object(s) */
1547 	err = msg ? msg_errcode(msg) : 0;
1548 	if (unlikely(err)) {
1549 		anc_data[0] = err;
1550 		anc_data[1] = msg_data_sz(msg);
1551 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1552 		if (res)
1553 			return res;
1554 		if (anc_data[1]) {
1555 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1556 				       msg_data(msg));
1557 			if (res)
1558 				return res;
1559 		}
1560 	}
1561 
1562 	/* Optionally capture message destination object */
1563 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1564 	switch (dest_type) {
1565 	case TIPC_NAMED_MSG:
1566 		has_name = 1;
1567 		anc_data[0] = msg_nametype(msg);
1568 		anc_data[1] = msg_namelower(msg);
1569 		anc_data[2] = msg_namelower(msg);
1570 		break;
1571 	case TIPC_MCAST_MSG:
1572 		has_name = 1;
1573 		anc_data[0] = msg_nametype(msg);
1574 		anc_data[1] = msg_namelower(msg);
1575 		anc_data[2] = msg_nameupper(msg);
1576 		break;
1577 	case TIPC_CONN_MSG:
1578 		has_name = (tsk->conn_type != 0);
1579 		anc_data[0] = tsk->conn_type;
1580 		anc_data[1] = tsk->conn_instance;
1581 		anc_data[2] = tsk->conn_instance;
1582 		break;
1583 	default:
1584 		has_name = 0;
1585 	}
1586 	if (has_name) {
1587 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1588 		if (res)
1589 			return res;
1590 	}
1591 
1592 	return 0;
1593 }
1594 
1595 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1596 {
1597 	struct sock *sk = &tsk->sk;
1598 	struct net *net = sock_net(sk);
1599 	struct sk_buff *skb = NULL;
1600 	struct tipc_msg *msg;
1601 	u32 peer_port = tsk_peer_port(tsk);
1602 	u32 dnode = tsk_peer_node(tsk);
1603 
1604 	if (!tipc_sk_connected(sk))
1605 		return;
1606 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1607 			      dnode, tsk_own_node(tsk), peer_port,
1608 			      tsk->portid, TIPC_OK);
1609 	if (!skb)
1610 		return;
1611 	msg = buf_msg(skb);
1612 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1613 	tsk->rcv_unacked = 0;
1614 
1615 	/* Adjust to and advertize the correct window limit */
1616 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1617 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1618 		msg_set_adv_win(msg, tsk->rcv_win);
1619 	}
1620 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1621 }
1622 
1623 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1624 {
1625 	struct sock *sk = sock->sk;
1626 	DEFINE_WAIT(wait);
1627 	long timeo = *timeop;
1628 	int err = sock_error(sk);
1629 
1630 	if (err)
1631 		return err;
1632 
1633 	for (;;) {
1634 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1635 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1636 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1637 				err = -ENOTCONN;
1638 				break;
1639 			}
1640 			release_sock(sk);
1641 			timeo = schedule_timeout(timeo);
1642 			lock_sock(sk);
1643 		}
1644 		err = 0;
1645 		if (!skb_queue_empty(&sk->sk_receive_queue))
1646 			break;
1647 		err = -EAGAIN;
1648 		if (!timeo)
1649 			break;
1650 		err = sock_intr_errno(timeo);
1651 		if (signal_pending(current))
1652 			break;
1653 
1654 		err = sock_error(sk);
1655 		if (err)
1656 			break;
1657 	}
1658 	finish_wait(sk_sleep(sk), &wait);
1659 	*timeop = timeo;
1660 	return err;
1661 }
1662 
1663 /**
1664  * tipc_recvmsg - receive packet-oriented message
1665  * @m: descriptor for message info
1666  * @buflen: length of user buffer area
1667  * @flags: receive flags
1668  *
1669  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1670  * If the complete message doesn't fit in user area, truncate it.
1671  *
1672  * Returns size of returned message data, errno otherwise
1673  */
1674 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1675 			size_t buflen,	int flags)
1676 {
1677 	struct sock *sk = sock->sk;
1678 	bool connected = !tipc_sk_type_connectionless(sk);
1679 	struct tipc_sock *tsk = tipc_sk(sk);
1680 	int rc, err, hlen, dlen, copy;
1681 	struct sk_buff_head xmitq;
1682 	struct tipc_msg *hdr;
1683 	struct sk_buff *skb;
1684 	bool grp_evt;
1685 	long timeout;
1686 
1687 	/* Catch invalid receive requests */
1688 	if (unlikely(!buflen))
1689 		return -EINVAL;
1690 
1691 	lock_sock(sk);
1692 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1693 		rc = -ENOTCONN;
1694 		goto exit;
1695 	}
1696 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1697 
1698 	/* Step rcv queue to first msg with data or error; wait if necessary */
1699 	do {
1700 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1701 		if (unlikely(rc))
1702 			goto exit;
1703 		skb = skb_peek(&sk->sk_receive_queue);
1704 		hdr = buf_msg(skb);
1705 		dlen = msg_data_sz(hdr);
1706 		hlen = msg_hdr_sz(hdr);
1707 		err = msg_errcode(hdr);
1708 		grp_evt = msg_is_grp_evt(hdr);
1709 		if (likely(dlen || err))
1710 			break;
1711 		tsk_advance_rx_queue(sk);
1712 	} while (1);
1713 
1714 	/* Collect msg meta data, including error code and rejected data */
1715 	tipc_sk_set_orig_addr(m, skb);
1716 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1717 	if (unlikely(rc))
1718 		goto exit;
1719 
1720 	/* Capture data if non-error msg, otherwise just set return value */
1721 	if (likely(!err)) {
1722 		copy = min_t(int, dlen, buflen);
1723 		if (unlikely(copy != dlen))
1724 			m->msg_flags |= MSG_TRUNC;
1725 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1726 	} else {
1727 		copy = 0;
1728 		rc = 0;
1729 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1730 			rc = -ECONNRESET;
1731 	}
1732 	if (unlikely(rc))
1733 		goto exit;
1734 
1735 	/* Mark message as group event if applicable */
1736 	if (unlikely(grp_evt)) {
1737 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1738 			m->msg_flags |= MSG_EOR;
1739 		m->msg_flags |= MSG_OOB;
1740 		copy = 0;
1741 	}
1742 
1743 	/* Caption of data or error code/rejected data was successful */
1744 	if (unlikely(flags & MSG_PEEK))
1745 		goto exit;
1746 
1747 	/* Send group flow control advertisement when applicable */
1748 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1749 		skb_queue_head_init(&xmitq);
1750 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1751 					  msg_orignode(hdr), msg_origport(hdr),
1752 					  &xmitq);
1753 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1754 	}
1755 
1756 	tsk_advance_rx_queue(sk);
1757 
1758 	if (likely(!connected))
1759 		goto exit;
1760 
1761 	/* Send connection flow control advertisement when applicable */
1762 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1763 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1764 		tipc_sk_send_ack(tsk);
1765 exit:
1766 	release_sock(sk);
1767 	return rc ? rc : copy;
1768 }
1769 
1770 /**
1771  * tipc_recvstream - receive stream-oriented data
1772  * @m: descriptor for message info
1773  * @buflen: total size of user buffer area
1774  * @flags: receive flags
1775  *
1776  * Used for SOCK_STREAM messages only.  If not enough data is available
1777  * will optionally wait for more; never truncates data.
1778  *
1779  * Returns size of returned message data, errno otherwise
1780  */
1781 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1782 			   size_t buflen, int flags)
1783 {
1784 	struct sock *sk = sock->sk;
1785 	struct tipc_sock *tsk = tipc_sk(sk);
1786 	struct sk_buff *skb;
1787 	struct tipc_msg *hdr;
1788 	struct tipc_skb_cb *skb_cb;
1789 	bool peek = flags & MSG_PEEK;
1790 	int offset, required, copy, copied = 0;
1791 	int hlen, dlen, err, rc;
1792 	long timeout;
1793 
1794 	/* Catch invalid receive attempts */
1795 	if (unlikely(!buflen))
1796 		return -EINVAL;
1797 
1798 	lock_sock(sk);
1799 
1800 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1801 		rc = -ENOTCONN;
1802 		goto exit;
1803 	}
1804 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1805 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1806 
1807 	do {
1808 		/* Look at first msg in receive queue; wait if necessary */
1809 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1810 		if (unlikely(rc))
1811 			break;
1812 		skb = skb_peek(&sk->sk_receive_queue);
1813 		skb_cb = TIPC_SKB_CB(skb);
1814 		hdr = buf_msg(skb);
1815 		dlen = msg_data_sz(hdr);
1816 		hlen = msg_hdr_sz(hdr);
1817 		err = msg_errcode(hdr);
1818 
1819 		/* Discard any empty non-errored (SYN-) message */
1820 		if (unlikely(!dlen && !err)) {
1821 			tsk_advance_rx_queue(sk);
1822 			continue;
1823 		}
1824 
1825 		/* Collect msg meta data, incl. error code and rejected data */
1826 		if (!copied) {
1827 			tipc_sk_set_orig_addr(m, skb);
1828 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1829 			if (rc)
1830 				break;
1831 		}
1832 
1833 		/* Copy data if msg ok, otherwise return error/partial data */
1834 		if (likely(!err)) {
1835 			offset = skb_cb->bytes_read;
1836 			copy = min_t(int, dlen - offset, buflen - copied);
1837 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1838 			if (unlikely(rc))
1839 				break;
1840 			copied += copy;
1841 			offset += copy;
1842 			if (unlikely(offset < dlen)) {
1843 				if (!peek)
1844 					skb_cb->bytes_read = offset;
1845 				break;
1846 			}
1847 		} else {
1848 			rc = 0;
1849 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1850 				rc = -ECONNRESET;
1851 			if (copied || rc)
1852 				break;
1853 		}
1854 
1855 		if (unlikely(peek))
1856 			break;
1857 
1858 		tsk_advance_rx_queue(sk);
1859 
1860 		/* Send connection flow control advertisement when applicable */
1861 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1862 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1863 			tipc_sk_send_ack(tsk);
1864 
1865 		/* Exit if all requested data or FIN/error received */
1866 		if (copied == buflen || err)
1867 			break;
1868 
1869 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1870 exit:
1871 	release_sock(sk);
1872 	return copied ? copied : rc;
1873 }
1874 
1875 /**
1876  * tipc_write_space - wake up thread if port congestion is released
1877  * @sk: socket
1878  */
1879 static void tipc_write_space(struct sock *sk)
1880 {
1881 	struct socket_wq *wq;
1882 
1883 	rcu_read_lock();
1884 	wq = rcu_dereference(sk->sk_wq);
1885 	if (skwq_has_sleeper(wq))
1886 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1887 						POLLWRNORM | POLLWRBAND);
1888 	rcu_read_unlock();
1889 }
1890 
1891 /**
1892  * tipc_data_ready - wake up threads to indicate messages have been received
1893  * @sk: socket
1894  * @len: the length of messages
1895  */
1896 static void tipc_data_ready(struct sock *sk)
1897 {
1898 	struct socket_wq *wq;
1899 
1900 	rcu_read_lock();
1901 	wq = rcu_dereference(sk->sk_wq);
1902 	if (skwq_has_sleeper(wq))
1903 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1904 						POLLRDNORM | POLLRDBAND);
1905 	rcu_read_unlock();
1906 }
1907 
1908 static void tipc_sock_destruct(struct sock *sk)
1909 {
1910 	__skb_queue_purge(&sk->sk_receive_queue);
1911 }
1912 
1913 static void tipc_sk_proto_rcv(struct sock *sk,
1914 			      struct sk_buff_head *inputq,
1915 			      struct sk_buff_head *xmitq)
1916 {
1917 	struct sk_buff *skb = __skb_dequeue(inputq);
1918 	struct tipc_sock *tsk = tipc_sk(sk);
1919 	struct tipc_msg *hdr = buf_msg(skb);
1920 	struct tipc_group *grp = tsk->group;
1921 	bool wakeup = false;
1922 
1923 	switch (msg_user(hdr)) {
1924 	case CONN_MANAGER:
1925 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1926 		return;
1927 	case SOCK_WAKEUP:
1928 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1929 		tsk->cong_link_cnt--;
1930 		wakeup = true;
1931 		break;
1932 	case GROUP_PROTOCOL:
1933 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1934 		break;
1935 	case TOP_SRV:
1936 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1937 				      skb, inputq, xmitq);
1938 		skb = NULL;
1939 		break;
1940 	default:
1941 		break;
1942 	}
1943 
1944 	if (wakeup)
1945 		sk->sk_write_space(sk);
1946 
1947 	kfree_skb(skb);
1948 }
1949 
1950 /**
1951  * tipc_filter_connect - Handle incoming message for a connection-based socket
1952  * @tsk: TIPC socket
1953  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1954  *
1955  * Returns true if everything ok, false otherwise
1956  */
1957 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1958 {
1959 	struct sock *sk = &tsk->sk;
1960 	struct net *net = sock_net(sk);
1961 	struct tipc_msg *hdr = buf_msg(skb);
1962 	u32 pport = msg_origport(hdr);
1963 	u32 pnode = msg_orignode(hdr);
1964 
1965 	if (unlikely(msg_mcast(hdr)))
1966 		return false;
1967 
1968 	switch (sk->sk_state) {
1969 	case TIPC_CONNECTING:
1970 		/* Accept only ACK or NACK message */
1971 		if (unlikely(!msg_connected(hdr))) {
1972 			if (pport != tsk_peer_port(tsk) ||
1973 			    pnode != tsk_peer_node(tsk))
1974 				return false;
1975 
1976 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1977 			sk->sk_err = ECONNREFUSED;
1978 			sk->sk_state_change(sk);
1979 			return true;
1980 		}
1981 
1982 		if (unlikely(msg_errcode(hdr))) {
1983 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1984 			sk->sk_err = ECONNREFUSED;
1985 			sk->sk_state_change(sk);
1986 			return true;
1987 		}
1988 
1989 		if (unlikely(!msg_isdata(hdr))) {
1990 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1991 			sk->sk_err = EINVAL;
1992 			sk->sk_state_change(sk);
1993 			return true;
1994 		}
1995 
1996 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1997 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
1998 
1999 		/* If 'ACK+' message, add to socket receive queue */
2000 		if (msg_data_sz(hdr))
2001 			return true;
2002 
2003 		/* If empty 'ACK-' message, wake up sleeping connect() */
2004 		sk->sk_data_ready(sk);
2005 
2006 		/* 'ACK-' message is neither accepted nor rejected: */
2007 		msg_set_dest_droppable(hdr, 1);
2008 		return false;
2009 
2010 	case TIPC_OPEN:
2011 	case TIPC_DISCONNECTING:
2012 		break;
2013 	case TIPC_LISTEN:
2014 		/* Accept only SYN message */
2015 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2016 			return true;
2017 		break;
2018 	case TIPC_ESTABLISHED:
2019 		/* Accept only connection-based messages sent by peer */
2020 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2021 			return false;
2022 
2023 		if (unlikely(msg_errcode(hdr))) {
2024 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2025 			/* Let timer expire on it's own */
2026 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2027 					      tsk->portid);
2028 			sk->sk_state_change(sk);
2029 		}
2030 		return true;
2031 	default:
2032 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2033 	}
2034 
2035 	return false;
2036 }
2037 
2038 /**
2039  * rcvbuf_limit - get proper overload limit of socket receive queue
2040  * @sk: socket
2041  * @skb: message
2042  *
2043  * For connection oriented messages, irrespective of importance,
2044  * default queue limit is 2 MB.
2045  *
2046  * For connectionless messages, queue limits are based on message
2047  * importance as follows:
2048  *
2049  * TIPC_LOW_IMPORTANCE       (2 MB)
2050  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2051  * TIPC_HIGH_IMPORTANCE      (8 MB)
2052  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2053  *
2054  * Returns overload limit according to corresponding message importance
2055  */
2056 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2057 {
2058 	struct tipc_sock *tsk = tipc_sk(sk);
2059 	struct tipc_msg *hdr = buf_msg(skb);
2060 
2061 	if (unlikely(msg_in_group(hdr)))
2062 		return sk->sk_rcvbuf;
2063 
2064 	if (unlikely(!msg_connected(hdr)))
2065 		return sk->sk_rcvbuf << msg_importance(hdr);
2066 
2067 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2068 		return sk->sk_rcvbuf;
2069 
2070 	return FLOWCTL_MSG_LIM;
2071 }
2072 
2073 /**
2074  * tipc_sk_filter_rcv - validate incoming message
2075  * @sk: socket
2076  * @skb: pointer to message.
2077  *
2078  * Enqueues message on receive queue if acceptable; optionally handles
2079  * disconnect indication for a connected socket.
2080  *
2081  * Called with socket lock already taken
2082  *
2083  */
2084 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2085 			       struct sk_buff_head *xmitq)
2086 {
2087 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2088 	struct tipc_sock *tsk = tipc_sk(sk);
2089 	struct tipc_group *grp = tsk->group;
2090 	struct tipc_msg *hdr = buf_msg(skb);
2091 	struct net *net = sock_net(sk);
2092 	struct sk_buff_head inputq;
2093 	int limit, err = TIPC_OK;
2094 
2095 	TIPC_SKB_CB(skb)->bytes_read = 0;
2096 	__skb_queue_head_init(&inputq);
2097 	__skb_queue_tail(&inputq, skb);
2098 
2099 	if (unlikely(!msg_isdata(hdr)))
2100 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2101 
2102 	if (unlikely(grp))
2103 		tipc_group_filter_msg(grp, &inputq, xmitq);
2104 
2105 	/* Validate and add to receive buffer if there is space */
2106 	while ((skb = __skb_dequeue(&inputq))) {
2107 		hdr = buf_msg(skb);
2108 		limit = rcvbuf_limit(sk, skb);
2109 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2110 		    (!sk_conn && msg_connected(hdr)) ||
2111 		    (!grp && msg_in_group(hdr)))
2112 			err = TIPC_ERR_NO_PORT;
2113 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit)
2114 			err = TIPC_ERR_OVERLOAD;
2115 
2116 		if (unlikely(err)) {
2117 			tipc_skb_reject(net, err, skb, xmitq);
2118 			err = TIPC_OK;
2119 			continue;
2120 		}
2121 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2122 		skb_set_owner_r(skb, sk);
2123 		sk->sk_data_ready(sk);
2124 	}
2125 }
2126 
2127 /**
2128  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2129  * @sk: socket
2130  * @skb: message
2131  *
2132  * Caller must hold socket lock
2133  */
2134 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2135 {
2136 	unsigned int before = sk_rmem_alloc_get(sk);
2137 	struct sk_buff_head xmitq;
2138 	unsigned int added;
2139 
2140 	__skb_queue_head_init(&xmitq);
2141 
2142 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2143 	added = sk_rmem_alloc_get(sk) - before;
2144 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2145 
2146 	/* Send pending response/rejected messages, if any */
2147 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2148 	return 0;
2149 }
2150 
2151 /**
2152  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2153  *                   inputq and try adding them to socket or backlog queue
2154  * @inputq: list of incoming buffers with potentially different destinations
2155  * @sk: socket where the buffers should be enqueued
2156  * @dport: port number for the socket
2157  *
2158  * Caller must hold socket lock
2159  */
2160 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2161 			    u32 dport, struct sk_buff_head *xmitq)
2162 {
2163 	unsigned long time_limit = jiffies + 2;
2164 	struct sk_buff *skb;
2165 	unsigned int lim;
2166 	atomic_t *dcnt;
2167 	u32 onode;
2168 
2169 	while (skb_queue_len(inputq)) {
2170 		if (unlikely(time_after_eq(jiffies, time_limit)))
2171 			return;
2172 
2173 		skb = tipc_skb_dequeue(inputq, dport);
2174 		if (unlikely(!skb))
2175 			return;
2176 
2177 		/* Add message directly to receive queue if possible */
2178 		if (!sock_owned_by_user(sk)) {
2179 			tipc_sk_filter_rcv(sk, skb, xmitq);
2180 			continue;
2181 		}
2182 
2183 		/* Try backlog, compensating for double-counted bytes */
2184 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2185 		if (!sk->sk_backlog.len)
2186 			atomic_set(dcnt, 0);
2187 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2188 		if (likely(!sk_add_backlog(sk, skb, lim)))
2189 			continue;
2190 
2191 		/* Overload => reject message back to sender */
2192 		onode = tipc_own_addr(sock_net(sk));
2193 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2194 			__skb_queue_tail(xmitq, skb);
2195 		break;
2196 	}
2197 }
2198 
2199 /**
2200  * tipc_sk_rcv - handle a chain of incoming buffers
2201  * @inputq: buffer list containing the buffers
2202  * Consumes all buffers in list until inputq is empty
2203  * Note: may be called in multiple threads referring to the same queue
2204  */
2205 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2206 {
2207 	struct sk_buff_head xmitq;
2208 	u32 dnode, dport = 0;
2209 	int err;
2210 	struct tipc_sock *tsk;
2211 	struct sock *sk;
2212 	struct sk_buff *skb;
2213 
2214 	__skb_queue_head_init(&xmitq);
2215 	while (skb_queue_len(inputq)) {
2216 		dport = tipc_skb_peek_port(inputq, dport);
2217 		tsk = tipc_sk_lookup(net, dport);
2218 
2219 		if (likely(tsk)) {
2220 			sk = &tsk->sk;
2221 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2222 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2223 				spin_unlock_bh(&sk->sk_lock.slock);
2224 			}
2225 			/* Send pending response/rejected messages, if any */
2226 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2227 			sock_put(sk);
2228 			continue;
2229 		}
2230 		/* No destination socket => dequeue skb if still there */
2231 		skb = tipc_skb_dequeue(inputq, dport);
2232 		if (!skb)
2233 			return;
2234 
2235 		/* Try secondary lookup if unresolved named message */
2236 		err = TIPC_ERR_NO_PORT;
2237 		if (tipc_msg_lookup_dest(net, skb, &err))
2238 			goto xmit;
2239 
2240 		/* Prepare for message rejection */
2241 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2242 			continue;
2243 xmit:
2244 		dnode = msg_destnode(buf_msg(skb));
2245 		tipc_node_xmit_skb(net, skb, dnode, dport);
2246 	}
2247 }
2248 
2249 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2250 {
2251 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2252 	struct sock *sk = sock->sk;
2253 	int done;
2254 
2255 	do {
2256 		int err = sock_error(sk);
2257 		if (err)
2258 			return err;
2259 		if (!*timeo_p)
2260 			return -ETIMEDOUT;
2261 		if (signal_pending(current))
2262 			return sock_intr_errno(*timeo_p);
2263 
2264 		add_wait_queue(sk_sleep(sk), &wait);
2265 		done = sk_wait_event(sk, timeo_p,
2266 				     sk->sk_state != TIPC_CONNECTING, &wait);
2267 		remove_wait_queue(sk_sleep(sk), &wait);
2268 	} while (!done);
2269 	return 0;
2270 }
2271 
2272 /**
2273  * tipc_connect - establish a connection to another TIPC port
2274  * @sock: socket structure
2275  * @dest: socket address for destination port
2276  * @destlen: size of socket address data structure
2277  * @flags: file-related flags associated with socket
2278  *
2279  * Returns 0 on success, errno otherwise
2280  */
2281 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2282 			int destlen, int flags)
2283 {
2284 	struct sock *sk = sock->sk;
2285 	struct tipc_sock *tsk = tipc_sk(sk);
2286 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2287 	struct msghdr m = {NULL,};
2288 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2289 	int previous;
2290 	int res = 0;
2291 
2292 	if (destlen != sizeof(struct sockaddr_tipc))
2293 		return -EINVAL;
2294 
2295 	lock_sock(sk);
2296 
2297 	if (tsk->group) {
2298 		res = -EINVAL;
2299 		goto exit;
2300 	}
2301 
2302 	if (dst->family == AF_UNSPEC) {
2303 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2304 		if (!tipc_sk_type_connectionless(sk))
2305 			res = -EINVAL;
2306 		goto exit;
2307 	} else if (dst->family != AF_TIPC) {
2308 		res = -EINVAL;
2309 	}
2310 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2311 		res = -EINVAL;
2312 	if (res)
2313 		goto exit;
2314 
2315 	/* DGRAM/RDM connect(), just save the destaddr */
2316 	if (tipc_sk_type_connectionless(sk)) {
2317 		memcpy(&tsk->peer, dest, destlen);
2318 		goto exit;
2319 	}
2320 
2321 	previous = sk->sk_state;
2322 
2323 	switch (sk->sk_state) {
2324 	case TIPC_OPEN:
2325 		/* Send a 'SYN-' to destination */
2326 		m.msg_name = dest;
2327 		m.msg_namelen = destlen;
2328 
2329 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2330 		 * indicate send_msg() is never blocked.
2331 		 */
2332 		if (!timeout)
2333 			m.msg_flags = MSG_DONTWAIT;
2334 
2335 		res = __tipc_sendmsg(sock, &m, 0);
2336 		if ((res < 0) && (res != -EWOULDBLOCK))
2337 			goto exit;
2338 
2339 		/* Just entered TIPC_CONNECTING state; the only
2340 		 * difference is that return value in non-blocking
2341 		 * case is EINPROGRESS, rather than EALREADY.
2342 		 */
2343 		res = -EINPROGRESS;
2344 		/* fall thru' */
2345 	case TIPC_CONNECTING:
2346 		if (!timeout) {
2347 			if (previous == TIPC_CONNECTING)
2348 				res = -EALREADY;
2349 			goto exit;
2350 		}
2351 		timeout = msecs_to_jiffies(timeout);
2352 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2353 		res = tipc_wait_for_connect(sock, &timeout);
2354 		break;
2355 	case TIPC_ESTABLISHED:
2356 		res = -EISCONN;
2357 		break;
2358 	default:
2359 		res = -EINVAL;
2360 	}
2361 
2362 exit:
2363 	release_sock(sk);
2364 	return res;
2365 }
2366 
2367 /**
2368  * tipc_listen - allow socket to listen for incoming connections
2369  * @sock: socket structure
2370  * @len: (unused)
2371  *
2372  * Returns 0 on success, errno otherwise
2373  */
2374 static int tipc_listen(struct socket *sock, int len)
2375 {
2376 	struct sock *sk = sock->sk;
2377 	int res;
2378 
2379 	lock_sock(sk);
2380 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2381 	release_sock(sk);
2382 
2383 	return res;
2384 }
2385 
2386 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2387 {
2388 	struct sock *sk = sock->sk;
2389 	DEFINE_WAIT(wait);
2390 	int err;
2391 
2392 	/* True wake-one mechanism for incoming connections: only
2393 	 * one process gets woken up, not the 'whole herd'.
2394 	 * Since we do not 'race & poll' for established sockets
2395 	 * anymore, the common case will execute the loop only once.
2396 	*/
2397 	for (;;) {
2398 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2399 					  TASK_INTERRUPTIBLE);
2400 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2401 			release_sock(sk);
2402 			timeo = schedule_timeout(timeo);
2403 			lock_sock(sk);
2404 		}
2405 		err = 0;
2406 		if (!skb_queue_empty(&sk->sk_receive_queue))
2407 			break;
2408 		err = -EAGAIN;
2409 		if (!timeo)
2410 			break;
2411 		err = sock_intr_errno(timeo);
2412 		if (signal_pending(current))
2413 			break;
2414 	}
2415 	finish_wait(sk_sleep(sk), &wait);
2416 	return err;
2417 }
2418 
2419 /**
2420  * tipc_accept - wait for connection request
2421  * @sock: listening socket
2422  * @newsock: new socket that is to be connected
2423  * @flags: file-related flags associated with socket
2424  *
2425  * Returns 0 on success, errno otherwise
2426  */
2427 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2428 		       bool kern)
2429 {
2430 	struct sock *new_sk, *sk = sock->sk;
2431 	struct sk_buff *buf;
2432 	struct tipc_sock *new_tsock;
2433 	struct tipc_msg *msg;
2434 	long timeo;
2435 	int res;
2436 
2437 	lock_sock(sk);
2438 
2439 	if (sk->sk_state != TIPC_LISTEN) {
2440 		res = -EINVAL;
2441 		goto exit;
2442 	}
2443 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2444 	res = tipc_wait_for_accept(sock, timeo);
2445 	if (res)
2446 		goto exit;
2447 
2448 	buf = skb_peek(&sk->sk_receive_queue);
2449 
2450 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2451 	if (res)
2452 		goto exit;
2453 	security_sk_clone(sock->sk, new_sock->sk);
2454 
2455 	new_sk = new_sock->sk;
2456 	new_tsock = tipc_sk(new_sk);
2457 	msg = buf_msg(buf);
2458 
2459 	/* we lock on new_sk; but lockdep sees the lock on sk */
2460 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2461 
2462 	/*
2463 	 * Reject any stray messages received by new socket
2464 	 * before the socket lock was taken (very, very unlikely)
2465 	 */
2466 	tsk_rej_rx_queue(new_sk);
2467 
2468 	/* Connect new socket to it's peer */
2469 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2470 
2471 	tsk_set_importance(new_tsock, msg_importance(msg));
2472 	if (msg_named(msg)) {
2473 		new_tsock->conn_type = msg_nametype(msg);
2474 		new_tsock->conn_instance = msg_nameinst(msg);
2475 	}
2476 
2477 	/*
2478 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2479 	 * Respond to 'SYN+' by queuing it on new socket.
2480 	 */
2481 	if (!msg_data_sz(msg)) {
2482 		struct msghdr m = {NULL,};
2483 
2484 		tsk_advance_rx_queue(sk);
2485 		__tipc_sendstream(new_sock, &m, 0);
2486 	} else {
2487 		__skb_dequeue(&sk->sk_receive_queue);
2488 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2489 		skb_set_owner_r(buf, new_sk);
2490 	}
2491 	release_sock(new_sk);
2492 exit:
2493 	release_sock(sk);
2494 	return res;
2495 }
2496 
2497 /**
2498  * tipc_shutdown - shutdown socket connection
2499  * @sock: socket structure
2500  * @how: direction to close (must be SHUT_RDWR)
2501  *
2502  * Terminates connection (if necessary), then purges socket's receive queue.
2503  *
2504  * Returns 0 on success, errno otherwise
2505  */
2506 static int tipc_shutdown(struct socket *sock, int how)
2507 {
2508 	struct sock *sk = sock->sk;
2509 	int res;
2510 
2511 	if (how != SHUT_RDWR)
2512 		return -EINVAL;
2513 
2514 	lock_sock(sk);
2515 
2516 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2517 	sk->sk_shutdown = SEND_SHUTDOWN;
2518 
2519 	if (sk->sk_state == TIPC_DISCONNECTING) {
2520 		/* Discard any unreceived messages */
2521 		__skb_queue_purge(&sk->sk_receive_queue);
2522 
2523 		/* Wake up anyone sleeping in poll */
2524 		sk->sk_state_change(sk);
2525 		res = 0;
2526 	} else {
2527 		res = -ENOTCONN;
2528 	}
2529 
2530 	release_sock(sk);
2531 	return res;
2532 }
2533 
2534 static void tipc_sk_timeout(unsigned long data)
2535 {
2536 	struct tipc_sock *tsk = (struct tipc_sock *)data;
2537 	struct sock *sk = &tsk->sk;
2538 	struct sk_buff *skb = NULL;
2539 	u32 peer_port, peer_node;
2540 	u32 own_node = tsk_own_node(tsk);
2541 
2542 	bh_lock_sock(sk);
2543 	if (!tipc_sk_connected(sk)) {
2544 		bh_unlock_sock(sk);
2545 		goto exit;
2546 	}
2547 	peer_port = tsk_peer_port(tsk);
2548 	peer_node = tsk_peer_node(tsk);
2549 
2550 	if (tsk->probe_unacked) {
2551 		if (!sock_owned_by_user(sk)) {
2552 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2553 			tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2554 					      tsk_peer_port(tsk));
2555 			sk->sk_state_change(sk);
2556 		} else {
2557 			/* Try again later */
2558 			sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2559 		}
2560 
2561 		bh_unlock_sock(sk);
2562 		goto exit;
2563 	}
2564 
2565 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2566 			      INT_H_SIZE, 0, peer_node, own_node,
2567 			      peer_port, tsk->portid, TIPC_OK);
2568 	tsk->probe_unacked = true;
2569 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
2570 	bh_unlock_sock(sk);
2571 	if (skb)
2572 		tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2573 exit:
2574 	sock_put(sk);
2575 }
2576 
2577 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2578 			   struct tipc_name_seq const *seq)
2579 {
2580 	struct sock *sk = &tsk->sk;
2581 	struct net *net = sock_net(sk);
2582 	struct publication *publ;
2583 	u32 key;
2584 
2585 	if (tipc_sk_connected(sk))
2586 		return -EINVAL;
2587 	key = tsk->portid + tsk->pub_count + 1;
2588 	if (key == tsk->portid)
2589 		return -EADDRINUSE;
2590 
2591 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2592 				    scope, tsk->portid, key);
2593 	if (unlikely(!publ))
2594 		return -EINVAL;
2595 
2596 	list_add(&publ->pport_list, &tsk->publications);
2597 	tsk->pub_count++;
2598 	tsk->published = 1;
2599 	return 0;
2600 }
2601 
2602 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2603 			    struct tipc_name_seq const *seq)
2604 {
2605 	struct net *net = sock_net(&tsk->sk);
2606 	struct publication *publ;
2607 	struct publication *safe;
2608 	int rc = -EINVAL;
2609 
2610 	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2611 		if (seq) {
2612 			if (publ->scope != scope)
2613 				continue;
2614 			if (publ->type != seq->type)
2615 				continue;
2616 			if (publ->lower != seq->lower)
2617 				continue;
2618 			if (publ->upper != seq->upper)
2619 				break;
2620 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2621 					      publ->ref, publ->key);
2622 			rc = 0;
2623 			break;
2624 		}
2625 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2626 				      publ->ref, publ->key);
2627 		rc = 0;
2628 	}
2629 	if (list_empty(&tsk->publications))
2630 		tsk->published = 0;
2631 	return rc;
2632 }
2633 
2634 /* tipc_sk_reinit: set non-zero address in all existing sockets
2635  *                 when we go from standalone to network mode.
2636  */
2637 void tipc_sk_reinit(struct net *net)
2638 {
2639 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2640 	struct rhashtable_iter iter;
2641 	struct tipc_sock *tsk;
2642 	struct tipc_msg *msg;
2643 
2644 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2645 
2646 	do {
2647 		tsk = ERR_PTR(rhashtable_walk_start(&iter));
2648 		if (IS_ERR(tsk))
2649 			goto walk_stop;
2650 
2651 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2652 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2653 			msg = &tsk->phdr;
2654 			msg_set_prevnode(msg, tn->own_addr);
2655 			msg_set_orignode(msg, tn->own_addr);
2656 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2657 		}
2658 walk_stop:
2659 		rhashtable_walk_stop(&iter);
2660 	} while (tsk == ERR_PTR(-EAGAIN));
2661 }
2662 
2663 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2664 {
2665 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2666 	struct tipc_sock *tsk;
2667 
2668 	rcu_read_lock();
2669 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2670 	if (tsk)
2671 		sock_hold(&tsk->sk);
2672 	rcu_read_unlock();
2673 
2674 	return tsk;
2675 }
2676 
2677 static int tipc_sk_insert(struct tipc_sock *tsk)
2678 {
2679 	struct sock *sk = &tsk->sk;
2680 	struct net *net = sock_net(sk);
2681 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2682 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2683 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2684 
2685 	while (remaining--) {
2686 		portid++;
2687 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2688 			portid = TIPC_MIN_PORT;
2689 		tsk->portid = portid;
2690 		sock_hold(&tsk->sk);
2691 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2692 						   tsk_rht_params))
2693 			return 0;
2694 		sock_put(&tsk->sk);
2695 	}
2696 
2697 	return -1;
2698 }
2699 
2700 static void tipc_sk_remove(struct tipc_sock *tsk)
2701 {
2702 	struct sock *sk = &tsk->sk;
2703 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2704 
2705 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2706 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2707 		__sock_put(sk);
2708 	}
2709 }
2710 
2711 static const struct rhashtable_params tsk_rht_params = {
2712 	.nelem_hint = 192,
2713 	.head_offset = offsetof(struct tipc_sock, node),
2714 	.key_offset = offsetof(struct tipc_sock, portid),
2715 	.key_len = sizeof(u32), /* portid */
2716 	.max_size = 1048576,
2717 	.min_size = 256,
2718 	.automatic_shrinking = true,
2719 };
2720 
2721 int tipc_sk_rht_init(struct net *net)
2722 {
2723 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2724 
2725 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2726 }
2727 
2728 void tipc_sk_rht_destroy(struct net *net)
2729 {
2730 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2731 
2732 	/* Wait for socket readers to complete */
2733 	synchronize_net();
2734 
2735 	rhashtable_destroy(&tn->sk_rht);
2736 }
2737 
2738 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2739 {
2740 	struct net *net = sock_net(&tsk->sk);
2741 	u32 domain = addr_domain(net, mreq->scope);
2742 	struct tipc_group *grp = tsk->group;
2743 	struct tipc_msg *hdr = &tsk->phdr;
2744 	struct tipc_name_seq seq;
2745 	int rc;
2746 
2747 	if (mreq->type < TIPC_RESERVED_TYPES)
2748 		return -EACCES;
2749 	if (grp)
2750 		return -EACCES;
2751 	grp = tipc_group_create(net, tsk->portid, mreq);
2752 	if (!grp)
2753 		return -ENOMEM;
2754 	tsk->group = grp;
2755 	msg_set_lookup_scope(hdr, mreq->scope);
2756 	msg_set_nametype(hdr, mreq->type);
2757 	msg_set_dest_droppable(hdr, true);
2758 	seq.type = mreq->type;
2759 	seq.lower = mreq->instance;
2760 	seq.upper = seq.lower;
2761 	tipc_nametbl_build_group(net, grp, mreq->type, domain);
2762 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2763 	if (rc)
2764 		tipc_group_delete(net, grp);
2765 
2766 	/* Eliminate any risk that a broadcast overtakes the sent JOIN */
2767 	tsk->mc_method.rcast = true;
2768 	tsk->mc_method.mandatory = true;
2769 	return rc;
2770 }
2771 
2772 static int tipc_sk_leave(struct tipc_sock *tsk)
2773 {
2774 	struct net *net = sock_net(&tsk->sk);
2775 	struct tipc_group *grp = tsk->group;
2776 	struct tipc_name_seq seq;
2777 	int scope;
2778 
2779 	if (!grp)
2780 		return -EINVAL;
2781 	tipc_group_self(grp, &seq, &scope);
2782 	tipc_group_delete(net, grp);
2783 	tsk->group = NULL;
2784 	tipc_sk_withdraw(tsk, scope, &seq);
2785 	return 0;
2786 }
2787 
2788 /**
2789  * tipc_setsockopt - set socket option
2790  * @sock: socket structure
2791  * @lvl: option level
2792  * @opt: option identifier
2793  * @ov: pointer to new option value
2794  * @ol: length of option value
2795  *
2796  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2797  * (to ease compatibility).
2798  *
2799  * Returns 0 on success, errno otherwise
2800  */
2801 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2802 			   char __user *ov, unsigned int ol)
2803 {
2804 	struct sock *sk = sock->sk;
2805 	struct tipc_sock *tsk = tipc_sk(sk);
2806 	struct tipc_group_req mreq;
2807 	u32 value = 0;
2808 	int res = 0;
2809 
2810 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2811 		return 0;
2812 	if (lvl != SOL_TIPC)
2813 		return -ENOPROTOOPT;
2814 
2815 	switch (opt) {
2816 	case TIPC_IMPORTANCE:
2817 	case TIPC_SRC_DROPPABLE:
2818 	case TIPC_DEST_DROPPABLE:
2819 	case TIPC_CONN_TIMEOUT:
2820 		if (ol < sizeof(value))
2821 			return -EINVAL;
2822 		if (get_user(value, (u32 __user *)ov))
2823 			return -EFAULT;
2824 		break;
2825 	case TIPC_GROUP_JOIN:
2826 		if (ol < sizeof(mreq))
2827 			return -EINVAL;
2828 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2829 			return -EFAULT;
2830 		break;
2831 	default:
2832 		if (ov || ol)
2833 			return -EINVAL;
2834 	}
2835 
2836 	lock_sock(sk);
2837 
2838 	switch (opt) {
2839 	case TIPC_IMPORTANCE:
2840 		res = tsk_set_importance(tsk, value);
2841 		break;
2842 	case TIPC_SRC_DROPPABLE:
2843 		if (sock->type != SOCK_STREAM)
2844 			tsk_set_unreliable(tsk, value);
2845 		else
2846 			res = -ENOPROTOOPT;
2847 		break;
2848 	case TIPC_DEST_DROPPABLE:
2849 		tsk_set_unreturnable(tsk, value);
2850 		break;
2851 	case TIPC_CONN_TIMEOUT:
2852 		tipc_sk(sk)->conn_timeout = value;
2853 		break;
2854 	case TIPC_MCAST_BROADCAST:
2855 		tsk->mc_method.rcast = false;
2856 		tsk->mc_method.mandatory = true;
2857 		break;
2858 	case TIPC_MCAST_REPLICAST:
2859 		tsk->mc_method.rcast = true;
2860 		tsk->mc_method.mandatory = true;
2861 		break;
2862 	case TIPC_GROUP_JOIN:
2863 		res = tipc_sk_join(tsk, &mreq);
2864 		break;
2865 	case TIPC_GROUP_LEAVE:
2866 		res = tipc_sk_leave(tsk);
2867 		break;
2868 	default:
2869 		res = -EINVAL;
2870 	}
2871 
2872 	release_sock(sk);
2873 
2874 	return res;
2875 }
2876 
2877 /**
2878  * tipc_getsockopt - get socket option
2879  * @sock: socket structure
2880  * @lvl: option level
2881  * @opt: option identifier
2882  * @ov: receptacle for option value
2883  * @ol: receptacle for length of option value
2884  *
2885  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2886  * (to ease compatibility).
2887  *
2888  * Returns 0 on success, errno otherwise
2889  */
2890 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2891 			   char __user *ov, int __user *ol)
2892 {
2893 	struct sock *sk = sock->sk;
2894 	struct tipc_sock *tsk = tipc_sk(sk);
2895 	struct tipc_name_seq seq;
2896 	int len, scope;
2897 	u32 value;
2898 	int res;
2899 
2900 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2901 		return put_user(0, ol);
2902 	if (lvl != SOL_TIPC)
2903 		return -ENOPROTOOPT;
2904 	res = get_user(len, ol);
2905 	if (res)
2906 		return res;
2907 
2908 	lock_sock(sk);
2909 
2910 	switch (opt) {
2911 	case TIPC_IMPORTANCE:
2912 		value = tsk_importance(tsk);
2913 		break;
2914 	case TIPC_SRC_DROPPABLE:
2915 		value = tsk_unreliable(tsk);
2916 		break;
2917 	case TIPC_DEST_DROPPABLE:
2918 		value = tsk_unreturnable(tsk);
2919 		break;
2920 	case TIPC_CONN_TIMEOUT:
2921 		value = tsk->conn_timeout;
2922 		/* no need to set "res", since already 0 at this point */
2923 		break;
2924 	case TIPC_NODE_RECVQ_DEPTH:
2925 		value = 0; /* was tipc_queue_size, now obsolete */
2926 		break;
2927 	case TIPC_SOCK_RECVQ_DEPTH:
2928 		value = skb_queue_len(&sk->sk_receive_queue);
2929 		break;
2930 	case TIPC_GROUP_JOIN:
2931 		seq.type = 0;
2932 		if (tsk->group)
2933 			tipc_group_self(tsk->group, &seq, &scope);
2934 		value = seq.type;
2935 		break;
2936 	default:
2937 		res = -EINVAL;
2938 	}
2939 
2940 	release_sock(sk);
2941 
2942 	if (res)
2943 		return res;	/* "get" failed */
2944 
2945 	if (len < sizeof(value))
2946 		return -EINVAL;
2947 
2948 	if (copy_to_user(ov, &value, sizeof(value)))
2949 		return -EFAULT;
2950 
2951 	return put_user(sizeof(value), ol);
2952 }
2953 
2954 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2955 {
2956 	struct sock *sk = sock->sk;
2957 	struct tipc_sioc_ln_req lnr;
2958 	void __user *argp = (void __user *)arg;
2959 
2960 	switch (cmd) {
2961 	case SIOCGETLINKNAME:
2962 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2963 			return -EFAULT;
2964 		if (!tipc_node_get_linkname(sock_net(sk),
2965 					    lnr.bearer_id & 0xffff, lnr.peer,
2966 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2967 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2968 				return -EFAULT;
2969 			return 0;
2970 		}
2971 		return -EADDRNOTAVAIL;
2972 	default:
2973 		return -ENOIOCTLCMD;
2974 	}
2975 }
2976 
2977 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
2978 {
2979 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
2980 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
2981 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
2982 
2983 	tsk1->peer.family = AF_TIPC;
2984 	tsk1->peer.addrtype = TIPC_ADDR_ID;
2985 	tsk1->peer.scope = TIPC_NODE_SCOPE;
2986 	tsk1->peer.addr.id.ref = tsk2->portid;
2987 	tsk1->peer.addr.id.node = onode;
2988 	tsk2->peer.family = AF_TIPC;
2989 	tsk2->peer.addrtype = TIPC_ADDR_ID;
2990 	tsk2->peer.scope = TIPC_NODE_SCOPE;
2991 	tsk2->peer.addr.id.ref = tsk1->portid;
2992 	tsk2->peer.addr.id.node = onode;
2993 
2994 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
2995 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
2996 	return 0;
2997 }
2998 
2999 /* Protocol switches for the various types of TIPC sockets */
3000 
3001 static const struct proto_ops msg_ops = {
3002 	.owner		= THIS_MODULE,
3003 	.family		= AF_TIPC,
3004 	.release	= tipc_release,
3005 	.bind		= tipc_bind,
3006 	.connect	= tipc_connect,
3007 	.socketpair	= tipc_socketpair,
3008 	.accept		= sock_no_accept,
3009 	.getname	= tipc_getname,
3010 	.poll		= tipc_poll,
3011 	.ioctl		= tipc_ioctl,
3012 	.listen		= sock_no_listen,
3013 	.shutdown	= tipc_shutdown,
3014 	.setsockopt	= tipc_setsockopt,
3015 	.getsockopt	= tipc_getsockopt,
3016 	.sendmsg	= tipc_sendmsg,
3017 	.recvmsg	= tipc_recvmsg,
3018 	.mmap		= sock_no_mmap,
3019 	.sendpage	= sock_no_sendpage
3020 };
3021 
3022 static const struct proto_ops packet_ops = {
3023 	.owner		= THIS_MODULE,
3024 	.family		= AF_TIPC,
3025 	.release	= tipc_release,
3026 	.bind		= tipc_bind,
3027 	.connect	= tipc_connect,
3028 	.socketpair	= tipc_socketpair,
3029 	.accept		= tipc_accept,
3030 	.getname	= tipc_getname,
3031 	.poll		= tipc_poll,
3032 	.ioctl		= tipc_ioctl,
3033 	.listen		= tipc_listen,
3034 	.shutdown	= tipc_shutdown,
3035 	.setsockopt	= tipc_setsockopt,
3036 	.getsockopt	= tipc_getsockopt,
3037 	.sendmsg	= tipc_send_packet,
3038 	.recvmsg	= tipc_recvmsg,
3039 	.mmap		= sock_no_mmap,
3040 	.sendpage	= sock_no_sendpage
3041 };
3042 
3043 static const struct proto_ops stream_ops = {
3044 	.owner		= THIS_MODULE,
3045 	.family		= AF_TIPC,
3046 	.release	= tipc_release,
3047 	.bind		= tipc_bind,
3048 	.connect	= tipc_connect,
3049 	.socketpair	= tipc_socketpair,
3050 	.accept		= tipc_accept,
3051 	.getname	= tipc_getname,
3052 	.poll		= tipc_poll,
3053 	.ioctl		= tipc_ioctl,
3054 	.listen		= tipc_listen,
3055 	.shutdown	= tipc_shutdown,
3056 	.setsockopt	= tipc_setsockopt,
3057 	.getsockopt	= tipc_getsockopt,
3058 	.sendmsg	= tipc_sendstream,
3059 	.recvmsg	= tipc_recvstream,
3060 	.mmap		= sock_no_mmap,
3061 	.sendpage	= sock_no_sendpage
3062 };
3063 
3064 static const struct net_proto_family tipc_family_ops = {
3065 	.owner		= THIS_MODULE,
3066 	.family		= AF_TIPC,
3067 	.create		= tipc_sk_create
3068 };
3069 
3070 static struct proto tipc_proto = {
3071 	.name		= "TIPC",
3072 	.owner		= THIS_MODULE,
3073 	.obj_size	= sizeof(struct tipc_sock),
3074 	.sysctl_rmem	= sysctl_tipc_rmem
3075 };
3076 
3077 /**
3078  * tipc_socket_init - initialize TIPC socket interface
3079  *
3080  * Returns 0 on success, errno otherwise
3081  */
3082 int tipc_socket_init(void)
3083 {
3084 	int res;
3085 
3086 	res = proto_register(&tipc_proto, 1);
3087 	if (res) {
3088 		pr_err("Failed to register TIPC protocol type\n");
3089 		goto out;
3090 	}
3091 
3092 	res = sock_register(&tipc_family_ops);
3093 	if (res) {
3094 		pr_err("Failed to register TIPC socket type\n");
3095 		proto_unregister(&tipc_proto);
3096 		goto out;
3097 	}
3098  out:
3099 	return res;
3100 }
3101 
3102 /**
3103  * tipc_socket_stop - stop TIPC socket interface
3104  */
3105 void tipc_socket_stop(void)
3106 {
3107 	sock_unregister(tipc_family_ops.family);
3108 	proto_unregister(&tipc_proto);
3109 }
3110 
3111 /* Caller should hold socket lock for the passed tipc socket. */
3112 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3113 {
3114 	u32 peer_node;
3115 	u32 peer_port;
3116 	struct nlattr *nest;
3117 
3118 	peer_node = tsk_peer_node(tsk);
3119 	peer_port = tsk_peer_port(tsk);
3120 
3121 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3122 
3123 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3124 		goto msg_full;
3125 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3126 		goto msg_full;
3127 
3128 	if (tsk->conn_type != 0) {
3129 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3130 			goto msg_full;
3131 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3132 			goto msg_full;
3133 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3134 			goto msg_full;
3135 	}
3136 	nla_nest_end(skb, nest);
3137 
3138 	return 0;
3139 
3140 msg_full:
3141 	nla_nest_cancel(skb, nest);
3142 
3143 	return -EMSGSIZE;
3144 }
3145 
3146 /* Caller should hold socket lock for the passed tipc socket. */
3147 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3148 			    struct tipc_sock *tsk)
3149 {
3150 	int err;
3151 	void *hdr;
3152 	struct nlattr *attrs;
3153 	struct net *net = sock_net(skb->sk);
3154 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3155 	struct sock *sk = &tsk->sk;
3156 
3157 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3158 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3159 	if (!hdr)
3160 		goto msg_cancel;
3161 
3162 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3163 	if (!attrs)
3164 		goto genlmsg_cancel;
3165 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
3166 		goto attr_msg_cancel;
3167 	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
3168 		goto attr_msg_cancel;
3169 
3170 	if (tipc_sk_connected(sk)) {
3171 		err = __tipc_nl_add_sk_con(skb, tsk);
3172 		if (err)
3173 			goto attr_msg_cancel;
3174 	} else if (!list_empty(&tsk->publications)) {
3175 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3176 			goto attr_msg_cancel;
3177 	}
3178 	nla_nest_end(skb, attrs);
3179 	genlmsg_end(skb, hdr);
3180 
3181 	return 0;
3182 
3183 attr_msg_cancel:
3184 	nla_nest_cancel(skb, attrs);
3185 genlmsg_cancel:
3186 	genlmsg_cancel(skb, hdr);
3187 msg_cancel:
3188 	return -EMSGSIZE;
3189 }
3190 
3191 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3192 {
3193 	int err;
3194 	struct tipc_sock *tsk;
3195 	const struct bucket_table *tbl;
3196 	struct rhash_head *pos;
3197 	struct net *net = sock_net(skb->sk);
3198 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3199 	u32 tbl_id = cb->args[0];
3200 	u32 prev_portid = cb->args[1];
3201 
3202 	rcu_read_lock();
3203 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3204 	for (; tbl_id < tbl->size; tbl_id++) {
3205 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3206 			spin_lock_bh(&tsk->sk.sk_lock.slock);
3207 			if (prev_portid && prev_portid != tsk->portid) {
3208 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3209 				continue;
3210 			}
3211 
3212 			err = __tipc_nl_add_sk(skb, cb, tsk);
3213 			if (err) {
3214 				prev_portid = tsk->portid;
3215 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3216 				goto out;
3217 			}
3218 			prev_portid = 0;
3219 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
3220 		}
3221 	}
3222 out:
3223 	rcu_read_unlock();
3224 	cb->args[0] = tbl_id;
3225 	cb->args[1] = prev_portid;
3226 
3227 	return skb->len;
3228 }
3229 
3230 /* Caller should hold socket lock for the passed tipc socket. */
3231 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3232 				 struct netlink_callback *cb,
3233 				 struct publication *publ)
3234 {
3235 	void *hdr;
3236 	struct nlattr *attrs;
3237 
3238 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3239 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3240 	if (!hdr)
3241 		goto msg_cancel;
3242 
3243 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3244 	if (!attrs)
3245 		goto genlmsg_cancel;
3246 
3247 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3248 		goto attr_msg_cancel;
3249 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3250 		goto attr_msg_cancel;
3251 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3252 		goto attr_msg_cancel;
3253 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3254 		goto attr_msg_cancel;
3255 
3256 	nla_nest_end(skb, attrs);
3257 	genlmsg_end(skb, hdr);
3258 
3259 	return 0;
3260 
3261 attr_msg_cancel:
3262 	nla_nest_cancel(skb, attrs);
3263 genlmsg_cancel:
3264 	genlmsg_cancel(skb, hdr);
3265 msg_cancel:
3266 	return -EMSGSIZE;
3267 }
3268 
3269 /* Caller should hold socket lock for the passed tipc socket. */
3270 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3271 				  struct netlink_callback *cb,
3272 				  struct tipc_sock *tsk, u32 *last_publ)
3273 {
3274 	int err;
3275 	struct publication *p;
3276 
3277 	if (*last_publ) {
3278 		list_for_each_entry(p, &tsk->publications, pport_list) {
3279 			if (p->key == *last_publ)
3280 				break;
3281 		}
3282 		if (p->key != *last_publ) {
3283 			/* We never set seq or call nl_dump_check_consistent()
3284 			 * this means that setting prev_seq here will cause the
3285 			 * consistence check to fail in the netlink callback
3286 			 * handler. Resulting in the last NLMSG_DONE message
3287 			 * having the NLM_F_DUMP_INTR flag set.
3288 			 */
3289 			cb->prev_seq = 1;
3290 			*last_publ = 0;
3291 			return -EPIPE;
3292 		}
3293 	} else {
3294 		p = list_first_entry(&tsk->publications, struct publication,
3295 				     pport_list);
3296 	}
3297 
3298 	list_for_each_entry_from(p, &tsk->publications, pport_list) {
3299 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3300 		if (err) {
3301 			*last_publ = p->key;
3302 			return err;
3303 		}
3304 	}
3305 	*last_publ = 0;
3306 
3307 	return 0;
3308 }
3309 
3310 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3311 {
3312 	int err;
3313 	u32 tsk_portid = cb->args[0];
3314 	u32 last_publ = cb->args[1];
3315 	u32 done = cb->args[2];
3316 	struct net *net = sock_net(skb->sk);
3317 	struct tipc_sock *tsk;
3318 
3319 	if (!tsk_portid) {
3320 		struct nlattr **attrs;
3321 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3322 
3323 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3324 		if (err)
3325 			return err;
3326 
3327 		if (!attrs[TIPC_NLA_SOCK])
3328 			return -EINVAL;
3329 
3330 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3331 				       attrs[TIPC_NLA_SOCK],
3332 				       tipc_nl_sock_policy, NULL);
3333 		if (err)
3334 			return err;
3335 
3336 		if (!sock[TIPC_NLA_SOCK_REF])
3337 			return -EINVAL;
3338 
3339 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3340 	}
3341 
3342 	if (done)
3343 		return 0;
3344 
3345 	tsk = tipc_sk_lookup(net, tsk_portid);
3346 	if (!tsk)
3347 		return -EINVAL;
3348 
3349 	lock_sock(&tsk->sk);
3350 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3351 	if (!err)
3352 		done = 1;
3353 	release_sock(&tsk->sk);
3354 	sock_put(&tsk->sk);
3355 
3356 	cb->args[0] = tsk_portid;
3357 	cb->args[1] = last_publ;
3358 	cb->args[2] = done;
3359 
3360 	return skb->len;
3361 }
3362