xref: /linux/net/tipc/socket.c (revision f87be8948185a4f1085aea38bccd0d2ce07f711c)
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_INTV	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 	bool group_is_open;
120 };
121 
122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123 static void tipc_data_ready(struct sock *sk);
124 static void tipc_write_space(struct sock *sk);
125 static void tipc_sock_destruct(struct sock *sk);
126 static int tipc_release(struct socket *sock);
127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
128 		       bool kern);
129 static void tipc_sk_timeout(struct timer_list *t);
130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 			   struct tipc_name_seq const *seq);
132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 			    struct tipc_name_seq const *seq);
134 static int tipc_sk_leave(struct tipc_sock *tsk);
135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136 static int tipc_sk_insert(struct tipc_sock *tsk);
137 static void tipc_sk_remove(struct tipc_sock *tsk);
138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
140 
141 static const struct proto_ops packet_ops;
142 static const struct proto_ops stream_ops;
143 static const struct proto_ops msg_ops;
144 static struct proto tipc_proto;
145 static const struct rhashtable_params tsk_rht_params;
146 
147 static u32 tsk_own_node(struct tipc_sock *tsk)
148 {
149 	return msg_prevnode(&tsk->phdr);
150 }
151 
152 static u32 tsk_peer_node(struct tipc_sock *tsk)
153 {
154 	return msg_destnode(&tsk->phdr);
155 }
156 
157 static u32 tsk_peer_port(struct tipc_sock *tsk)
158 {
159 	return msg_destport(&tsk->phdr);
160 }
161 
162 static  bool tsk_unreliable(struct tipc_sock *tsk)
163 {
164 	return msg_src_droppable(&tsk->phdr) != 0;
165 }
166 
167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
168 {
169 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
170 }
171 
172 static bool tsk_unreturnable(struct tipc_sock *tsk)
173 {
174 	return msg_dest_droppable(&tsk->phdr) != 0;
175 }
176 
177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
178 {
179 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
180 }
181 
182 static int tsk_importance(struct tipc_sock *tsk)
183 {
184 	return msg_importance(&tsk->phdr);
185 }
186 
187 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
188 {
189 	if (imp > TIPC_CRITICAL_IMPORTANCE)
190 		return -EINVAL;
191 	msg_set_importance(&tsk->phdr, (u32)imp);
192 	return 0;
193 }
194 
195 static struct tipc_sock *tipc_sk(const struct sock *sk)
196 {
197 	return container_of(sk, struct tipc_sock, sk);
198 }
199 
200 static bool tsk_conn_cong(struct tipc_sock *tsk)
201 {
202 	return tsk->snt_unacked > tsk->snd_win;
203 }
204 
205 static u16 tsk_blocks(int len)
206 {
207 	return ((len / FLOWCTL_BLK_SZ) + 1);
208 }
209 
210 /* tsk_blocks(): translate a buffer size in bytes to number of
211  * advertisable blocks, taking into account the ratio truesize(len)/len
212  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
213  */
214 static u16 tsk_adv_blocks(int len)
215 {
216 	return len / FLOWCTL_BLK_SZ / 4;
217 }
218 
219 /* tsk_inc(): increment counter for sent or received data
220  * - If block based flow control is not supported by peer we
221  *   fall back to message based ditto, incrementing the counter
222  */
223 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
224 {
225 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
226 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
227 	return 1;
228 }
229 
230 /**
231  * tsk_advance_rx_queue - discard first buffer in socket receive queue
232  *
233  * Caller must hold socket lock
234  */
235 static void tsk_advance_rx_queue(struct sock *sk)
236 {
237 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
238 }
239 
240 /* tipc_sk_respond() : send response message back to sender
241  */
242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
243 {
244 	u32 selector;
245 	u32 dnode;
246 	u32 onode = tipc_own_addr(sock_net(sk));
247 
248 	if (!tipc_msg_reverse(onode, &skb, err))
249 		return;
250 
251 	dnode = msg_destnode(buf_msg(skb));
252 	selector = msg_origport(buf_msg(skb));
253 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
254 }
255 
256 /**
257  * tsk_rej_rx_queue - reject all buffers in socket receive queue
258  *
259  * Caller must hold socket lock
260  */
261 static void tsk_rej_rx_queue(struct sock *sk)
262 {
263 	struct sk_buff *skb;
264 
265 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
266 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
267 }
268 
269 static bool tipc_sk_connected(struct sock *sk)
270 {
271 	return sk->sk_state == TIPC_ESTABLISHED;
272 }
273 
274 /* tipc_sk_type_connectionless - check if the socket is datagram socket
275  * @sk: socket
276  *
277  * Returns true if connection less, false otherwise
278  */
279 static bool tipc_sk_type_connectionless(struct sock *sk)
280 {
281 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
282 }
283 
284 /* tsk_peer_msg - verify if message was sent by connected port's peer
285  *
286  * Handles cases where the node's network address has changed from
287  * the default of <0.0.0> to its configured setting.
288  */
289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
290 {
291 	struct sock *sk = &tsk->sk;
292 	u32 self = tipc_own_addr(sock_net(sk));
293 	u32 peer_port = tsk_peer_port(tsk);
294 	u32 orig_node, 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 == self)
309 		return true;
310 
311 	if (!peer_node && orig_node == self)
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(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
464 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
465 
466 	msg_set_origport(msg, tsk->portid);
467 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
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 	tsk->group_is_open = true;
476 	atomic_set(&tsk->dupl_rcvcnt, 0);
477 
478 	/* Start out with safe limits until we receive an advertised window */
479 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
480 	tsk->rcv_win = tsk->snd_win;
481 
482 	if (tipc_sk_type_connectionless(sk)) {
483 		tsk_set_unreturnable(tsk, true);
484 		if (sock->type == SOCK_DGRAM)
485 			tsk_set_unreliable(tsk, true);
486 	}
487 
488 	return 0;
489 }
490 
491 static void tipc_sk_callback(struct rcu_head *head)
492 {
493 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
494 
495 	sock_put(&tsk->sk);
496 }
497 
498 /* Caller should hold socket lock for the socket. */
499 static void __tipc_shutdown(struct socket *sock, int error)
500 {
501 	struct sock *sk = sock->sk;
502 	struct tipc_sock *tsk = tipc_sk(sk);
503 	struct net *net = sock_net(sk);
504 	long timeout = CONN_TIMEOUT_DEFAULT;
505 	u32 dnode = tsk_peer_node(tsk);
506 	struct sk_buff *skb;
507 
508 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
509 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
510 					    !tsk_conn_cong(tsk)));
511 
512 	/* Reject all unreceived messages, except on an active connection
513 	 * (which disconnects locally & sends a 'FIN+' to peer).
514 	 */
515 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
516 		if (TIPC_SKB_CB(skb)->bytes_read) {
517 			kfree_skb(skb);
518 			continue;
519 		}
520 		if (!tipc_sk_type_connectionless(sk) &&
521 		    sk->sk_state != TIPC_DISCONNECTING) {
522 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
523 			tipc_node_remove_conn(net, dnode, tsk->portid);
524 		}
525 		tipc_sk_respond(sk, skb, error);
526 	}
527 
528 	if (tipc_sk_type_connectionless(sk))
529 		return;
530 
531 	if (sk->sk_state != TIPC_DISCONNECTING) {
532 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
533 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
534 				      tsk_own_node(tsk), tsk_peer_port(tsk),
535 				      tsk->portid, error);
536 		if (skb)
537 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
538 		tipc_node_remove_conn(net, dnode, tsk->portid);
539 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
540 	}
541 }
542 
543 /**
544  * tipc_release - destroy a TIPC socket
545  * @sock: socket to destroy
546  *
547  * This routine cleans up any messages that are still queued on the socket.
548  * For DGRAM and RDM socket types, all queued messages are rejected.
549  * For SEQPACKET and STREAM socket types, the first message is rejected
550  * and any others are discarded.  (If the first message on a STREAM socket
551  * is partially-read, it is discarded and the next one is rejected instead.)
552  *
553  * NOTE: Rejected messages are not necessarily returned to the sender!  They
554  * are returned or discarded according to the "destination droppable" setting
555  * specified for the message by the sender.
556  *
557  * Returns 0 on success, errno otherwise
558  */
559 static int tipc_release(struct socket *sock)
560 {
561 	struct sock *sk = sock->sk;
562 	struct tipc_sock *tsk;
563 
564 	/*
565 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
566 	 * releases a pre-allocated child socket that was never used)
567 	 */
568 	if (sk == NULL)
569 		return 0;
570 
571 	tsk = tipc_sk(sk);
572 	lock_sock(sk);
573 
574 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
575 	sk->sk_shutdown = SHUTDOWN_MASK;
576 	tipc_sk_leave(tsk);
577 	tipc_sk_withdraw(tsk, 0, NULL);
578 	sk_stop_timer(sk, &sk->sk_timer);
579 	tipc_sk_remove(tsk);
580 
581 	/* Reject any messages that accumulated in backlog queue */
582 	release_sock(sk);
583 	tipc_dest_list_purge(&tsk->cong_links);
584 	tsk->cong_link_cnt = 0;
585 	call_rcu(&tsk->rcu, tipc_sk_callback);
586 	sock->sk = NULL;
587 
588 	return 0;
589 }
590 
591 /**
592  * tipc_bind - associate or disassocate TIPC name(s) with a socket
593  * @sock: socket structure
594  * @uaddr: socket address describing name(s) and desired operation
595  * @uaddr_len: size of socket address data structure
596  *
597  * Name and name sequence binding is indicated using a positive scope value;
598  * a negative scope value unbinds the specified name.  Specifying no name
599  * (i.e. a socket address length of 0) unbinds all names from the socket.
600  *
601  * Returns 0 on success, errno otherwise
602  *
603  * NOTE: This routine doesn't need to take the socket lock since it doesn't
604  *       access any non-constant socket information.
605  */
606 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
607 		     int uaddr_len)
608 {
609 	struct sock *sk = sock->sk;
610 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
611 	struct tipc_sock *tsk = tipc_sk(sk);
612 	int res = -EINVAL;
613 
614 	lock_sock(sk);
615 	if (unlikely(!uaddr_len)) {
616 		res = tipc_sk_withdraw(tsk, 0, NULL);
617 		goto exit;
618 	}
619 	if (tsk->group) {
620 		res = -EACCES;
621 		goto exit;
622 	}
623 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
624 		res = -EINVAL;
625 		goto exit;
626 	}
627 	if (addr->family != AF_TIPC) {
628 		res = -EAFNOSUPPORT;
629 		goto exit;
630 	}
631 
632 	if (addr->addrtype == TIPC_ADDR_NAME)
633 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
634 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
635 		res = -EAFNOSUPPORT;
636 		goto exit;
637 	}
638 
639 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
640 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
641 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
642 		res = -EACCES;
643 		goto exit;
644 	}
645 
646 	res = (addr->scope >= 0) ?
647 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
648 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
649 exit:
650 	release_sock(sk);
651 	return res;
652 }
653 
654 /**
655  * tipc_getname - get port ID of socket or peer socket
656  * @sock: socket structure
657  * @uaddr: area for returned socket address
658  * @uaddr_len: area for returned length of socket address
659  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
660  *
661  * Returns 0 on success, errno otherwise
662  *
663  * NOTE: This routine doesn't need to take the socket lock since it only
664  *       accesses socket information that is unchanging (or which changes in
665  *       a completely predictable manner).
666  */
667 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
668 			int peer)
669 {
670 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
671 	struct sock *sk = sock->sk;
672 	struct tipc_sock *tsk = tipc_sk(sk);
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 = tipc_own_addr(sock_net(sk));
684 	}
685 
686 	addr->addrtype = TIPC_ADDR_ID;
687 	addr->family = AF_TIPC;
688 	addr->scope = 0;
689 	addr->addr.name.domain = 0;
690 
691 	return sizeof(*addr);
692 }
693 
694 /**
695  * tipc_poll - read pollmask
696  * @file: file structure associated with the socket
697  * @sock: socket for which to calculate the poll bits
698  *
699  * Returns pollmask value
700  *
701  * COMMENTARY:
702  * It appears that the usual socket locking mechanisms are not useful here
703  * since the pollmask info is potentially out-of-date the moment this routine
704  * exits.  TCP and other protocols seem to rely on higher level poll routines
705  * to handle any preventable race conditions, so TIPC will do the same ...
706  *
707  * IMPORTANT: The fact that a read or write operation is indicated does NOT
708  * imply that the operation will succeed, merely that it should be performed
709  * and will not block.
710  */
711 static __poll_t tipc_poll_mask(struct socket *sock, __poll_t events)
712 {
713 	struct sock *sk = sock->sk;
714 	struct tipc_sock *tsk = tipc_sk(sk);
715 	__poll_t revents = 0;
716 
717 	if (sk->sk_shutdown & RCV_SHUTDOWN)
718 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
719 	if (sk->sk_shutdown == SHUTDOWN_MASK)
720 		revents |= EPOLLHUP;
721 
722 	switch (sk->sk_state) {
723 	case TIPC_ESTABLISHED:
724 	case TIPC_CONNECTING:
725 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
726 			revents |= EPOLLOUT;
727 		/* fall thru' */
728 	case TIPC_LISTEN:
729 		if (!skb_queue_empty(&sk->sk_receive_queue))
730 			revents |= EPOLLIN | EPOLLRDNORM;
731 		break;
732 	case TIPC_OPEN:
733 		if (tsk->group_is_open && !tsk->cong_link_cnt)
734 			revents |= EPOLLOUT;
735 		if (!tipc_sk_type_connectionless(sk))
736 			break;
737 		if (skb_queue_empty(&sk->sk_receive_queue))
738 			break;
739 		revents |= EPOLLIN | EPOLLRDNORM;
740 		break;
741 	case TIPC_DISCONNECTING:
742 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
743 		break;
744 	}
745 	return revents;
746 }
747 
748 /**
749  * tipc_sendmcast - send multicast message
750  * @sock: socket structure
751  * @seq: destination address
752  * @msg: message to send
753  * @dlen: length of data to send
754  * @timeout: timeout to wait for wakeup
755  *
756  * Called from function tipc_sendmsg(), which has done all sanity checks
757  * Returns the number of bytes sent on success, or errno
758  */
759 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
760 			  struct msghdr *msg, size_t dlen, long timeout)
761 {
762 	struct sock *sk = sock->sk;
763 	struct tipc_sock *tsk = tipc_sk(sk);
764 	struct tipc_msg *hdr = &tsk->phdr;
765 	struct net *net = sock_net(sk);
766 	int mtu = tipc_bcast_get_mtu(net);
767 	struct tipc_mc_method *method = &tsk->mc_method;
768 	struct sk_buff_head pkts;
769 	struct tipc_nlist dsts;
770 	int rc;
771 
772 	if (tsk->group)
773 		return -EACCES;
774 
775 	/* Block or return if any destination link is congested */
776 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
777 	if (unlikely(rc))
778 		return rc;
779 
780 	/* Lookup destination nodes */
781 	tipc_nlist_init(&dsts, tipc_own_addr(net));
782 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
783 				      seq->upper, &dsts);
784 	if (!dsts.local && !dsts.remote)
785 		return -EHOSTUNREACH;
786 
787 	/* Build message header */
788 	msg_set_type(hdr, TIPC_MCAST_MSG);
789 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
790 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
791 	msg_set_destport(hdr, 0);
792 	msg_set_destnode(hdr, 0);
793 	msg_set_nametype(hdr, seq->type);
794 	msg_set_namelower(hdr, seq->lower);
795 	msg_set_nameupper(hdr, seq->upper);
796 
797 	/* Build message as chain of buffers */
798 	skb_queue_head_init(&pkts);
799 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
800 
801 	/* Send message if build was successful */
802 	if (unlikely(rc == dlen))
803 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
804 				     &tsk->cong_link_cnt);
805 
806 	tipc_nlist_purge(&dsts);
807 
808 	return rc ? rc : dlen;
809 }
810 
811 /**
812  * tipc_send_group_msg - send a message to a member in the group
813  * @net: network namespace
814  * @m: message to send
815  * @mb: group member
816  * @dnode: destination node
817  * @dport: destination port
818  * @dlen: total length of message data
819  */
820 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
821 			       struct msghdr *m, struct tipc_member *mb,
822 			       u32 dnode, u32 dport, int dlen)
823 {
824 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
825 	struct tipc_mc_method *method = &tsk->mc_method;
826 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
827 	struct tipc_msg *hdr = &tsk->phdr;
828 	struct sk_buff_head pkts;
829 	int mtu, rc;
830 
831 	/* Complete message header */
832 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
833 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
834 	msg_set_destport(hdr, dport);
835 	msg_set_destnode(hdr, dnode);
836 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
837 
838 	/* Build message as chain of buffers */
839 	skb_queue_head_init(&pkts);
840 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
841 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
842 	if (unlikely(rc != dlen))
843 		return rc;
844 
845 	/* Send message */
846 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
847 	if (unlikely(rc == -ELINKCONG)) {
848 		tipc_dest_push(&tsk->cong_links, dnode, 0);
849 		tsk->cong_link_cnt++;
850 	}
851 
852 	/* Update send window */
853 	tipc_group_update_member(mb, blks);
854 
855 	/* A broadcast sent within next EXPIRE period must follow same path */
856 	method->rcast = true;
857 	method->mandatory = true;
858 	return dlen;
859 }
860 
861 /**
862  * tipc_send_group_unicast - send message to a member in the group
863  * @sock: socket structure
864  * @m: message to send
865  * @dlen: total length of message data
866  * @timeout: timeout to wait for wakeup
867  *
868  * Called from function tipc_sendmsg(), which has done all sanity checks
869  * Returns the number of bytes sent on success, or errno
870  */
871 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
872 				   int dlen, long timeout)
873 {
874 	struct sock *sk = sock->sk;
875 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
876 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
877 	struct tipc_sock *tsk = tipc_sk(sk);
878 	struct tipc_group *grp = tsk->group;
879 	struct net *net = sock_net(sk);
880 	struct tipc_member *mb = NULL;
881 	u32 node, port;
882 	int rc;
883 
884 	node = dest->addr.id.node;
885 	port = dest->addr.id.ref;
886 	if (!port && !node)
887 		return -EHOSTUNREACH;
888 
889 	/* Block or return if destination link or member is congested */
890 	rc = tipc_wait_for_cond(sock, &timeout,
891 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
892 				!tipc_group_cong(grp, node, port, blks, &mb));
893 	if (unlikely(rc))
894 		return rc;
895 
896 	if (unlikely(!mb))
897 		return -EHOSTUNREACH;
898 
899 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
900 
901 	return rc ? rc : dlen;
902 }
903 
904 /**
905  * tipc_send_group_anycast - send message to any member with given identity
906  * @sock: socket structure
907  * @m: message to send
908  * @dlen: total length of message data
909  * @timeout: timeout to wait for wakeup
910  *
911  * Called from function tipc_sendmsg(), which has done all sanity checks
912  * Returns the number of bytes sent on success, or errno
913  */
914 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
915 				   int dlen, long timeout)
916 {
917 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
918 	struct sock *sk = sock->sk;
919 	struct tipc_sock *tsk = tipc_sk(sk);
920 	struct list_head *cong_links = &tsk->cong_links;
921 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
922 	struct tipc_group *grp = tsk->group;
923 	struct tipc_msg *hdr = &tsk->phdr;
924 	struct tipc_member *first = NULL;
925 	struct tipc_member *mbr = NULL;
926 	struct net *net = sock_net(sk);
927 	u32 node, port, exclude;
928 	struct list_head dsts;
929 	u32 type, inst, scope;
930 	int lookups = 0;
931 	int dstcnt, rc;
932 	bool cong;
933 
934 	INIT_LIST_HEAD(&dsts);
935 
936 	type = msg_nametype(hdr);
937 	inst = dest->addr.name.name.instance;
938 	scope = msg_lookup_scope(hdr);
939 	exclude = tipc_group_exclude(grp);
940 
941 	while (++lookups < 4) {
942 		first = NULL;
943 
944 		/* Look for a non-congested destination member, if any */
945 		while (1) {
946 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
947 						 &dstcnt, exclude, false))
948 				return -EHOSTUNREACH;
949 			tipc_dest_pop(&dsts, &node, &port);
950 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
951 			if (!cong)
952 				break;
953 			if (mbr == first)
954 				break;
955 			if (!first)
956 				first = mbr;
957 		}
958 
959 		/* Start over if destination was not in member list */
960 		if (unlikely(!mbr))
961 			continue;
962 
963 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
964 			break;
965 
966 		/* Block or return if destination link or member is congested */
967 		rc = tipc_wait_for_cond(sock, &timeout,
968 					!tipc_dest_find(cong_links, node, 0) &&
969 					!tipc_group_cong(grp, node, port,
970 							 blks, &mbr));
971 		if (unlikely(rc))
972 			return rc;
973 
974 		/* Send, unless destination disappeared while waiting */
975 		if (likely(mbr))
976 			break;
977 	}
978 
979 	if (unlikely(lookups >= 4))
980 		return -EHOSTUNREACH;
981 
982 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
983 
984 	return rc ? rc : dlen;
985 }
986 
987 /**
988  * tipc_send_group_bcast - send message to all members in communication group
989  * @sk: socket structure
990  * @m: message to send
991  * @dlen: total length of message data
992  * @timeout: timeout to wait for wakeup
993  *
994  * Called from function tipc_sendmsg(), which has done all sanity checks
995  * Returns the number of bytes sent on success, or errno
996  */
997 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
998 				 int dlen, long timeout)
999 {
1000 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1001 	struct sock *sk = sock->sk;
1002 	struct net *net = sock_net(sk);
1003 	struct tipc_sock *tsk = tipc_sk(sk);
1004 	struct tipc_group *grp = tsk->group;
1005 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1006 	struct tipc_mc_method *method = &tsk->mc_method;
1007 	bool ack = method->mandatory && method->rcast;
1008 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1009 	struct tipc_msg *hdr = &tsk->phdr;
1010 	int mtu = tipc_bcast_get_mtu(net);
1011 	struct sk_buff_head pkts;
1012 	int rc = -EHOSTUNREACH;
1013 
1014 	if (!dsts->local && !dsts->remote)
1015 		return -EHOSTUNREACH;
1016 
1017 	/* Block or return if any destination link or member is congested */
1018 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1019 				!tipc_group_bc_cong(grp, blks));
1020 	if (unlikely(rc))
1021 		return rc;
1022 
1023 	/* Complete message header */
1024 	if (dest) {
1025 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1026 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1027 	} else {
1028 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1029 		msg_set_nameinst(hdr, 0);
1030 	}
1031 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1032 	msg_set_destport(hdr, 0);
1033 	msg_set_destnode(hdr, 0);
1034 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1035 
1036 	/* Avoid getting stuck with repeated forced replicasts */
1037 	msg_set_grp_bc_ack_req(hdr, ack);
1038 
1039 	/* Build message as chain of buffers */
1040 	skb_queue_head_init(&pkts);
1041 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1042 	if (unlikely(rc != dlen))
1043 		return rc;
1044 
1045 	/* Send message */
1046 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1047 	if (unlikely(rc))
1048 		return rc;
1049 
1050 	/* Update broadcast sequence number and send windows */
1051 	tipc_group_update_bc_members(tsk->group, blks, ack);
1052 
1053 	/* Broadcast link is now free to choose method for next broadcast */
1054 	method->mandatory = false;
1055 	method->expires = jiffies;
1056 
1057 	return dlen;
1058 }
1059 
1060 /**
1061  * tipc_send_group_mcast - send message to all members with given identity
1062  * @sock: socket structure
1063  * @m: message to send
1064  * @dlen: total length of message data
1065  * @timeout: timeout to wait for wakeup
1066  *
1067  * Called from function tipc_sendmsg(), which has done all sanity checks
1068  * Returns the number of bytes sent on success, or errno
1069  */
1070 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1071 				 int dlen, long timeout)
1072 {
1073 	struct sock *sk = sock->sk;
1074 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1075 	struct tipc_sock *tsk = tipc_sk(sk);
1076 	struct tipc_group *grp = tsk->group;
1077 	struct tipc_msg *hdr = &tsk->phdr;
1078 	struct net *net = sock_net(sk);
1079 	u32 type, inst, scope, exclude;
1080 	struct list_head dsts;
1081 	u32 dstcnt;
1082 
1083 	INIT_LIST_HEAD(&dsts);
1084 
1085 	type = msg_nametype(hdr);
1086 	inst = dest->addr.name.name.instance;
1087 	scope = msg_lookup_scope(hdr);
1088 	exclude = tipc_group_exclude(grp);
1089 
1090 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1091 				 &dstcnt, exclude, true))
1092 		return -EHOSTUNREACH;
1093 
1094 	if (dstcnt == 1) {
1095 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1096 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1097 	}
1098 
1099 	tipc_dest_list_purge(&dsts);
1100 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1101 }
1102 
1103 /**
1104  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1105  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1106  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1107  *
1108  * Multi-threaded: parallel calls with reference to same queues may occur
1109  */
1110 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1111 		       struct sk_buff_head *inputq)
1112 {
1113 	u32 self = tipc_own_addr(net);
1114 	u32 type, lower, upper, scope;
1115 	struct sk_buff *skb, *_skb;
1116 	u32 portid, oport, onode;
1117 	struct sk_buff_head tmpq;
1118 	struct list_head dports;
1119 	struct tipc_msg *hdr;
1120 	int user, mtyp, hlen;
1121 	bool exact;
1122 
1123 	__skb_queue_head_init(&tmpq);
1124 	INIT_LIST_HEAD(&dports);
1125 
1126 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1127 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1128 		hdr = buf_msg(skb);
1129 		user = msg_user(hdr);
1130 		mtyp = msg_type(hdr);
1131 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1132 		oport = msg_origport(hdr);
1133 		onode = msg_orignode(hdr);
1134 		type = msg_nametype(hdr);
1135 
1136 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1137 			spin_lock_bh(&inputq->lock);
1138 			if (skb_peek(arrvq) == skb) {
1139 				__skb_dequeue(arrvq);
1140 				__skb_queue_tail(inputq, skb);
1141 			}
1142 			kfree_skb(skb);
1143 			spin_unlock_bh(&inputq->lock);
1144 			continue;
1145 		}
1146 
1147 		/* Group messages require exact scope match */
1148 		if (msg_in_group(hdr)) {
1149 			lower = 0;
1150 			upper = ~0;
1151 			scope = msg_lookup_scope(hdr);
1152 			exact = true;
1153 		} else {
1154 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1155 			if (onode == self)
1156 				scope = TIPC_NODE_SCOPE;
1157 			else
1158 				scope = TIPC_CLUSTER_SCOPE;
1159 			exact = false;
1160 			lower = msg_namelower(hdr);
1161 			upper = msg_nameupper(hdr);
1162 		}
1163 
1164 		/* Create destination port list: */
1165 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1166 				       scope, exact, &dports);
1167 
1168 		/* Clone message per destination */
1169 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1170 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1171 			if (_skb) {
1172 				msg_set_destport(buf_msg(_skb), portid);
1173 				__skb_queue_tail(&tmpq, _skb);
1174 				continue;
1175 			}
1176 			pr_warn("Failed to clone mcast rcv buffer\n");
1177 		}
1178 		/* Append to inputq if not already done by other thread */
1179 		spin_lock_bh(&inputq->lock);
1180 		if (skb_peek(arrvq) == skb) {
1181 			skb_queue_splice_tail_init(&tmpq, inputq);
1182 			kfree_skb(__skb_dequeue(arrvq));
1183 		}
1184 		spin_unlock_bh(&inputq->lock);
1185 		__skb_queue_purge(&tmpq);
1186 		kfree_skb(skb);
1187 	}
1188 	tipc_sk_rcv(net, inputq);
1189 }
1190 
1191 /**
1192  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1193  * @tsk: receiving socket
1194  * @skb: pointer to message buffer.
1195  */
1196 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1197 				   struct sk_buff_head *xmitq)
1198 {
1199 	struct tipc_msg *hdr = buf_msg(skb);
1200 	u32 onode = tsk_own_node(tsk);
1201 	struct sock *sk = &tsk->sk;
1202 	int mtyp = msg_type(hdr);
1203 	bool conn_cong;
1204 
1205 	/* Ignore if connection cannot be validated: */
1206 	if (!tsk_peer_msg(tsk, hdr))
1207 		goto exit;
1208 
1209 	if (unlikely(msg_errcode(hdr))) {
1210 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1211 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1212 				      tsk_peer_port(tsk));
1213 		sk->sk_state_change(sk);
1214 		goto exit;
1215 	}
1216 
1217 	tsk->probe_unacked = false;
1218 
1219 	if (mtyp == CONN_PROBE) {
1220 		msg_set_type(hdr, CONN_PROBE_REPLY);
1221 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1222 			__skb_queue_tail(xmitq, skb);
1223 		return;
1224 	} else if (mtyp == CONN_ACK) {
1225 		conn_cong = tsk_conn_cong(tsk);
1226 		tsk->snt_unacked -= msg_conn_ack(hdr);
1227 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1228 			tsk->snd_win = msg_adv_win(hdr);
1229 		if (conn_cong)
1230 			sk->sk_write_space(sk);
1231 	} else if (mtyp != CONN_PROBE_REPLY) {
1232 		pr_warn("Received unknown CONN_PROTO msg\n");
1233 	}
1234 exit:
1235 	kfree_skb(skb);
1236 }
1237 
1238 /**
1239  * tipc_sendmsg - send message in connectionless manner
1240  * @sock: socket structure
1241  * @m: message to send
1242  * @dsz: amount of user data to be sent
1243  *
1244  * Message must have an destination specified explicitly.
1245  * Used for SOCK_RDM and SOCK_DGRAM messages,
1246  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1247  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1248  *
1249  * Returns the number of bytes sent on success, or errno otherwise
1250  */
1251 static int tipc_sendmsg(struct socket *sock,
1252 			struct msghdr *m, size_t dsz)
1253 {
1254 	struct sock *sk = sock->sk;
1255 	int ret;
1256 
1257 	lock_sock(sk);
1258 	ret = __tipc_sendmsg(sock, m, dsz);
1259 	release_sock(sk);
1260 
1261 	return ret;
1262 }
1263 
1264 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1265 {
1266 	struct sock *sk = sock->sk;
1267 	struct net *net = sock_net(sk);
1268 	struct tipc_sock *tsk = tipc_sk(sk);
1269 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1270 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1271 	struct list_head *clinks = &tsk->cong_links;
1272 	bool syn = !tipc_sk_type_connectionless(sk);
1273 	struct tipc_group *grp = tsk->group;
1274 	struct tipc_msg *hdr = &tsk->phdr;
1275 	struct tipc_name_seq *seq;
1276 	struct sk_buff_head pkts;
1277 	u32 dport, dnode = 0;
1278 	u32 type, inst;
1279 	int mtu, rc;
1280 
1281 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1282 		return -EMSGSIZE;
1283 
1284 	if (likely(dest)) {
1285 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1286 			return -EINVAL;
1287 		if (unlikely(dest->family != AF_TIPC))
1288 			return -EINVAL;
1289 	}
1290 
1291 	if (grp) {
1292 		if (!dest)
1293 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1294 		if (dest->addrtype == TIPC_ADDR_NAME)
1295 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1296 		if (dest->addrtype == TIPC_ADDR_ID)
1297 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1298 		if (dest->addrtype == TIPC_ADDR_MCAST)
1299 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1300 		return -EINVAL;
1301 	}
1302 
1303 	if (unlikely(!dest)) {
1304 		dest = &tsk->peer;
1305 		if (!syn || dest->family != AF_TIPC)
1306 			return -EDESTADDRREQ;
1307 	}
1308 
1309 	if (unlikely(syn)) {
1310 		if (sk->sk_state == TIPC_LISTEN)
1311 			return -EPIPE;
1312 		if (sk->sk_state != TIPC_OPEN)
1313 			return -EISCONN;
1314 		if (tsk->published)
1315 			return -EOPNOTSUPP;
1316 		if (dest->addrtype == TIPC_ADDR_NAME) {
1317 			tsk->conn_type = dest->addr.name.name.type;
1318 			tsk->conn_instance = dest->addr.name.name.instance;
1319 		}
1320 	}
1321 
1322 	seq = &dest->addr.nameseq;
1323 	if (dest->addrtype == TIPC_ADDR_MCAST)
1324 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1325 
1326 	if (dest->addrtype == TIPC_ADDR_NAME) {
1327 		type = dest->addr.name.name.type;
1328 		inst = dest->addr.name.name.instance;
1329 		dnode = dest->addr.name.domain;
1330 		msg_set_type(hdr, TIPC_NAMED_MSG);
1331 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1332 		msg_set_nametype(hdr, type);
1333 		msg_set_nameinst(hdr, inst);
1334 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1335 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1336 		msg_set_destnode(hdr, dnode);
1337 		msg_set_destport(hdr, dport);
1338 		if (unlikely(!dport && !dnode))
1339 			return -EHOSTUNREACH;
1340 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1341 		dnode = dest->addr.id.node;
1342 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1343 		msg_set_lookup_scope(hdr, 0);
1344 		msg_set_destnode(hdr, dnode);
1345 		msg_set_destport(hdr, dest->addr.id.ref);
1346 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1347 	} else {
1348 		return -EINVAL;
1349 	}
1350 
1351 	/* Block or return if destination link is congested */
1352 	rc = tipc_wait_for_cond(sock, &timeout,
1353 				!tipc_dest_find(clinks, dnode, 0));
1354 	if (unlikely(rc))
1355 		return rc;
1356 
1357 	skb_queue_head_init(&pkts);
1358 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1359 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1360 	if (unlikely(rc != dlen))
1361 		return rc;
1362 
1363 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1364 	if (unlikely(rc == -ELINKCONG)) {
1365 		tipc_dest_push(clinks, dnode, 0);
1366 		tsk->cong_link_cnt++;
1367 		rc = 0;
1368 	}
1369 
1370 	if (unlikely(syn && !rc))
1371 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1372 
1373 	return rc ? rc : dlen;
1374 }
1375 
1376 /**
1377  * tipc_sendstream - send stream-oriented data
1378  * @sock: socket structure
1379  * @m: data to send
1380  * @dsz: total length of data to be transmitted
1381  *
1382  * Used for SOCK_STREAM data.
1383  *
1384  * Returns the number of bytes sent on success (or partial success),
1385  * or errno if no data sent
1386  */
1387 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1388 {
1389 	struct sock *sk = sock->sk;
1390 	int ret;
1391 
1392 	lock_sock(sk);
1393 	ret = __tipc_sendstream(sock, m, dsz);
1394 	release_sock(sk);
1395 
1396 	return ret;
1397 }
1398 
1399 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1400 {
1401 	struct sock *sk = sock->sk;
1402 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1403 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1404 	struct tipc_sock *tsk = tipc_sk(sk);
1405 	struct tipc_msg *hdr = &tsk->phdr;
1406 	struct net *net = sock_net(sk);
1407 	struct sk_buff_head pkts;
1408 	u32 dnode = tsk_peer_node(tsk);
1409 	int send, sent = 0;
1410 	int rc = 0;
1411 
1412 	skb_queue_head_init(&pkts);
1413 
1414 	if (unlikely(dlen > INT_MAX))
1415 		return -EMSGSIZE;
1416 
1417 	/* Handle implicit connection setup */
1418 	if (unlikely(dest)) {
1419 		rc = __tipc_sendmsg(sock, m, dlen);
1420 		if (dlen && (dlen == rc))
1421 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1422 		return rc;
1423 	}
1424 
1425 	do {
1426 		rc = tipc_wait_for_cond(sock, &timeout,
1427 					(!tsk->cong_link_cnt &&
1428 					 !tsk_conn_cong(tsk) &&
1429 					 tipc_sk_connected(sk)));
1430 		if (unlikely(rc))
1431 			break;
1432 
1433 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1434 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1435 		if (unlikely(rc != send))
1436 			break;
1437 
1438 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1439 		if (unlikely(rc == -ELINKCONG)) {
1440 			tsk->cong_link_cnt = 1;
1441 			rc = 0;
1442 		}
1443 		if (likely(!rc)) {
1444 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1445 			sent += send;
1446 		}
1447 	} while (sent < dlen && !rc);
1448 
1449 	return sent ? sent : rc;
1450 }
1451 
1452 /**
1453  * tipc_send_packet - send a connection-oriented message
1454  * @sock: socket structure
1455  * @m: message to send
1456  * @dsz: length of data to be transmitted
1457  *
1458  * Used for SOCK_SEQPACKET messages.
1459  *
1460  * Returns the number of bytes sent on success, or errno otherwise
1461  */
1462 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1463 {
1464 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1465 		return -EMSGSIZE;
1466 
1467 	return tipc_sendstream(sock, m, dsz);
1468 }
1469 
1470 /* tipc_sk_finish_conn - complete the setup of a connection
1471  */
1472 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1473 				u32 peer_node)
1474 {
1475 	struct sock *sk = &tsk->sk;
1476 	struct net *net = sock_net(sk);
1477 	struct tipc_msg *msg = &tsk->phdr;
1478 
1479 	msg_set_destnode(msg, peer_node);
1480 	msg_set_destport(msg, peer_port);
1481 	msg_set_type(msg, TIPC_CONN_MSG);
1482 	msg_set_lookup_scope(msg, 0);
1483 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1484 
1485 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1486 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1487 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1488 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1489 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1490 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1491 		return;
1492 
1493 	/* Fall back to message based flow control */
1494 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1495 	tsk->snd_win = FLOWCTL_MSG_WIN;
1496 }
1497 
1498 /**
1499  * tipc_sk_set_orig_addr - capture sender's address for received message
1500  * @m: descriptor for message info
1501  * @hdr: received message header
1502  *
1503  * Note: Address is not captured if not requested by receiver.
1504  */
1505 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1506 {
1507 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1508 	struct tipc_msg *hdr = buf_msg(skb);
1509 
1510 	if (!srcaddr)
1511 		return;
1512 
1513 	srcaddr->sock.family = AF_TIPC;
1514 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1515 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1516 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1517 	srcaddr->sock.addr.name.domain = 0;
1518 	srcaddr->sock.scope = 0;
1519 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1520 
1521 	if (!msg_in_group(hdr))
1522 		return;
1523 
1524 	/* Group message users may also want to know sending member's id */
1525 	srcaddr->member.family = AF_TIPC;
1526 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1527 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1528 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1529 	srcaddr->member.addr.name.domain = 0;
1530 	m->msg_namelen = sizeof(*srcaddr);
1531 }
1532 
1533 /**
1534  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1535  * @m: descriptor for message info
1536  * @msg: received message header
1537  * @tsk: TIPC port associated with message
1538  *
1539  * Note: Ancillary data is not captured if not requested by receiver.
1540  *
1541  * Returns 0 if successful, otherwise errno
1542  */
1543 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1544 				 struct tipc_sock *tsk)
1545 {
1546 	u32 anc_data[3];
1547 	u32 err;
1548 	u32 dest_type;
1549 	int has_name;
1550 	int res;
1551 
1552 	if (likely(m->msg_controllen == 0))
1553 		return 0;
1554 
1555 	/* Optionally capture errored message object(s) */
1556 	err = msg ? msg_errcode(msg) : 0;
1557 	if (unlikely(err)) {
1558 		anc_data[0] = err;
1559 		anc_data[1] = msg_data_sz(msg);
1560 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1561 		if (res)
1562 			return res;
1563 		if (anc_data[1]) {
1564 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1565 				       msg_data(msg));
1566 			if (res)
1567 				return res;
1568 		}
1569 	}
1570 
1571 	/* Optionally capture message destination object */
1572 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1573 	switch (dest_type) {
1574 	case TIPC_NAMED_MSG:
1575 		has_name = 1;
1576 		anc_data[0] = msg_nametype(msg);
1577 		anc_data[1] = msg_namelower(msg);
1578 		anc_data[2] = msg_namelower(msg);
1579 		break;
1580 	case TIPC_MCAST_MSG:
1581 		has_name = 1;
1582 		anc_data[0] = msg_nametype(msg);
1583 		anc_data[1] = msg_namelower(msg);
1584 		anc_data[2] = msg_nameupper(msg);
1585 		break;
1586 	case TIPC_CONN_MSG:
1587 		has_name = (tsk->conn_type != 0);
1588 		anc_data[0] = tsk->conn_type;
1589 		anc_data[1] = tsk->conn_instance;
1590 		anc_data[2] = tsk->conn_instance;
1591 		break;
1592 	default:
1593 		has_name = 0;
1594 	}
1595 	if (has_name) {
1596 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1597 		if (res)
1598 			return res;
1599 	}
1600 
1601 	return 0;
1602 }
1603 
1604 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1605 {
1606 	struct sock *sk = &tsk->sk;
1607 	struct net *net = sock_net(sk);
1608 	struct sk_buff *skb = NULL;
1609 	struct tipc_msg *msg;
1610 	u32 peer_port = tsk_peer_port(tsk);
1611 	u32 dnode = tsk_peer_node(tsk);
1612 
1613 	if (!tipc_sk_connected(sk))
1614 		return;
1615 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1616 			      dnode, tsk_own_node(tsk), peer_port,
1617 			      tsk->portid, TIPC_OK);
1618 	if (!skb)
1619 		return;
1620 	msg = buf_msg(skb);
1621 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1622 	tsk->rcv_unacked = 0;
1623 
1624 	/* Adjust to and advertize the correct window limit */
1625 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1626 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1627 		msg_set_adv_win(msg, tsk->rcv_win);
1628 	}
1629 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1630 }
1631 
1632 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1633 {
1634 	struct sock *sk = sock->sk;
1635 	DEFINE_WAIT(wait);
1636 	long timeo = *timeop;
1637 	int err = sock_error(sk);
1638 
1639 	if (err)
1640 		return err;
1641 
1642 	for (;;) {
1643 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1644 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1645 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1646 				err = -ENOTCONN;
1647 				break;
1648 			}
1649 			release_sock(sk);
1650 			timeo = schedule_timeout(timeo);
1651 			lock_sock(sk);
1652 		}
1653 		err = 0;
1654 		if (!skb_queue_empty(&sk->sk_receive_queue))
1655 			break;
1656 		err = -EAGAIN;
1657 		if (!timeo)
1658 			break;
1659 		err = sock_intr_errno(timeo);
1660 		if (signal_pending(current))
1661 			break;
1662 
1663 		err = sock_error(sk);
1664 		if (err)
1665 			break;
1666 	}
1667 	finish_wait(sk_sleep(sk), &wait);
1668 	*timeop = timeo;
1669 	return err;
1670 }
1671 
1672 /**
1673  * tipc_recvmsg - receive packet-oriented message
1674  * @m: descriptor for message info
1675  * @buflen: length of user buffer area
1676  * @flags: receive flags
1677  *
1678  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1679  * If the complete message doesn't fit in user area, truncate it.
1680  *
1681  * Returns size of returned message data, errno otherwise
1682  */
1683 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1684 			size_t buflen,	int flags)
1685 {
1686 	struct sock *sk = sock->sk;
1687 	bool connected = !tipc_sk_type_connectionless(sk);
1688 	struct tipc_sock *tsk = tipc_sk(sk);
1689 	int rc, err, hlen, dlen, copy;
1690 	struct sk_buff_head xmitq;
1691 	struct tipc_msg *hdr;
1692 	struct sk_buff *skb;
1693 	bool grp_evt;
1694 	long timeout;
1695 
1696 	/* Catch invalid receive requests */
1697 	if (unlikely(!buflen))
1698 		return -EINVAL;
1699 
1700 	lock_sock(sk);
1701 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1702 		rc = -ENOTCONN;
1703 		goto exit;
1704 	}
1705 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1706 
1707 	/* Step rcv queue to first msg with data or error; wait if necessary */
1708 	do {
1709 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1710 		if (unlikely(rc))
1711 			goto exit;
1712 		skb = skb_peek(&sk->sk_receive_queue);
1713 		hdr = buf_msg(skb);
1714 		dlen = msg_data_sz(hdr);
1715 		hlen = msg_hdr_sz(hdr);
1716 		err = msg_errcode(hdr);
1717 		grp_evt = msg_is_grp_evt(hdr);
1718 		if (likely(dlen || err))
1719 			break;
1720 		tsk_advance_rx_queue(sk);
1721 	} while (1);
1722 
1723 	/* Collect msg meta data, including error code and rejected data */
1724 	tipc_sk_set_orig_addr(m, skb);
1725 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1726 	if (unlikely(rc))
1727 		goto exit;
1728 
1729 	/* Capture data if non-error msg, otherwise just set return value */
1730 	if (likely(!err)) {
1731 		copy = min_t(int, dlen, buflen);
1732 		if (unlikely(copy != dlen))
1733 			m->msg_flags |= MSG_TRUNC;
1734 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1735 	} else {
1736 		copy = 0;
1737 		rc = 0;
1738 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1739 			rc = -ECONNRESET;
1740 	}
1741 	if (unlikely(rc))
1742 		goto exit;
1743 
1744 	/* Mark message as group event if applicable */
1745 	if (unlikely(grp_evt)) {
1746 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1747 			m->msg_flags |= MSG_EOR;
1748 		m->msg_flags |= MSG_OOB;
1749 		copy = 0;
1750 	}
1751 
1752 	/* Caption of data or error code/rejected data was successful */
1753 	if (unlikely(flags & MSG_PEEK))
1754 		goto exit;
1755 
1756 	/* Send group flow control advertisement when applicable */
1757 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1758 		skb_queue_head_init(&xmitq);
1759 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1760 					  msg_orignode(hdr), msg_origport(hdr),
1761 					  &xmitq);
1762 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1763 	}
1764 
1765 	tsk_advance_rx_queue(sk);
1766 
1767 	if (likely(!connected))
1768 		goto exit;
1769 
1770 	/* Send connection flow control advertisement when applicable */
1771 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1772 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1773 		tipc_sk_send_ack(tsk);
1774 exit:
1775 	release_sock(sk);
1776 	return rc ? rc : copy;
1777 }
1778 
1779 /**
1780  * tipc_recvstream - receive stream-oriented data
1781  * @m: descriptor for message info
1782  * @buflen: total size of user buffer area
1783  * @flags: receive flags
1784  *
1785  * Used for SOCK_STREAM messages only.  If not enough data is available
1786  * will optionally wait for more; never truncates data.
1787  *
1788  * Returns size of returned message data, errno otherwise
1789  */
1790 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1791 			   size_t buflen, int flags)
1792 {
1793 	struct sock *sk = sock->sk;
1794 	struct tipc_sock *tsk = tipc_sk(sk);
1795 	struct sk_buff *skb;
1796 	struct tipc_msg *hdr;
1797 	struct tipc_skb_cb *skb_cb;
1798 	bool peek = flags & MSG_PEEK;
1799 	int offset, required, copy, copied = 0;
1800 	int hlen, dlen, err, rc;
1801 	long timeout;
1802 
1803 	/* Catch invalid receive attempts */
1804 	if (unlikely(!buflen))
1805 		return -EINVAL;
1806 
1807 	lock_sock(sk);
1808 
1809 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1810 		rc = -ENOTCONN;
1811 		goto exit;
1812 	}
1813 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1814 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1815 
1816 	do {
1817 		/* Look at first msg in receive queue; wait if necessary */
1818 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1819 		if (unlikely(rc))
1820 			break;
1821 		skb = skb_peek(&sk->sk_receive_queue);
1822 		skb_cb = TIPC_SKB_CB(skb);
1823 		hdr = buf_msg(skb);
1824 		dlen = msg_data_sz(hdr);
1825 		hlen = msg_hdr_sz(hdr);
1826 		err = msg_errcode(hdr);
1827 
1828 		/* Discard any empty non-errored (SYN-) message */
1829 		if (unlikely(!dlen && !err)) {
1830 			tsk_advance_rx_queue(sk);
1831 			continue;
1832 		}
1833 
1834 		/* Collect msg meta data, incl. error code and rejected data */
1835 		if (!copied) {
1836 			tipc_sk_set_orig_addr(m, skb);
1837 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1838 			if (rc)
1839 				break;
1840 		}
1841 
1842 		/* Copy data if msg ok, otherwise return error/partial data */
1843 		if (likely(!err)) {
1844 			offset = skb_cb->bytes_read;
1845 			copy = min_t(int, dlen - offset, buflen - copied);
1846 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1847 			if (unlikely(rc))
1848 				break;
1849 			copied += copy;
1850 			offset += copy;
1851 			if (unlikely(offset < dlen)) {
1852 				if (!peek)
1853 					skb_cb->bytes_read = offset;
1854 				break;
1855 			}
1856 		} else {
1857 			rc = 0;
1858 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1859 				rc = -ECONNRESET;
1860 			if (copied || rc)
1861 				break;
1862 		}
1863 
1864 		if (unlikely(peek))
1865 			break;
1866 
1867 		tsk_advance_rx_queue(sk);
1868 
1869 		/* Send connection flow control advertisement when applicable */
1870 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1871 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1872 			tipc_sk_send_ack(tsk);
1873 
1874 		/* Exit if all requested data or FIN/error received */
1875 		if (copied == buflen || err)
1876 			break;
1877 
1878 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1879 exit:
1880 	release_sock(sk);
1881 	return copied ? copied : rc;
1882 }
1883 
1884 /**
1885  * tipc_write_space - wake up thread if port congestion is released
1886  * @sk: socket
1887  */
1888 static void tipc_write_space(struct sock *sk)
1889 {
1890 	struct socket_wq *wq;
1891 
1892 	rcu_read_lock();
1893 	wq = rcu_dereference(sk->sk_wq);
1894 	if (skwq_has_sleeper(wq))
1895 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1896 						EPOLLWRNORM | EPOLLWRBAND);
1897 	rcu_read_unlock();
1898 }
1899 
1900 /**
1901  * tipc_data_ready - wake up threads to indicate messages have been received
1902  * @sk: socket
1903  * @len: the length of messages
1904  */
1905 static void tipc_data_ready(struct sock *sk)
1906 {
1907 	struct socket_wq *wq;
1908 
1909 	rcu_read_lock();
1910 	wq = rcu_dereference(sk->sk_wq);
1911 	if (skwq_has_sleeper(wq))
1912 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1913 						EPOLLRDNORM | EPOLLRDBAND);
1914 	rcu_read_unlock();
1915 }
1916 
1917 static void tipc_sock_destruct(struct sock *sk)
1918 {
1919 	__skb_queue_purge(&sk->sk_receive_queue);
1920 }
1921 
1922 static void tipc_sk_proto_rcv(struct sock *sk,
1923 			      struct sk_buff_head *inputq,
1924 			      struct sk_buff_head *xmitq)
1925 {
1926 	struct sk_buff *skb = __skb_dequeue(inputq);
1927 	struct tipc_sock *tsk = tipc_sk(sk);
1928 	struct tipc_msg *hdr = buf_msg(skb);
1929 	struct tipc_group *grp = tsk->group;
1930 	bool wakeup = false;
1931 
1932 	switch (msg_user(hdr)) {
1933 	case CONN_MANAGER:
1934 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1935 		return;
1936 	case SOCK_WAKEUP:
1937 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1938 		tsk->cong_link_cnt--;
1939 		wakeup = true;
1940 		break;
1941 	case GROUP_PROTOCOL:
1942 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1943 		break;
1944 	case TOP_SRV:
1945 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1946 				      hdr, inputq, xmitq);
1947 		break;
1948 	default:
1949 		break;
1950 	}
1951 
1952 	if (wakeup)
1953 		sk->sk_write_space(sk);
1954 
1955 	kfree_skb(skb);
1956 }
1957 
1958 /**
1959  * tipc_filter_connect - Handle incoming message for a connection-based socket
1960  * @tsk: TIPC socket
1961  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1962  *
1963  * Returns true if everything ok, false otherwise
1964  */
1965 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1966 {
1967 	struct sock *sk = &tsk->sk;
1968 	struct net *net = sock_net(sk);
1969 	struct tipc_msg *hdr = buf_msg(skb);
1970 	u32 pport = msg_origport(hdr);
1971 	u32 pnode = msg_orignode(hdr);
1972 
1973 	if (unlikely(msg_mcast(hdr)))
1974 		return false;
1975 
1976 	switch (sk->sk_state) {
1977 	case TIPC_CONNECTING:
1978 		/* Accept only ACK or NACK message */
1979 		if (unlikely(!msg_connected(hdr))) {
1980 			if (pport != tsk_peer_port(tsk) ||
1981 			    pnode != tsk_peer_node(tsk))
1982 				return false;
1983 
1984 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1985 			sk->sk_err = ECONNREFUSED;
1986 			sk->sk_state_change(sk);
1987 			return true;
1988 		}
1989 
1990 		if (unlikely(msg_errcode(hdr))) {
1991 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1992 			sk->sk_err = ECONNREFUSED;
1993 			sk->sk_state_change(sk);
1994 			return true;
1995 		}
1996 
1997 		if (unlikely(!msg_isdata(hdr))) {
1998 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1999 			sk->sk_err = EINVAL;
2000 			sk->sk_state_change(sk);
2001 			return true;
2002 		}
2003 
2004 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2005 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2006 
2007 		/* If 'ACK+' message, add to socket receive queue */
2008 		if (msg_data_sz(hdr))
2009 			return true;
2010 
2011 		/* If empty 'ACK-' message, wake up sleeping connect() */
2012 		sk->sk_data_ready(sk);
2013 
2014 		/* 'ACK-' message is neither accepted nor rejected: */
2015 		msg_set_dest_droppable(hdr, 1);
2016 		return false;
2017 
2018 	case TIPC_OPEN:
2019 	case TIPC_DISCONNECTING:
2020 		break;
2021 	case TIPC_LISTEN:
2022 		/* Accept only SYN message */
2023 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2024 			return true;
2025 		break;
2026 	case TIPC_ESTABLISHED:
2027 		/* Accept only connection-based messages sent by peer */
2028 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2029 			return false;
2030 
2031 		if (unlikely(msg_errcode(hdr))) {
2032 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2033 			/* Let timer expire on it's own */
2034 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2035 					      tsk->portid);
2036 			sk->sk_state_change(sk);
2037 		}
2038 		return true;
2039 	default:
2040 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2041 	}
2042 
2043 	return false;
2044 }
2045 
2046 /**
2047  * rcvbuf_limit - get proper overload limit of socket receive queue
2048  * @sk: socket
2049  * @skb: message
2050  *
2051  * For connection oriented messages, irrespective of importance,
2052  * default queue limit is 2 MB.
2053  *
2054  * For connectionless messages, queue limits are based on message
2055  * importance as follows:
2056  *
2057  * TIPC_LOW_IMPORTANCE       (2 MB)
2058  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2059  * TIPC_HIGH_IMPORTANCE      (8 MB)
2060  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2061  *
2062  * Returns overload limit according to corresponding message importance
2063  */
2064 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2065 {
2066 	struct tipc_sock *tsk = tipc_sk(sk);
2067 	struct tipc_msg *hdr = buf_msg(skb);
2068 
2069 	if (unlikely(msg_in_group(hdr)))
2070 		return sk->sk_rcvbuf;
2071 
2072 	if (unlikely(!msg_connected(hdr)))
2073 		return sk->sk_rcvbuf << msg_importance(hdr);
2074 
2075 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2076 		return sk->sk_rcvbuf;
2077 
2078 	return FLOWCTL_MSG_LIM;
2079 }
2080 
2081 /**
2082  * tipc_sk_filter_rcv - validate incoming message
2083  * @sk: socket
2084  * @skb: pointer to message.
2085  *
2086  * Enqueues message on receive queue if acceptable; optionally handles
2087  * disconnect indication for a connected socket.
2088  *
2089  * Called with socket lock already taken
2090  *
2091  */
2092 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2093 			       struct sk_buff_head *xmitq)
2094 {
2095 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2096 	struct tipc_sock *tsk = tipc_sk(sk);
2097 	struct tipc_group *grp = tsk->group;
2098 	struct tipc_msg *hdr = buf_msg(skb);
2099 	struct net *net = sock_net(sk);
2100 	struct sk_buff_head inputq;
2101 	int limit, err = TIPC_OK;
2102 
2103 	TIPC_SKB_CB(skb)->bytes_read = 0;
2104 	__skb_queue_head_init(&inputq);
2105 	__skb_queue_tail(&inputq, skb);
2106 
2107 	if (unlikely(!msg_isdata(hdr)))
2108 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2109 
2110 	if (unlikely(grp))
2111 		tipc_group_filter_msg(grp, &inputq, xmitq);
2112 
2113 	/* Validate and add to receive buffer if there is space */
2114 	while ((skb = __skb_dequeue(&inputq))) {
2115 		hdr = buf_msg(skb);
2116 		limit = rcvbuf_limit(sk, skb);
2117 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2118 		    (!sk_conn && msg_connected(hdr)) ||
2119 		    (!grp && msg_in_group(hdr)))
2120 			err = TIPC_ERR_NO_PORT;
2121 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2122 			atomic_inc(&sk->sk_drops);
2123 			err = TIPC_ERR_OVERLOAD;
2124 		}
2125 
2126 		if (unlikely(err)) {
2127 			tipc_skb_reject(net, err, skb, xmitq);
2128 			err = TIPC_OK;
2129 			continue;
2130 		}
2131 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2132 		skb_set_owner_r(skb, sk);
2133 		sk->sk_data_ready(sk);
2134 	}
2135 }
2136 
2137 /**
2138  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2139  * @sk: socket
2140  * @skb: message
2141  *
2142  * Caller must hold socket lock
2143  */
2144 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2145 {
2146 	unsigned int before = sk_rmem_alloc_get(sk);
2147 	struct sk_buff_head xmitq;
2148 	unsigned int added;
2149 
2150 	__skb_queue_head_init(&xmitq);
2151 
2152 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2153 	added = sk_rmem_alloc_get(sk) - before;
2154 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2155 
2156 	/* Send pending response/rejected messages, if any */
2157 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2158 	return 0;
2159 }
2160 
2161 /**
2162  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2163  *                   inputq and try adding them to socket or backlog queue
2164  * @inputq: list of incoming buffers with potentially different destinations
2165  * @sk: socket where the buffers should be enqueued
2166  * @dport: port number for the socket
2167  *
2168  * Caller must hold socket lock
2169  */
2170 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2171 			    u32 dport, struct sk_buff_head *xmitq)
2172 {
2173 	unsigned long time_limit = jiffies + 2;
2174 	struct sk_buff *skb;
2175 	unsigned int lim;
2176 	atomic_t *dcnt;
2177 	u32 onode;
2178 
2179 	while (skb_queue_len(inputq)) {
2180 		if (unlikely(time_after_eq(jiffies, time_limit)))
2181 			return;
2182 
2183 		skb = tipc_skb_dequeue(inputq, dport);
2184 		if (unlikely(!skb))
2185 			return;
2186 
2187 		/* Add message directly to receive queue if possible */
2188 		if (!sock_owned_by_user(sk)) {
2189 			tipc_sk_filter_rcv(sk, skb, xmitq);
2190 			continue;
2191 		}
2192 
2193 		/* Try backlog, compensating for double-counted bytes */
2194 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2195 		if (!sk->sk_backlog.len)
2196 			atomic_set(dcnt, 0);
2197 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2198 		if (likely(!sk_add_backlog(sk, skb, lim)))
2199 			continue;
2200 
2201 		/* Overload => reject message back to sender */
2202 		onode = tipc_own_addr(sock_net(sk));
2203 		atomic_inc(&sk->sk_drops);
2204 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2205 			__skb_queue_tail(xmitq, skb);
2206 		break;
2207 	}
2208 }
2209 
2210 /**
2211  * tipc_sk_rcv - handle a chain of incoming buffers
2212  * @inputq: buffer list containing the buffers
2213  * Consumes all buffers in list until inputq is empty
2214  * Note: may be called in multiple threads referring to the same queue
2215  */
2216 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2217 {
2218 	struct sk_buff_head xmitq;
2219 	u32 dnode, dport = 0;
2220 	int err;
2221 	struct tipc_sock *tsk;
2222 	struct sock *sk;
2223 	struct sk_buff *skb;
2224 
2225 	__skb_queue_head_init(&xmitq);
2226 	while (skb_queue_len(inputq)) {
2227 		dport = tipc_skb_peek_port(inputq, dport);
2228 		tsk = tipc_sk_lookup(net, dport);
2229 
2230 		if (likely(tsk)) {
2231 			sk = &tsk->sk;
2232 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2233 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2234 				spin_unlock_bh(&sk->sk_lock.slock);
2235 			}
2236 			/* Send pending response/rejected messages, if any */
2237 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2238 			sock_put(sk);
2239 			continue;
2240 		}
2241 		/* No destination socket => dequeue skb if still there */
2242 		skb = tipc_skb_dequeue(inputq, dport);
2243 		if (!skb)
2244 			return;
2245 
2246 		/* Try secondary lookup if unresolved named message */
2247 		err = TIPC_ERR_NO_PORT;
2248 		if (tipc_msg_lookup_dest(net, skb, &err))
2249 			goto xmit;
2250 
2251 		/* Prepare for message rejection */
2252 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2253 			continue;
2254 xmit:
2255 		dnode = msg_destnode(buf_msg(skb));
2256 		tipc_node_xmit_skb(net, skb, dnode, dport);
2257 	}
2258 }
2259 
2260 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2261 {
2262 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2263 	struct sock *sk = sock->sk;
2264 	int done;
2265 
2266 	do {
2267 		int err = sock_error(sk);
2268 		if (err)
2269 			return err;
2270 		if (!*timeo_p)
2271 			return -ETIMEDOUT;
2272 		if (signal_pending(current))
2273 			return sock_intr_errno(*timeo_p);
2274 
2275 		add_wait_queue(sk_sleep(sk), &wait);
2276 		done = sk_wait_event(sk, timeo_p,
2277 				     sk->sk_state != TIPC_CONNECTING, &wait);
2278 		remove_wait_queue(sk_sleep(sk), &wait);
2279 	} while (!done);
2280 	return 0;
2281 }
2282 
2283 /**
2284  * tipc_connect - establish a connection to another TIPC port
2285  * @sock: socket structure
2286  * @dest: socket address for destination port
2287  * @destlen: size of socket address data structure
2288  * @flags: file-related flags associated with socket
2289  *
2290  * Returns 0 on success, errno otherwise
2291  */
2292 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2293 			int destlen, int flags)
2294 {
2295 	struct sock *sk = sock->sk;
2296 	struct tipc_sock *tsk = tipc_sk(sk);
2297 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2298 	struct msghdr m = {NULL,};
2299 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2300 	int previous;
2301 	int res = 0;
2302 
2303 	if (destlen != sizeof(struct sockaddr_tipc))
2304 		return -EINVAL;
2305 
2306 	lock_sock(sk);
2307 
2308 	if (tsk->group) {
2309 		res = -EINVAL;
2310 		goto exit;
2311 	}
2312 
2313 	if (dst->family == AF_UNSPEC) {
2314 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2315 		if (!tipc_sk_type_connectionless(sk))
2316 			res = -EINVAL;
2317 		goto exit;
2318 	} else if (dst->family != AF_TIPC) {
2319 		res = -EINVAL;
2320 	}
2321 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2322 		res = -EINVAL;
2323 	if (res)
2324 		goto exit;
2325 
2326 	/* DGRAM/RDM connect(), just save the destaddr */
2327 	if (tipc_sk_type_connectionless(sk)) {
2328 		memcpy(&tsk->peer, dest, destlen);
2329 		goto exit;
2330 	}
2331 
2332 	previous = sk->sk_state;
2333 
2334 	switch (sk->sk_state) {
2335 	case TIPC_OPEN:
2336 		/* Send a 'SYN-' to destination */
2337 		m.msg_name = dest;
2338 		m.msg_namelen = destlen;
2339 
2340 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2341 		 * indicate send_msg() is never blocked.
2342 		 */
2343 		if (!timeout)
2344 			m.msg_flags = MSG_DONTWAIT;
2345 
2346 		res = __tipc_sendmsg(sock, &m, 0);
2347 		if ((res < 0) && (res != -EWOULDBLOCK))
2348 			goto exit;
2349 
2350 		/* Just entered TIPC_CONNECTING state; the only
2351 		 * difference is that return value in non-blocking
2352 		 * case is EINPROGRESS, rather than EALREADY.
2353 		 */
2354 		res = -EINPROGRESS;
2355 		/* fall thru' */
2356 	case TIPC_CONNECTING:
2357 		if (!timeout) {
2358 			if (previous == TIPC_CONNECTING)
2359 				res = -EALREADY;
2360 			goto exit;
2361 		}
2362 		timeout = msecs_to_jiffies(timeout);
2363 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2364 		res = tipc_wait_for_connect(sock, &timeout);
2365 		break;
2366 	case TIPC_ESTABLISHED:
2367 		res = -EISCONN;
2368 		break;
2369 	default:
2370 		res = -EINVAL;
2371 	}
2372 
2373 exit:
2374 	release_sock(sk);
2375 	return res;
2376 }
2377 
2378 /**
2379  * tipc_listen - allow socket to listen for incoming connections
2380  * @sock: socket structure
2381  * @len: (unused)
2382  *
2383  * Returns 0 on success, errno otherwise
2384  */
2385 static int tipc_listen(struct socket *sock, int len)
2386 {
2387 	struct sock *sk = sock->sk;
2388 	int res;
2389 
2390 	lock_sock(sk);
2391 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2392 	release_sock(sk);
2393 
2394 	return res;
2395 }
2396 
2397 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2398 {
2399 	struct sock *sk = sock->sk;
2400 	DEFINE_WAIT(wait);
2401 	int err;
2402 
2403 	/* True wake-one mechanism for incoming connections: only
2404 	 * one process gets woken up, not the 'whole herd'.
2405 	 * Since we do not 'race & poll' for established sockets
2406 	 * anymore, the common case will execute the loop only once.
2407 	*/
2408 	for (;;) {
2409 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2410 					  TASK_INTERRUPTIBLE);
2411 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2412 			release_sock(sk);
2413 			timeo = schedule_timeout(timeo);
2414 			lock_sock(sk);
2415 		}
2416 		err = 0;
2417 		if (!skb_queue_empty(&sk->sk_receive_queue))
2418 			break;
2419 		err = -EAGAIN;
2420 		if (!timeo)
2421 			break;
2422 		err = sock_intr_errno(timeo);
2423 		if (signal_pending(current))
2424 			break;
2425 	}
2426 	finish_wait(sk_sleep(sk), &wait);
2427 	return err;
2428 }
2429 
2430 /**
2431  * tipc_accept - wait for connection request
2432  * @sock: listening socket
2433  * @newsock: new socket that is to be connected
2434  * @flags: file-related flags associated with socket
2435  *
2436  * Returns 0 on success, errno otherwise
2437  */
2438 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2439 		       bool kern)
2440 {
2441 	struct sock *new_sk, *sk = sock->sk;
2442 	struct sk_buff *buf;
2443 	struct tipc_sock *new_tsock;
2444 	struct tipc_msg *msg;
2445 	long timeo;
2446 	int res;
2447 
2448 	lock_sock(sk);
2449 
2450 	if (sk->sk_state != TIPC_LISTEN) {
2451 		res = -EINVAL;
2452 		goto exit;
2453 	}
2454 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2455 	res = tipc_wait_for_accept(sock, timeo);
2456 	if (res)
2457 		goto exit;
2458 
2459 	buf = skb_peek(&sk->sk_receive_queue);
2460 
2461 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2462 	if (res)
2463 		goto exit;
2464 	security_sk_clone(sock->sk, new_sock->sk);
2465 
2466 	new_sk = new_sock->sk;
2467 	new_tsock = tipc_sk(new_sk);
2468 	msg = buf_msg(buf);
2469 
2470 	/* we lock on new_sk; but lockdep sees the lock on sk */
2471 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2472 
2473 	/*
2474 	 * Reject any stray messages received by new socket
2475 	 * before the socket lock was taken (very, very unlikely)
2476 	 */
2477 	tsk_rej_rx_queue(new_sk);
2478 
2479 	/* Connect new socket to it's peer */
2480 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2481 
2482 	tsk_set_importance(new_tsock, msg_importance(msg));
2483 	if (msg_named(msg)) {
2484 		new_tsock->conn_type = msg_nametype(msg);
2485 		new_tsock->conn_instance = msg_nameinst(msg);
2486 	}
2487 
2488 	/*
2489 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2490 	 * Respond to 'SYN+' by queuing it on new socket.
2491 	 */
2492 	if (!msg_data_sz(msg)) {
2493 		struct msghdr m = {NULL,};
2494 
2495 		tsk_advance_rx_queue(sk);
2496 		__tipc_sendstream(new_sock, &m, 0);
2497 	} else {
2498 		__skb_dequeue(&sk->sk_receive_queue);
2499 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2500 		skb_set_owner_r(buf, new_sk);
2501 	}
2502 	release_sock(new_sk);
2503 exit:
2504 	release_sock(sk);
2505 	return res;
2506 }
2507 
2508 /**
2509  * tipc_shutdown - shutdown socket connection
2510  * @sock: socket structure
2511  * @how: direction to close (must be SHUT_RDWR)
2512  *
2513  * Terminates connection (if necessary), then purges socket's receive queue.
2514  *
2515  * Returns 0 on success, errno otherwise
2516  */
2517 static int tipc_shutdown(struct socket *sock, int how)
2518 {
2519 	struct sock *sk = sock->sk;
2520 	int res;
2521 
2522 	if (how != SHUT_RDWR)
2523 		return -EINVAL;
2524 
2525 	lock_sock(sk);
2526 
2527 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2528 	sk->sk_shutdown = SEND_SHUTDOWN;
2529 
2530 	if (sk->sk_state == TIPC_DISCONNECTING) {
2531 		/* Discard any unreceived messages */
2532 		__skb_queue_purge(&sk->sk_receive_queue);
2533 
2534 		/* Wake up anyone sleeping in poll */
2535 		sk->sk_state_change(sk);
2536 		res = 0;
2537 	} else {
2538 		res = -ENOTCONN;
2539 	}
2540 
2541 	release_sock(sk);
2542 	return res;
2543 }
2544 
2545 static void tipc_sk_timeout(struct timer_list *t)
2546 {
2547 	struct sock *sk = from_timer(sk, t, sk_timer);
2548 	struct tipc_sock *tsk = tipc_sk(sk);
2549 	u32 peer_port = tsk_peer_port(tsk);
2550 	u32 peer_node = tsk_peer_node(tsk);
2551 	u32 own_node = tsk_own_node(tsk);
2552 	u32 own_port = tsk->portid;
2553 	struct net *net = sock_net(sk);
2554 	struct sk_buff *skb = NULL;
2555 
2556 	bh_lock_sock(sk);
2557 	if (!tipc_sk_connected(sk))
2558 		goto exit;
2559 
2560 	/* Try again later if socket is busy */
2561 	if (sock_owned_by_user(sk)) {
2562 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2563 		goto exit;
2564 	}
2565 
2566 	if (tsk->probe_unacked) {
2567 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2568 		tipc_node_remove_conn(net, peer_node, peer_port);
2569 		sk->sk_state_change(sk);
2570 		goto exit;
2571 	}
2572 	/* Send new probe */
2573 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2574 			      peer_node, own_node, peer_port, own_port,
2575 			      TIPC_OK);
2576 	tsk->probe_unacked = true;
2577 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2578 exit:
2579 	bh_unlock_sock(sk);
2580 	if (skb)
2581 		tipc_node_xmit_skb(net, skb, peer_node, own_port);
2582 	sock_put(sk);
2583 }
2584 
2585 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2586 			   struct tipc_name_seq const *seq)
2587 {
2588 	struct sock *sk = &tsk->sk;
2589 	struct net *net = sock_net(sk);
2590 	struct publication *publ;
2591 	u32 key;
2592 
2593 	if (scope != TIPC_NODE_SCOPE)
2594 		scope = TIPC_CLUSTER_SCOPE;
2595 
2596 	if (tipc_sk_connected(sk))
2597 		return -EINVAL;
2598 	key = tsk->portid + tsk->pub_count + 1;
2599 	if (key == tsk->portid)
2600 		return -EADDRINUSE;
2601 
2602 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2603 				    scope, tsk->portid, key);
2604 	if (unlikely(!publ))
2605 		return -EINVAL;
2606 
2607 	list_add(&publ->binding_sock, &tsk->publications);
2608 	tsk->pub_count++;
2609 	tsk->published = 1;
2610 	return 0;
2611 }
2612 
2613 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2614 			    struct tipc_name_seq const *seq)
2615 {
2616 	struct net *net = sock_net(&tsk->sk);
2617 	struct publication *publ;
2618 	struct publication *safe;
2619 	int rc = -EINVAL;
2620 
2621 	if (scope != TIPC_NODE_SCOPE)
2622 		scope = TIPC_CLUSTER_SCOPE;
2623 
2624 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2625 		if (seq) {
2626 			if (publ->scope != scope)
2627 				continue;
2628 			if (publ->type != seq->type)
2629 				continue;
2630 			if (publ->lower != seq->lower)
2631 				continue;
2632 			if (publ->upper != seq->upper)
2633 				break;
2634 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2635 					      publ->upper, publ->key);
2636 			rc = 0;
2637 			break;
2638 		}
2639 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2640 				      publ->upper, publ->key);
2641 		rc = 0;
2642 	}
2643 	if (list_empty(&tsk->publications))
2644 		tsk->published = 0;
2645 	return rc;
2646 }
2647 
2648 /* tipc_sk_reinit: set non-zero address in all existing sockets
2649  *                 when we go from standalone to network mode.
2650  */
2651 void tipc_sk_reinit(struct net *net)
2652 {
2653 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2654 	struct rhashtable_iter iter;
2655 	struct tipc_sock *tsk;
2656 	struct tipc_msg *msg;
2657 
2658 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2659 
2660 	do {
2661 		rhashtable_walk_start(&iter);
2662 
2663 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2664 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2665 			msg = &tsk->phdr;
2666 			msg_set_prevnode(msg, tipc_own_addr(net));
2667 			msg_set_orignode(msg, tipc_own_addr(net));
2668 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2669 		}
2670 
2671 		rhashtable_walk_stop(&iter);
2672 	} while (tsk == ERR_PTR(-EAGAIN));
2673 }
2674 
2675 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2676 {
2677 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2678 	struct tipc_sock *tsk;
2679 
2680 	rcu_read_lock();
2681 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2682 	if (tsk)
2683 		sock_hold(&tsk->sk);
2684 	rcu_read_unlock();
2685 
2686 	return tsk;
2687 }
2688 
2689 static int tipc_sk_insert(struct tipc_sock *tsk)
2690 {
2691 	struct sock *sk = &tsk->sk;
2692 	struct net *net = sock_net(sk);
2693 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2694 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2695 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2696 
2697 	while (remaining--) {
2698 		portid++;
2699 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2700 			portid = TIPC_MIN_PORT;
2701 		tsk->portid = portid;
2702 		sock_hold(&tsk->sk);
2703 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2704 						   tsk_rht_params))
2705 			return 0;
2706 		sock_put(&tsk->sk);
2707 	}
2708 
2709 	return -1;
2710 }
2711 
2712 static void tipc_sk_remove(struct tipc_sock *tsk)
2713 {
2714 	struct sock *sk = &tsk->sk;
2715 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2716 
2717 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2718 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2719 		__sock_put(sk);
2720 	}
2721 }
2722 
2723 static const struct rhashtable_params tsk_rht_params = {
2724 	.nelem_hint = 192,
2725 	.head_offset = offsetof(struct tipc_sock, node),
2726 	.key_offset = offsetof(struct tipc_sock, portid),
2727 	.key_len = sizeof(u32), /* portid */
2728 	.max_size = 1048576,
2729 	.min_size = 256,
2730 	.automatic_shrinking = true,
2731 };
2732 
2733 int tipc_sk_rht_init(struct net *net)
2734 {
2735 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2736 
2737 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2738 }
2739 
2740 void tipc_sk_rht_destroy(struct net *net)
2741 {
2742 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2743 
2744 	/* Wait for socket readers to complete */
2745 	synchronize_net();
2746 
2747 	rhashtable_destroy(&tn->sk_rht);
2748 }
2749 
2750 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2751 {
2752 	struct net *net = sock_net(&tsk->sk);
2753 	struct tipc_group *grp = tsk->group;
2754 	struct tipc_msg *hdr = &tsk->phdr;
2755 	struct tipc_name_seq seq;
2756 	int rc;
2757 
2758 	if (mreq->type < TIPC_RESERVED_TYPES)
2759 		return -EACCES;
2760 	if (mreq->scope > TIPC_NODE_SCOPE)
2761 		return -EINVAL;
2762 	if (grp)
2763 		return -EACCES;
2764 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2765 	if (!grp)
2766 		return -ENOMEM;
2767 	tsk->group = grp;
2768 	msg_set_lookup_scope(hdr, mreq->scope);
2769 	msg_set_nametype(hdr, mreq->type);
2770 	msg_set_dest_droppable(hdr, true);
2771 	seq.type = mreq->type;
2772 	seq.lower = mreq->instance;
2773 	seq.upper = seq.lower;
2774 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2775 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2776 	if (rc) {
2777 		tipc_group_delete(net, grp);
2778 		tsk->group = NULL;
2779 		return rc;
2780 	}
2781 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2782 	tsk->mc_method.rcast = true;
2783 	tsk->mc_method.mandatory = true;
2784 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2785 	return rc;
2786 }
2787 
2788 static int tipc_sk_leave(struct tipc_sock *tsk)
2789 {
2790 	struct net *net = sock_net(&tsk->sk);
2791 	struct tipc_group *grp = tsk->group;
2792 	struct tipc_name_seq seq;
2793 	int scope;
2794 
2795 	if (!grp)
2796 		return -EINVAL;
2797 	tipc_group_self(grp, &seq, &scope);
2798 	tipc_group_delete(net, grp);
2799 	tsk->group = NULL;
2800 	tipc_sk_withdraw(tsk, scope, &seq);
2801 	return 0;
2802 }
2803 
2804 /**
2805  * tipc_setsockopt - set socket option
2806  * @sock: socket structure
2807  * @lvl: option level
2808  * @opt: option identifier
2809  * @ov: pointer to new option value
2810  * @ol: length of option value
2811  *
2812  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2813  * (to ease compatibility).
2814  *
2815  * Returns 0 on success, errno otherwise
2816  */
2817 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2818 			   char __user *ov, unsigned int ol)
2819 {
2820 	struct sock *sk = sock->sk;
2821 	struct tipc_sock *tsk = tipc_sk(sk);
2822 	struct tipc_group_req mreq;
2823 	u32 value = 0;
2824 	int res = 0;
2825 
2826 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2827 		return 0;
2828 	if (lvl != SOL_TIPC)
2829 		return -ENOPROTOOPT;
2830 
2831 	switch (opt) {
2832 	case TIPC_IMPORTANCE:
2833 	case TIPC_SRC_DROPPABLE:
2834 	case TIPC_DEST_DROPPABLE:
2835 	case TIPC_CONN_TIMEOUT:
2836 		if (ol < sizeof(value))
2837 			return -EINVAL;
2838 		if (get_user(value, (u32 __user *)ov))
2839 			return -EFAULT;
2840 		break;
2841 	case TIPC_GROUP_JOIN:
2842 		if (ol < sizeof(mreq))
2843 			return -EINVAL;
2844 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2845 			return -EFAULT;
2846 		break;
2847 	default:
2848 		if (ov || ol)
2849 			return -EINVAL;
2850 	}
2851 
2852 	lock_sock(sk);
2853 
2854 	switch (opt) {
2855 	case TIPC_IMPORTANCE:
2856 		res = tsk_set_importance(tsk, value);
2857 		break;
2858 	case TIPC_SRC_DROPPABLE:
2859 		if (sock->type != SOCK_STREAM)
2860 			tsk_set_unreliable(tsk, value);
2861 		else
2862 			res = -ENOPROTOOPT;
2863 		break;
2864 	case TIPC_DEST_DROPPABLE:
2865 		tsk_set_unreturnable(tsk, value);
2866 		break;
2867 	case TIPC_CONN_TIMEOUT:
2868 		tipc_sk(sk)->conn_timeout = value;
2869 		break;
2870 	case TIPC_MCAST_BROADCAST:
2871 		tsk->mc_method.rcast = false;
2872 		tsk->mc_method.mandatory = true;
2873 		break;
2874 	case TIPC_MCAST_REPLICAST:
2875 		tsk->mc_method.rcast = true;
2876 		tsk->mc_method.mandatory = true;
2877 		break;
2878 	case TIPC_GROUP_JOIN:
2879 		res = tipc_sk_join(tsk, &mreq);
2880 		break;
2881 	case TIPC_GROUP_LEAVE:
2882 		res = tipc_sk_leave(tsk);
2883 		break;
2884 	default:
2885 		res = -EINVAL;
2886 	}
2887 
2888 	release_sock(sk);
2889 
2890 	return res;
2891 }
2892 
2893 /**
2894  * tipc_getsockopt - get socket option
2895  * @sock: socket structure
2896  * @lvl: option level
2897  * @opt: option identifier
2898  * @ov: receptacle for option value
2899  * @ol: receptacle for length of option value
2900  *
2901  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2902  * (to ease compatibility).
2903  *
2904  * Returns 0 on success, errno otherwise
2905  */
2906 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2907 			   char __user *ov, int __user *ol)
2908 {
2909 	struct sock *sk = sock->sk;
2910 	struct tipc_sock *tsk = tipc_sk(sk);
2911 	struct tipc_name_seq seq;
2912 	int len, scope;
2913 	u32 value;
2914 	int res;
2915 
2916 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2917 		return put_user(0, ol);
2918 	if (lvl != SOL_TIPC)
2919 		return -ENOPROTOOPT;
2920 	res = get_user(len, ol);
2921 	if (res)
2922 		return res;
2923 
2924 	lock_sock(sk);
2925 
2926 	switch (opt) {
2927 	case TIPC_IMPORTANCE:
2928 		value = tsk_importance(tsk);
2929 		break;
2930 	case TIPC_SRC_DROPPABLE:
2931 		value = tsk_unreliable(tsk);
2932 		break;
2933 	case TIPC_DEST_DROPPABLE:
2934 		value = tsk_unreturnable(tsk);
2935 		break;
2936 	case TIPC_CONN_TIMEOUT:
2937 		value = tsk->conn_timeout;
2938 		/* no need to set "res", since already 0 at this point */
2939 		break;
2940 	case TIPC_NODE_RECVQ_DEPTH:
2941 		value = 0; /* was tipc_queue_size, now obsolete */
2942 		break;
2943 	case TIPC_SOCK_RECVQ_DEPTH:
2944 		value = skb_queue_len(&sk->sk_receive_queue);
2945 		break;
2946 	case TIPC_GROUP_JOIN:
2947 		seq.type = 0;
2948 		if (tsk->group)
2949 			tipc_group_self(tsk->group, &seq, &scope);
2950 		value = seq.type;
2951 		break;
2952 	default:
2953 		res = -EINVAL;
2954 	}
2955 
2956 	release_sock(sk);
2957 
2958 	if (res)
2959 		return res;	/* "get" failed */
2960 
2961 	if (len < sizeof(value))
2962 		return -EINVAL;
2963 
2964 	if (copy_to_user(ov, &value, sizeof(value)))
2965 		return -EFAULT;
2966 
2967 	return put_user(sizeof(value), ol);
2968 }
2969 
2970 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2971 {
2972 	struct sock *sk = sock->sk;
2973 	struct tipc_sioc_ln_req lnr;
2974 	void __user *argp = (void __user *)arg;
2975 
2976 	switch (cmd) {
2977 	case SIOCGETLINKNAME:
2978 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2979 			return -EFAULT;
2980 		if (!tipc_node_get_linkname(sock_net(sk),
2981 					    lnr.bearer_id & 0xffff, lnr.peer,
2982 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2983 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2984 				return -EFAULT;
2985 			return 0;
2986 		}
2987 		return -EADDRNOTAVAIL;
2988 	default:
2989 		return -ENOIOCTLCMD;
2990 	}
2991 }
2992 
2993 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
2994 {
2995 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
2996 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
2997 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
2998 
2999 	tsk1->peer.family = AF_TIPC;
3000 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3001 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3002 	tsk1->peer.addr.id.ref = tsk2->portid;
3003 	tsk1->peer.addr.id.node = onode;
3004 	tsk2->peer.family = AF_TIPC;
3005 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3006 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3007 	tsk2->peer.addr.id.ref = tsk1->portid;
3008 	tsk2->peer.addr.id.node = onode;
3009 
3010 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3011 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3012 	return 0;
3013 }
3014 
3015 /* Protocol switches for the various types of TIPC sockets */
3016 
3017 static const struct proto_ops msg_ops = {
3018 	.owner		= THIS_MODULE,
3019 	.family		= AF_TIPC,
3020 	.release	= tipc_release,
3021 	.bind		= tipc_bind,
3022 	.connect	= tipc_connect,
3023 	.socketpair	= tipc_socketpair,
3024 	.accept		= sock_no_accept,
3025 	.getname	= tipc_getname,
3026 	.poll_mask	= tipc_poll_mask,
3027 	.ioctl		= tipc_ioctl,
3028 	.listen		= sock_no_listen,
3029 	.shutdown	= tipc_shutdown,
3030 	.setsockopt	= tipc_setsockopt,
3031 	.getsockopt	= tipc_getsockopt,
3032 	.sendmsg	= tipc_sendmsg,
3033 	.recvmsg	= tipc_recvmsg,
3034 	.mmap		= sock_no_mmap,
3035 	.sendpage	= sock_no_sendpage
3036 };
3037 
3038 static const struct proto_ops packet_ops = {
3039 	.owner		= THIS_MODULE,
3040 	.family		= AF_TIPC,
3041 	.release	= tipc_release,
3042 	.bind		= tipc_bind,
3043 	.connect	= tipc_connect,
3044 	.socketpair	= tipc_socketpair,
3045 	.accept		= tipc_accept,
3046 	.getname	= tipc_getname,
3047 	.poll_mask	= tipc_poll_mask,
3048 	.ioctl		= tipc_ioctl,
3049 	.listen		= tipc_listen,
3050 	.shutdown	= tipc_shutdown,
3051 	.setsockopt	= tipc_setsockopt,
3052 	.getsockopt	= tipc_getsockopt,
3053 	.sendmsg	= tipc_send_packet,
3054 	.recvmsg	= tipc_recvmsg,
3055 	.mmap		= sock_no_mmap,
3056 	.sendpage	= sock_no_sendpage
3057 };
3058 
3059 static const struct proto_ops stream_ops = {
3060 	.owner		= THIS_MODULE,
3061 	.family		= AF_TIPC,
3062 	.release	= tipc_release,
3063 	.bind		= tipc_bind,
3064 	.connect	= tipc_connect,
3065 	.socketpair	= tipc_socketpair,
3066 	.accept		= tipc_accept,
3067 	.getname	= tipc_getname,
3068 	.poll_mask	= tipc_poll_mask,
3069 	.ioctl		= tipc_ioctl,
3070 	.listen		= tipc_listen,
3071 	.shutdown	= tipc_shutdown,
3072 	.setsockopt	= tipc_setsockopt,
3073 	.getsockopt	= tipc_getsockopt,
3074 	.sendmsg	= tipc_sendstream,
3075 	.recvmsg	= tipc_recvstream,
3076 	.mmap		= sock_no_mmap,
3077 	.sendpage	= sock_no_sendpage
3078 };
3079 
3080 static const struct net_proto_family tipc_family_ops = {
3081 	.owner		= THIS_MODULE,
3082 	.family		= AF_TIPC,
3083 	.create		= tipc_sk_create
3084 };
3085 
3086 static struct proto tipc_proto = {
3087 	.name		= "TIPC",
3088 	.owner		= THIS_MODULE,
3089 	.obj_size	= sizeof(struct tipc_sock),
3090 	.sysctl_rmem	= sysctl_tipc_rmem
3091 };
3092 
3093 /**
3094  * tipc_socket_init - initialize TIPC socket interface
3095  *
3096  * Returns 0 on success, errno otherwise
3097  */
3098 int tipc_socket_init(void)
3099 {
3100 	int res;
3101 
3102 	res = proto_register(&tipc_proto, 1);
3103 	if (res) {
3104 		pr_err("Failed to register TIPC protocol type\n");
3105 		goto out;
3106 	}
3107 
3108 	res = sock_register(&tipc_family_ops);
3109 	if (res) {
3110 		pr_err("Failed to register TIPC socket type\n");
3111 		proto_unregister(&tipc_proto);
3112 		goto out;
3113 	}
3114  out:
3115 	return res;
3116 }
3117 
3118 /**
3119  * tipc_socket_stop - stop TIPC socket interface
3120  */
3121 void tipc_socket_stop(void)
3122 {
3123 	sock_unregister(tipc_family_ops.family);
3124 	proto_unregister(&tipc_proto);
3125 }
3126 
3127 /* Caller should hold socket lock for the passed tipc socket. */
3128 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3129 {
3130 	u32 peer_node;
3131 	u32 peer_port;
3132 	struct nlattr *nest;
3133 
3134 	peer_node = tsk_peer_node(tsk);
3135 	peer_port = tsk_peer_port(tsk);
3136 
3137 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3138 
3139 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3140 		goto msg_full;
3141 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3142 		goto msg_full;
3143 
3144 	if (tsk->conn_type != 0) {
3145 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3146 			goto msg_full;
3147 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3148 			goto msg_full;
3149 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3150 			goto msg_full;
3151 	}
3152 	nla_nest_end(skb, nest);
3153 
3154 	return 0;
3155 
3156 msg_full:
3157 	nla_nest_cancel(skb, nest);
3158 
3159 	return -EMSGSIZE;
3160 }
3161 
3162 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3163 			  *tsk)
3164 {
3165 	struct net *net = sock_net(skb->sk);
3166 	struct sock *sk = &tsk->sk;
3167 
3168 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3169 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3170 		return -EMSGSIZE;
3171 
3172 	if (tipc_sk_connected(sk)) {
3173 		if (__tipc_nl_add_sk_con(skb, tsk))
3174 			return -EMSGSIZE;
3175 	} else if (!list_empty(&tsk->publications)) {
3176 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3177 			return -EMSGSIZE;
3178 	}
3179 	return 0;
3180 }
3181 
3182 /* Caller should hold socket lock for the passed tipc socket. */
3183 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3184 			    struct tipc_sock *tsk)
3185 {
3186 	struct nlattr *attrs;
3187 	void *hdr;
3188 
3189 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3190 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3191 	if (!hdr)
3192 		goto msg_cancel;
3193 
3194 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3195 	if (!attrs)
3196 		goto genlmsg_cancel;
3197 
3198 	if (__tipc_nl_add_sk_info(skb, tsk))
3199 		goto attr_msg_cancel;
3200 
3201 	nla_nest_end(skb, attrs);
3202 	genlmsg_end(skb, hdr);
3203 
3204 	return 0;
3205 
3206 attr_msg_cancel:
3207 	nla_nest_cancel(skb, attrs);
3208 genlmsg_cancel:
3209 	genlmsg_cancel(skb, hdr);
3210 msg_cancel:
3211 	return -EMSGSIZE;
3212 }
3213 
3214 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3215 		    int (*skb_handler)(struct sk_buff *skb,
3216 				       struct netlink_callback *cb,
3217 				       struct tipc_sock *tsk))
3218 {
3219 	struct net *net = sock_net(skb->sk);
3220 	struct tipc_net *tn = tipc_net(net);
3221 	const struct bucket_table *tbl;
3222 	u32 prev_portid = cb->args[1];
3223 	u32 tbl_id = cb->args[0];
3224 	struct rhash_head *pos;
3225 	struct tipc_sock *tsk;
3226 	int err;
3227 
3228 	rcu_read_lock();
3229 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3230 	for (; tbl_id < tbl->size; tbl_id++) {
3231 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3232 			spin_lock_bh(&tsk->sk.sk_lock.slock);
3233 			if (prev_portid && prev_portid != tsk->portid) {
3234 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3235 				continue;
3236 			}
3237 
3238 			err = skb_handler(skb, cb, tsk);
3239 			if (err) {
3240 				prev_portid = tsk->portid;
3241 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3242 				goto out;
3243 			}
3244 
3245 			prev_portid = 0;
3246 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
3247 		}
3248 	}
3249 out:
3250 	rcu_read_unlock();
3251 	cb->args[0] = tbl_id;
3252 	cb->args[1] = prev_portid;
3253 
3254 	return skb->len;
3255 }
3256 EXPORT_SYMBOL(tipc_nl_sk_walk);
3257 
3258 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3259 			   struct tipc_sock *tsk, u32 sk_filter_state,
3260 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3261 {
3262 	struct sock *sk = &tsk->sk;
3263 	struct nlattr *attrs;
3264 	struct nlattr *stat;
3265 
3266 	/*filter response w.r.t sk_state*/
3267 	if (!(sk_filter_state & (1 << sk->sk_state)))
3268 		return 0;
3269 
3270 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3271 	if (!attrs)
3272 		goto msg_cancel;
3273 
3274 	if (__tipc_nl_add_sk_info(skb, tsk))
3275 		goto attr_msg_cancel;
3276 
3277 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3278 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3279 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3280 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3281 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3282 					 sock_i_uid(sk))) ||
3283 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3284 			      tipc_diag_gen_cookie(sk),
3285 			      TIPC_NLA_SOCK_PAD))
3286 		goto attr_msg_cancel;
3287 
3288 	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3289 	if (!stat)
3290 		goto attr_msg_cancel;
3291 
3292 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3293 			skb_queue_len(&sk->sk_receive_queue)) ||
3294 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3295 			skb_queue_len(&sk->sk_write_queue)) ||
3296 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3297 			atomic_read(&sk->sk_drops)))
3298 		goto stat_msg_cancel;
3299 
3300 	if (tsk->cong_link_cnt &&
3301 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3302 		goto stat_msg_cancel;
3303 
3304 	if (tsk_conn_cong(tsk) &&
3305 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3306 		goto stat_msg_cancel;
3307 
3308 	nla_nest_end(skb, stat);
3309 	nla_nest_end(skb, attrs);
3310 
3311 	return 0;
3312 
3313 stat_msg_cancel:
3314 	nla_nest_cancel(skb, stat);
3315 attr_msg_cancel:
3316 	nla_nest_cancel(skb, attrs);
3317 msg_cancel:
3318 	return -EMSGSIZE;
3319 }
3320 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3321 
3322 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3323 {
3324 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3325 }
3326 
3327 /* Caller should hold socket lock for the passed tipc socket. */
3328 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3329 				 struct netlink_callback *cb,
3330 				 struct publication *publ)
3331 {
3332 	void *hdr;
3333 	struct nlattr *attrs;
3334 
3335 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3336 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3337 	if (!hdr)
3338 		goto msg_cancel;
3339 
3340 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3341 	if (!attrs)
3342 		goto genlmsg_cancel;
3343 
3344 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3345 		goto attr_msg_cancel;
3346 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3347 		goto attr_msg_cancel;
3348 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3349 		goto attr_msg_cancel;
3350 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3351 		goto attr_msg_cancel;
3352 
3353 	nla_nest_end(skb, attrs);
3354 	genlmsg_end(skb, hdr);
3355 
3356 	return 0;
3357 
3358 attr_msg_cancel:
3359 	nla_nest_cancel(skb, attrs);
3360 genlmsg_cancel:
3361 	genlmsg_cancel(skb, hdr);
3362 msg_cancel:
3363 	return -EMSGSIZE;
3364 }
3365 
3366 /* Caller should hold socket lock for the passed tipc socket. */
3367 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3368 				  struct netlink_callback *cb,
3369 				  struct tipc_sock *tsk, u32 *last_publ)
3370 {
3371 	int err;
3372 	struct publication *p;
3373 
3374 	if (*last_publ) {
3375 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3376 			if (p->key == *last_publ)
3377 				break;
3378 		}
3379 		if (p->key != *last_publ) {
3380 			/* We never set seq or call nl_dump_check_consistent()
3381 			 * this means that setting prev_seq here will cause the
3382 			 * consistence check to fail in the netlink callback
3383 			 * handler. Resulting in the last NLMSG_DONE message
3384 			 * having the NLM_F_DUMP_INTR flag set.
3385 			 */
3386 			cb->prev_seq = 1;
3387 			*last_publ = 0;
3388 			return -EPIPE;
3389 		}
3390 	} else {
3391 		p = list_first_entry(&tsk->publications, struct publication,
3392 				     binding_sock);
3393 	}
3394 
3395 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3396 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3397 		if (err) {
3398 			*last_publ = p->key;
3399 			return err;
3400 		}
3401 	}
3402 	*last_publ = 0;
3403 
3404 	return 0;
3405 }
3406 
3407 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3408 {
3409 	int err;
3410 	u32 tsk_portid = cb->args[0];
3411 	u32 last_publ = cb->args[1];
3412 	u32 done = cb->args[2];
3413 	struct net *net = sock_net(skb->sk);
3414 	struct tipc_sock *tsk;
3415 
3416 	if (!tsk_portid) {
3417 		struct nlattr **attrs;
3418 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3419 
3420 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3421 		if (err)
3422 			return err;
3423 
3424 		if (!attrs[TIPC_NLA_SOCK])
3425 			return -EINVAL;
3426 
3427 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3428 				       attrs[TIPC_NLA_SOCK],
3429 				       tipc_nl_sock_policy, NULL);
3430 		if (err)
3431 			return err;
3432 
3433 		if (!sock[TIPC_NLA_SOCK_REF])
3434 			return -EINVAL;
3435 
3436 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3437 	}
3438 
3439 	if (done)
3440 		return 0;
3441 
3442 	tsk = tipc_sk_lookup(net, tsk_portid);
3443 	if (!tsk)
3444 		return -EINVAL;
3445 
3446 	lock_sock(&tsk->sk);
3447 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3448 	if (!err)
3449 		done = 1;
3450 	release_sock(&tsk->sk);
3451 	sock_put(&tsk->sk);
3452 
3453 	cb->args[0] = tsk_portid;
3454 	cb->args[1] = last_publ;
3455 	cb->args[2] = done;
3456 
3457 	return skb->len;
3458 }
3459