xref: /linux/net/tipc/socket.c (revision 81fa7a69c2174ed8de314b9c231ef30a8718e5e1)
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  * @conn_timeout: the time we can wait for an unresponded setup request
84  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
85  * @cong_link_cnt: number of congested links
86  * @snt_unacked: # messages sent by socket, and not yet acked by peer
87  * @rcv_unacked: # messages read by user, but not yet acked back to peer
88  * @peer: 'connected' peer for dgram/rdm
89  * @node: hash table node
90  * @mc_method: cookie for use between socket and broadcast layer
91  * @rcu: rcu struct for tipc_sock
92  */
93 struct tipc_sock {
94 	struct sock sk;
95 	u32 conn_type;
96 	u32 conn_instance;
97 	int published;
98 	u32 max_pkt;
99 	u32 portid;
100 	struct tipc_msg phdr;
101 	struct list_head cong_links;
102 	struct list_head publications;
103 	u32 pub_count;
104 	atomic_t dupl_rcvcnt;
105 	u16 conn_timeout;
106 	bool probe_unacked;
107 	u16 cong_link_cnt;
108 	u16 snt_unacked;
109 	u16 snd_win;
110 	u16 peer_caps;
111 	u16 rcv_unacked;
112 	u16 rcv_win;
113 	struct sockaddr_tipc peer;
114 	struct rhash_head node;
115 	struct tipc_mc_method mc_method;
116 	struct rcu_head rcu;
117 	struct tipc_group *group;
118 	bool group_is_open;
119 };
120 
121 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
122 static void tipc_data_ready(struct sock *sk);
123 static void tipc_write_space(struct sock *sk);
124 static void tipc_sock_destruct(struct sock *sk);
125 static int tipc_release(struct socket *sock);
126 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
127 		       bool kern);
128 static void tipc_sk_timeout(struct timer_list *t);
129 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
130 			   struct tipc_name_seq const *seq);
131 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
132 			    struct tipc_name_seq const *seq);
133 static int tipc_sk_leave(struct tipc_sock *tsk);
134 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
135 static int tipc_sk_insert(struct tipc_sock *tsk);
136 static void tipc_sk_remove(struct tipc_sock *tsk);
137 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
138 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
139 
140 static const struct proto_ops packet_ops;
141 static const struct proto_ops stream_ops;
142 static const struct proto_ops msg_ops;
143 static struct proto tipc_proto;
144 static const struct rhashtable_params tsk_rht_params;
145 
146 static u32 tsk_own_node(struct tipc_sock *tsk)
147 {
148 	return msg_prevnode(&tsk->phdr);
149 }
150 
151 static u32 tsk_peer_node(struct tipc_sock *tsk)
152 {
153 	return msg_destnode(&tsk->phdr);
154 }
155 
156 static u32 tsk_peer_port(struct tipc_sock *tsk)
157 {
158 	return msg_destport(&tsk->phdr);
159 }
160 
161 static  bool tsk_unreliable(struct tipc_sock *tsk)
162 {
163 	return msg_src_droppable(&tsk->phdr) != 0;
164 }
165 
166 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
167 {
168 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
169 }
170 
171 static bool tsk_unreturnable(struct tipc_sock *tsk)
172 {
173 	return msg_dest_droppable(&tsk->phdr) != 0;
174 }
175 
176 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
177 {
178 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
179 }
180 
181 static int tsk_importance(struct tipc_sock *tsk)
182 {
183 	return msg_importance(&tsk->phdr);
184 }
185 
186 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
187 {
188 	if (imp > TIPC_CRITICAL_IMPORTANCE)
189 		return -EINVAL;
190 	msg_set_importance(&tsk->phdr, (u32)imp);
191 	return 0;
192 }
193 
194 static struct tipc_sock *tipc_sk(const struct sock *sk)
195 {
196 	return container_of(sk, struct tipc_sock, sk);
197 }
198 
199 static bool tsk_conn_cong(struct tipc_sock *tsk)
200 {
201 	return tsk->snt_unacked > tsk->snd_win;
202 }
203 
204 static u16 tsk_blocks(int len)
205 {
206 	return ((len / FLOWCTL_BLK_SZ) + 1);
207 }
208 
209 /* tsk_blocks(): translate a buffer size in bytes to number of
210  * advertisable blocks, taking into account the ratio truesize(len)/len
211  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
212  */
213 static u16 tsk_adv_blocks(int len)
214 {
215 	return len / FLOWCTL_BLK_SZ / 4;
216 }
217 
218 /* tsk_inc(): increment counter for sent or received data
219  * - If block based flow control is not supported by peer we
220  *   fall back to message based ditto, incrementing the counter
221  */
222 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
223 {
224 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
225 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
226 	return 1;
227 }
228 
229 /**
230  * tsk_advance_rx_queue - discard first buffer in socket receive queue
231  *
232  * Caller must hold socket lock
233  */
234 static void tsk_advance_rx_queue(struct sock *sk)
235 {
236 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
237 }
238 
239 /* tipc_sk_respond() : send response message back to sender
240  */
241 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
242 {
243 	u32 selector;
244 	u32 dnode;
245 	u32 onode = tipc_own_addr(sock_net(sk));
246 
247 	if (!tipc_msg_reverse(onode, &skb, err))
248 		return;
249 
250 	dnode = msg_destnode(buf_msg(skb));
251 	selector = msg_origport(buf_msg(skb));
252 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
253 }
254 
255 /**
256  * tsk_rej_rx_queue - reject all buffers in socket receive queue
257  *
258  * Caller must hold socket lock
259  */
260 static void tsk_rej_rx_queue(struct sock *sk)
261 {
262 	struct sk_buff *skb;
263 
264 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
265 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
266 }
267 
268 static bool tipc_sk_connected(struct sock *sk)
269 {
270 	return sk->sk_state == TIPC_ESTABLISHED;
271 }
272 
273 /* tipc_sk_type_connectionless - check if the socket is datagram socket
274  * @sk: socket
275  *
276  * Returns true if connection less, false otherwise
277  */
278 static bool tipc_sk_type_connectionless(struct sock *sk)
279 {
280 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
281 }
282 
283 /* tsk_peer_msg - verify if message was sent by connected port's peer
284  *
285  * Handles cases where the node's network address has changed from
286  * the default of <0.0.0> to its configured setting.
287  */
288 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
289 {
290 	struct sock *sk = &tsk->sk;
291 	u32 self = tipc_own_addr(sock_net(sk));
292 	u32 peer_port = tsk_peer_port(tsk);
293 	u32 orig_node, peer_node;
294 
295 	if (unlikely(!tipc_sk_connected(sk)))
296 		return false;
297 
298 	if (unlikely(msg_origport(msg) != peer_port))
299 		return false;
300 
301 	orig_node = msg_orignode(msg);
302 	peer_node = tsk_peer_node(tsk);
303 
304 	if (likely(orig_node == peer_node))
305 		return true;
306 
307 	if (!orig_node && peer_node == self)
308 		return true;
309 
310 	if (!peer_node && orig_node == self)
311 		return true;
312 
313 	return false;
314 }
315 
316 /* tipc_set_sk_state - set the sk_state of the socket
317  * @sk: socket
318  *
319  * Caller must hold socket lock
320  *
321  * Returns 0 on success, errno otherwise
322  */
323 static int tipc_set_sk_state(struct sock *sk, int state)
324 {
325 	int oldsk_state = sk->sk_state;
326 	int res = -EINVAL;
327 
328 	switch (state) {
329 	case TIPC_OPEN:
330 		res = 0;
331 		break;
332 	case TIPC_LISTEN:
333 	case TIPC_CONNECTING:
334 		if (oldsk_state == TIPC_OPEN)
335 			res = 0;
336 		break;
337 	case TIPC_ESTABLISHED:
338 		if (oldsk_state == TIPC_CONNECTING ||
339 		    oldsk_state == TIPC_OPEN)
340 			res = 0;
341 		break;
342 	case TIPC_DISCONNECTING:
343 		if (oldsk_state == TIPC_CONNECTING ||
344 		    oldsk_state == TIPC_ESTABLISHED)
345 			res = 0;
346 		break;
347 	}
348 
349 	if (!res)
350 		sk->sk_state = state;
351 
352 	return res;
353 }
354 
355 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
356 {
357 	struct sock *sk = sock->sk;
358 	int err = sock_error(sk);
359 	int typ = sock->type;
360 
361 	if (err)
362 		return err;
363 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
364 		if (sk->sk_state == TIPC_DISCONNECTING)
365 			return -EPIPE;
366 		else if (!tipc_sk_connected(sk))
367 			return -ENOTCONN;
368 	}
369 	if (!*timeout)
370 		return -EAGAIN;
371 	if (signal_pending(current))
372 		return sock_intr_errno(*timeout);
373 
374 	return 0;
375 }
376 
377 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
378 ({                                                                             \
379 	struct sock *sk_;						       \
380 	int rc_;							       \
381 									       \
382 	while ((rc_ = !(condition_))) {					       \
383 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
384 		sk_ = (sock_)->sk;					       \
385 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
386 		if (rc_)						       \
387 			break;						       \
388 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
389 		release_sock(sk_);					       \
390 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
391 		sched_annotate_sleep();				               \
392 		lock_sock(sk_);						       \
393 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
394 	}								       \
395 	rc_;								       \
396 })
397 
398 /**
399  * tipc_sk_create - create a TIPC socket
400  * @net: network namespace (must be default network)
401  * @sock: pre-allocated socket structure
402  * @protocol: protocol indicator (must be 0)
403  * @kern: caused by kernel or by userspace?
404  *
405  * This routine creates additional data structures used by the TIPC socket,
406  * initializes them, and links them together.
407  *
408  * Returns 0 on success, errno otherwise
409  */
410 static int tipc_sk_create(struct net *net, struct socket *sock,
411 			  int protocol, int kern)
412 {
413 	const struct proto_ops *ops;
414 	struct sock *sk;
415 	struct tipc_sock *tsk;
416 	struct tipc_msg *msg;
417 
418 	/* Validate arguments */
419 	if (unlikely(protocol != 0))
420 		return -EPROTONOSUPPORT;
421 
422 	switch (sock->type) {
423 	case SOCK_STREAM:
424 		ops = &stream_ops;
425 		break;
426 	case SOCK_SEQPACKET:
427 		ops = &packet_ops;
428 		break;
429 	case SOCK_DGRAM:
430 	case SOCK_RDM:
431 		ops = &msg_ops;
432 		break;
433 	default:
434 		return -EPROTOTYPE;
435 	}
436 
437 	/* Allocate socket's protocol area */
438 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
439 	if (sk == NULL)
440 		return -ENOMEM;
441 
442 	tsk = tipc_sk(sk);
443 	tsk->max_pkt = MAX_PKT_DEFAULT;
444 	INIT_LIST_HEAD(&tsk->publications);
445 	INIT_LIST_HEAD(&tsk->cong_links);
446 	msg = &tsk->phdr;
447 
448 	/* Finish initializing socket data structures */
449 	sock->ops = ops;
450 	sock_init_data(sock, sk);
451 	tipc_set_sk_state(sk, TIPC_OPEN);
452 	if (tipc_sk_insert(tsk)) {
453 		pr_warn("Socket create failed; port number exhausted\n");
454 		return -EINVAL;
455 	}
456 
457 	/* Ensure tsk is visible before we read own_addr. */
458 	smp_mb();
459 
460 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
461 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
462 
463 	msg_set_origport(msg, tsk->portid);
464 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
465 	sk->sk_shutdown = 0;
466 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
467 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
468 	sk->sk_data_ready = tipc_data_ready;
469 	sk->sk_write_space = tipc_write_space;
470 	sk->sk_destruct = tipc_sock_destruct;
471 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
472 	tsk->group_is_open = true;
473 	atomic_set(&tsk->dupl_rcvcnt, 0);
474 
475 	/* Start out with safe limits until we receive an advertised window */
476 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
477 	tsk->rcv_win = tsk->snd_win;
478 
479 	if (tipc_sk_type_connectionless(sk)) {
480 		tsk_set_unreturnable(tsk, true);
481 		if (sock->type == SOCK_DGRAM)
482 			tsk_set_unreliable(tsk, true);
483 	}
484 
485 	return 0;
486 }
487 
488 static void tipc_sk_callback(struct rcu_head *head)
489 {
490 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
491 
492 	sock_put(&tsk->sk);
493 }
494 
495 /* Caller should hold socket lock for the socket. */
496 static void __tipc_shutdown(struct socket *sock, int error)
497 {
498 	struct sock *sk = sock->sk;
499 	struct tipc_sock *tsk = tipc_sk(sk);
500 	struct net *net = sock_net(sk);
501 	long timeout = CONN_TIMEOUT_DEFAULT;
502 	u32 dnode = tsk_peer_node(tsk);
503 	struct sk_buff *skb;
504 
505 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
506 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
507 					    !tsk_conn_cong(tsk)));
508 
509 	/* Remove any pending SYN message */
510 	__skb_queue_purge(&sk->sk_write_queue);
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 	sock_orphan(sk);
582 	/* Reject any messages that accumulated in backlog queue */
583 	release_sock(sk);
584 	tipc_dest_list_purge(&tsk->cong_links);
585 	tsk->cong_link_cnt = 0;
586 	call_rcu(&tsk->rcu, tipc_sk_callback);
587 	sock->sk = NULL;
588 
589 	return 0;
590 }
591 
592 /**
593  * tipc_bind - associate or disassocate TIPC name(s) with a socket
594  * @sock: socket structure
595  * @uaddr: socket address describing name(s) and desired operation
596  * @uaddr_len: size of socket address data structure
597  *
598  * Name and name sequence binding is indicated using a positive scope value;
599  * a negative scope value unbinds the specified name.  Specifying no name
600  * (i.e. a socket address length of 0) unbinds all names from the socket.
601  *
602  * Returns 0 on success, errno otherwise
603  *
604  * NOTE: This routine doesn't need to take the socket lock since it doesn't
605  *       access any non-constant socket information.
606  */
607 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
608 		     int uaddr_len)
609 {
610 	struct sock *sk = sock->sk;
611 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
612 	struct tipc_sock *tsk = tipc_sk(sk);
613 	int res = -EINVAL;
614 
615 	lock_sock(sk);
616 	if (unlikely(!uaddr_len)) {
617 		res = tipc_sk_withdraw(tsk, 0, NULL);
618 		goto exit;
619 	}
620 	if (tsk->group) {
621 		res = -EACCES;
622 		goto exit;
623 	}
624 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
625 		res = -EINVAL;
626 		goto exit;
627 	}
628 	if (addr->family != AF_TIPC) {
629 		res = -EAFNOSUPPORT;
630 		goto exit;
631 	}
632 
633 	if (addr->addrtype == TIPC_ADDR_NAME)
634 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
635 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
636 		res = -EAFNOSUPPORT;
637 		goto exit;
638 	}
639 
640 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
641 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
642 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
643 		res = -EACCES;
644 		goto exit;
645 	}
646 
647 	res = (addr->scope >= 0) ?
648 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
649 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
650 exit:
651 	release_sock(sk);
652 	return res;
653 }
654 
655 /**
656  * tipc_getname - get port ID of socket or peer socket
657  * @sock: socket structure
658  * @uaddr: area for returned socket address
659  * @uaddr_len: area for returned length of socket address
660  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
661  *
662  * Returns 0 on success, errno otherwise
663  *
664  * NOTE: This routine doesn't need to take the socket lock since it only
665  *       accesses socket information that is unchanging (or which changes in
666  *       a completely predictable manner).
667  */
668 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
669 			int peer)
670 {
671 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
672 	struct sock *sk = sock->sk;
673 	struct tipc_sock *tsk = tipc_sk(sk);
674 
675 	memset(addr, 0, sizeof(*addr));
676 	if (peer) {
677 		if ((!tipc_sk_connected(sk)) &&
678 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
679 			return -ENOTCONN;
680 		addr->addr.id.ref = tsk_peer_port(tsk);
681 		addr->addr.id.node = tsk_peer_node(tsk);
682 	} else {
683 		addr->addr.id.ref = tsk->portid;
684 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
685 	}
686 
687 	addr->addrtype = TIPC_ADDR_ID;
688 	addr->family = AF_TIPC;
689 	addr->scope = 0;
690 	addr->addr.name.domain = 0;
691 
692 	return sizeof(*addr);
693 }
694 
695 /**
696  * tipc_poll - read and possibly block on pollmask
697  * @file: file structure associated with the socket
698  * @sock: socket for which to calculate the poll bits
699  * @wait: ???
700  *
701  * Returns pollmask value
702  *
703  * COMMENTARY:
704  * It appears that the usual socket locking mechanisms are not useful here
705  * since the pollmask info is potentially out-of-date the moment this routine
706  * exits.  TCP and other protocols seem to rely on higher level poll routines
707  * to handle any preventable race conditions, so TIPC will do the same ...
708  *
709  * IMPORTANT: The fact that a read or write operation is indicated does NOT
710  * imply that the operation will succeed, merely that it should be performed
711  * and will not block.
712  */
713 static __poll_t tipc_poll(struct file *file, struct socket *sock,
714 			      poll_table *wait)
715 {
716 	struct sock *sk = sock->sk;
717 	struct tipc_sock *tsk = tipc_sk(sk);
718 	__poll_t revents = 0;
719 
720 	sock_poll_wait(file, wait);
721 
722 	if (sk->sk_shutdown & RCV_SHUTDOWN)
723 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
724 	if (sk->sk_shutdown == SHUTDOWN_MASK)
725 		revents |= EPOLLHUP;
726 
727 	switch (sk->sk_state) {
728 	case TIPC_ESTABLISHED:
729 	case TIPC_CONNECTING:
730 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
731 			revents |= EPOLLOUT;
732 		/* fall thru' */
733 	case TIPC_LISTEN:
734 		if (!skb_queue_empty(&sk->sk_receive_queue))
735 			revents |= EPOLLIN | EPOLLRDNORM;
736 		break;
737 	case TIPC_OPEN:
738 		if (tsk->group_is_open && !tsk->cong_link_cnt)
739 			revents |= EPOLLOUT;
740 		if (!tipc_sk_type_connectionless(sk))
741 			break;
742 		if (skb_queue_empty(&sk->sk_receive_queue))
743 			break;
744 		revents |= EPOLLIN | EPOLLRDNORM;
745 		break;
746 	case TIPC_DISCONNECTING:
747 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
748 		break;
749 	}
750 	return revents;
751 }
752 
753 /**
754  * tipc_sendmcast - send multicast message
755  * @sock: socket structure
756  * @seq: destination address
757  * @msg: message to send
758  * @dlen: length of data to send
759  * @timeout: timeout to wait for wakeup
760  *
761  * Called from function tipc_sendmsg(), which has done all sanity checks
762  * Returns the number of bytes sent on success, or errno
763  */
764 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
765 			  struct msghdr *msg, size_t dlen, long timeout)
766 {
767 	struct sock *sk = sock->sk;
768 	struct tipc_sock *tsk = tipc_sk(sk);
769 	struct tipc_msg *hdr = &tsk->phdr;
770 	struct net *net = sock_net(sk);
771 	int mtu = tipc_bcast_get_mtu(net);
772 	struct tipc_mc_method *method = &tsk->mc_method;
773 	struct sk_buff_head pkts;
774 	struct tipc_nlist dsts;
775 	int rc;
776 
777 	if (tsk->group)
778 		return -EACCES;
779 
780 	/* Block or return if any destination link is congested */
781 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
782 	if (unlikely(rc))
783 		return rc;
784 
785 	/* Lookup destination nodes */
786 	tipc_nlist_init(&dsts, tipc_own_addr(net));
787 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
788 				      seq->upper, &dsts);
789 	if (!dsts.local && !dsts.remote)
790 		return -EHOSTUNREACH;
791 
792 	/* Build message header */
793 	msg_set_type(hdr, TIPC_MCAST_MSG);
794 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
795 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
796 	msg_set_destport(hdr, 0);
797 	msg_set_destnode(hdr, 0);
798 	msg_set_nametype(hdr, seq->type);
799 	msg_set_namelower(hdr, seq->lower);
800 	msg_set_nameupper(hdr, seq->upper);
801 
802 	/* Build message as chain of buffers */
803 	skb_queue_head_init(&pkts);
804 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
805 
806 	/* Send message if build was successful */
807 	if (unlikely(rc == dlen))
808 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
809 				     &tsk->cong_link_cnt);
810 
811 	tipc_nlist_purge(&dsts);
812 
813 	return rc ? rc : dlen;
814 }
815 
816 /**
817  * tipc_send_group_msg - send a message to a member in the group
818  * @net: network namespace
819  * @m: message to send
820  * @mb: group member
821  * @dnode: destination node
822  * @dport: destination port
823  * @dlen: total length of message data
824  */
825 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
826 			       struct msghdr *m, struct tipc_member *mb,
827 			       u32 dnode, u32 dport, int dlen)
828 {
829 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
830 	struct tipc_mc_method *method = &tsk->mc_method;
831 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
832 	struct tipc_msg *hdr = &tsk->phdr;
833 	struct sk_buff_head pkts;
834 	int mtu, rc;
835 
836 	/* Complete message header */
837 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
838 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
839 	msg_set_destport(hdr, dport);
840 	msg_set_destnode(hdr, dnode);
841 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
842 
843 	/* Build message as chain of buffers */
844 	skb_queue_head_init(&pkts);
845 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
846 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
847 	if (unlikely(rc != dlen))
848 		return rc;
849 
850 	/* Send message */
851 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
852 	if (unlikely(rc == -ELINKCONG)) {
853 		tipc_dest_push(&tsk->cong_links, dnode, 0);
854 		tsk->cong_link_cnt++;
855 	}
856 
857 	/* Update send window */
858 	tipc_group_update_member(mb, blks);
859 
860 	/* A broadcast sent within next EXPIRE period must follow same path */
861 	method->rcast = true;
862 	method->mandatory = true;
863 	return dlen;
864 }
865 
866 /**
867  * tipc_send_group_unicast - send message to a member in the group
868  * @sock: socket structure
869  * @m: message to send
870  * @dlen: total length of message data
871  * @timeout: timeout to wait for wakeup
872  *
873  * Called from function tipc_sendmsg(), which has done all sanity checks
874  * Returns the number of bytes sent on success, or errno
875  */
876 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
877 				   int dlen, long timeout)
878 {
879 	struct sock *sk = sock->sk;
880 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
881 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
882 	struct tipc_sock *tsk = tipc_sk(sk);
883 	struct tipc_group *grp = tsk->group;
884 	struct net *net = sock_net(sk);
885 	struct tipc_member *mb = NULL;
886 	u32 node, port;
887 	int rc;
888 
889 	node = dest->addr.id.node;
890 	port = dest->addr.id.ref;
891 	if (!port && !node)
892 		return -EHOSTUNREACH;
893 
894 	/* Block or return if destination link or member is congested */
895 	rc = tipc_wait_for_cond(sock, &timeout,
896 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
897 				!tipc_group_cong(grp, node, port, blks, &mb));
898 	if (unlikely(rc))
899 		return rc;
900 
901 	if (unlikely(!mb))
902 		return -EHOSTUNREACH;
903 
904 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
905 
906 	return rc ? rc : dlen;
907 }
908 
909 /**
910  * tipc_send_group_anycast - send message to any member with given identity
911  * @sock: socket structure
912  * @m: message to send
913  * @dlen: total length of message data
914  * @timeout: timeout to wait for wakeup
915  *
916  * Called from function tipc_sendmsg(), which has done all sanity checks
917  * Returns the number of bytes sent on success, or errno
918  */
919 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
920 				   int dlen, long timeout)
921 {
922 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
923 	struct sock *sk = sock->sk;
924 	struct tipc_sock *tsk = tipc_sk(sk);
925 	struct list_head *cong_links = &tsk->cong_links;
926 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
927 	struct tipc_group *grp = tsk->group;
928 	struct tipc_msg *hdr = &tsk->phdr;
929 	struct tipc_member *first = NULL;
930 	struct tipc_member *mbr = NULL;
931 	struct net *net = sock_net(sk);
932 	u32 node, port, exclude;
933 	struct list_head dsts;
934 	u32 type, inst, scope;
935 	int lookups = 0;
936 	int dstcnt, rc;
937 	bool cong;
938 
939 	INIT_LIST_HEAD(&dsts);
940 
941 	type = msg_nametype(hdr);
942 	inst = dest->addr.name.name.instance;
943 	scope = msg_lookup_scope(hdr);
944 	exclude = tipc_group_exclude(grp);
945 
946 	while (++lookups < 4) {
947 		first = NULL;
948 
949 		/* Look for a non-congested destination member, if any */
950 		while (1) {
951 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
952 						 &dstcnt, exclude, false))
953 				return -EHOSTUNREACH;
954 			tipc_dest_pop(&dsts, &node, &port);
955 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
956 			if (!cong)
957 				break;
958 			if (mbr == first)
959 				break;
960 			if (!first)
961 				first = mbr;
962 		}
963 
964 		/* Start over if destination was not in member list */
965 		if (unlikely(!mbr))
966 			continue;
967 
968 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
969 			break;
970 
971 		/* Block or return if destination link or member is congested */
972 		rc = tipc_wait_for_cond(sock, &timeout,
973 					!tipc_dest_find(cong_links, node, 0) &&
974 					!tipc_group_cong(grp, node, port,
975 							 blks, &mbr));
976 		if (unlikely(rc))
977 			return rc;
978 
979 		/* Send, unless destination disappeared while waiting */
980 		if (likely(mbr))
981 			break;
982 	}
983 
984 	if (unlikely(lookups >= 4))
985 		return -EHOSTUNREACH;
986 
987 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
988 
989 	return rc ? rc : dlen;
990 }
991 
992 /**
993  * tipc_send_group_bcast - send message to all members in communication group
994  * @sk: socket structure
995  * @m: message to send
996  * @dlen: total length of message data
997  * @timeout: timeout to wait for wakeup
998  *
999  * Called from function tipc_sendmsg(), which has done all sanity checks
1000  * Returns the number of bytes sent on success, or errno
1001  */
1002 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1003 				 int dlen, long timeout)
1004 {
1005 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1006 	struct sock *sk = sock->sk;
1007 	struct net *net = sock_net(sk);
1008 	struct tipc_sock *tsk = tipc_sk(sk);
1009 	struct tipc_group *grp = tsk->group;
1010 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1011 	struct tipc_mc_method *method = &tsk->mc_method;
1012 	bool ack = method->mandatory && method->rcast;
1013 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1014 	struct tipc_msg *hdr = &tsk->phdr;
1015 	int mtu = tipc_bcast_get_mtu(net);
1016 	struct sk_buff_head pkts;
1017 	int rc = -EHOSTUNREACH;
1018 
1019 	if (!dsts->local && !dsts->remote)
1020 		return -EHOSTUNREACH;
1021 
1022 	/* Block or return if any destination link or member is congested */
1023 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1024 				!tipc_group_bc_cong(grp, blks));
1025 	if (unlikely(rc))
1026 		return rc;
1027 
1028 	/* Complete message header */
1029 	if (dest) {
1030 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1031 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1032 	} else {
1033 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1034 		msg_set_nameinst(hdr, 0);
1035 	}
1036 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1037 	msg_set_destport(hdr, 0);
1038 	msg_set_destnode(hdr, 0);
1039 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1040 
1041 	/* Avoid getting stuck with repeated forced replicasts */
1042 	msg_set_grp_bc_ack_req(hdr, ack);
1043 
1044 	/* Build message as chain of buffers */
1045 	skb_queue_head_init(&pkts);
1046 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1047 	if (unlikely(rc != dlen))
1048 		return rc;
1049 
1050 	/* Send message */
1051 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1052 	if (unlikely(rc))
1053 		return rc;
1054 
1055 	/* Update broadcast sequence number and send windows */
1056 	tipc_group_update_bc_members(tsk->group, blks, ack);
1057 
1058 	/* Broadcast link is now free to choose method for next broadcast */
1059 	method->mandatory = false;
1060 	method->expires = jiffies;
1061 
1062 	return dlen;
1063 }
1064 
1065 /**
1066  * tipc_send_group_mcast - send message to all members with given identity
1067  * @sock: socket structure
1068  * @m: message to send
1069  * @dlen: total length of message data
1070  * @timeout: timeout to wait for wakeup
1071  *
1072  * Called from function tipc_sendmsg(), which has done all sanity checks
1073  * Returns the number of bytes sent on success, or errno
1074  */
1075 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1076 				 int dlen, long timeout)
1077 {
1078 	struct sock *sk = sock->sk;
1079 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1080 	struct tipc_sock *tsk = tipc_sk(sk);
1081 	struct tipc_group *grp = tsk->group;
1082 	struct tipc_msg *hdr = &tsk->phdr;
1083 	struct net *net = sock_net(sk);
1084 	u32 type, inst, scope, exclude;
1085 	struct list_head dsts;
1086 	u32 dstcnt;
1087 
1088 	INIT_LIST_HEAD(&dsts);
1089 
1090 	type = msg_nametype(hdr);
1091 	inst = dest->addr.name.name.instance;
1092 	scope = msg_lookup_scope(hdr);
1093 	exclude = tipc_group_exclude(grp);
1094 
1095 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1096 				 &dstcnt, exclude, true))
1097 		return -EHOSTUNREACH;
1098 
1099 	if (dstcnt == 1) {
1100 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1101 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1102 	}
1103 
1104 	tipc_dest_list_purge(&dsts);
1105 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1106 }
1107 
1108 /**
1109  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1110  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1111  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1112  *
1113  * Multi-threaded: parallel calls with reference to same queues may occur
1114  */
1115 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1116 		       struct sk_buff_head *inputq)
1117 {
1118 	u32 self = tipc_own_addr(net);
1119 	u32 type, lower, upper, scope;
1120 	struct sk_buff *skb, *_skb;
1121 	u32 portid, onode;
1122 	struct sk_buff_head tmpq;
1123 	struct list_head dports;
1124 	struct tipc_msg *hdr;
1125 	int user, mtyp, hlen;
1126 	bool exact;
1127 
1128 	__skb_queue_head_init(&tmpq);
1129 	INIT_LIST_HEAD(&dports);
1130 
1131 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1132 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1133 		hdr = buf_msg(skb);
1134 		user = msg_user(hdr);
1135 		mtyp = msg_type(hdr);
1136 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1137 		onode = msg_orignode(hdr);
1138 		type = msg_nametype(hdr);
1139 
1140 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1141 			spin_lock_bh(&inputq->lock);
1142 			if (skb_peek(arrvq) == skb) {
1143 				__skb_dequeue(arrvq);
1144 				__skb_queue_tail(inputq, skb);
1145 			}
1146 			kfree_skb(skb);
1147 			spin_unlock_bh(&inputq->lock);
1148 			continue;
1149 		}
1150 
1151 		/* Group messages require exact scope match */
1152 		if (msg_in_group(hdr)) {
1153 			lower = 0;
1154 			upper = ~0;
1155 			scope = msg_lookup_scope(hdr);
1156 			exact = true;
1157 		} else {
1158 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1159 			if (onode == self)
1160 				scope = TIPC_NODE_SCOPE;
1161 			else
1162 				scope = TIPC_CLUSTER_SCOPE;
1163 			exact = false;
1164 			lower = msg_namelower(hdr);
1165 			upper = msg_nameupper(hdr);
1166 		}
1167 
1168 		/* Create destination port list: */
1169 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1170 				       scope, exact, &dports);
1171 
1172 		/* Clone message per destination */
1173 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1174 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1175 			if (_skb) {
1176 				msg_set_destport(buf_msg(_skb), portid);
1177 				__skb_queue_tail(&tmpq, _skb);
1178 				continue;
1179 			}
1180 			pr_warn("Failed to clone mcast rcv buffer\n");
1181 		}
1182 		/* Append to inputq if not already done by other thread */
1183 		spin_lock_bh(&inputq->lock);
1184 		if (skb_peek(arrvq) == skb) {
1185 			skb_queue_splice_tail_init(&tmpq, inputq);
1186 			kfree_skb(__skb_dequeue(arrvq));
1187 		}
1188 		spin_unlock_bh(&inputq->lock);
1189 		__skb_queue_purge(&tmpq);
1190 		kfree_skb(skb);
1191 	}
1192 	tipc_sk_rcv(net, inputq);
1193 }
1194 
1195 /**
1196  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1197  * @tsk: receiving socket
1198  * @skb: pointer to message buffer.
1199  */
1200 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1201 				   struct sk_buff_head *inputq,
1202 				   struct sk_buff_head *xmitq)
1203 {
1204 	struct tipc_msg *hdr = buf_msg(skb);
1205 	u32 onode = tsk_own_node(tsk);
1206 	struct sock *sk = &tsk->sk;
1207 	int mtyp = msg_type(hdr);
1208 	bool conn_cong;
1209 
1210 	/* Ignore if connection cannot be validated: */
1211 	if (!tsk_peer_msg(tsk, hdr))
1212 		goto exit;
1213 
1214 	if (unlikely(msg_errcode(hdr))) {
1215 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1216 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1217 				      tsk_peer_port(tsk));
1218 		sk->sk_state_change(sk);
1219 
1220 		/* State change is ignored if socket already awake,
1221 		 * - convert msg to abort msg and add to inqueue
1222 		 */
1223 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1224 		msg_set_type(hdr, TIPC_CONN_MSG);
1225 		msg_set_size(hdr, BASIC_H_SIZE);
1226 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1227 		__skb_queue_tail(inputq, skb);
1228 		return;
1229 	}
1230 
1231 	tsk->probe_unacked = false;
1232 
1233 	if (mtyp == CONN_PROBE) {
1234 		msg_set_type(hdr, CONN_PROBE_REPLY);
1235 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1236 			__skb_queue_tail(xmitq, skb);
1237 		return;
1238 	} else if (mtyp == CONN_ACK) {
1239 		conn_cong = tsk_conn_cong(tsk);
1240 		tsk->snt_unacked -= msg_conn_ack(hdr);
1241 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1242 			tsk->snd_win = msg_adv_win(hdr);
1243 		if (conn_cong)
1244 			sk->sk_write_space(sk);
1245 	} else if (mtyp != CONN_PROBE_REPLY) {
1246 		pr_warn("Received unknown CONN_PROTO msg\n");
1247 	}
1248 exit:
1249 	kfree_skb(skb);
1250 }
1251 
1252 /**
1253  * tipc_sendmsg - send message in connectionless manner
1254  * @sock: socket structure
1255  * @m: message to send
1256  * @dsz: amount of user data to be sent
1257  *
1258  * Message must have an destination specified explicitly.
1259  * Used for SOCK_RDM and SOCK_DGRAM messages,
1260  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1261  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1262  *
1263  * Returns the number of bytes sent on success, or errno otherwise
1264  */
1265 static int tipc_sendmsg(struct socket *sock,
1266 			struct msghdr *m, size_t dsz)
1267 {
1268 	struct sock *sk = sock->sk;
1269 	int ret;
1270 
1271 	lock_sock(sk);
1272 	ret = __tipc_sendmsg(sock, m, dsz);
1273 	release_sock(sk);
1274 
1275 	return ret;
1276 }
1277 
1278 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1279 {
1280 	struct sock *sk = sock->sk;
1281 	struct net *net = sock_net(sk);
1282 	struct tipc_sock *tsk = tipc_sk(sk);
1283 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1284 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1285 	struct list_head *clinks = &tsk->cong_links;
1286 	bool syn = !tipc_sk_type_connectionless(sk);
1287 	struct tipc_group *grp = tsk->group;
1288 	struct tipc_msg *hdr = &tsk->phdr;
1289 	struct tipc_name_seq *seq;
1290 	struct sk_buff_head pkts;
1291 	u32 dport, dnode = 0;
1292 	u32 type, inst;
1293 	int mtu, rc;
1294 
1295 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1296 		return -EMSGSIZE;
1297 
1298 	if (likely(dest)) {
1299 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1300 			return -EINVAL;
1301 		if (unlikely(dest->family != AF_TIPC))
1302 			return -EINVAL;
1303 	}
1304 
1305 	if (grp) {
1306 		if (!dest)
1307 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1308 		if (dest->addrtype == TIPC_ADDR_NAME)
1309 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1310 		if (dest->addrtype == TIPC_ADDR_ID)
1311 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1312 		if (dest->addrtype == TIPC_ADDR_MCAST)
1313 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1314 		return -EINVAL;
1315 	}
1316 
1317 	if (unlikely(!dest)) {
1318 		dest = &tsk->peer;
1319 		if (!syn || dest->family != AF_TIPC)
1320 			return -EDESTADDRREQ;
1321 	}
1322 
1323 	if (unlikely(syn)) {
1324 		if (sk->sk_state == TIPC_LISTEN)
1325 			return -EPIPE;
1326 		if (sk->sk_state != TIPC_OPEN)
1327 			return -EISCONN;
1328 		if (tsk->published)
1329 			return -EOPNOTSUPP;
1330 		if (dest->addrtype == TIPC_ADDR_NAME) {
1331 			tsk->conn_type = dest->addr.name.name.type;
1332 			tsk->conn_instance = dest->addr.name.name.instance;
1333 		}
1334 		msg_set_syn(hdr, 1);
1335 	}
1336 
1337 	seq = &dest->addr.nameseq;
1338 	if (dest->addrtype == TIPC_ADDR_MCAST)
1339 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1340 
1341 	if (dest->addrtype == TIPC_ADDR_NAME) {
1342 		type = dest->addr.name.name.type;
1343 		inst = dest->addr.name.name.instance;
1344 		dnode = dest->addr.name.domain;
1345 		msg_set_type(hdr, TIPC_NAMED_MSG);
1346 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1347 		msg_set_nametype(hdr, type);
1348 		msg_set_nameinst(hdr, inst);
1349 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1350 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1351 		msg_set_destnode(hdr, dnode);
1352 		msg_set_destport(hdr, dport);
1353 		if (unlikely(!dport && !dnode))
1354 			return -EHOSTUNREACH;
1355 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1356 		dnode = dest->addr.id.node;
1357 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1358 		msg_set_lookup_scope(hdr, 0);
1359 		msg_set_destnode(hdr, dnode);
1360 		msg_set_destport(hdr, dest->addr.id.ref);
1361 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1362 	} else {
1363 		return -EINVAL;
1364 	}
1365 
1366 	/* Block or return if destination link is congested */
1367 	rc = tipc_wait_for_cond(sock, &timeout,
1368 				!tipc_dest_find(clinks, dnode, 0));
1369 	if (unlikely(rc))
1370 		return rc;
1371 
1372 	skb_queue_head_init(&pkts);
1373 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1374 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1375 	if (unlikely(rc != dlen))
1376 		return rc;
1377 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1378 		return -ENOMEM;
1379 
1380 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1381 	if (unlikely(rc == -ELINKCONG)) {
1382 		tipc_dest_push(clinks, dnode, 0);
1383 		tsk->cong_link_cnt++;
1384 		rc = 0;
1385 	}
1386 
1387 	if (unlikely(syn && !rc))
1388 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1389 
1390 	return rc ? rc : dlen;
1391 }
1392 
1393 /**
1394  * tipc_sendstream - send stream-oriented data
1395  * @sock: socket structure
1396  * @m: data to send
1397  * @dsz: total length of data to be transmitted
1398  *
1399  * Used for SOCK_STREAM data.
1400  *
1401  * Returns the number of bytes sent on success (or partial success),
1402  * or errno if no data sent
1403  */
1404 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1405 {
1406 	struct sock *sk = sock->sk;
1407 	int ret;
1408 
1409 	lock_sock(sk);
1410 	ret = __tipc_sendstream(sock, m, dsz);
1411 	release_sock(sk);
1412 
1413 	return ret;
1414 }
1415 
1416 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1417 {
1418 	struct sock *sk = sock->sk;
1419 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1420 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1421 	struct tipc_sock *tsk = tipc_sk(sk);
1422 	struct tipc_msg *hdr = &tsk->phdr;
1423 	struct net *net = sock_net(sk);
1424 	struct sk_buff_head pkts;
1425 	u32 dnode = tsk_peer_node(tsk);
1426 	int send, sent = 0;
1427 	int rc = 0;
1428 
1429 	skb_queue_head_init(&pkts);
1430 
1431 	if (unlikely(dlen > INT_MAX))
1432 		return -EMSGSIZE;
1433 
1434 	/* Handle implicit connection setup */
1435 	if (unlikely(dest)) {
1436 		rc = __tipc_sendmsg(sock, m, dlen);
1437 		if (dlen && dlen == rc) {
1438 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1439 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1440 		}
1441 		return rc;
1442 	}
1443 
1444 	do {
1445 		rc = tipc_wait_for_cond(sock, &timeout,
1446 					(!tsk->cong_link_cnt &&
1447 					 !tsk_conn_cong(tsk) &&
1448 					 tipc_sk_connected(sk)));
1449 		if (unlikely(rc))
1450 			break;
1451 
1452 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1453 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1454 		if (unlikely(rc != send))
1455 			break;
1456 
1457 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1458 		if (unlikely(rc == -ELINKCONG)) {
1459 			tsk->cong_link_cnt = 1;
1460 			rc = 0;
1461 		}
1462 		if (likely(!rc)) {
1463 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1464 			sent += send;
1465 		}
1466 	} while (sent < dlen && !rc);
1467 
1468 	return sent ? sent : rc;
1469 }
1470 
1471 /**
1472  * tipc_send_packet - send a connection-oriented message
1473  * @sock: socket structure
1474  * @m: message to send
1475  * @dsz: length of data to be transmitted
1476  *
1477  * Used for SOCK_SEQPACKET messages.
1478  *
1479  * Returns the number of bytes sent on success, or errno otherwise
1480  */
1481 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1482 {
1483 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1484 		return -EMSGSIZE;
1485 
1486 	return tipc_sendstream(sock, m, dsz);
1487 }
1488 
1489 /* tipc_sk_finish_conn - complete the setup of a connection
1490  */
1491 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1492 				u32 peer_node)
1493 {
1494 	struct sock *sk = &tsk->sk;
1495 	struct net *net = sock_net(sk);
1496 	struct tipc_msg *msg = &tsk->phdr;
1497 
1498 	msg_set_syn(msg, 0);
1499 	msg_set_destnode(msg, peer_node);
1500 	msg_set_destport(msg, peer_port);
1501 	msg_set_type(msg, TIPC_CONN_MSG);
1502 	msg_set_lookup_scope(msg, 0);
1503 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1504 
1505 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1506 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1507 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1508 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1509 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1510 	__skb_queue_purge(&sk->sk_write_queue);
1511 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1512 		return;
1513 
1514 	/* Fall back to message based flow control */
1515 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1516 	tsk->snd_win = FLOWCTL_MSG_WIN;
1517 }
1518 
1519 /**
1520  * tipc_sk_set_orig_addr - capture sender's address for received message
1521  * @m: descriptor for message info
1522  * @hdr: received message header
1523  *
1524  * Note: Address is not captured if not requested by receiver.
1525  */
1526 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1527 {
1528 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1529 	struct tipc_msg *hdr = buf_msg(skb);
1530 
1531 	if (!srcaddr)
1532 		return;
1533 
1534 	srcaddr->sock.family = AF_TIPC;
1535 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1536 	srcaddr->sock.scope = 0;
1537 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1538 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1539 	srcaddr->sock.addr.name.domain = 0;
1540 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1541 
1542 	if (!msg_in_group(hdr))
1543 		return;
1544 
1545 	/* Group message users may also want to know sending member's id */
1546 	srcaddr->member.family = AF_TIPC;
1547 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1548 	srcaddr->member.scope = 0;
1549 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1550 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1551 	srcaddr->member.addr.name.domain = 0;
1552 	m->msg_namelen = sizeof(*srcaddr);
1553 }
1554 
1555 /**
1556  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1557  * @m: descriptor for message info
1558  * @msg: received message header
1559  * @tsk: TIPC port associated with message
1560  *
1561  * Note: Ancillary data is not captured if not requested by receiver.
1562  *
1563  * Returns 0 if successful, otherwise errno
1564  */
1565 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1566 				 struct tipc_sock *tsk)
1567 {
1568 	u32 anc_data[3];
1569 	u32 err;
1570 	u32 dest_type;
1571 	int has_name;
1572 	int res;
1573 
1574 	if (likely(m->msg_controllen == 0))
1575 		return 0;
1576 
1577 	/* Optionally capture errored message object(s) */
1578 	err = msg ? msg_errcode(msg) : 0;
1579 	if (unlikely(err)) {
1580 		anc_data[0] = err;
1581 		anc_data[1] = msg_data_sz(msg);
1582 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1583 		if (res)
1584 			return res;
1585 		if (anc_data[1]) {
1586 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1587 				       msg_data(msg));
1588 			if (res)
1589 				return res;
1590 		}
1591 	}
1592 
1593 	/* Optionally capture message destination object */
1594 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1595 	switch (dest_type) {
1596 	case TIPC_NAMED_MSG:
1597 		has_name = 1;
1598 		anc_data[0] = msg_nametype(msg);
1599 		anc_data[1] = msg_namelower(msg);
1600 		anc_data[2] = msg_namelower(msg);
1601 		break;
1602 	case TIPC_MCAST_MSG:
1603 		has_name = 1;
1604 		anc_data[0] = msg_nametype(msg);
1605 		anc_data[1] = msg_namelower(msg);
1606 		anc_data[2] = msg_nameupper(msg);
1607 		break;
1608 	case TIPC_CONN_MSG:
1609 		has_name = (tsk->conn_type != 0);
1610 		anc_data[0] = tsk->conn_type;
1611 		anc_data[1] = tsk->conn_instance;
1612 		anc_data[2] = tsk->conn_instance;
1613 		break;
1614 	default:
1615 		has_name = 0;
1616 	}
1617 	if (has_name) {
1618 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1619 		if (res)
1620 			return res;
1621 	}
1622 
1623 	return 0;
1624 }
1625 
1626 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1627 {
1628 	struct sock *sk = &tsk->sk;
1629 	struct net *net = sock_net(sk);
1630 	struct sk_buff *skb = NULL;
1631 	struct tipc_msg *msg;
1632 	u32 peer_port = tsk_peer_port(tsk);
1633 	u32 dnode = tsk_peer_node(tsk);
1634 
1635 	if (!tipc_sk_connected(sk))
1636 		return;
1637 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1638 			      dnode, tsk_own_node(tsk), peer_port,
1639 			      tsk->portid, TIPC_OK);
1640 	if (!skb)
1641 		return;
1642 	msg = buf_msg(skb);
1643 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1644 	tsk->rcv_unacked = 0;
1645 
1646 	/* Adjust to and advertize the correct window limit */
1647 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1648 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1649 		msg_set_adv_win(msg, tsk->rcv_win);
1650 	}
1651 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1652 }
1653 
1654 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1655 {
1656 	struct sock *sk = sock->sk;
1657 	DEFINE_WAIT(wait);
1658 	long timeo = *timeop;
1659 	int err = sock_error(sk);
1660 
1661 	if (err)
1662 		return err;
1663 
1664 	for (;;) {
1665 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1666 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1667 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1668 				err = -ENOTCONN;
1669 				break;
1670 			}
1671 			release_sock(sk);
1672 			timeo = schedule_timeout(timeo);
1673 			lock_sock(sk);
1674 		}
1675 		err = 0;
1676 		if (!skb_queue_empty(&sk->sk_receive_queue))
1677 			break;
1678 		err = -EAGAIN;
1679 		if (!timeo)
1680 			break;
1681 		err = sock_intr_errno(timeo);
1682 		if (signal_pending(current))
1683 			break;
1684 
1685 		err = sock_error(sk);
1686 		if (err)
1687 			break;
1688 	}
1689 	finish_wait(sk_sleep(sk), &wait);
1690 	*timeop = timeo;
1691 	return err;
1692 }
1693 
1694 /**
1695  * tipc_recvmsg - receive packet-oriented message
1696  * @m: descriptor for message info
1697  * @buflen: length of user buffer area
1698  * @flags: receive flags
1699  *
1700  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1701  * If the complete message doesn't fit in user area, truncate it.
1702  *
1703  * Returns size of returned message data, errno otherwise
1704  */
1705 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1706 			size_t buflen,	int flags)
1707 {
1708 	struct sock *sk = sock->sk;
1709 	bool connected = !tipc_sk_type_connectionless(sk);
1710 	struct tipc_sock *tsk = tipc_sk(sk);
1711 	int rc, err, hlen, dlen, copy;
1712 	struct sk_buff_head xmitq;
1713 	struct tipc_msg *hdr;
1714 	struct sk_buff *skb;
1715 	bool grp_evt;
1716 	long timeout;
1717 
1718 	/* Catch invalid receive requests */
1719 	if (unlikely(!buflen))
1720 		return -EINVAL;
1721 
1722 	lock_sock(sk);
1723 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1724 		rc = -ENOTCONN;
1725 		goto exit;
1726 	}
1727 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1728 
1729 	/* Step rcv queue to first msg with data or error; wait if necessary */
1730 	do {
1731 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1732 		if (unlikely(rc))
1733 			goto exit;
1734 		skb = skb_peek(&sk->sk_receive_queue);
1735 		hdr = buf_msg(skb);
1736 		dlen = msg_data_sz(hdr);
1737 		hlen = msg_hdr_sz(hdr);
1738 		err = msg_errcode(hdr);
1739 		grp_evt = msg_is_grp_evt(hdr);
1740 		if (likely(dlen || err))
1741 			break;
1742 		tsk_advance_rx_queue(sk);
1743 	} while (1);
1744 
1745 	/* Collect msg meta data, including error code and rejected data */
1746 	tipc_sk_set_orig_addr(m, skb);
1747 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1748 	if (unlikely(rc))
1749 		goto exit;
1750 
1751 	/* Capture data if non-error msg, otherwise just set return value */
1752 	if (likely(!err)) {
1753 		copy = min_t(int, dlen, buflen);
1754 		if (unlikely(copy != dlen))
1755 			m->msg_flags |= MSG_TRUNC;
1756 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1757 	} else {
1758 		copy = 0;
1759 		rc = 0;
1760 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1761 			rc = -ECONNRESET;
1762 	}
1763 	if (unlikely(rc))
1764 		goto exit;
1765 
1766 	/* Mark message as group event if applicable */
1767 	if (unlikely(grp_evt)) {
1768 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1769 			m->msg_flags |= MSG_EOR;
1770 		m->msg_flags |= MSG_OOB;
1771 		copy = 0;
1772 	}
1773 
1774 	/* Caption of data or error code/rejected data was successful */
1775 	if (unlikely(flags & MSG_PEEK))
1776 		goto exit;
1777 
1778 	/* Send group flow control advertisement when applicable */
1779 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1780 		skb_queue_head_init(&xmitq);
1781 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1782 					  msg_orignode(hdr), msg_origport(hdr),
1783 					  &xmitq);
1784 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1785 	}
1786 
1787 	tsk_advance_rx_queue(sk);
1788 
1789 	if (likely(!connected))
1790 		goto exit;
1791 
1792 	/* Send connection flow control advertisement when applicable */
1793 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1794 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1795 		tipc_sk_send_ack(tsk);
1796 exit:
1797 	release_sock(sk);
1798 	return rc ? rc : copy;
1799 }
1800 
1801 /**
1802  * tipc_recvstream - receive stream-oriented data
1803  * @m: descriptor for message info
1804  * @buflen: total size of user buffer area
1805  * @flags: receive flags
1806  *
1807  * Used for SOCK_STREAM messages only.  If not enough data is available
1808  * will optionally wait for more; never truncates data.
1809  *
1810  * Returns size of returned message data, errno otherwise
1811  */
1812 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1813 			   size_t buflen, int flags)
1814 {
1815 	struct sock *sk = sock->sk;
1816 	struct tipc_sock *tsk = tipc_sk(sk);
1817 	struct sk_buff *skb;
1818 	struct tipc_msg *hdr;
1819 	struct tipc_skb_cb *skb_cb;
1820 	bool peek = flags & MSG_PEEK;
1821 	int offset, required, copy, copied = 0;
1822 	int hlen, dlen, err, rc;
1823 	long timeout;
1824 
1825 	/* Catch invalid receive attempts */
1826 	if (unlikely(!buflen))
1827 		return -EINVAL;
1828 
1829 	lock_sock(sk);
1830 
1831 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1832 		rc = -ENOTCONN;
1833 		goto exit;
1834 	}
1835 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1836 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1837 
1838 	do {
1839 		/* Look at first msg in receive queue; wait if necessary */
1840 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1841 		if (unlikely(rc))
1842 			break;
1843 		skb = skb_peek(&sk->sk_receive_queue);
1844 		skb_cb = TIPC_SKB_CB(skb);
1845 		hdr = buf_msg(skb);
1846 		dlen = msg_data_sz(hdr);
1847 		hlen = msg_hdr_sz(hdr);
1848 		err = msg_errcode(hdr);
1849 
1850 		/* Discard any empty non-errored (SYN-) message */
1851 		if (unlikely(!dlen && !err)) {
1852 			tsk_advance_rx_queue(sk);
1853 			continue;
1854 		}
1855 
1856 		/* Collect msg meta data, incl. error code and rejected data */
1857 		if (!copied) {
1858 			tipc_sk_set_orig_addr(m, skb);
1859 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1860 			if (rc)
1861 				break;
1862 		}
1863 
1864 		/* Copy data if msg ok, otherwise return error/partial data */
1865 		if (likely(!err)) {
1866 			offset = skb_cb->bytes_read;
1867 			copy = min_t(int, dlen - offset, buflen - copied);
1868 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1869 			if (unlikely(rc))
1870 				break;
1871 			copied += copy;
1872 			offset += copy;
1873 			if (unlikely(offset < dlen)) {
1874 				if (!peek)
1875 					skb_cb->bytes_read = offset;
1876 				break;
1877 			}
1878 		} else {
1879 			rc = 0;
1880 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1881 				rc = -ECONNRESET;
1882 			if (copied || rc)
1883 				break;
1884 		}
1885 
1886 		if (unlikely(peek))
1887 			break;
1888 
1889 		tsk_advance_rx_queue(sk);
1890 
1891 		/* Send connection flow control advertisement when applicable */
1892 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1893 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1894 			tipc_sk_send_ack(tsk);
1895 
1896 		/* Exit if all requested data or FIN/error received */
1897 		if (copied == buflen || err)
1898 			break;
1899 
1900 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1901 exit:
1902 	release_sock(sk);
1903 	return copied ? copied : rc;
1904 }
1905 
1906 /**
1907  * tipc_write_space - wake up thread if port congestion is released
1908  * @sk: socket
1909  */
1910 static void tipc_write_space(struct sock *sk)
1911 {
1912 	struct socket_wq *wq;
1913 
1914 	rcu_read_lock();
1915 	wq = rcu_dereference(sk->sk_wq);
1916 	if (skwq_has_sleeper(wq))
1917 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1918 						EPOLLWRNORM | EPOLLWRBAND);
1919 	rcu_read_unlock();
1920 }
1921 
1922 /**
1923  * tipc_data_ready - wake up threads to indicate messages have been received
1924  * @sk: socket
1925  * @len: the length of messages
1926  */
1927 static void tipc_data_ready(struct sock *sk)
1928 {
1929 	struct socket_wq *wq;
1930 
1931 	rcu_read_lock();
1932 	wq = rcu_dereference(sk->sk_wq);
1933 	if (skwq_has_sleeper(wq))
1934 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1935 						EPOLLRDNORM | EPOLLRDBAND);
1936 	rcu_read_unlock();
1937 }
1938 
1939 static void tipc_sock_destruct(struct sock *sk)
1940 {
1941 	__skb_queue_purge(&sk->sk_receive_queue);
1942 }
1943 
1944 static void tipc_sk_proto_rcv(struct sock *sk,
1945 			      struct sk_buff_head *inputq,
1946 			      struct sk_buff_head *xmitq)
1947 {
1948 	struct sk_buff *skb = __skb_dequeue(inputq);
1949 	struct tipc_sock *tsk = tipc_sk(sk);
1950 	struct tipc_msg *hdr = buf_msg(skb);
1951 	struct tipc_group *grp = tsk->group;
1952 	bool wakeup = false;
1953 
1954 	switch (msg_user(hdr)) {
1955 	case CONN_MANAGER:
1956 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1957 		return;
1958 	case SOCK_WAKEUP:
1959 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1960 		tsk->cong_link_cnt--;
1961 		wakeup = true;
1962 		break;
1963 	case GROUP_PROTOCOL:
1964 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1965 		break;
1966 	case TOP_SRV:
1967 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1968 				      hdr, inputq, xmitq);
1969 		break;
1970 	default:
1971 		break;
1972 	}
1973 
1974 	if (wakeup)
1975 		sk->sk_write_space(sk);
1976 
1977 	kfree_skb(skb);
1978 }
1979 
1980 /**
1981  * tipc_sk_filter_connect - check incoming message for a connection-based socket
1982  * @tsk: TIPC socket
1983  * @skb: pointer to message buffer.
1984  * Returns true if message should be added to receive queue, false otherwise
1985  */
1986 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1987 {
1988 	struct sock *sk = &tsk->sk;
1989 	struct net *net = sock_net(sk);
1990 	struct tipc_msg *hdr = buf_msg(skb);
1991 	bool con_msg = msg_connected(hdr);
1992 	u32 pport = tsk_peer_port(tsk);
1993 	u32 pnode = tsk_peer_node(tsk);
1994 	u32 oport = msg_origport(hdr);
1995 	u32 onode = msg_orignode(hdr);
1996 	int err = msg_errcode(hdr);
1997 	unsigned long delay;
1998 
1999 	if (unlikely(msg_mcast(hdr)))
2000 		return false;
2001 
2002 	switch (sk->sk_state) {
2003 	case TIPC_CONNECTING:
2004 		/* Setup ACK */
2005 		if (likely(con_msg)) {
2006 			if (err)
2007 				break;
2008 			tipc_sk_finish_conn(tsk, oport, onode);
2009 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2010 			/* ACK+ message with data is added to receive queue */
2011 			if (msg_data_sz(hdr))
2012 				return true;
2013 			/* Empty ACK-, - wake up sleeping connect() and drop */
2014 			sk->sk_data_ready(sk);
2015 			msg_set_dest_droppable(hdr, 1);
2016 			return false;
2017 		}
2018 		/* Ignore connectionless message if not from listening socket */
2019 		if (oport != pport || onode != pnode)
2020 			return false;
2021 
2022 		/* Rejected SYN */
2023 		if (err != TIPC_ERR_OVERLOAD)
2024 			break;
2025 
2026 		/* Prepare for new setup attempt if we have a SYN clone */
2027 		if (skb_queue_empty(&sk->sk_write_queue))
2028 			break;
2029 		get_random_bytes(&delay, 2);
2030 		delay %= (tsk->conn_timeout / 4);
2031 		delay = msecs_to_jiffies(delay + 100);
2032 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2033 		return false;
2034 	case TIPC_OPEN:
2035 	case TIPC_DISCONNECTING:
2036 		return false;
2037 	case TIPC_LISTEN:
2038 		/* Accept only SYN message */
2039 		if (!msg_is_syn(hdr) &&
2040 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2041 			return false;
2042 		if (!con_msg && !err)
2043 			return true;
2044 		return false;
2045 	case TIPC_ESTABLISHED:
2046 		/* Accept only connection-based messages sent by peer */
2047 		if (likely(con_msg && !err && pport == oport && pnode == onode))
2048 			return true;
2049 		if (!tsk_peer_msg(tsk, hdr))
2050 			return false;
2051 		if (!err)
2052 			return true;
2053 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2054 		tipc_node_remove_conn(net, pnode, tsk->portid);
2055 		sk->sk_state_change(sk);
2056 		return true;
2057 	default:
2058 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2059 	}
2060 	/* Abort connection setup attempt */
2061 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2062 	sk->sk_err = ECONNREFUSED;
2063 	sk->sk_state_change(sk);
2064 	return true;
2065 }
2066 
2067 /**
2068  * rcvbuf_limit - get proper overload limit of socket receive queue
2069  * @sk: socket
2070  * @skb: message
2071  *
2072  * For connection oriented messages, irrespective of importance,
2073  * default queue limit is 2 MB.
2074  *
2075  * For connectionless messages, queue limits are based on message
2076  * importance as follows:
2077  *
2078  * TIPC_LOW_IMPORTANCE       (2 MB)
2079  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2080  * TIPC_HIGH_IMPORTANCE      (8 MB)
2081  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2082  *
2083  * Returns overload limit according to corresponding message importance
2084  */
2085 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2086 {
2087 	struct tipc_sock *tsk = tipc_sk(sk);
2088 	struct tipc_msg *hdr = buf_msg(skb);
2089 
2090 	if (unlikely(msg_in_group(hdr)))
2091 		return sk->sk_rcvbuf;
2092 
2093 	if (unlikely(!msg_connected(hdr)))
2094 		return sk->sk_rcvbuf << msg_importance(hdr);
2095 
2096 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2097 		return sk->sk_rcvbuf;
2098 
2099 	return FLOWCTL_MSG_LIM;
2100 }
2101 
2102 /**
2103  * tipc_sk_filter_rcv - validate incoming message
2104  * @sk: socket
2105  * @skb: pointer to message.
2106  *
2107  * Enqueues message on receive queue if acceptable; optionally handles
2108  * disconnect indication for a connected socket.
2109  *
2110  * Called with socket lock already taken
2111  *
2112  */
2113 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2114 			       struct sk_buff_head *xmitq)
2115 {
2116 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2117 	struct tipc_sock *tsk = tipc_sk(sk);
2118 	struct tipc_group *grp = tsk->group;
2119 	struct tipc_msg *hdr = buf_msg(skb);
2120 	struct net *net = sock_net(sk);
2121 	struct sk_buff_head inputq;
2122 	int limit, err = TIPC_OK;
2123 
2124 	TIPC_SKB_CB(skb)->bytes_read = 0;
2125 	__skb_queue_head_init(&inputq);
2126 	__skb_queue_tail(&inputq, skb);
2127 
2128 	if (unlikely(!msg_isdata(hdr)))
2129 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2130 
2131 	if (unlikely(grp))
2132 		tipc_group_filter_msg(grp, &inputq, xmitq);
2133 
2134 	/* Validate and add to receive buffer if there is space */
2135 	while ((skb = __skb_dequeue(&inputq))) {
2136 		hdr = buf_msg(skb);
2137 		limit = rcvbuf_limit(sk, skb);
2138 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2139 		    (!sk_conn && msg_connected(hdr)) ||
2140 		    (!grp && msg_in_group(hdr)))
2141 			err = TIPC_ERR_NO_PORT;
2142 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2143 			atomic_inc(&sk->sk_drops);
2144 			err = TIPC_ERR_OVERLOAD;
2145 		}
2146 
2147 		if (unlikely(err)) {
2148 			tipc_skb_reject(net, err, skb, xmitq);
2149 			err = TIPC_OK;
2150 			continue;
2151 		}
2152 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2153 		skb_set_owner_r(skb, sk);
2154 		sk->sk_data_ready(sk);
2155 	}
2156 }
2157 
2158 /**
2159  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2160  * @sk: socket
2161  * @skb: message
2162  *
2163  * Caller must hold socket lock
2164  */
2165 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2166 {
2167 	unsigned int before = sk_rmem_alloc_get(sk);
2168 	struct sk_buff_head xmitq;
2169 	unsigned int added;
2170 
2171 	__skb_queue_head_init(&xmitq);
2172 
2173 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2174 	added = sk_rmem_alloc_get(sk) - before;
2175 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2176 
2177 	/* Send pending response/rejected messages, if any */
2178 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2179 	return 0;
2180 }
2181 
2182 /**
2183  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2184  *                   inputq and try adding them to socket or backlog queue
2185  * @inputq: list of incoming buffers with potentially different destinations
2186  * @sk: socket where the buffers should be enqueued
2187  * @dport: port number for the socket
2188  *
2189  * Caller must hold socket lock
2190  */
2191 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2192 			    u32 dport, struct sk_buff_head *xmitq)
2193 {
2194 	unsigned long time_limit = jiffies + 2;
2195 	struct sk_buff *skb;
2196 	unsigned int lim;
2197 	atomic_t *dcnt;
2198 	u32 onode;
2199 
2200 	while (skb_queue_len(inputq)) {
2201 		if (unlikely(time_after_eq(jiffies, time_limit)))
2202 			return;
2203 
2204 		skb = tipc_skb_dequeue(inputq, dport);
2205 		if (unlikely(!skb))
2206 			return;
2207 
2208 		/* Add message directly to receive queue if possible */
2209 		if (!sock_owned_by_user(sk)) {
2210 			tipc_sk_filter_rcv(sk, skb, xmitq);
2211 			continue;
2212 		}
2213 
2214 		/* Try backlog, compensating for double-counted bytes */
2215 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2216 		if (!sk->sk_backlog.len)
2217 			atomic_set(dcnt, 0);
2218 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2219 		if (likely(!sk_add_backlog(sk, skb, lim)))
2220 			continue;
2221 
2222 		/* Overload => reject message back to sender */
2223 		onode = tipc_own_addr(sock_net(sk));
2224 		atomic_inc(&sk->sk_drops);
2225 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2226 			__skb_queue_tail(xmitq, skb);
2227 		break;
2228 	}
2229 }
2230 
2231 /**
2232  * tipc_sk_rcv - handle a chain of incoming buffers
2233  * @inputq: buffer list containing the buffers
2234  * Consumes all buffers in list until inputq is empty
2235  * Note: may be called in multiple threads referring to the same queue
2236  */
2237 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2238 {
2239 	struct sk_buff_head xmitq;
2240 	u32 dnode, dport = 0;
2241 	int err;
2242 	struct tipc_sock *tsk;
2243 	struct sock *sk;
2244 	struct sk_buff *skb;
2245 
2246 	__skb_queue_head_init(&xmitq);
2247 	while (skb_queue_len(inputq)) {
2248 		dport = tipc_skb_peek_port(inputq, dport);
2249 		tsk = tipc_sk_lookup(net, dport);
2250 
2251 		if (likely(tsk)) {
2252 			sk = &tsk->sk;
2253 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2254 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2255 				spin_unlock_bh(&sk->sk_lock.slock);
2256 			}
2257 			/* Send pending response/rejected messages, if any */
2258 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2259 			sock_put(sk);
2260 			continue;
2261 		}
2262 		/* No destination socket => dequeue skb if still there */
2263 		skb = tipc_skb_dequeue(inputq, dport);
2264 		if (!skb)
2265 			return;
2266 
2267 		/* Try secondary lookup if unresolved named message */
2268 		err = TIPC_ERR_NO_PORT;
2269 		if (tipc_msg_lookup_dest(net, skb, &err))
2270 			goto xmit;
2271 
2272 		/* Prepare for message rejection */
2273 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2274 			continue;
2275 xmit:
2276 		dnode = msg_destnode(buf_msg(skb));
2277 		tipc_node_xmit_skb(net, skb, dnode, dport);
2278 	}
2279 }
2280 
2281 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2282 {
2283 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2284 	struct sock *sk = sock->sk;
2285 	int done;
2286 
2287 	do {
2288 		int err = sock_error(sk);
2289 		if (err)
2290 			return err;
2291 		if (!*timeo_p)
2292 			return -ETIMEDOUT;
2293 		if (signal_pending(current))
2294 			return sock_intr_errno(*timeo_p);
2295 
2296 		add_wait_queue(sk_sleep(sk), &wait);
2297 		done = sk_wait_event(sk, timeo_p,
2298 				     sk->sk_state != TIPC_CONNECTING, &wait);
2299 		remove_wait_queue(sk_sleep(sk), &wait);
2300 	} while (!done);
2301 	return 0;
2302 }
2303 
2304 /**
2305  * tipc_connect - establish a connection to another TIPC port
2306  * @sock: socket structure
2307  * @dest: socket address for destination port
2308  * @destlen: size of socket address data structure
2309  * @flags: file-related flags associated with socket
2310  *
2311  * Returns 0 on success, errno otherwise
2312  */
2313 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2314 			int destlen, int flags)
2315 {
2316 	struct sock *sk = sock->sk;
2317 	struct tipc_sock *tsk = tipc_sk(sk);
2318 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2319 	struct msghdr m = {NULL,};
2320 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2321 	int previous;
2322 	int res = 0;
2323 
2324 	if (destlen != sizeof(struct sockaddr_tipc))
2325 		return -EINVAL;
2326 
2327 	lock_sock(sk);
2328 
2329 	if (tsk->group) {
2330 		res = -EINVAL;
2331 		goto exit;
2332 	}
2333 
2334 	if (dst->family == AF_UNSPEC) {
2335 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2336 		if (!tipc_sk_type_connectionless(sk))
2337 			res = -EINVAL;
2338 		goto exit;
2339 	} else if (dst->family != AF_TIPC) {
2340 		res = -EINVAL;
2341 	}
2342 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2343 		res = -EINVAL;
2344 	if (res)
2345 		goto exit;
2346 
2347 	/* DGRAM/RDM connect(), just save the destaddr */
2348 	if (tipc_sk_type_connectionless(sk)) {
2349 		memcpy(&tsk->peer, dest, destlen);
2350 		goto exit;
2351 	}
2352 
2353 	previous = sk->sk_state;
2354 
2355 	switch (sk->sk_state) {
2356 	case TIPC_OPEN:
2357 		/* Send a 'SYN-' to destination */
2358 		m.msg_name = dest;
2359 		m.msg_namelen = destlen;
2360 
2361 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2362 		 * indicate send_msg() is never blocked.
2363 		 */
2364 		if (!timeout)
2365 			m.msg_flags = MSG_DONTWAIT;
2366 
2367 		res = __tipc_sendmsg(sock, &m, 0);
2368 		if ((res < 0) && (res != -EWOULDBLOCK))
2369 			goto exit;
2370 
2371 		/* Just entered TIPC_CONNECTING state; the only
2372 		 * difference is that return value in non-blocking
2373 		 * case is EINPROGRESS, rather than EALREADY.
2374 		 */
2375 		res = -EINPROGRESS;
2376 		/* fall thru' */
2377 	case TIPC_CONNECTING:
2378 		if (!timeout) {
2379 			if (previous == TIPC_CONNECTING)
2380 				res = -EALREADY;
2381 			goto exit;
2382 		}
2383 		timeout = msecs_to_jiffies(timeout);
2384 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2385 		res = tipc_wait_for_connect(sock, &timeout);
2386 		break;
2387 	case TIPC_ESTABLISHED:
2388 		res = -EISCONN;
2389 		break;
2390 	default:
2391 		res = -EINVAL;
2392 	}
2393 
2394 exit:
2395 	release_sock(sk);
2396 	return res;
2397 }
2398 
2399 /**
2400  * tipc_listen - allow socket to listen for incoming connections
2401  * @sock: socket structure
2402  * @len: (unused)
2403  *
2404  * Returns 0 on success, errno otherwise
2405  */
2406 static int tipc_listen(struct socket *sock, int len)
2407 {
2408 	struct sock *sk = sock->sk;
2409 	int res;
2410 
2411 	lock_sock(sk);
2412 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2413 	release_sock(sk);
2414 
2415 	return res;
2416 }
2417 
2418 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2419 {
2420 	struct sock *sk = sock->sk;
2421 	DEFINE_WAIT(wait);
2422 	int err;
2423 
2424 	/* True wake-one mechanism for incoming connections: only
2425 	 * one process gets woken up, not the 'whole herd'.
2426 	 * Since we do not 'race & poll' for established sockets
2427 	 * anymore, the common case will execute the loop only once.
2428 	*/
2429 	for (;;) {
2430 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2431 					  TASK_INTERRUPTIBLE);
2432 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2433 			release_sock(sk);
2434 			timeo = schedule_timeout(timeo);
2435 			lock_sock(sk);
2436 		}
2437 		err = 0;
2438 		if (!skb_queue_empty(&sk->sk_receive_queue))
2439 			break;
2440 		err = -EAGAIN;
2441 		if (!timeo)
2442 			break;
2443 		err = sock_intr_errno(timeo);
2444 		if (signal_pending(current))
2445 			break;
2446 	}
2447 	finish_wait(sk_sleep(sk), &wait);
2448 	return err;
2449 }
2450 
2451 /**
2452  * tipc_accept - wait for connection request
2453  * @sock: listening socket
2454  * @newsock: new socket that is to be connected
2455  * @flags: file-related flags associated with socket
2456  *
2457  * Returns 0 on success, errno otherwise
2458  */
2459 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2460 		       bool kern)
2461 {
2462 	struct sock *new_sk, *sk = sock->sk;
2463 	struct sk_buff *buf;
2464 	struct tipc_sock *new_tsock;
2465 	struct tipc_msg *msg;
2466 	long timeo;
2467 	int res;
2468 
2469 	lock_sock(sk);
2470 
2471 	if (sk->sk_state != TIPC_LISTEN) {
2472 		res = -EINVAL;
2473 		goto exit;
2474 	}
2475 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2476 	res = tipc_wait_for_accept(sock, timeo);
2477 	if (res)
2478 		goto exit;
2479 
2480 	buf = skb_peek(&sk->sk_receive_queue);
2481 
2482 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2483 	if (res)
2484 		goto exit;
2485 	security_sk_clone(sock->sk, new_sock->sk);
2486 
2487 	new_sk = new_sock->sk;
2488 	new_tsock = tipc_sk(new_sk);
2489 	msg = buf_msg(buf);
2490 
2491 	/* we lock on new_sk; but lockdep sees the lock on sk */
2492 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2493 
2494 	/*
2495 	 * Reject any stray messages received by new socket
2496 	 * before the socket lock was taken (very, very unlikely)
2497 	 */
2498 	tsk_rej_rx_queue(new_sk);
2499 
2500 	/* Connect new socket to it's peer */
2501 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2502 
2503 	tsk_set_importance(new_tsock, msg_importance(msg));
2504 	if (msg_named(msg)) {
2505 		new_tsock->conn_type = msg_nametype(msg);
2506 		new_tsock->conn_instance = msg_nameinst(msg);
2507 	}
2508 
2509 	/*
2510 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2511 	 * Respond to 'SYN+' by queuing it on new socket.
2512 	 */
2513 	if (!msg_data_sz(msg)) {
2514 		struct msghdr m = {NULL,};
2515 
2516 		tsk_advance_rx_queue(sk);
2517 		__tipc_sendstream(new_sock, &m, 0);
2518 	} else {
2519 		__skb_dequeue(&sk->sk_receive_queue);
2520 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2521 		skb_set_owner_r(buf, new_sk);
2522 	}
2523 	release_sock(new_sk);
2524 exit:
2525 	release_sock(sk);
2526 	return res;
2527 }
2528 
2529 /**
2530  * tipc_shutdown - shutdown socket connection
2531  * @sock: socket structure
2532  * @how: direction to close (must be SHUT_RDWR)
2533  *
2534  * Terminates connection (if necessary), then purges socket's receive queue.
2535  *
2536  * Returns 0 on success, errno otherwise
2537  */
2538 static int tipc_shutdown(struct socket *sock, int how)
2539 {
2540 	struct sock *sk = sock->sk;
2541 	int res;
2542 
2543 	if (how != SHUT_RDWR)
2544 		return -EINVAL;
2545 
2546 	lock_sock(sk);
2547 
2548 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2549 	sk->sk_shutdown = SEND_SHUTDOWN;
2550 
2551 	if (sk->sk_state == TIPC_DISCONNECTING) {
2552 		/* Discard any unreceived messages */
2553 		__skb_queue_purge(&sk->sk_receive_queue);
2554 
2555 		/* Wake up anyone sleeping in poll */
2556 		sk->sk_state_change(sk);
2557 		res = 0;
2558 	} else {
2559 		res = -ENOTCONN;
2560 	}
2561 
2562 	release_sock(sk);
2563 	return res;
2564 }
2565 
2566 static void tipc_sk_check_probing_state(struct sock *sk,
2567 					struct sk_buff_head *list)
2568 {
2569 	struct tipc_sock *tsk = tipc_sk(sk);
2570 	u32 pnode = tsk_peer_node(tsk);
2571 	u32 pport = tsk_peer_port(tsk);
2572 	u32 self = tsk_own_node(tsk);
2573 	u32 oport = tsk->portid;
2574 	struct sk_buff *skb;
2575 
2576 	if (tsk->probe_unacked) {
2577 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2578 		sk->sk_err = ECONNABORTED;
2579 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2580 		sk->sk_state_change(sk);
2581 		return;
2582 	}
2583 	/* Prepare new probe */
2584 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2585 			      pnode, self, pport, oport, TIPC_OK);
2586 	if (skb)
2587 		__skb_queue_tail(list, skb);
2588 	tsk->probe_unacked = true;
2589 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2590 }
2591 
2592 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2593 {
2594 	struct tipc_sock *tsk = tipc_sk(sk);
2595 
2596 	/* Try again later if dest link is congested */
2597 	if (tsk->cong_link_cnt) {
2598 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2599 		return;
2600 	}
2601 	/* Prepare SYN for retransmit */
2602 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2603 }
2604 
2605 static void tipc_sk_timeout(struct timer_list *t)
2606 {
2607 	struct sock *sk = from_timer(sk, t, sk_timer);
2608 	struct tipc_sock *tsk = tipc_sk(sk);
2609 	u32 pnode = tsk_peer_node(tsk);
2610 	struct sk_buff_head list;
2611 	int rc = 0;
2612 
2613 	skb_queue_head_init(&list);
2614 	bh_lock_sock(sk);
2615 
2616 	/* Try again later if socket is busy */
2617 	if (sock_owned_by_user(sk)) {
2618 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2619 		bh_unlock_sock(sk);
2620 		return;
2621 	}
2622 
2623 	if (sk->sk_state == TIPC_ESTABLISHED)
2624 		tipc_sk_check_probing_state(sk, &list);
2625 	else if (sk->sk_state == TIPC_CONNECTING)
2626 		tipc_sk_retry_connect(sk, &list);
2627 
2628 	bh_unlock_sock(sk);
2629 
2630 	if (!skb_queue_empty(&list))
2631 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2632 
2633 	/* SYN messages may cause link congestion */
2634 	if (rc == -ELINKCONG) {
2635 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2636 		tsk->cong_link_cnt = 1;
2637 	}
2638 	sock_put(sk);
2639 }
2640 
2641 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2642 			   struct tipc_name_seq const *seq)
2643 {
2644 	struct sock *sk = &tsk->sk;
2645 	struct net *net = sock_net(sk);
2646 	struct publication *publ;
2647 	u32 key;
2648 
2649 	if (scope != TIPC_NODE_SCOPE)
2650 		scope = TIPC_CLUSTER_SCOPE;
2651 
2652 	if (tipc_sk_connected(sk))
2653 		return -EINVAL;
2654 	key = tsk->portid + tsk->pub_count + 1;
2655 	if (key == tsk->portid)
2656 		return -EADDRINUSE;
2657 
2658 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2659 				    scope, tsk->portid, key);
2660 	if (unlikely(!publ))
2661 		return -EINVAL;
2662 
2663 	list_add(&publ->binding_sock, &tsk->publications);
2664 	tsk->pub_count++;
2665 	tsk->published = 1;
2666 	return 0;
2667 }
2668 
2669 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2670 			    struct tipc_name_seq const *seq)
2671 {
2672 	struct net *net = sock_net(&tsk->sk);
2673 	struct publication *publ;
2674 	struct publication *safe;
2675 	int rc = -EINVAL;
2676 
2677 	if (scope != TIPC_NODE_SCOPE)
2678 		scope = TIPC_CLUSTER_SCOPE;
2679 
2680 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2681 		if (seq) {
2682 			if (publ->scope != scope)
2683 				continue;
2684 			if (publ->type != seq->type)
2685 				continue;
2686 			if (publ->lower != seq->lower)
2687 				continue;
2688 			if (publ->upper != seq->upper)
2689 				break;
2690 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2691 					      publ->upper, publ->key);
2692 			rc = 0;
2693 			break;
2694 		}
2695 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2696 				      publ->upper, publ->key);
2697 		rc = 0;
2698 	}
2699 	if (list_empty(&tsk->publications))
2700 		tsk->published = 0;
2701 	return rc;
2702 }
2703 
2704 /* tipc_sk_reinit: set non-zero address in all existing sockets
2705  *                 when we go from standalone to network mode.
2706  */
2707 void tipc_sk_reinit(struct net *net)
2708 {
2709 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2710 	struct rhashtable_iter iter;
2711 	struct tipc_sock *tsk;
2712 	struct tipc_msg *msg;
2713 
2714 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2715 
2716 	do {
2717 		rhashtable_walk_start(&iter);
2718 
2719 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2720 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2721 			msg = &tsk->phdr;
2722 			msg_set_prevnode(msg, tipc_own_addr(net));
2723 			msg_set_orignode(msg, tipc_own_addr(net));
2724 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2725 		}
2726 
2727 		rhashtable_walk_stop(&iter);
2728 	} while (tsk == ERR_PTR(-EAGAIN));
2729 
2730 	rhashtable_walk_exit(&iter);
2731 }
2732 
2733 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2734 {
2735 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2736 	struct tipc_sock *tsk;
2737 
2738 	rcu_read_lock();
2739 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2740 	if (tsk)
2741 		sock_hold(&tsk->sk);
2742 	rcu_read_unlock();
2743 
2744 	return tsk;
2745 }
2746 
2747 static int tipc_sk_insert(struct tipc_sock *tsk)
2748 {
2749 	struct sock *sk = &tsk->sk;
2750 	struct net *net = sock_net(sk);
2751 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2752 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2753 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2754 
2755 	while (remaining--) {
2756 		portid++;
2757 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2758 			portid = TIPC_MIN_PORT;
2759 		tsk->portid = portid;
2760 		sock_hold(&tsk->sk);
2761 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2762 						   tsk_rht_params))
2763 			return 0;
2764 		sock_put(&tsk->sk);
2765 	}
2766 
2767 	return -1;
2768 }
2769 
2770 static void tipc_sk_remove(struct tipc_sock *tsk)
2771 {
2772 	struct sock *sk = &tsk->sk;
2773 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2774 
2775 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2776 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2777 		__sock_put(sk);
2778 	}
2779 }
2780 
2781 static const struct rhashtable_params tsk_rht_params = {
2782 	.nelem_hint = 192,
2783 	.head_offset = offsetof(struct tipc_sock, node),
2784 	.key_offset = offsetof(struct tipc_sock, portid),
2785 	.key_len = sizeof(u32), /* portid */
2786 	.max_size = 1048576,
2787 	.min_size = 256,
2788 	.automatic_shrinking = true,
2789 };
2790 
2791 int tipc_sk_rht_init(struct net *net)
2792 {
2793 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2794 
2795 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2796 }
2797 
2798 void tipc_sk_rht_destroy(struct net *net)
2799 {
2800 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2801 
2802 	/* Wait for socket readers to complete */
2803 	synchronize_net();
2804 
2805 	rhashtable_destroy(&tn->sk_rht);
2806 }
2807 
2808 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2809 {
2810 	struct net *net = sock_net(&tsk->sk);
2811 	struct tipc_group *grp = tsk->group;
2812 	struct tipc_msg *hdr = &tsk->phdr;
2813 	struct tipc_name_seq seq;
2814 	int rc;
2815 
2816 	if (mreq->type < TIPC_RESERVED_TYPES)
2817 		return -EACCES;
2818 	if (mreq->scope > TIPC_NODE_SCOPE)
2819 		return -EINVAL;
2820 	if (grp)
2821 		return -EACCES;
2822 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2823 	if (!grp)
2824 		return -ENOMEM;
2825 	tsk->group = grp;
2826 	msg_set_lookup_scope(hdr, mreq->scope);
2827 	msg_set_nametype(hdr, mreq->type);
2828 	msg_set_dest_droppable(hdr, true);
2829 	seq.type = mreq->type;
2830 	seq.lower = mreq->instance;
2831 	seq.upper = seq.lower;
2832 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2833 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2834 	if (rc) {
2835 		tipc_group_delete(net, grp);
2836 		tsk->group = NULL;
2837 		return rc;
2838 	}
2839 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2840 	tsk->mc_method.rcast = true;
2841 	tsk->mc_method.mandatory = true;
2842 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2843 	return rc;
2844 }
2845 
2846 static int tipc_sk_leave(struct tipc_sock *tsk)
2847 {
2848 	struct net *net = sock_net(&tsk->sk);
2849 	struct tipc_group *grp = tsk->group;
2850 	struct tipc_name_seq seq;
2851 	int scope;
2852 
2853 	if (!grp)
2854 		return -EINVAL;
2855 	tipc_group_self(grp, &seq, &scope);
2856 	tipc_group_delete(net, grp);
2857 	tsk->group = NULL;
2858 	tipc_sk_withdraw(tsk, scope, &seq);
2859 	return 0;
2860 }
2861 
2862 /**
2863  * tipc_setsockopt - set socket option
2864  * @sock: socket structure
2865  * @lvl: option level
2866  * @opt: option identifier
2867  * @ov: pointer to new option value
2868  * @ol: length of option value
2869  *
2870  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2871  * (to ease compatibility).
2872  *
2873  * Returns 0 on success, errno otherwise
2874  */
2875 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2876 			   char __user *ov, unsigned int ol)
2877 {
2878 	struct sock *sk = sock->sk;
2879 	struct tipc_sock *tsk = tipc_sk(sk);
2880 	struct tipc_group_req mreq;
2881 	u32 value = 0;
2882 	int res = 0;
2883 
2884 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2885 		return 0;
2886 	if (lvl != SOL_TIPC)
2887 		return -ENOPROTOOPT;
2888 
2889 	switch (opt) {
2890 	case TIPC_IMPORTANCE:
2891 	case TIPC_SRC_DROPPABLE:
2892 	case TIPC_DEST_DROPPABLE:
2893 	case TIPC_CONN_TIMEOUT:
2894 		if (ol < sizeof(value))
2895 			return -EINVAL;
2896 		if (get_user(value, (u32 __user *)ov))
2897 			return -EFAULT;
2898 		break;
2899 	case TIPC_GROUP_JOIN:
2900 		if (ol < sizeof(mreq))
2901 			return -EINVAL;
2902 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2903 			return -EFAULT;
2904 		break;
2905 	default:
2906 		if (ov || ol)
2907 			return -EINVAL;
2908 	}
2909 
2910 	lock_sock(sk);
2911 
2912 	switch (opt) {
2913 	case TIPC_IMPORTANCE:
2914 		res = tsk_set_importance(tsk, value);
2915 		break;
2916 	case TIPC_SRC_DROPPABLE:
2917 		if (sock->type != SOCK_STREAM)
2918 			tsk_set_unreliable(tsk, value);
2919 		else
2920 			res = -ENOPROTOOPT;
2921 		break;
2922 	case TIPC_DEST_DROPPABLE:
2923 		tsk_set_unreturnable(tsk, value);
2924 		break;
2925 	case TIPC_CONN_TIMEOUT:
2926 		tipc_sk(sk)->conn_timeout = value;
2927 		break;
2928 	case TIPC_MCAST_BROADCAST:
2929 		tsk->mc_method.rcast = false;
2930 		tsk->mc_method.mandatory = true;
2931 		break;
2932 	case TIPC_MCAST_REPLICAST:
2933 		tsk->mc_method.rcast = true;
2934 		tsk->mc_method.mandatory = true;
2935 		break;
2936 	case TIPC_GROUP_JOIN:
2937 		res = tipc_sk_join(tsk, &mreq);
2938 		break;
2939 	case TIPC_GROUP_LEAVE:
2940 		res = tipc_sk_leave(tsk);
2941 		break;
2942 	default:
2943 		res = -EINVAL;
2944 	}
2945 
2946 	release_sock(sk);
2947 
2948 	return res;
2949 }
2950 
2951 /**
2952  * tipc_getsockopt - get socket option
2953  * @sock: socket structure
2954  * @lvl: option level
2955  * @opt: option identifier
2956  * @ov: receptacle for option value
2957  * @ol: receptacle for length of option value
2958  *
2959  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2960  * (to ease compatibility).
2961  *
2962  * Returns 0 on success, errno otherwise
2963  */
2964 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2965 			   char __user *ov, int __user *ol)
2966 {
2967 	struct sock *sk = sock->sk;
2968 	struct tipc_sock *tsk = tipc_sk(sk);
2969 	struct tipc_name_seq seq;
2970 	int len, scope;
2971 	u32 value;
2972 	int res;
2973 
2974 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2975 		return put_user(0, ol);
2976 	if (lvl != SOL_TIPC)
2977 		return -ENOPROTOOPT;
2978 	res = get_user(len, ol);
2979 	if (res)
2980 		return res;
2981 
2982 	lock_sock(sk);
2983 
2984 	switch (opt) {
2985 	case TIPC_IMPORTANCE:
2986 		value = tsk_importance(tsk);
2987 		break;
2988 	case TIPC_SRC_DROPPABLE:
2989 		value = tsk_unreliable(tsk);
2990 		break;
2991 	case TIPC_DEST_DROPPABLE:
2992 		value = tsk_unreturnable(tsk);
2993 		break;
2994 	case TIPC_CONN_TIMEOUT:
2995 		value = tsk->conn_timeout;
2996 		/* no need to set "res", since already 0 at this point */
2997 		break;
2998 	case TIPC_NODE_RECVQ_DEPTH:
2999 		value = 0; /* was tipc_queue_size, now obsolete */
3000 		break;
3001 	case TIPC_SOCK_RECVQ_DEPTH:
3002 		value = skb_queue_len(&sk->sk_receive_queue);
3003 		break;
3004 	case TIPC_GROUP_JOIN:
3005 		seq.type = 0;
3006 		if (tsk->group)
3007 			tipc_group_self(tsk->group, &seq, &scope);
3008 		value = seq.type;
3009 		break;
3010 	default:
3011 		res = -EINVAL;
3012 	}
3013 
3014 	release_sock(sk);
3015 
3016 	if (res)
3017 		return res;	/* "get" failed */
3018 
3019 	if (len < sizeof(value))
3020 		return -EINVAL;
3021 
3022 	if (copy_to_user(ov, &value, sizeof(value)))
3023 		return -EFAULT;
3024 
3025 	return put_user(sizeof(value), ol);
3026 }
3027 
3028 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3029 {
3030 	struct net *net = sock_net(sock->sk);
3031 	struct tipc_sioc_nodeid_req nr = {0};
3032 	struct tipc_sioc_ln_req lnr;
3033 	void __user *argp = (void __user *)arg;
3034 
3035 	switch (cmd) {
3036 	case SIOCGETLINKNAME:
3037 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3038 			return -EFAULT;
3039 		if (!tipc_node_get_linkname(net,
3040 					    lnr.bearer_id & 0xffff, lnr.peer,
3041 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3042 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3043 				return -EFAULT;
3044 			return 0;
3045 		}
3046 		return -EADDRNOTAVAIL;
3047 	case SIOCGETNODEID:
3048 		if (copy_from_user(&nr, argp, sizeof(nr)))
3049 			return -EFAULT;
3050 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3051 			return -EADDRNOTAVAIL;
3052 		if (copy_to_user(argp, &nr, sizeof(nr)))
3053 			return -EFAULT;
3054 		return 0;
3055 	default:
3056 		return -ENOIOCTLCMD;
3057 	}
3058 }
3059 
3060 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3061 {
3062 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3063 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3064 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3065 
3066 	tsk1->peer.family = AF_TIPC;
3067 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3068 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3069 	tsk1->peer.addr.id.ref = tsk2->portid;
3070 	tsk1->peer.addr.id.node = onode;
3071 	tsk2->peer.family = AF_TIPC;
3072 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3073 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3074 	tsk2->peer.addr.id.ref = tsk1->portid;
3075 	tsk2->peer.addr.id.node = onode;
3076 
3077 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3078 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3079 	return 0;
3080 }
3081 
3082 /* Protocol switches for the various types of TIPC sockets */
3083 
3084 static const struct proto_ops msg_ops = {
3085 	.owner		= THIS_MODULE,
3086 	.family		= AF_TIPC,
3087 	.release	= tipc_release,
3088 	.bind		= tipc_bind,
3089 	.connect	= tipc_connect,
3090 	.socketpair	= tipc_socketpair,
3091 	.accept		= sock_no_accept,
3092 	.getname	= tipc_getname,
3093 	.poll		= tipc_poll,
3094 	.ioctl		= tipc_ioctl,
3095 	.listen		= sock_no_listen,
3096 	.shutdown	= tipc_shutdown,
3097 	.setsockopt	= tipc_setsockopt,
3098 	.getsockopt	= tipc_getsockopt,
3099 	.sendmsg	= tipc_sendmsg,
3100 	.recvmsg	= tipc_recvmsg,
3101 	.mmap		= sock_no_mmap,
3102 	.sendpage	= sock_no_sendpage
3103 };
3104 
3105 static const struct proto_ops packet_ops = {
3106 	.owner		= THIS_MODULE,
3107 	.family		= AF_TIPC,
3108 	.release	= tipc_release,
3109 	.bind		= tipc_bind,
3110 	.connect	= tipc_connect,
3111 	.socketpair	= tipc_socketpair,
3112 	.accept		= tipc_accept,
3113 	.getname	= tipc_getname,
3114 	.poll		= tipc_poll,
3115 	.ioctl		= tipc_ioctl,
3116 	.listen		= tipc_listen,
3117 	.shutdown	= tipc_shutdown,
3118 	.setsockopt	= tipc_setsockopt,
3119 	.getsockopt	= tipc_getsockopt,
3120 	.sendmsg	= tipc_send_packet,
3121 	.recvmsg	= tipc_recvmsg,
3122 	.mmap		= sock_no_mmap,
3123 	.sendpage	= sock_no_sendpage
3124 };
3125 
3126 static const struct proto_ops stream_ops = {
3127 	.owner		= THIS_MODULE,
3128 	.family		= AF_TIPC,
3129 	.release	= tipc_release,
3130 	.bind		= tipc_bind,
3131 	.connect	= tipc_connect,
3132 	.socketpair	= tipc_socketpair,
3133 	.accept		= tipc_accept,
3134 	.getname	= tipc_getname,
3135 	.poll		= tipc_poll,
3136 	.ioctl		= tipc_ioctl,
3137 	.listen		= tipc_listen,
3138 	.shutdown	= tipc_shutdown,
3139 	.setsockopt	= tipc_setsockopt,
3140 	.getsockopt	= tipc_getsockopt,
3141 	.sendmsg	= tipc_sendstream,
3142 	.recvmsg	= tipc_recvstream,
3143 	.mmap		= sock_no_mmap,
3144 	.sendpage	= sock_no_sendpage
3145 };
3146 
3147 static const struct net_proto_family tipc_family_ops = {
3148 	.owner		= THIS_MODULE,
3149 	.family		= AF_TIPC,
3150 	.create		= tipc_sk_create
3151 };
3152 
3153 static struct proto tipc_proto = {
3154 	.name		= "TIPC",
3155 	.owner		= THIS_MODULE,
3156 	.obj_size	= sizeof(struct tipc_sock),
3157 	.sysctl_rmem	= sysctl_tipc_rmem
3158 };
3159 
3160 /**
3161  * tipc_socket_init - initialize TIPC socket interface
3162  *
3163  * Returns 0 on success, errno otherwise
3164  */
3165 int tipc_socket_init(void)
3166 {
3167 	int res;
3168 
3169 	res = proto_register(&tipc_proto, 1);
3170 	if (res) {
3171 		pr_err("Failed to register TIPC protocol type\n");
3172 		goto out;
3173 	}
3174 
3175 	res = sock_register(&tipc_family_ops);
3176 	if (res) {
3177 		pr_err("Failed to register TIPC socket type\n");
3178 		proto_unregister(&tipc_proto);
3179 		goto out;
3180 	}
3181  out:
3182 	return res;
3183 }
3184 
3185 /**
3186  * tipc_socket_stop - stop TIPC socket interface
3187  */
3188 void tipc_socket_stop(void)
3189 {
3190 	sock_unregister(tipc_family_ops.family);
3191 	proto_unregister(&tipc_proto);
3192 }
3193 
3194 /* Caller should hold socket lock for the passed tipc socket. */
3195 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3196 {
3197 	u32 peer_node;
3198 	u32 peer_port;
3199 	struct nlattr *nest;
3200 
3201 	peer_node = tsk_peer_node(tsk);
3202 	peer_port = tsk_peer_port(tsk);
3203 
3204 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3205 
3206 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3207 		goto msg_full;
3208 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3209 		goto msg_full;
3210 
3211 	if (tsk->conn_type != 0) {
3212 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3213 			goto msg_full;
3214 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3215 			goto msg_full;
3216 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3217 			goto msg_full;
3218 	}
3219 	nla_nest_end(skb, nest);
3220 
3221 	return 0;
3222 
3223 msg_full:
3224 	nla_nest_cancel(skb, nest);
3225 
3226 	return -EMSGSIZE;
3227 }
3228 
3229 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3230 			  *tsk)
3231 {
3232 	struct net *net = sock_net(skb->sk);
3233 	struct sock *sk = &tsk->sk;
3234 
3235 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3236 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3237 		return -EMSGSIZE;
3238 
3239 	if (tipc_sk_connected(sk)) {
3240 		if (__tipc_nl_add_sk_con(skb, tsk))
3241 			return -EMSGSIZE;
3242 	} else if (!list_empty(&tsk->publications)) {
3243 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3244 			return -EMSGSIZE;
3245 	}
3246 	return 0;
3247 }
3248 
3249 /* Caller should hold socket lock for the passed tipc socket. */
3250 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3251 			    struct tipc_sock *tsk)
3252 {
3253 	struct nlattr *attrs;
3254 	void *hdr;
3255 
3256 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3257 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3258 	if (!hdr)
3259 		goto msg_cancel;
3260 
3261 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3262 	if (!attrs)
3263 		goto genlmsg_cancel;
3264 
3265 	if (__tipc_nl_add_sk_info(skb, tsk))
3266 		goto attr_msg_cancel;
3267 
3268 	nla_nest_end(skb, attrs);
3269 	genlmsg_end(skb, hdr);
3270 
3271 	return 0;
3272 
3273 attr_msg_cancel:
3274 	nla_nest_cancel(skb, attrs);
3275 genlmsg_cancel:
3276 	genlmsg_cancel(skb, hdr);
3277 msg_cancel:
3278 	return -EMSGSIZE;
3279 }
3280 
3281 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3282 		    int (*skb_handler)(struct sk_buff *skb,
3283 				       struct netlink_callback *cb,
3284 				       struct tipc_sock *tsk))
3285 {
3286 	struct rhashtable_iter *iter = (void *)cb->args[4];
3287 	struct tipc_sock *tsk;
3288 	int err;
3289 
3290 	rhashtable_walk_start(iter);
3291 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3292 		if (IS_ERR(tsk)) {
3293 			err = PTR_ERR(tsk);
3294 			if (err == -EAGAIN) {
3295 				err = 0;
3296 				continue;
3297 			}
3298 			break;
3299 		}
3300 
3301 		sock_hold(&tsk->sk);
3302 		rhashtable_walk_stop(iter);
3303 		lock_sock(&tsk->sk);
3304 		err = skb_handler(skb, cb, tsk);
3305 		if (err) {
3306 			release_sock(&tsk->sk);
3307 			sock_put(&tsk->sk);
3308 			goto out;
3309 		}
3310 		release_sock(&tsk->sk);
3311 		rhashtable_walk_start(iter);
3312 		sock_put(&tsk->sk);
3313 	}
3314 	rhashtable_walk_stop(iter);
3315 out:
3316 	return skb->len;
3317 }
3318 EXPORT_SYMBOL(tipc_nl_sk_walk);
3319 
3320 int tipc_dump_start(struct netlink_callback *cb)
3321 {
3322 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3323 }
3324 EXPORT_SYMBOL(tipc_dump_start);
3325 
3326 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3327 {
3328 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3329 	struct rhashtable_iter *iter = (void *)cb->args[4];
3330 	struct tipc_net *tn = tipc_net(net);
3331 
3332 	if (!iter) {
3333 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3334 		if (!iter)
3335 			return -ENOMEM;
3336 
3337 		cb->args[4] = (long)iter;
3338 	}
3339 
3340 	rhashtable_walk_enter(&tn->sk_rht, iter);
3341 	return 0;
3342 }
3343 
3344 int tipc_dump_done(struct netlink_callback *cb)
3345 {
3346 	struct rhashtable_iter *hti = (void *)cb->args[4];
3347 
3348 	rhashtable_walk_exit(hti);
3349 	kfree(hti);
3350 	return 0;
3351 }
3352 EXPORT_SYMBOL(tipc_dump_done);
3353 
3354 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3355 			   struct tipc_sock *tsk, u32 sk_filter_state,
3356 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3357 {
3358 	struct sock *sk = &tsk->sk;
3359 	struct nlattr *attrs;
3360 	struct nlattr *stat;
3361 
3362 	/*filter response w.r.t sk_state*/
3363 	if (!(sk_filter_state & (1 << sk->sk_state)))
3364 		return 0;
3365 
3366 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3367 	if (!attrs)
3368 		goto msg_cancel;
3369 
3370 	if (__tipc_nl_add_sk_info(skb, tsk))
3371 		goto attr_msg_cancel;
3372 
3373 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3374 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3375 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3376 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3377 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3378 					 sock_i_uid(sk))) ||
3379 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3380 			      tipc_diag_gen_cookie(sk),
3381 			      TIPC_NLA_SOCK_PAD))
3382 		goto attr_msg_cancel;
3383 
3384 	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3385 	if (!stat)
3386 		goto attr_msg_cancel;
3387 
3388 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3389 			skb_queue_len(&sk->sk_receive_queue)) ||
3390 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3391 			skb_queue_len(&sk->sk_write_queue)) ||
3392 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3393 			atomic_read(&sk->sk_drops)))
3394 		goto stat_msg_cancel;
3395 
3396 	if (tsk->cong_link_cnt &&
3397 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3398 		goto stat_msg_cancel;
3399 
3400 	if (tsk_conn_cong(tsk) &&
3401 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3402 		goto stat_msg_cancel;
3403 
3404 	nla_nest_end(skb, stat);
3405 
3406 	if (tsk->group)
3407 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3408 			goto stat_msg_cancel;
3409 
3410 	nla_nest_end(skb, attrs);
3411 
3412 	return 0;
3413 
3414 stat_msg_cancel:
3415 	nla_nest_cancel(skb, stat);
3416 attr_msg_cancel:
3417 	nla_nest_cancel(skb, attrs);
3418 msg_cancel:
3419 	return -EMSGSIZE;
3420 }
3421 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3422 
3423 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3424 {
3425 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3426 }
3427 
3428 /* Caller should hold socket lock for the passed tipc socket. */
3429 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3430 				 struct netlink_callback *cb,
3431 				 struct publication *publ)
3432 {
3433 	void *hdr;
3434 	struct nlattr *attrs;
3435 
3436 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3437 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3438 	if (!hdr)
3439 		goto msg_cancel;
3440 
3441 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3442 	if (!attrs)
3443 		goto genlmsg_cancel;
3444 
3445 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3446 		goto attr_msg_cancel;
3447 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3448 		goto attr_msg_cancel;
3449 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3450 		goto attr_msg_cancel;
3451 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3452 		goto attr_msg_cancel;
3453 
3454 	nla_nest_end(skb, attrs);
3455 	genlmsg_end(skb, hdr);
3456 
3457 	return 0;
3458 
3459 attr_msg_cancel:
3460 	nla_nest_cancel(skb, attrs);
3461 genlmsg_cancel:
3462 	genlmsg_cancel(skb, hdr);
3463 msg_cancel:
3464 	return -EMSGSIZE;
3465 }
3466 
3467 /* Caller should hold socket lock for the passed tipc socket. */
3468 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3469 				  struct netlink_callback *cb,
3470 				  struct tipc_sock *tsk, u32 *last_publ)
3471 {
3472 	int err;
3473 	struct publication *p;
3474 
3475 	if (*last_publ) {
3476 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3477 			if (p->key == *last_publ)
3478 				break;
3479 		}
3480 		if (p->key != *last_publ) {
3481 			/* We never set seq or call nl_dump_check_consistent()
3482 			 * this means that setting prev_seq here will cause the
3483 			 * consistence check to fail in the netlink callback
3484 			 * handler. Resulting in the last NLMSG_DONE message
3485 			 * having the NLM_F_DUMP_INTR flag set.
3486 			 */
3487 			cb->prev_seq = 1;
3488 			*last_publ = 0;
3489 			return -EPIPE;
3490 		}
3491 	} else {
3492 		p = list_first_entry(&tsk->publications, struct publication,
3493 				     binding_sock);
3494 	}
3495 
3496 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3497 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3498 		if (err) {
3499 			*last_publ = p->key;
3500 			return err;
3501 		}
3502 	}
3503 	*last_publ = 0;
3504 
3505 	return 0;
3506 }
3507 
3508 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3509 {
3510 	int err;
3511 	u32 tsk_portid = cb->args[0];
3512 	u32 last_publ = cb->args[1];
3513 	u32 done = cb->args[2];
3514 	struct net *net = sock_net(skb->sk);
3515 	struct tipc_sock *tsk;
3516 
3517 	if (!tsk_portid) {
3518 		struct nlattr **attrs;
3519 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3520 
3521 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3522 		if (err)
3523 			return err;
3524 
3525 		if (!attrs[TIPC_NLA_SOCK])
3526 			return -EINVAL;
3527 
3528 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3529 				       attrs[TIPC_NLA_SOCK],
3530 				       tipc_nl_sock_policy, NULL);
3531 		if (err)
3532 			return err;
3533 
3534 		if (!sock[TIPC_NLA_SOCK_REF])
3535 			return -EINVAL;
3536 
3537 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3538 	}
3539 
3540 	if (done)
3541 		return 0;
3542 
3543 	tsk = tipc_sk_lookup(net, tsk_portid);
3544 	if (!tsk)
3545 		return -EINVAL;
3546 
3547 	lock_sock(&tsk->sk);
3548 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3549 	if (!err)
3550 		done = 1;
3551 	release_sock(&tsk->sk);
3552 	sock_put(&tsk->sk);
3553 
3554 	cb->args[0] = tsk_portid;
3555 	cb->args[1] = last_publ;
3556 	cb->args[2] = done;
3557 
3558 	return skb->len;
3559 }
3560