xref: /linux/net/tipc/socket.c (revision a8b70ccf10e38775785d9cb12ead916474549f99)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include <linux/rhashtable.h>
38 #include <linux/sched/signal.h>
39 
40 #include "core.h"
41 #include "name_table.h"
42 #include "node.h"
43 #include "link.h"
44 #include "name_distr.h"
45 #include "socket.h"
46 #include "bcast.h"
47 #include "netlink.h"
48 #include "group.h"
49 
50 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
51 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
52 #define TIPC_FWD_MSG		1
53 #define TIPC_MAX_PORT		0xffffffff
54 #define TIPC_MIN_PORT		1
55 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
56 
57 enum {
58 	TIPC_LISTEN = TCP_LISTEN,
59 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
60 	TIPC_OPEN = TCP_CLOSE,
61 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
62 	TIPC_CONNECTING = TCP_SYN_SENT,
63 };
64 
65 struct sockaddr_pair {
66 	struct sockaddr_tipc sock;
67 	struct sockaddr_tipc member;
68 };
69 
70 /**
71  * struct tipc_sock - TIPC socket structure
72  * @sk: socket - interacts with 'port' and with user via the socket API
73  * @conn_type: TIPC type used when connection was established
74  * @conn_instance: TIPC instance used when connection was established
75  * @published: non-zero if port has one or more associated names
76  * @max_pkt: maximum packet size "hint" used when building messages sent by port
77  * @portid: unique port identity in TIPC socket hash table
78  * @phdr: preformatted message header used when sending messages
79  * #cong_links: list of congested links
80  * @publications: list of publications for port
81  * @blocking_link: address of the congested link we are currently sleeping on
82  * @pub_count: total # of publications port has made during its lifetime
83  * @probing_state:
84  * @conn_timeout: the time we can wait for an unresponded setup request
85  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86  * @cong_link_cnt: number of congested links
87  * @snt_unacked: # messages sent by socket, and not yet acked by peer
88  * @rcv_unacked: # messages read by user, but not yet acked back to peer
89  * @peer: 'connected' peer for dgram/rdm
90  * @node: hash table node
91  * @mc_method: cookie for use between socket and broadcast layer
92  * @rcu: rcu struct for tipc_sock
93  */
94 struct tipc_sock {
95 	struct sock sk;
96 	u32 conn_type;
97 	u32 conn_instance;
98 	int published;
99 	u32 max_pkt;
100 	u32 portid;
101 	struct tipc_msg phdr;
102 	struct list_head cong_links;
103 	struct list_head publications;
104 	u32 pub_count;
105 	uint conn_timeout;
106 	atomic_t dupl_rcvcnt;
107 	bool probe_unacked;
108 	u16 cong_link_cnt;
109 	u16 snt_unacked;
110 	u16 snd_win;
111 	u16 peer_caps;
112 	u16 rcv_unacked;
113 	u16 rcv_win;
114 	struct sockaddr_tipc peer;
115 	struct rhash_head node;
116 	struct tipc_mc_method mc_method;
117 	struct rcu_head rcu;
118 	struct tipc_group *group;
119 	bool group_is_open;
120 };
121 
122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123 static void tipc_data_ready(struct sock *sk);
124 static void tipc_write_space(struct sock *sk);
125 static void tipc_sock_destruct(struct sock *sk);
126 static int tipc_release(struct socket *sock);
127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
128 		       bool kern);
129 static void tipc_sk_timeout(struct timer_list *t);
130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 			   struct tipc_name_seq const *seq);
132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 			    struct tipc_name_seq const *seq);
134 static int tipc_sk_leave(struct tipc_sock *tsk);
135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136 static int tipc_sk_insert(struct tipc_sock *tsk);
137 static void tipc_sk_remove(struct tipc_sock *tsk);
138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
140 
141 static const struct proto_ops packet_ops;
142 static const struct proto_ops stream_ops;
143 static const struct proto_ops msg_ops;
144 static struct proto tipc_proto;
145 static const struct rhashtable_params tsk_rht_params;
146 
147 static u32 tsk_own_node(struct tipc_sock *tsk)
148 {
149 	return msg_prevnode(&tsk->phdr);
150 }
151 
152 static u32 tsk_peer_node(struct tipc_sock *tsk)
153 {
154 	return msg_destnode(&tsk->phdr);
155 }
156 
157 static u32 tsk_peer_port(struct tipc_sock *tsk)
158 {
159 	return msg_destport(&tsk->phdr);
160 }
161 
162 static  bool tsk_unreliable(struct tipc_sock *tsk)
163 {
164 	return msg_src_droppable(&tsk->phdr) != 0;
165 }
166 
167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
168 {
169 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
170 }
171 
172 static bool tsk_unreturnable(struct tipc_sock *tsk)
173 {
174 	return msg_dest_droppable(&tsk->phdr) != 0;
175 }
176 
177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
178 {
179 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
180 }
181 
182 static int tsk_importance(struct tipc_sock *tsk)
183 {
184 	return msg_importance(&tsk->phdr);
185 }
186 
187 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
188 {
189 	if (imp > TIPC_CRITICAL_IMPORTANCE)
190 		return -EINVAL;
191 	msg_set_importance(&tsk->phdr, (u32)imp);
192 	return 0;
193 }
194 
195 static struct tipc_sock *tipc_sk(const struct sock *sk)
196 {
197 	return container_of(sk, struct tipc_sock, sk);
198 }
199 
200 static bool tsk_conn_cong(struct tipc_sock *tsk)
201 {
202 	return tsk->snt_unacked > tsk->snd_win;
203 }
204 
205 static u16 tsk_blocks(int len)
206 {
207 	return ((len / FLOWCTL_BLK_SZ) + 1);
208 }
209 
210 /* tsk_blocks(): translate a buffer size in bytes to number of
211  * advertisable blocks, taking into account the ratio truesize(len)/len
212  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
213  */
214 static u16 tsk_adv_blocks(int len)
215 {
216 	return len / FLOWCTL_BLK_SZ / 4;
217 }
218 
219 /* tsk_inc(): increment counter for sent or received data
220  * - If block based flow control is not supported by peer we
221  *   fall back to message based ditto, incrementing the counter
222  */
223 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
224 {
225 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
226 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
227 	return 1;
228 }
229 
230 /**
231  * tsk_advance_rx_queue - discard first buffer in socket receive queue
232  *
233  * Caller must hold socket lock
234  */
235 static void tsk_advance_rx_queue(struct sock *sk)
236 {
237 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
238 }
239 
240 /* tipc_sk_respond() : send response message back to sender
241  */
242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
243 {
244 	u32 selector;
245 	u32 dnode;
246 	u32 onode = tipc_own_addr(sock_net(sk));
247 
248 	if (!tipc_msg_reverse(onode, &skb, err))
249 		return;
250 
251 	dnode = msg_destnode(buf_msg(skb));
252 	selector = msg_origport(buf_msg(skb));
253 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
254 }
255 
256 /**
257  * tsk_rej_rx_queue - reject all buffers in socket receive queue
258  *
259  * Caller must hold socket lock
260  */
261 static void tsk_rej_rx_queue(struct sock *sk)
262 {
263 	struct sk_buff *skb;
264 
265 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
266 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
267 }
268 
269 static bool tipc_sk_connected(struct sock *sk)
270 {
271 	return sk->sk_state == TIPC_ESTABLISHED;
272 }
273 
274 /* tipc_sk_type_connectionless - check if the socket is datagram socket
275  * @sk: socket
276  *
277  * Returns true if connection less, false otherwise
278  */
279 static bool tipc_sk_type_connectionless(struct sock *sk)
280 {
281 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
282 }
283 
284 /* tsk_peer_msg - verify if message was sent by connected port's peer
285  *
286  * Handles cases where the node's network address has changed from
287  * the default of <0.0.0> to its configured setting.
288  */
289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
290 {
291 	struct sock *sk = &tsk->sk;
292 	u32 self = tipc_own_addr(sock_net(sk));
293 	u32 peer_port = tsk_peer_port(tsk);
294 	u32 orig_node, peer_node;
295 
296 	if (unlikely(!tipc_sk_connected(sk)))
297 		return false;
298 
299 	if (unlikely(msg_origport(msg) != peer_port))
300 		return false;
301 
302 	orig_node = msg_orignode(msg);
303 	peer_node = tsk_peer_node(tsk);
304 
305 	if (likely(orig_node == peer_node))
306 		return true;
307 
308 	if (!orig_node && peer_node == self)
309 		return true;
310 
311 	if (!peer_node && orig_node == self)
312 		return true;
313 
314 	return false;
315 }
316 
317 /* tipc_set_sk_state - set the sk_state of the socket
318  * @sk: socket
319  *
320  * Caller must hold socket lock
321  *
322  * Returns 0 on success, errno otherwise
323  */
324 static int tipc_set_sk_state(struct sock *sk, int state)
325 {
326 	int oldsk_state = sk->sk_state;
327 	int res = -EINVAL;
328 
329 	switch (state) {
330 	case TIPC_OPEN:
331 		res = 0;
332 		break;
333 	case TIPC_LISTEN:
334 	case TIPC_CONNECTING:
335 		if (oldsk_state == TIPC_OPEN)
336 			res = 0;
337 		break;
338 	case TIPC_ESTABLISHED:
339 		if (oldsk_state == TIPC_CONNECTING ||
340 		    oldsk_state == TIPC_OPEN)
341 			res = 0;
342 		break;
343 	case TIPC_DISCONNECTING:
344 		if (oldsk_state == TIPC_CONNECTING ||
345 		    oldsk_state == TIPC_ESTABLISHED)
346 			res = 0;
347 		break;
348 	}
349 
350 	if (!res)
351 		sk->sk_state = state;
352 
353 	return res;
354 }
355 
356 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
357 {
358 	struct sock *sk = sock->sk;
359 	int err = sock_error(sk);
360 	int typ = sock->type;
361 
362 	if (err)
363 		return err;
364 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
365 		if (sk->sk_state == TIPC_DISCONNECTING)
366 			return -EPIPE;
367 		else if (!tipc_sk_connected(sk))
368 			return -ENOTCONN;
369 	}
370 	if (!*timeout)
371 		return -EAGAIN;
372 	if (signal_pending(current))
373 		return sock_intr_errno(*timeout);
374 
375 	return 0;
376 }
377 
378 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
379 ({                                                                             \
380 	struct sock *sk_;						       \
381 	int rc_;							       \
382 									       \
383 	while ((rc_ = !(condition_))) {					       \
384 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
385 		sk_ = (sock_)->sk;					       \
386 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
387 		if (rc_)						       \
388 			break;						       \
389 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
390 		release_sock(sk_);					       \
391 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
392 		sched_annotate_sleep();				               \
393 		lock_sock(sk_);						       \
394 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
395 	}								       \
396 	rc_;								       \
397 })
398 
399 /**
400  * tipc_sk_create - create a TIPC socket
401  * @net: network namespace (must be default network)
402  * @sock: pre-allocated socket structure
403  * @protocol: protocol indicator (must be 0)
404  * @kern: caused by kernel or by userspace?
405  *
406  * This routine creates additional data structures used by the TIPC socket,
407  * initializes them, and links them together.
408  *
409  * Returns 0 on success, errno otherwise
410  */
411 static int tipc_sk_create(struct net *net, struct socket *sock,
412 			  int protocol, int kern)
413 {
414 	struct tipc_net *tn;
415 	const struct proto_ops *ops;
416 	struct sock *sk;
417 	struct tipc_sock *tsk;
418 	struct tipc_msg *msg;
419 
420 	/* Validate arguments */
421 	if (unlikely(protocol != 0))
422 		return -EPROTONOSUPPORT;
423 
424 	switch (sock->type) {
425 	case SOCK_STREAM:
426 		ops = &stream_ops;
427 		break;
428 	case SOCK_SEQPACKET:
429 		ops = &packet_ops;
430 		break;
431 	case SOCK_DGRAM:
432 	case SOCK_RDM:
433 		ops = &msg_ops;
434 		break;
435 	default:
436 		return -EPROTOTYPE;
437 	}
438 
439 	/* Allocate socket's protocol area */
440 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
441 	if (sk == NULL)
442 		return -ENOMEM;
443 
444 	tsk = tipc_sk(sk);
445 	tsk->max_pkt = MAX_PKT_DEFAULT;
446 	INIT_LIST_HEAD(&tsk->publications);
447 	INIT_LIST_HEAD(&tsk->cong_links);
448 	msg = &tsk->phdr;
449 	tn = net_generic(sock_net(sk), tipc_net_id);
450 
451 	/* Finish initializing socket data structures */
452 	sock->ops = ops;
453 	sock_init_data(sock, sk);
454 	tipc_set_sk_state(sk, TIPC_OPEN);
455 	if (tipc_sk_insert(tsk)) {
456 		pr_warn("Socket create failed; port number exhausted\n");
457 		return -EINVAL;
458 	}
459 
460 	/* Ensure tsk is visible before we read own_addr. */
461 	smp_mb();
462 
463 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
464 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
465 
466 	msg_set_origport(msg, tsk->portid);
467 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
468 	sk->sk_shutdown = 0;
469 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
470 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
471 	sk->sk_data_ready = tipc_data_ready;
472 	sk->sk_write_space = tipc_write_space;
473 	sk->sk_destruct = tipc_sock_destruct;
474 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
475 	tsk->group_is_open = true;
476 	atomic_set(&tsk->dupl_rcvcnt, 0);
477 
478 	/* Start out with safe limits until we receive an advertised window */
479 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
480 	tsk->rcv_win = tsk->snd_win;
481 
482 	if (tipc_sk_type_connectionless(sk)) {
483 		tsk_set_unreturnable(tsk, true);
484 		if (sock->type == SOCK_DGRAM)
485 			tsk_set_unreliable(tsk, true);
486 	}
487 
488 	return 0;
489 }
490 
491 static void tipc_sk_callback(struct rcu_head *head)
492 {
493 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
494 
495 	sock_put(&tsk->sk);
496 }
497 
498 /* Caller should hold socket lock for the socket. */
499 static void __tipc_shutdown(struct socket *sock, int error)
500 {
501 	struct sock *sk = sock->sk;
502 	struct tipc_sock *tsk = tipc_sk(sk);
503 	struct net *net = sock_net(sk);
504 	long timeout = CONN_TIMEOUT_DEFAULT;
505 	u32 dnode = tsk_peer_node(tsk);
506 	struct sk_buff *skb;
507 
508 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
509 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
510 					    !tsk_conn_cong(tsk)));
511 
512 	/* Reject all unreceived messages, except on an active connection
513 	 * (which disconnects locally & sends a 'FIN+' to peer).
514 	 */
515 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
516 		if (TIPC_SKB_CB(skb)->bytes_read) {
517 			kfree_skb(skb);
518 			continue;
519 		}
520 		if (!tipc_sk_type_connectionless(sk) &&
521 		    sk->sk_state != TIPC_DISCONNECTING) {
522 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
523 			tipc_node_remove_conn(net, dnode, tsk->portid);
524 		}
525 		tipc_sk_respond(sk, skb, error);
526 	}
527 
528 	if (tipc_sk_type_connectionless(sk))
529 		return;
530 
531 	if (sk->sk_state != TIPC_DISCONNECTING) {
532 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
533 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
534 				      tsk_own_node(tsk), tsk_peer_port(tsk),
535 				      tsk->portid, error);
536 		if (skb)
537 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
538 		tipc_node_remove_conn(net, dnode, tsk->portid);
539 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
540 	}
541 }
542 
543 /**
544  * tipc_release - destroy a TIPC socket
545  * @sock: socket to destroy
546  *
547  * This routine cleans up any messages that are still queued on the socket.
548  * For DGRAM and RDM socket types, all queued messages are rejected.
549  * For SEQPACKET and STREAM socket types, the first message is rejected
550  * and any others are discarded.  (If the first message on a STREAM socket
551  * is partially-read, it is discarded and the next one is rejected instead.)
552  *
553  * NOTE: Rejected messages are not necessarily returned to the sender!  They
554  * are returned or discarded according to the "destination droppable" setting
555  * specified for the message by the sender.
556  *
557  * Returns 0 on success, errno otherwise
558  */
559 static int tipc_release(struct socket *sock)
560 {
561 	struct sock *sk = sock->sk;
562 	struct tipc_sock *tsk;
563 
564 	/*
565 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
566 	 * releases a pre-allocated child socket that was never used)
567 	 */
568 	if (sk == NULL)
569 		return 0;
570 
571 	tsk = tipc_sk(sk);
572 	lock_sock(sk);
573 
574 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
575 	sk->sk_shutdown = SHUTDOWN_MASK;
576 	tipc_sk_leave(tsk);
577 	tipc_sk_withdraw(tsk, 0, NULL);
578 	sk_stop_timer(sk, &sk->sk_timer);
579 	tipc_sk_remove(tsk);
580 
581 	/* Reject any messages that accumulated in backlog queue */
582 	release_sock(sk);
583 	tipc_dest_list_purge(&tsk->cong_links);
584 	tsk->cong_link_cnt = 0;
585 	call_rcu(&tsk->rcu, tipc_sk_callback);
586 	sock->sk = NULL;
587 
588 	return 0;
589 }
590 
591 /**
592  * tipc_bind - associate or disassocate TIPC name(s) with a socket
593  * @sock: socket structure
594  * @uaddr: socket address describing name(s) and desired operation
595  * @uaddr_len: size of socket address data structure
596  *
597  * Name and name sequence binding is indicated using a positive scope value;
598  * a negative scope value unbinds the specified name.  Specifying no name
599  * (i.e. a socket address length of 0) unbinds all names from the socket.
600  *
601  * Returns 0 on success, errno otherwise
602  *
603  * NOTE: This routine doesn't need to take the socket lock since it doesn't
604  *       access any non-constant socket information.
605  */
606 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
607 		     int uaddr_len)
608 {
609 	struct sock *sk = sock->sk;
610 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
611 	struct tipc_sock *tsk = tipc_sk(sk);
612 	int res = -EINVAL;
613 
614 	lock_sock(sk);
615 	if (unlikely(!uaddr_len)) {
616 		res = tipc_sk_withdraw(tsk, 0, NULL);
617 		goto exit;
618 	}
619 	if (tsk->group) {
620 		res = -EACCES;
621 		goto exit;
622 	}
623 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
624 		res = -EINVAL;
625 		goto exit;
626 	}
627 	if (addr->family != AF_TIPC) {
628 		res = -EAFNOSUPPORT;
629 		goto exit;
630 	}
631 
632 	if (addr->addrtype == TIPC_ADDR_NAME)
633 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
634 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
635 		res = -EAFNOSUPPORT;
636 		goto exit;
637 	}
638 
639 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
640 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
641 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
642 		res = -EACCES;
643 		goto exit;
644 	}
645 
646 	res = (addr->scope >= 0) ?
647 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
648 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
649 exit:
650 	release_sock(sk);
651 	return res;
652 }
653 
654 /**
655  * tipc_getname - get port ID of socket or peer socket
656  * @sock: socket structure
657  * @uaddr: area for returned socket address
658  * @uaddr_len: area for returned length of socket address
659  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
660  *
661  * Returns 0 on success, errno otherwise
662  *
663  * NOTE: This routine doesn't need to take the socket lock since it only
664  *       accesses socket information that is unchanging (or which changes in
665  *       a completely predictable manner).
666  */
667 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
668 			int peer)
669 {
670 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
671 	struct sock *sk = sock->sk;
672 	struct tipc_sock *tsk = tipc_sk(sk);
673 
674 	memset(addr, 0, sizeof(*addr));
675 	if (peer) {
676 		if ((!tipc_sk_connected(sk)) &&
677 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
678 			return -ENOTCONN;
679 		addr->addr.id.ref = tsk_peer_port(tsk);
680 		addr->addr.id.node = tsk_peer_node(tsk);
681 	} else {
682 		addr->addr.id.ref = tsk->portid;
683 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
684 	}
685 
686 	addr->addrtype = TIPC_ADDR_ID;
687 	addr->family = AF_TIPC;
688 	addr->scope = 0;
689 	addr->addr.name.domain = 0;
690 
691 	return sizeof(*addr);
692 }
693 
694 /**
695  * tipc_poll - read and possibly block on pollmask
696  * @file: file structure associated with the socket
697  * @sock: socket for which to calculate the poll bits
698  * @wait: ???
699  *
700  * Returns pollmask value
701  *
702  * COMMENTARY:
703  * It appears that the usual socket locking mechanisms are not useful here
704  * since the pollmask info is potentially out-of-date the moment this routine
705  * exits.  TCP and other protocols seem to rely on higher level poll routines
706  * to handle any preventable race conditions, so TIPC will do the same ...
707  *
708  * IMPORTANT: The fact that a read or write operation is indicated does NOT
709  * imply that the operation will succeed, merely that it should be performed
710  * and will not block.
711  */
712 static __poll_t tipc_poll(struct file *file, struct socket *sock,
713 			      poll_table *wait)
714 {
715 	struct sock *sk = sock->sk;
716 	struct tipc_sock *tsk = tipc_sk(sk);
717 	__poll_t revents = 0;
718 
719 	sock_poll_wait(file, sk_sleep(sk), wait);
720 
721 	if (sk->sk_shutdown & RCV_SHUTDOWN)
722 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
723 	if (sk->sk_shutdown == SHUTDOWN_MASK)
724 		revents |= EPOLLHUP;
725 
726 	switch (sk->sk_state) {
727 	case TIPC_ESTABLISHED:
728 	case TIPC_CONNECTING:
729 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
730 			revents |= EPOLLOUT;
731 		/* fall thru' */
732 	case TIPC_LISTEN:
733 		if (!skb_queue_empty(&sk->sk_receive_queue))
734 			revents |= EPOLLIN | EPOLLRDNORM;
735 		break;
736 	case TIPC_OPEN:
737 		if (tsk->group_is_open && !tsk->cong_link_cnt)
738 			revents |= EPOLLOUT;
739 		if (!tipc_sk_type_connectionless(sk))
740 			break;
741 		if (skb_queue_empty(&sk->sk_receive_queue))
742 			break;
743 		revents |= EPOLLIN | EPOLLRDNORM;
744 		break;
745 	case TIPC_DISCONNECTING:
746 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
747 		break;
748 	}
749 	return revents;
750 }
751 
752 /**
753  * tipc_sendmcast - send multicast message
754  * @sock: socket structure
755  * @seq: destination address
756  * @msg: message to send
757  * @dlen: length of data to send
758  * @timeout: timeout to wait for wakeup
759  *
760  * Called from function tipc_sendmsg(), which has done all sanity checks
761  * Returns the number of bytes sent on success, or errno
762  */
763 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
764 			  struct msghdr *msg, size_t dlen, long timeout)
765 {
766 	struct sock *sk = sock->sk;
767 	struct tipc_sock *tsk = tipc_sk(sk);
768 	struct tipc_msg *hdr = &tsk->phdr;
769 	struct net *net = sock_net(sk);
770 	int mtu = tipc_bcast_get_mtu(net);
771 	struct tipc_mc_method *method = &tsk->mc_method;
772 	struct sk_buff_head pkts;
773 	struct tipc_nlist dsts;
774 	int rc;
775 
776 	if (tsk->group)
777 		return -EACCES;
778 
779 	/* Block or return if any destination link is congested */
780 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
781 	if (unlikely(rc))
782 		return rc;
783 
784 	/* Lookup destination nodes */
785 	tipc_nlist_init(&dsts, tipc_own_addr(net));
786 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
787 				      seq->upper, &dsts);
788 	if (!dsts.local && !dsts.remote)
789 		return -EHOSTUNREACH;
790 
791 	/* Build message header */
792 	msg_set_type(hdr, TIPC_MCAST_MSG);
793 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
794 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
795 	msg_set_destport(hdr, 0);
796 	msg_set_destnode(hdr, 0);
797 	msg_set_nametype(hdr, seq->type);
798 	msg_set_namelower(hdr, seq->lower);
799 	msg_set_nameupper(hdr, seq->upper);
800 
801 	/* Build message as chain of buffers */
802 	skb_queue_head_init(&pkts);
803 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
804 
805 	/* Send message if build was successful */
806 	if (unlikely(rc == dlen))
807 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
808 				     &tsk->cong_link_cnt);
809 
810 	tipc_nlist_purge(&dsts);
811 
812 	return rc ? rc : dlen;
813 }
814 
815 /**
816  * tipc_send_group_msg - send a message to a member in the group
817  * @net: network namespace
818  * @m: message to send
819  * @mb: group member
820  * @dnode: destination node
821  * @dport: destination port
822  * @dlen: total length of message data
823  */
824 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
825 			       struct msghdr *m, struct tipc_member *mb,
826 			       u32 dnode, u32 dport, int dlen)
827 {
828 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
829 	struct tipc_mc_method *method = &tsk->mc_method;
830 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
831 	struct tipc_msg *hdr = &tsk->phdr;
832 	struct sk_buff_head pkts;
833 	int mtu, rc;
834 
835 	/* Complete message header */
836 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
837 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
838 	msg_set_destport(hdr, dport);
839 	msg_set_destnode(hdr, dnode);
840 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
841 
842 	/* Build message as chain of buffers */
843 	skb_queue_head_init(&pkts);
844 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
845 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
846 	if (unlikely(rc != dlen))
847 		return rc;
848 
849 	/* Send message */
850 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
851 	if (unlikely(rc == -ELINKCONG)) {
852 		tipc_dest_push(&tsk->cong_links, dnode, 0);
853 		tsk->cong_link_cnt++;
854 	}
855 
856 	/* Update send window */
857 	tipc_group_update_member(mb, blks);
858 
859 	/* A broadcast sent within next EXPIRE period must follow same path */
860 	method->rcast = true;
861 	method->mandatory = true;
862 	return dlen;
863 }
864 
865 /**
866  * tipc_send_group_unicast - send message to a member in the group
867  * @sock: socket structure
868  * @m: message to send
869  * @dlen: total length of message data
870  * @timeout: timeout to wait for wakeup
871  *
872  * Called from function tipc_sendmsg(), which has done all sanity checks
873  * Returns the number of bytes sent on success, or errno
874  */
875 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
876 				   int dlen, long timeout)
877 {
878 	struct sock *sk = sock->sk;
879 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
880 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
881 	struct tipc_sock *tsk = tipc_sk(sk);
882 	struct tipc_group *grp = tsk->group;
883 	struct net *net = sock_net(sk);
884 	struct tipc_member *mb = NULL;
885 	u32 node, port;
886 	int rc;
887 
888 	node = dest->addr.id.node;
889 	port = dest->addr.id.ref;
890 	if (!port && !node)
891 		return -EHOSTUNREACH;
892 
893 	/* Block or return if destination link or member is congested */
894 	rc = tipc_wait_for_cond(sock, &timeout,
895 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
896 				!tipc_group_cong(grp, node, port, blks, &mb));
897 	if (unlikely(rc))
898 		return rc;
899 
900 	if (unlikely(!mb))
901 		return -EHOSTUNREACH;
902 
903 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
904 
905 	return rc ? rc : dlen;
906 }
907 
908 /**
909  * tipc_send_group_anycast - send message to any member with given identity
910  * @sock: socket structure
911  * @m: message to send
912  * @dlen: total length of message data
913  * @timeout: timeout to wait for wakeup
914  *
915  * Called from function tipc_sendmsg(), which has done all sanity checks
916  * Returns the number of bytes sent on success, or errno
917  */
918 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
919 				   int dlen, long timeout)
920 {
921 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
922 	struct sock *sk = sock->sk;
923 	struct tipc_sock *tsk = tipc_sk(sk);
924 	struct list_head *cong_links = &tsk->cong_links;
925 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
926 	struct tipc_group *grp = tsk->group;
927 	struct tipc_msg *hdr = &tsk->phdr;
928 	struct tipc_member *first = NULL;
929 	struct tipc_member *mbr = NULL;
930 	struct net *net = sock_net(sk);
931 	u32 node, port, exclude;
932 	struct list_head dsts;
933 	u32 type, inst, scope;
934 	int lookups = 0;
935 	int dstcnt, rc;
936 	bool cong;
937 
938 	INIT_LIST_HEAD(&dsts);
939 
940 	type = msg_nametype(hdr);
941 	inst = dest->addr.name.name.instance;
942 	scope = msg_lookup_scope(hdr);
943 	exclude = tipc_group_exclude(grp);
944 
945 	while (++lookups < 4) {
946 		first = NULL;
947 
948 		/* Look for a non-congested destination member, if any */
949 		while (1) {
950 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
951 						 &dstcnt, exclude, false))
952 				return -EHOSTUNREACH;
953 			tipc_dest_pop(&dsts, &node, &port);
954 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
955 			if (!cong)
956 				break;
957 			if (mbr == first)
958 				break;
959 			if (!first)
960 				first = mbr;
961 		}
962 
963 		/* Start over if destination was not in member list */
964 		if (unlikely(!mbr))
965 			continue;
966 
967 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
968 			break;
969 
970 		/* Block or return if destination link or member is congested */
971 		rc = tipc_wait_for_cond(sock, &timeout,
972 					!tipc_dest_find(cong_links, node, 0) &&
973 					!tipc_group_cong(grp, node, port,
974 							 blks, &mbr));
975 		if (unlikely(rc))
976 			return rc;
977 
978 		/* Send, unless destination disappeared while waiting */
979 		if (likely(mbr))
980 			break;
981 	}
982 
983 	if (unlikely(lookups >= 4))
984 		return -EHOSTUNREACH;
985 
986 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
987 
988 	return rc ? rc : dlen;
989 }
990 
991 /**
992  * tipc_send_group_bcast - send message to all members in communication group
993  * @sk: socket structure
994  * @m: message to send
995  * @dlen: total length of message data
996  * @timeout: timeout to wait for wakeup
997  *
998  * Called from function tipc_sendmsg(), which has done all sanity checks
999  * Returns the number of bytes sent on success, or errno
1000  */
1001 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1002 				 int dlen, long timeout)
1003 {
1004 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1005 	struct sock *sk = sock->sk;
1006 	struct net *net = sock_net(sk);
1007 	struct tipc_sock *tsk = tipc_sk(sk);
1008 	struct tipc_group *grp = tsk->group;
1009 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1010 	struct tipc_mc_method *method = &tsk->mc_method;
1011 	bool ack = method->mandatory && method->rcast;
1012 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1013 	struct tipc_msg *hdr = &tsk->phdr;
1014 	int mtu = tipc_bcast_get_mtu(net);
1015 	struct sk_buff_head pkts;
1016 	int rc = -EHOSTUNREACH;
1017 
1018 	if (!dsts->local && !dsts->remote)
1019 		return -EHOSTUNREACH;
1020 
1021 	/* Block or return if any destination link or member is congested */
1022 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1023 				!tipc_group_bc_cong(grp, blks));
1024 	if (unlikely(rc))
1025 		return rc;
1026 
1027 	/* Complete message header */
1028 	if (dest) {
1029 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1030 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1031 	} else {
1032 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1033 		msg_set_nameinst(hdr, 0);
1034 	}
1035 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1036 	msg_set_destport(hdr, 0);
1037 	msg_set_destnode(hdr, 0);
1038 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1039 
1040 	/* Avoid getting stuck with repeated forced replicasts */
1041 	msg_set_grp_bc_ack_req(hdr, ack);
1042 
1043 	/* Build message as chain of buffers */
1044 	skb_queue_head_init(&pkts);
1045 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1046 	if (unlikely(rc != dlen))
1047 		return rc;
1048 
1049 	/* Send message */
1050 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1051 	if (unlikely(rc))
1052 		return rc;
1053 
1054 	/* Update broadcast sequence number and send windows */
1055 	tipc_group_update_bc_members(tsk->group, blks, ack);
1056 
1057 	/* Broadcast link is now free to choose method for next broadcast */
1058 	method->mandatory = false;
1059 	method->expires = jiffies;
1060 
1061 	return dlen;
1062 }
1063 
1064 /**
1065  * tipc_send_group_mcast - send message to all members with given identity
1066  * @sock: socket structure
1067  * @m: message to send
1068  * @dlen: total length of message data
1069  * @timeout: timeout to wait for wakeup
1070  *
1071  * Called from function tipc_sendmsg(), which has done all sanity checks
1072  * Returns the number of bytes sent on success, or errno
1073  */
1074 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1075 				 int dlen, long timeout)
1076 {
1077 	struct sock *sk = sock->sk;
1078 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1079 	struct tipc_sock *tsk = tipc_sk(sk);
1080 	struct tipc_group *grp = tsk->group;
1081 	struct tipc_msg *hdr = &tsk->phdr;
1082 	struct net *net = sock_net(sk);
1083 	u32 type, inst, scope, exclude;
1084 	struct list_head dsts;
1085 	u32 dstcnt;
1086 
1087 	INIT_LIST_HEAD(&dsts);
1088 
1089 	type = msg_nametype(hdr);
1090 	inst = dest->addr.name.name.instance;
1091 	scope = msg_lookup_scope(hdr);
1092 	exclude = tipc_group_exclude(grp);
1093 
1094 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1095 				 &dstcnt, exclude, true))
1096 		return -EHOSTUNREACH;
1097 
1098 	if (dstcnt == 1) {
1099 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1100 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1101 	}
1102 
1103 	tipc_dest_list_purge(&dsts);
1104 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1105 }
1106 
1107 /**
1108  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1109  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1110  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1111  *
1112  * Multi-threaded: parallel calls with reference to same queues may occur
1113  */
1114 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1115 		       struct sk_buff_head *inputq)
1116 {
1117 	u32 self = tipc_own_addr(net);
1118 	u32 type, lower, upper, scope;
1119 	struct sk_buff *skb, *_skb;
1120 	u32 portid, oport, onode;
1121 	struct sk_buff_head tmpq;
1122 	struct list_head dports;
1123 	struct tipc_msg *hdr;
1124 	int user, mtyp, hlen;
1125 	bool exact;
1126 
1127 	__skb_queue_head_init(&tmpq);
1128 	INIT_LIST_HEAD(&dports);
1129 
1130 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1131 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1132 		hdr = buf_msg(skb);
1133 		user = msg_user(hdr);
1134 		mtyp = msg_type(hdr);
1135 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1136 		oport = msg_origport(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 *xmitq)
1202 {
1203 	struct tipc_msg *hdr = buf_msg(skb);
1204 	u32 onode = tsk_own_node(tsk);
1205 	struct sock *sk = &tsk->sk;
1206 	int mtyp = msg_type(hdr);
1207 	bool conn_cong;
1208 
1209 	/* Ignore if connection cannot be validated: */
1210 	if (!tsk_peer_msg(tsk, hdr))
1211 		goto exit;
1212 
1213 	if (unlikely(msg_errcode(hdr))) {
1214 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1215 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1216 				      tsk_peer_port(tsk));
1217 		sk->sk_state_change(sk);
1218 		goto exit;
1219 	}
1220 
1221 	tsk->probe_unacked = false;
1222 
1223 	if (mtyp == CONN_PROBE) {
1224 		msg_set_type(hdr, CONN_PROBE_REPLY);
1225 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1226 			__skb_queue_tail(xmitq, skb);
1227 		return;
1228 	} else if (mtyp == CONN_ACK) {
1229 		conn_cong = tsk_conn_cong(tsk);
1230 		tsk->snt_unacked -= msg_conn_ack(hdr);
1231 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1232 			tsk->snd_win = msg_adv_win(hdr);
1233 		if (conn_cong)
1234 			sk->sk_write_space(sk);
1235 	} else if (mtyp != CONN_PROBE_REPLY) {
1236 		pr_warn("Received unknown CONN_PROTO msg\n");
1237 	}
1238 exit:
1239 	kfree_skb(skb);
1240 }
1241 
1242 /**
1243  * tipc_sendmsg - send message in connectionless manner
1244  * @sock: socket structure
1245  * @m: message to send
1246  * @dsz: amount of user data to be sent
1247  *
1248  * Message must have an destination specified explicitly.
1249  * Used for SOCK_RDM and SOCK_DGRAM messages,
1250  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1251  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1252  *
1253  * Returns the number of bytes sent on success, or errno otherwise
1254  */
1255 static int tipc_sendmsg(struct socket *sock,
1256 			struct msghdr *m, size_t dsz)
1257 {
1258 	struct sock *sk = sock->sk;
1259 	int ret;
1260 
1261 	lock_sock(sk);
1262 	ret = __tipc_sendmsg(sock, m, dsz);
1263 	release_sock(sk);
1264 
1265 	return ret;
1266 }
1267 
1268 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1269 {
1270 	struct sock *sk = sock->sk;
1271 	struct net *net = sock_net(sk);
1272 	struct tipc_sock *tsk = tipc_sk(sk);
1273 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1274 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1275 	struct list_head *clinks = &tsk->cong_links;
1276 	bool syn = !tipc_sk_type_connectionless(sk);
1277 	struct tipc_group *grp = tsk->group;
1278 	struct tipc_msg *hdr = &tsk->phdr;
1279 	struct tipc_name_seq *seq;
1280 	struct sk_buff_head pkts;
1281 	u32 dnode, dport;
1282 	u32 type, inst;
1283 	int mtu, rc;
1284 
1285 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1286 		return -EMSGSIZE;
1287 
1288 	if (likely(dest)) {
1289 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1290 			return -EINVAL;
1291 		if (unlikely(dest->family != AF_TIPC))
1292 			return -EINVAL;
1293 	}
1294 
1295 	if (grp) {
1296 		if (!dest)
1297 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1298 		if (dest->addrtype == TIPC_ADDR_NAME)
1299 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1300 		if (dest->addrtype == TIPC_ADDR_ID)
1301 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1302 		if (dest->addrtype == TIPC_ADDR_MCAST)
1303 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1304 		return -EINVAL;
1305 	}
1306 
1307 	if (unlikely(!dest)) {
1308 		dest = &tsk->peer;
1309 		if (!syn || dest->family != AF_TIPC)
1310 			return -EDESTADDRREQ;
1311 	}
1312 
1313 	if (unlikely(syn)) {
1314 		if (sk->sk_state == TIPC_LISTEN)
1315 			return -EPIPE;
1316 		if (sk->sk_state != TIPC_OPEN)
1317 			return -EISCONN;
1318 		if (tsk->published)
1319 			return -EOPNOTSUPP;
1320 		if (dest->addrtype == TIPC_ADDR_NAME) {
1321 			tsk->conn_type = dest->addr.name.name.type;
1322 			tsk->conn_instance = dest->addr.name.name.instance;
1323 		}
1324 	}
1325 
1326 	seq = &dest->addr.nameseq;
1327 	if (dest->addrtype == TIPC_ADDR_MCAST)
1328 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1329 
1330 	if (dest->addrtype == TIPC_ADDR_NAME) {
1331 		type = dest->addr.name.name.type;
1332 		inst = dest->addr.name.name.instance;
1333 		dnode = dest->addr.name.domain;
1334 		msg_set_type(hdr, TIPC_NAMED_MSG);
1335 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1336 		msg_set_nametype(hdr, type);
1337 		msg_set_nameinst(hdr, inst);
1338 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1339 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1340 		msg_set_destnode(hdr, dnode);
1341 		msg_set_destport(hdr, dport);
1342 		if (unlikely(!dport && !dnode))
1343 			return -EHOSTUNREACH;
1344 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1345 		dnode = dest->addr.id.node;
1346 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1347 		msg_set_lookup_scope(hdr, 0);
1348 		msg_set_destnode(hdr, dnode);
1349 		msg_set_destport(hdr, dest->addr.id.ref);
1350 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1351 	}
1352 
1353 	/* Block or return if destination link is congested */
1354 	rc = tipc_wait_for_cond(sock, &timeout,
1355 				!tipc_dest_find(clinks, dnode, 0));
1356 	if (unlikely(rc))
1357 		return rc;
1358 
1359 	skb_queue_head_init(&pkts);
1360 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1361 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1362 	if (unlikely(rc != dlen))
1363 		return rc;
1364 
1365 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1366 	if (unlikely(rc == -ELINKCONG)) {
1367 		tipc_dest_push(clinks, dnode, 0);
1368 		tsk->cong_link_cnt++;
1369 		rc = 0;
1370 	}
1371 
1372 	if (unlikely(syn && !rc))
1373 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1374 
1375 	return rc ? rc : dlen;
1376 }
1377 
1378 /**
1379  * tipc_sendstream - send stream-oriented data
1380  * @sock: socket structure
1381  * @m: data to send
1382  * @dsz: total length of data to be transmitted
1383  *
1384  * Used for SOCK_STREAM data.
1385  *
1386  * Returns the number of bytes sent on success (or partial success),
1387  * or errno if no data sent
1388  */
1389 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1390 {
1391 	struct sock *sk = sock->sk;
1392 	int ret;
1393 
1394 	lock_sock(sk);
1395 	ret = __tipc_sendstream(sock, m, dsz);
1396 	release_sock(sk);
1397 
1398 	return ret;
1399 }
1400 
1401 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1402 {
1403 	struct sock *sk = sock->sk;
1404 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1405 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1406 	struct tipc_sock *tsk = tipc_sk(sk);
1407 	struct tipc_msg *hdr = &tsk->phdr;
1408 	struct net *net = sock_net(sk);
1409 	struct sk_buff_head pkts;
1410 	u32 dnode = tsk_peer_node(tsk);
1411 	int send, sent = 0;
1412 	int rc = 0;
1413 
1414 	skb_queue_head_init(&pkts);
1415 
1416 	if (unlikely(dlen > INT_MAX))
1417 		return -EMSGSIZE;
1418 
1419 	/* Handle implicit connection setup */
1420 	if (unlikely(dest)) {
1421 		rc = __tipc_sendmsg(sock, m, dlen);
1422 		if (dlen && (dlen == rc))
1423 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1424 		return rc;
1425 	}
1426 
1427 	do {
1428 		rc = tipc_wait_for_cond(sock, &timeout,
1429 					(!tsk->cong_link_cnt &&
1430 					 !tsk_conn_cong(tsk) &&
1431 					 tipc_sk_connected(sk)));
1432 		if (unlikely(rc))
1433 			break;
1434 
1435 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1436 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1437 		if (unlikely(rc != send))
1438 			break;
1439 
1440 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1441 		if (unlikely(rc == -ELINKCONG)) {
1442 			tsk->cong_link_cnt = 1;
1443 			rc = 0;
1444 		}
1445 		if (likely(!rc)) {
1446 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1447 			sent += send;
1448 		}
1449 	} while (sent < dlen && !rc);
1450 
1451 	return sent ? sent : rc;
1452 }
1453 
1454 /**
1455  * tipc_send_packet - send a connection-oriented message
1456  * @sock: socket structure
1457  * @m: message to send
1458  * @dsz: length of data to be transmitted
1459  *
1460  * Used for SOCK_SEQPACKET messages.
1461  *
1462  * Returns the number of bytes sent on success, or errno otherwise
1463  */
1464 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1465 {
1466 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1467 		return -EMSGSIZE;
1468 
1469 	return tipc_sendstream(sock, m, dsz);
1470 }
1471 
1472 /* tipc_sk_finish_conn - complete the setup of a connection
1473  */
1474 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1475 				u32 peer_node)
1476 {
1477 	struct sock *sk = &tsk->sk;
1478 	struct net *net = sock_net(sk);
1479 	struct tipc_msg *msg = &tsk->phdr;
1480 
1481 	msg_set_destnode(msg, peer_node);
1482 	msg_set_destport(msg, peer_port);
1483 	msg_set_type(msg, TIPC_CONN_MSG);
1484 	msg_set_lookup_scope(msg, 0);
1485 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1486 
1487 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1488 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1489 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1490 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1491 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1492 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1493 		return;
1494 
1495 	/* Fall back to message based flow control */
1496 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1497 	tsk->snd_win = FLOWCTL_MSG_WIN;
1498 }
1499 
1500 /**
1501  * tipc_sk_set_orig_addr - capture sender's address for received message
1502  * @m: descriptor for message info
1503  * @hdr: received message header
1504  *
1505  * Note: Address is not captured if not requested by receiver.
1506  */
1507 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1508 {
1509 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1510 	struct tipc_msg *hdr = buf_msg(skb);
1511 
1512 	if (!srcaddr)
1513 		return;
1514 
1515 	srcaddr->sock.family = AF_TIPC;
1516 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1517 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1518 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1519 	srcaddr->sock.addr.name.domain = 0;
1520 	srcaddr->sock.scope = 0;
1521 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1522 
1523 	if (!msg_in_group(hdr))
1524 		return;
1525 
1526 	/* Group message users may also want to know sending member's id */
1527 	srcaddr->member.family = AF_TIPC;
1528 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1529 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1530 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1531 	srcaddr->member.addr.name.domain = 0;
1532 	m->msg_namelen = sizeof(*srcaddr);
1533 }
1534 
1535 /**
1536  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1537  * @m: descriptor for message info
1538  * @msg: received message header
1539  * @tsk: TIPC port associated with message
1540  *
1541  * Note: Ancillary data is not captured if not requested by receiver.
1542  *
1543  * Returns 0 if successful, otherwise errno
1544  */
1545 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1546 				 struct tipc_sock *tsk)
1547 {
1548 	u32 anc_data[3];
1549 	u32 err;
1550 	u32 dest_type;
1551 	int has_name;
1552 	int res;
1553 
1554 	if (likely(m->msg_controllen == 0))
1555 		return 0;
1556 
1557 	/* Optionally capture errored message object(s) */
1558 	err = msg ? msg_errcode(msg) : 0;
1559 	if (unlikely(err)) {
1560 		anc_data[0] = err;
1561 		anc_data[1] = msg_data_sz(msg);
1562 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1563 		if (res)
1564 			return res;
1565 		if (anc_data[1]) {
1566 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1567 				       msg_data(msg));
1568 			if (res)
1569 				return res;
1570 		}
1571 	}
1572 
1573 	/* Optionally capture message destination object */
1574 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1575 	switch (dest_type) {
1576 	case TIPC_NAMED_MSG:
1577 		has_name = 1;
1578 		anc_data[0] = msg_nametype(msg);
1579 		anc_data[1] = msg_namelower(msg);
1580 		anc_data[2] = msg_namelower(msg);
1581 		break;
1582 	case TIPC_MCAST_MSG:
1583 		has_name = 1;
1584 		anc_data[0] = msg_nametype(msg);
1585 		anc_data[1] = msg_namelower(msg);
1586 		anc_data[2] = msg_nameupper(msg);
1587 		break;
1588 	case TIPC_CONN_MSG:
1589 		has_name = (tsk->conn_type != 0);
1590 		anc_data[0] = tsk->conn_type;
1591 		anc_data[1] = tsk->conn_instance;
1592 		anc_data[2] = tsk->conn_instance;
1593 		break;
1594 	default:
1595 		has_name = 0;
1596 	}
1597 	if (has_name) {
1598 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1599 		if (res)
1600 			return res;
1601 	}
1602 
1603 	return 0;
1604 }
1605 
1606 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1607 {
1608 	struct sock *sk = &tsk->sk;
1609 	struct net *net = sock_net(sk);
1610 	struct sk_buff *skb = NULL;
1611 	struct tipc_msg *msg;
1612 	u32 peer_port = tsk_peer_port(tsk);
1613 	u32 dnode = tsk_peer_node(tsk);
1614 
1615 	if (!tipc_sk_connected(sk))
1616 		return;
1617 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1618 			      dnode, tsk_own_node(tsk), peer_port,
1619 			      tsk->portid, TIPC_OK);
1620 	if (!skb)
1621 		return;
1622 	msg = buf_msg(skb);
1623 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1624 	tsk->rcv_unacked = 0;
1625 
1626 	/* Adjust to and advertize the correct window limit */
1627 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1628 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1629 		msg_set_adv_win(msg, tsk->rcv_win);
1630 	}
1631 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1632 }
1633 
1634 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1635 {
1636 	struct sock *sk = sock->sk;
1637 	DEFINE_WAIT(wait);
1638 	long timeo = *timeop;
1639 	int err = sock_error(sk);
1640 
1641 	if (err)
1642 		return err;
1643 
1644 	for (;;) {
1645 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1646 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1647 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1648 				err = -ENOTCONN;
1649 				break;
1650 			}
1651 			release_sock(sk);
1652 			timeo = schedule_timeout(timeo);
1653 			lock_sock(sk);
1654 		}
1655 		err = 0;
1656 		if (!skb_queue_empty(&sk->sk_receive_queue))
1657 			break;
1658 		err = -EAGAIN;
1659 		if (!timeo)
1660 			break;
1661 		err = sock_intr_errno(timeo);
1662 		if (signal_pending(current))
1663 			break;
1664 
1665 		err = sock_error(sk);
1666 		if (err)
1667 			break;
1668 	}
1669 	finish_wait(sk_sleep(sk), &wait);
1670 	*timeop = timeo;
1671 	return err;
1672 }
1673 
1674 /**
1675  * tipc_recvmsg - receive packet-oriented message
1676  * @m: descriptor for message info
1677  * @buflen: length of user buffer area
1678  * @flags: receive flags
1679  *
1680  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1681  * If the complete message doesn't fit in user area, truncate it.
1682  *
1683  * Returns size of returned message data, errno otherwise
1684  */
1685 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1686 			size_t buflen,	int flags)
1687 {
1688 	struct sock *sk = sock->sk;
1689 	bool connected = !tipc_sk_type_connectionless(sk);
1690 	struct tipc_sock *tsk = tipc_sk(sk);
1691 	int rc, err, hlen, dlen, copy;
1692 	struct sk_buff_head xmitq;
1693 	struct tipc_msg *hdr;
1694 	struct sk_buff *skb;
1695 	bool grp_evt;
1696 	long timeout;
1697 
1698 	/* Catch invalid receive requests */
1699 	if (unlikely(!buflen))
1700 		return -EINVAL;
1701 
1702 	lock_sock(sk);
1703 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1704 		rc = -ENOTCONN;
1705 		goto exit;
1706 	}
1707 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1708 
1709 	/* Step rcv queue to first msg with data or error; wait if necessary */
1710 	do {
1711 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1712 		if (unlikely(rc))
1713 			goto exit;
1714 		skb = skb_peek(&sk->sk_receive_queue);
1715 		hdr = buf_msg(skb);
1716 		dlen = msg_data_sz(hdr);
1717 		hlen = msg_hdr_sz(hdr);
1718 		err = msg_errcode(hdr);
1719 		grp_evt = msg_is_grp_evt(hdr);
1720 		if (likely(dlen || err))
1721 			break;
1722 		tsk_advance_rx_queue(sk);
1723 	} while (1);
1724 
1725 	/* Collect msg meta data, including error code and rejected data */
1726 	tipc_sk_set_orig_addr(m, skb);
1727 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1728 	if (unlikely(rc))
1729 		goto exit;
1730 
1731 	/* Capture data if non-error msg, otherwise just set return value */
1732 	if (likely(!err)) {
1733 		copy = min_t(int, dlen, buflen);
1734 		if (unlikely(copy != dlen))
1735 			m->msg_flags |= MSG_TRUNC;
1736 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1737 	} else {
1738 		copy = 0;
1739 		rc = 0;
1740 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1741 			rc = -ECONNRESET;
1742 	}
1743 	if (unlikely(rc))
1744 		goto exit;
1745 
1746 	/* Mark message as group event if applicable */
1747 	if (unlikely(grp_evt)) {
1748 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1749 			m->msg_flags |= MSG_EOR;
1750 		m->msg_flags |= MSG_OOB;
1751 		copy = 0;
1752 	}
1753 
1754 	/* Caption of data or error code/rejected data was successful */
1755 	if (unlikely(flags & MSG_PEEK))
1756 		goto exit;
1757 
1758 	/* Send group flow control advertisement when applicable */
1759 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1760 		skb_queue_head_init(&xmitq);
1761 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1762 					  msg_orignode(hdr), msg_origport(hdr),
1763 					  &xmitq);
1764 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1765 	}
1766 
1767 	tsk_advance_rx_queue(sk);
1768 
1769 	if (likely(!connected))
1770 		goto exit;
1771 
1772 	/* Send connection flow control advertisement when applicable */
1773 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1774 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1775 		tipc_sk_send_ack(tsk);
1776 exit:
1777 	release_sock(sk);
1778 	return rc ? rc : copy;
1779 }
1780 
1781 /**
1782  * tipc_recvstream - receive stream-oriented data
1783  * @m: descriptor for message info
1784  * @buflen: total size of user buffer area
1785  * @flags: receive flags
1786  *
1787  * Used for SOCK_STREAM messages only.  If not enough data is available
1788  * will optionally wait for more; never truncates data.
1789  *
1790  * Returns size of returned message data, errno otherwise
1791  */
1792 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1793 			   size_t buflen, int flags)
1794 {
1795 	struct sock *sk = sock->sk;
1796 	struct tipc_sock *tsk = tipc_sk(sk);
1797 	struct sk_buff *skb;
1798 	struct tipc_msg *hdr;
1799 	struct tipc_skb_cb *skb_cb;
1800 	bool peek = flags & MSG_PEEK;
1801 	int offset, required, copy, copied = 0;
1802 	int hlen, dlen, err, rc;
1803 	long timeout;
1804 
1805 	/* Catch invalid receive attempts */
1806 	if (unlikely(!buflen))
1807 		return -EINVAL;
1808 
1809 	lock_sock(sk);
1810 
1811 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1812 		rc = -ENOTCONN;
1813 		goto exit;
1814 	}
1815 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1816 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1817 
1818 	do {
1819 		/* Look at first msg in receive queue; wait if necessary */
1820 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1821 		if (unlikely(rc))
1822 			break;
1823 		skb = skb_peek(&sk->sk_receive_queue);
1824 		skb_cb = TIPC_SKB_CB(skb);
1825 		hdr = buf_msg(skb);
1826 		dlen = msg_data_sz(hdr);
1827 		hlen = msg_hdr_sz(hdr);
1828 		err = msg_errcode(hdr);
1829 
1830 		/* Discard any empty non-errored (SYN-) message */
1831 		if (unlikely(!dlen && !err)) {
1832 			tsk_advance_rx_queue(sk);
1833 			continue;
1834 		}
1835 
1836 		/* Collect msg meta data, incl. error code and rejected data */
1837 		if (!copied) {
1838 			tipc_sk_set_orig_addr(m, skb);
1839 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1840 			if (rc)
1841 				break;
1842 		}
1843 
1844 		/* Copy data if msg ok, otherwise return error/partial data */
1845 		if (likely(!err)) {
1846 			offset = skb_cb->bytes_read;
1847 			copy = min_t(int, dlen - offset, buflen - copied);
1848 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1849 			if (unlikely(rc))
1850 				break;
1851 			copied += copy;
1852 			offset += copy;
1853 			if (unlikely(offset < dlen)) {
1854 				if (!peek)
1855 					skb_cb->bytes_read = offset;
1856 				break;
1857 			}
1858 		} else {
1859 			rc = 0;
1860 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1861 				rc = -ECONNRESET;
1862 			if (copied || rc)
1863 				break;
1864 		}
1865 
1866 		if (unlikely(peek))
1867 			break;
1868 
1869 		tsk_advance_rx_queue(sk);
1870 
1871 		/* Send connection flow control advertisement when applicable */
1872 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1873 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1874 			tipc_sk_send_ack(tsk);
1875 
1876 		/* Exit if all requested data or FIN/error received */
1877 		if (copied == buflen || err)
1878 			break;
1879 
1880 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1881 exit:
1882 	release_sock(sk);
1883 	return copied ? copied : rc;
1884 }
1885 
1886 /**
1887  * tipc_write_space - wake up thread if port congestion is released
1888  * @sk: socket
1889  */
1890 static void tipc_write_space(struct sock *sk)
1891 {
1892 	struct socket_wq *wq;
1893 
1894 	rcu_read_lock();
1895 	wq = rcu_dereference(sk->sk_wq);
1896 	if (skwq_has_sleeper(wq))
1897 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1898 						EPOLLWRNORM | EPOLLWRBAND);
1899 	rcu_read_unlock();
1900 }
1901 
1902 /**
1903  * tipc_data_ready - wake up threads to indicate messages have been received
1904  * @sk: socket
1905  * @len: the length of messages
1906  */
1907 static void tipc_data_ready(struct sock *sk)
1908 {
1909 	struct socket_wq *wq;
1910 
1911 	rcu_read_lock();
1912 	wq = rcu_dereference(sk->sk_wq);
1913 	if (skwq_has_sleeper(wq))
1914 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1915 						EPOLLRDNORM | EPOLLRDBAND);
1916 	rcu_read_unlock();
1917 }
1918 
1919 static void tipc_sock_destruct(struct sock *sk)
1920 {
1921 	__skb_queue_purge(&sk->sk_receive_queue);
1922 }
1923 
1924 static void tipc_sk_proto_rcv(struct sock *sk,
1925 			      struct sk_buff_head *inputq,
1926 			      struct sk_buff_head *xmitq)
1927 {
1928 	struct sk_buff *skb = __skb_dequeue(inputq);
1929 	struct tipc_sock *tsk = tipc_sk(sk);
1930 	struct tipc_msg *hdr = buf_msg(skb);
1931 	struct tipc_group *grp = tsk->group;
1932 	bool wakeup = false;
1933 
1934 	switch (msg_user(hdr)) {
1935 	case CONN_MANAGER:
1936 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1937 		return;
1938 	case SOCK_WAKEUP:
1939 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1940 		tsk->cong_link_cnt--;
1941 		wakeup = true;
1942 		break;
1943 	case GROUP_PROTOCOL:
1944 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1945 		break;
1946 	case TOP_SRV:
1947 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1948 				      hdr, inputq, xmitq);
1949 		break;
1950 	default:
1951 		break;
1952 	}
1953 
1954 	if (wakeup)
1955 		sk->sk_write_space(sk);
1956 
1957 	kfree_skb(skb);
1958 }
1959 
1960 /**
1961  * tipc_filter_connect - Handle incoming message for a connection-based socket
1962  * @tsk: TIPC socket
1963  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1964  *
1965  * Returns true if everything ok, false otherwise
1966  */
1967 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1968 {
1969 	struct sock *sk = &tsk->sk;
1970 	struct net *net = sock_net(sk);
1971 	struct tipc_msg *hdr = buf_msg(skb);
1972 	u32 pport = msg_origport(hdr);
1973 	u32 pnode = msg_orignode(hdr);
1974 
1975 	if (unlikely(msg_mcast(hdr)))
1976 		return false;
1977 
1978 	switch (sk->sk_state) {
1979 	case TIPC_CONNECTING:
1980 		/* Accept only ACK or NACK message */
1981 		if (unlikely(!msg_connected(hdr))) {
1982 			if (pport != tsk_peer_port(tsk) ||
1983 			    pnode != tsk_peer_node(tsk))
1984 				return false;
1985 
1986 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1987 			sk->sk_err = ECONNREFUSED;
1988 			sk->sk_state_change(sk);
1989 			return true;
1990 		}
1991 
1992 		if (unlikely(msg_errcode(hdr))) {
1993 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1994 			sk->sk_err = ECONNREFUSED;
1995 			sk->sk_state_change(sk);
1996 			return true;
1997 		}
1998 
1999 		if (unlikely(!msg_isdata(hdr))) {
2000 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2001 			sk->sk_err = EINVAL;
2002 			sk->sk_state_change(sk);
2003 			return true;
2004 		}
2005 
2006 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2007 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2008 
2009 		/* If 'ACK+' message, add to socket receive queue */
2010 		if (msg_data_sz(hdr))
2011 			return true;
2012 
2013 		/* If empty 'ACK-' message, wake up sleeping connect() */
2014 		sk->sk_data_ready(sk);
2015 
2016 		/* 'ACK-' message is neither accepted nor rejected: */
2017 		msg_set_dest_droppable(hdr, 1);
2018 		return false;
2019 
2020 	case TIPC_OPEN:
2021 	case TIPC_DISCONNECTING:
2022 		break;
2023 	case TIPC_LISTEN:
2024 		/* Accept only SYN message */
2025 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2026 			return true;
2027 		break;
2028 	case TIPC_ESTABLISHED:
2029 		/* Accept only connection-based messages sent by peer */
2030 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2031 			return false;
2032 
2033 		if (unlikely(msg_errcode(hdr))) {
2034 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2035 			/* Let timer expire on it's own */
2036 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2037 					      tsk->portid);
2038 			sk->sk_state_change(sk);
2039 		}
2040 		return true;
2041 	default:
2042 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2043 	}
2044 
2045 	return false;
2046 }
2047 
2048 /**
2049  * rcvbuf_limit - get proper overload limit of socket receive queue
2050  * @sk: socket
2051  * @skb: message
2052  *
2053  * For connection oriented messages, irrespective of importance,
2054  * default queue limit is 2 MB.
2055  *
2056  * For connectionless messages, queue limits are based on message
2057  * importance as follows:
2058  *
2059  * TIPC_LOW_IMPORTANCE       (2 MB)
2060  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2061  * TIPC_HIGH_IMPORTANCE      (8 MB)
2062  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2063  *
2064  * Returns overload limit according to corresponding message importance
2065  */
2066 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2067 {
2068 	struct tipc_sock *tsk = tipc_sk(sk);
2069 	struct tipc_msg *hdr = buf_msg(skb);
2070 
2071 	if (unlikely(msg_in_group(hdr)))
2072 		return sk->sk_rcvbuf;
2073 
2074 	if (unlikely(!msg_connected(hdr)))
2075 		return sk->sk_rcvbuf << msg_importance(hdr);
2076 
2077 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2078 		return sk->sk_rcvbuf;
2079 
2080 	return FLOWCTL_MSG_LIM;
2081 }
2082 
2083 /**
2084  * tipc_sk_filter_rcv - validate incoming message
2085  * @sk: socket
2086  * @skb: pointer to message.
2087  *
2088  * Enqueues message on receive queue if acceptable; optionally handles
2089  * disconnect indication for a connected socket.
2090  *
2091  * Called with socket lock already taken
2092  *
2093  */
2094 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2095 			       struct sk_buff_head *xmitq)
2096 {
2097 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2098 	struct tipc_sock *tsk = tipc_sk(sk);
2099 	struct tipc_group *grp = tsk->group;
2100 	struct tipc_msg *hdr = buf_msg(skb);
2101 	struct net *net = sock_net(sk);
2102 	struct sk_buff_head inputq;
2103 	int limit, err = TIPC_OK;
2104 
2105 	TIPC_SKB_CB(skb)->bytes_read = 0;
2106 	__skb_queue_head_init(&inputq);
2107 	__skb_queue_tail(&inputq, skb);
2108 
2109 	if (unlikely(!msg_isdata(hdr)))
2110 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2111 
2112 	if (unlikely(grp))
2113 		tipc_group_filter_msg(grp, &inputq, xmitq);
2114 
2115 	/* Validate and add to receive buffer if there is space */
2116 	while ((skb = __skb_dequeue(&inputq))) {
2117 		hdr = buf_msg(skb);
2118 		limit = rcvbuf_limit(sk, skb);
2119 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2120 		    (!sk_conn && msg_connected(hdr)) ||
2121 		    (!grp && msg_in_group(hdr)))
2122 			err = TIPC_ERR_NO_PORT;
2123 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2124 			atomic_inc(&sk->sk_drops);
2125 			err = TIPC_ERR_OVERLOAD;
2126 		}
2127 
2128 		if (unlikely(err)) {
2129 			tipc_skb_reject(net, err, skb, xmitq);
2130 			err = TIPC_OK;
2131 			continue;
2132 		}
2133 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2134 		skb_set_owner_r(skb, sk);
2135 		sk->sk_data_ready(sk);
2136 	}
2137 }
2138 
2139 /**
2140  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2141  * @sk: socket
2142  * @skb: message
2143  *
2144  * Caller must hold socket lock
2145  */
2146 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2147 {
2148 	unsigned int before = sk_rmem_alloc_get(sk);
2149 	struct sk_buff_head xmitq;
2150 	unsigned int added;
2151 
2152 	__skb_queue_head_init(&xmitq);
2153 
2154 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2155 	added = sk_rmem_alloc_get(sk) - before;
2156 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2157 
2158 	/* Send pending response/rejected messages, if any */
2159 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2160 	return 0;
2161 }
2162 
2163 /**
2164  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2165  *                   inputq and try adding them to socket or backlog queue
2166  * @inputq: list of incoming buffers with potentially different destinations
2167  * @sk: socket where the buffers should be enqueued
2168  * @dport: port number for the socket
2169  *
2170  * Caller must hold socket lock
2171  */
2172 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2173 			    u32 dport, struct sk_buff_head *xmitq)
2174 {
2175 	unsigned long time_limit = jiffies + 2;
2176 	struct sk_buff *skb;
2177 	unsigned int lim;
2178 	atomic_t *dcnt;
2179 	u32 onode;
2180 
2181 	while (skb_queue_len(inputq)) {
2182 		if (unlikely(time_after_eq(jiffies, time_limit)))
2183 			return;
2184 
2185 		skb = tipc_skb_dequeue(inputq, dport);
2186 		if (unlikely(!skb))
2187 			return;
2188 
2189 		/* Add message directly to receive queue if possible */
2190 		if (!sock_owned_by_user(sk)) {
2191 			tipc_sk_filter_rcv(sk, skb, xmitq);
2192 			continue;
2193 		}
2194 
2195 		/* Try backlog, compensating for double-counted bytes */
2196 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2197 		if (!sk->sk_backlog.len)
2198 			atomic_set(dcnt, 0);
2199 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2200 		if (likely(!sk_add_backlog(sk, skb, lim)))
2201 			continue;
2202 
2203 		/* Overload => reject message back to sender */
2204 		onode = tipc_own_addr(sock_net(sk));
2205 		atomic_inc(&sk->sk_drops);
2206 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2207 			__skb_queue_tail(xmitq, skb);
2208 		break;
2209 	}
2210 }
2211 
2212 /**
2213  * tipc_sk_rcv - handle a chain of incoming buffers
2214  * @inputq: buffer list containing the buffers
2215  * Consumes all buffers in list until inputq is empty
2216  * Note: may be called in multiple threads referring to the same queue
2217  */
2218 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2219 {
2220 	struct sk_buff_head xmitq;
2221 	u32 dnode, dport = 0;
2222 	int err;
2223 	struct tipc_sock *tsk;
2224 	struct sock *sk;
2225 	struct sk_buff *skb;
2226 
2227 	__skb_queue_head_init(&xmitq);
2228 	while (skb_queue_len(inputq)) {
2229 		dport = tipc_skb_peek_port(inputq, dport);
2230 		tsk = tipc_sk_lookup(net, dport);
2231 
2232 		if (likely(tsk)) {
2233 			sk = &tsk->sk;
2234 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2235 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2236 				spin_unlock_bh(&sk->sk_lock.slock);
2237 			}
2238 			/* Send pending response/rejected messages, if any */
2239 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2240 			sock_put(sk);
2241 			continue;
2242 		}
2243 		/* No destination socket => dequeue skb if still there */
2244 		skb = tipc_skb_dequeue(inputq, dport);
2245 		if (!skb)
2246 			return;
2247 
2248 		/* Try secondary lookup if unresolved named message */
2249 		err = TIPC_ERR_NO_PORT;
2250 		if (tipc_msg_lookup_dest(net, skb, &err))
2251 			goto xmit;
2252 
2253 		/* Prepare for message rejection */
2254 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2255 			continue;
2256 xmit:
2257 		dnode = msg_destnode(buf_msg(skb));
2258 		tipc_node_xmit_skb(net, skb, dnode, dport);
2259 	}
2260 }
2261 
2262 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2263 {
2264 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2265 	struct sock *sk = sock->sk;
2266 	int done;
2267 
2268 	do {
2269 		int err = sock_error(sk);
2270 		if (err)
2271 			return err;
2272 		if (!*timeo_p)
2273 			return -ETIMEDOUT;
2274 		if (signal_pending(current))
2275 			return sock_intr_errno(*timeo_p);
2276 
2277 		add_wait_queue(sk_sleep(sk), &wait);
2278 		done = sk_wait_event(sk, timeo_p,
2279 				     sk->sk_state != TIPC_CONNECTING, &wait);
2280 		remove_wait_queue(sk_sleep(sk), &wait);
2281 	} while (!done);
2282 	return 0;
2283 }
2284 
2285 /**
2286  * tipc_connect - establish a connection to another TIPC port
2287  * @sock: socket structure
2288  * @dest: socket address for destination port
2289  * @destlen: size of socket address data structure
2290  * @flags: file-related flags associated with socket
2291  *
2292  * Returns 0 on success, errno otherwise
2293  */
2294 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2295 			int destlen, int flags)
2296 {
2297 	struct sock *sk = sock->sk;
2298 	struct tipc_sock *tsk = tipc_sk(sk);
2299 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2300 	struct msghdr m = {NULL,};
2301 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2302 	int previous;
2303 	int res = 0;
2304 
2305 	if (destlen != sizeof(struct sockaddr_tipc))
2306 		return -EINVAL;
2307 
2308 	lock_sock(sk);
2309 
2310 	if (tsk->group) {
2311 		res = -EINVAL;
2312 		goto exit;
2313 	}
2314 
2315 	if (dst->family == AF_UNSPEC) {
2316 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2317 		if (!tipc_sk_type_connectionless(sk))
2318 			res = -EINVAL;
2319 		goto exit;
2320 	} else if (dst->family != AF_TIPC) {
2321 		res = -EINVAL;
2322 	}
2323 	if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2324 		res = -EINVAL;
2325 	if (res)
2326 		goto exit;
2327 
2328 	/* DGRAM/RDM connect(), just save the destaddr */
2329 	if (tipc_sk_type_connectionless(sk)) {
2330 		memcpy(&tsk->peer, dest, destlen);
2331 		goto exit;
2332 	}
2333 
2334 	previous = sk->sk_state;
2335 
2336 	switch (sk->sk_state) {
2337 	case TIPC_OPEN:
2338 		/* Send a 'SYN-' to destination */
2339 		m.msg_name = dest;
2340 		m.msg_namelen = destlen;
2341 
2342 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2343 		 * indicate send_msg() is never blocked.
2344 		 */
2345 		if (!timeout)
2346 			m.msg_flags = MSG_DONTWAIT;
2347 
2348 		res = __tipc_sendmsg(sock, &m, 0);
2349 		if ((res < 0) && (res != -EWOULDBLOCK))
2350 			goto exit;
2351 
2352 		/* Just entered TIPC_CONNECTING state; the only
2353 		 * difference is that return value in non-blocking
2354 		 * case is EINPROGRESS, rather than EALREADY.
2355 		 */
2356 		res = -EINPROGRESS;
2357 		/* fall thru' */
2358 	case TIPC_CONNECTING:
2359 		if (!timeout) {
2360 			if (previous == TIPC_CONNECTING)
2361 				res = -EALREADY;
2362 			goto exit;
2363 		}
2364 		timeout = msecs_to_jiffies(timeout);
2365 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2366 		res = tipc_wait_for_connect(sock, &timeout);
2367 		break;
2368 	case TIPC_ESTABLISHED:
2369 		res = -EISCONN;
2370 		break;
2371 	default:
2372 		res = -EINVAL;
2373 	}
2374 
2375 exit:
2376 	release_sock(sk);
2377 	return res;
2378 }
2379 
2380 /**
2381  * tipc_listen - allow socket to listen for incoming connections
2382  * @sock: socket structure
2383  * @len: (unused)
2384  *
2385  * Returns 0 on success, errno otherwise
2386  */
2387 static int tipc_listen(struct socket *sock, int len)
2388 {
2389 	struct sock *sk = sock->sk;
2390 	int res;
2391 
2392 	lock_sock(sk);
2393 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2394 	release_sock(sk);
2395 
2396 	return res;
2397 }
2398 
2399 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2400 {
2401 	struct sock *sk = sock->sk;
2402 	DEFINE_WAIT(wait);
2403 	int err;
2404 
2405 	/* True wake-one mechanism for incoming connections: only
2406 	 * one process gets woken up, not the 'whole herd'.
2407 	 * Since we do not 'race & poll' for established sockets
2408 	 * anymore, the common case will execute the loop only once.
2409 	*/
2410 	for (;;) {
2411 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2412 					  TASK_INTERRUPTIBLE);
2413 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2414 			release_sock(sk);
2415 			timeo = schedule_timeout(timeo);
2416 			lock_sock(sk);
2417 		}
2418 		err = 0;
2419 		if (!skb_queue_empty(&sk->sk_receive_queue))
2420 			break;
2421 		err = -EAGAIN;
2422 		if (!timeo)
2423 			break;
2424 		err = sock_intr_errno(timeo);
2425 		if (signal_pending(current))
2426 			break;
2427 	}
2428 	finish_wait(sk_sleep(sk), &wait);
2429 	return err;
2430 }
2431 
2432 /**
2433  * tipc_accept - wait for connection request
2434  * @sock: listening socket
2435  * @newsock: new socket that is to be connected
2436  * @flags: file-related flags associated with socket
2437  *
2438  * Returns 0 on success, errno otherwise
2439  */
2440 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2441 		       bool kern)
2442 {
2443 	struct sock *new_sk, *sk = sock->sk;
2444 	struct sk_buff *buf;
2445 	struct tipc_sock *new_tsock;
2446 	struct tipc_msg *msg;
2447 	long timeo;
2448 	int res;
2449 
2450 	lock_sock(sk);
2451 
2452 	if (sk->sk_state != TIPC_LISTEN) {
2453 		res = -EINVAL;
2454 		goto exit;
2455 	}
2456 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2457 	res = tipc_wait_for_accept(sock, timeo);
2458 	if (res)
2459 		goto exit;
2460 
2461 	buf = skb_peek(&sk->sk_receive_queue);
2462 
2463 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2464 	if (res)
2465 		goto exit;
2466 	security_sk_clone(sock->sk, new_sock->sk);
2467 
2468 	new_sk = new_sock->sk;
2469 	new_tsock = tipc_sk(new_sk);
2470 	msg = buf_msg(buf);
2471 
2472 	/* we lock on new_sk; but lockdep sees the lock on sk */
2473 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2474 
2475 	/*
2476 	 * Reject any stray messages received by new socket
2477 	 * before the socket lock was taken (very, very unlikely)
2478 	 */
2479 	tsk_rej_rx_queue(new_sk);
2480 
2481 	/* Connect new socket to it's peer */
2482 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2483 
2484 	tsk_set_importance(new_tsock, msg_importance(msg));
2485 	if (msg_named(msg)) {
2486 		new_tsock->conn_type = msg_nametype(msg);
2487 		new_tsock->conn_instance = msg_nameinst(msg);
2488 	}
2489 
2490 	/*
2491 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2492 	 * Respond to 'SYN+' by queuing it on new socket.
2493 	 */
2494 	if (!msg_data_sz(msg)) {
2495 		struct msghdr m = {NULL,};
2496 
2497 		tsk_advance_rx_queue(sk);
2498 		__tipc_sendstream(new_sock, &m, 0);
2499 	} else {
2500 		__skb_dequeue(&sk->sk_receive_queue);
2501 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2502 		skb_set_owner_r(buf, new_sk);
2503 	}
2504 	release_sock(new_sk);
2505 exit:
2506 	release_sock(sk);
2507 	return res;
2508 }
2509 
2510 /**
2511  * tipc_shutdown - shutdown socket connection
2512  * @sock: socket structure
2513  * @how: direction to close (must be SHUT_RDWR)
2514  *
2515  * Terminates connection (if necessary), then purges socket's receive queue.
2516  *
2517  * Returns 0 on success, errno otherwise
2518  */
2519 static int tipc_shutdown(struct socket *sock, int how)
2520 {
2521 	struct sock *sk = sock->sk;
2522 	int res;
2523 
2524 	if (how != SHUT_RDWR)
2525 		return -EINVAL;
2526 
2527 	lock_sock(sk);
2528 
2529 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2530 	sk->sk_shutdown = SEND_SHUTDOWN;
2531 
2532 	if (sk->sk_state == TIPC_DISCONNECTING) {
2533 		/* Discard any unreceived messages */
2534 		__skb_queue_purge(&sk->sk_receive_queue);
2535 
2536 		/* Wake up anyone sleeping in poll */
2537 		sk->sk_state_change(sk);
2538 		res = 0;
2539 	} else {
2540 		res = -ENOTCONN;
2541 	}
2542 
2543 	release_sock(sk);
2544 	return res;
2545 }
2546 
2547 static void tipc_sk_timeout(struct timer_list *t)
2548 {
2549 	struct sock *sk = from_timer(sk, t, sk_timer);
2550 	struct tipc_sock *tsk = tipc_sk(sk);
2551 	u32 peer_port = tsk_peer_port(tsk);
2552 	u32 peer_node = tsk_peer_node(tsk);
2553 	u32 own_node = tsk_own_node(tsk);
2554 	u32 own_port = tsk->portid;
2555 	struct net *net = sock_net(sk);
2556 	struct sk_buff *skb = NULL;
2557 
2558 	bh_lock_sock(sk);
2559 	if (!tipc_sk_connected(sk))
2560 		goto exit;
2561 
2562 	/* Try again later if socket is busy */
2563 	if (sock_owned_by_user(sk)) {
2564 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2565 		goto exit;
2566 	}
2567 
2568 	if (tsk->probe_unacked) {
2569 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2570 		tipc_node_remove_conn(net, peer_node, peer_port);
2571 		sk->sk_state_change(sk);
2572 		goto exit;
2573 	}
2574 	/* Send new probe */
2575 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2576 			      peer_node, own_node, peer_port, own_port,
2577 			      TIPC_OK);
2578 	tsk->probe_unacked = true;
2579 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2580 exit:
2581 	bh_unlock_sock(sk);
2582 	if (skb)
2583 		tipc_node_xmit_skb(net, skb, peer_node, own_port);
2584 	sock_put(sk);
2585 }
2586 
2587 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2588 			   struct tipc_name_seq const *seq)
2589 {
2590 	struct sock *sk = &tsk->sk;
2591 	struct net *net = sock_net(sk);
2592 	struct publication *publ;
2593 	u32 key;
2594 
2595 	if (scope != TIPC_NODE_SCOPE)
2596 		scope = TIPC_CLUSTER_SCOPE;
2597 
2598 	if (tipc_sk_connected(sk))
2599 		return -EINVAL;
2600 	key = tsk->portid + tsk->pub_count + 1;
2601 	if (key == tsk->portid)
2602 		return -EADDRINUSE;
2603 
2604 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2605 				    scope, tsk->portid, key);
2606 	if (unlikely(!publ))
2607 		return -EINVAL;
2608 
2609 	list_add(&publ->binding_sock, &tsk->publications);
2610 	tsk->pub_count++;
2611 	tsk->published = 1;
2612 	return 0;
2613 }
2614 
2615 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2616 			    struct tipc_name_seq const *seq)
2617 {
2618 	struct net *net = sock_net(&tsk->sk);
2619 	struct publication *publ;
2620 	struct publication *safe;
2621 	int rc = -EINVAL;
2622 
2623 	if (scope != TIPC_NODE_SCOPE)
2624 		scope = TIPC_CLUSTER_SCOPE;
2625 
2626 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2627 		if (seq) {
2628 			if (publ->scope != scope)
2629 				continue;
2630 			if (publ->type != seq->type)
2631 				continue;
2632 			if (publ->lower != seq->lower)
2633 				continue;
2634 			if (publ->upper != seq->upper)
2635 				break;
2636 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2637 					      publ->upper, publ->key);
2638 			rc = 0;
2639 			break;
2640 		}
2641 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2642 				      publ->upper, publ->key);
2643 		rc = 0;
2644 	}
2645 	if (list_empty(&tsk->publications))
2646 		tsk->published = 0;
2647 	return rc;
2648 }
2649 
2650 /* tipc_sk_reinit: set non-zero address in all existing sockets
2651  *                 when we go from standalone to network mode.
2652  */
2653 void tipc_sk_reinit(struct net *net)
2654 {
2655 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2656 	struct rhashtable_iter iter;
2657 	struct tipc_sock *tsk;
2658 	struct tipc_msg *msg;
2659 
2660 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2661 
2662 	do {
2663 		rhashtable_walk_start(&iter);
2664 
2665 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2666 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2667 			msg = &tsk->phdr;
2668 			msg_set_prevnode(msg, tipc_own_addr(net));
2669 			msg_set_orignode(msg, tipc_own_addr(net));
2670 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2671 		}
2672 
2673 		rhashtable_walk_stop(&iter);
2674 	} while (tsk == ERR_PTR(-EAGAIN));
2675 }
2676 
2677 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2678 {
2679 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2680 	struct tipc_sock *tsk;
2681 
2682 	rcu_read_lock();
2683 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2684 	if (tsk)
2685 		sock_hold(&tsk->sk);
2686 	rcu_read_unlock();
2687 
2688 	return tsk;
2689 }
2690 
2691 static int tipc_sk_insert(struct tipc_sock *tsk)
2692 {
2693 	struct sock *sk = &tsk->sk;
2694 	struct net *net = sock_net(sk);
2695 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2696 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2697 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2698 
2699 	while (remaining--) {
2700 		portid++;
2701 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2702 			portid = TIPC_MIN_PORT;
2703 		tsk->portid = portid;
2704 		sock_hold(&tsk->sk);
2705 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2706 						   tsk_rht_params))
2707 			return 0;
2708 		sock_put(&tsk->sk);
2709 	}
2710 
2711 	return -1;
2712 }
2713 
2714 static void tipc_sk_remove(struct tipc_sock *tsk)
2715 {
2716 	struct sock *sk = &tsk->sk;
2717 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2718 
2719 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2720 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2721 		__sock_put(sk);
2722 	}
2723 }
2724 
2725 static const struct rhashtable_params tsk_rht_params = {
2726 	.nelem_hint = 192,
2727 	.head_offset = offsetof(struct tipc_sock, node),
2728 	.key_offset = offsetof(struct tipc_sock, portid),
2729 	.key_len = sizeof(u32), /* portid */
2730 	.max_size = 1048576,
2731 	.min_size = 256,
2732 	.automatic_shrinking = true,
2733 };
2734 
2735 int tipc_sk_rht_init(struct net *net)
2736 {
2737 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2738 
2739 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2740 }
2741 
2742 void tipc_sk_rht_destroy(struct net *net)
2743 {
2744 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2745 
2746 	/* Wait for socket readers to complete */
2747 	synchronize_net();
2748 
2749 	rhashtable_destroy(&tn->sk_rht);
2750 }
2751 
2752 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2753 {
2754 	struct net *net = sock_net(&tsk->sk);
2755 	struct tipc_group *grp = tsk->group;
2756 	struct tipc_msg *hdr = &tsk->phdr;
2757 	struct tipc_name_seq seq;
2758 	int rc;
2759 
2760 	if (mreq->type < TIPC_RESERVED_TYPES)
2761 		return -EACCES;
2762 	if (mreq->scope > TIPC_NODE_SCOPE)
2763 		return -EINVAL;
2764 	if (grp)
2765 		return -EACCES;
2766 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2767 	if (!grp)
2768 		return -ENOMEM;
2769 	tsk->group = grp;
2770 	msg_set_lookup_scope(hdr, mreq->scope);
2771 	msg_set_nametype(hdr, mreq->type);
2772 	msg_set_dest_droppable(hdr, true);
2773 	seq.type = mreq->type;
2774 	seq.lower = mreq->instance;
2775 	seq.upper = seq.lower;
2776 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2777 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2778 	if (rc) {
2779 		tipc_group_delete(net, grp);
2780 		tsk->group = NULL;
2781 		return rc;
2782 	}
2783 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2784 	tsk->mc_method.rcast = true;
2785 	tsk->mc_method.mandatory = true;
2786 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2787 	return rc;
2788 }
2789 
2790 static int tipc_sk_leave(struct tipc_sock *tsk)
2791 {
2792 	struct net *net = sock_net(&tsk->sk);
2793 	struct tipc_group *grp = tsk->group;
2794 	struct tipc_name_seq seq;
2795 	int scope;
2796 
2797 	if (!grp)
2798 		return -EINVAL;
2799 	tipc_group_self(grp, &seq, &scope);
2800 	tipc_group_delete(net, grp);
2801 	tsk->group = NULL;
2802 	tipc_sk_withdraw(tsk, scope, &seq);
2803 	return 0;
2804 }
2805 
2806 /**
2807  * tipc_setsockopt - set socket option
2808  * @sock: socket structure
2809  * @lvl: option level
2810  * @opt: option identifier
2811  * @ov: pointer to new option value
2812  * @ol: length of option value
2813  *
2814  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2815  * (to ease compatibility).
2816  *
2817  * Returns 0 on success, errno otherwise
2818  */
2819 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2820 			   char __user *ov, unsigned int ol)
2821 {
2822 	struct sock *sk = sock->sk;
2823 	struct tipc_sock *tsk = tipc_sk(sk);
2824 	struct tipc_group_req mreq;
2825 	u32 value = 0;
2826 	int res = 0;
2827 
2828 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2829 		return 0;
2830 	if (lvl != SOL_TIPC)
2831 		return -ENOPROTOOPT;
2832 
2833 	switch (opt) {
2834 	case TIPC_IMPORTANCE:
2835 	case TIPC_SRC_DROPPABLE:
2836 	case TIPC_DEST_DROPPABLE:
2837 	case TIPC_CONN_TIMEOUT:
2838 		if (ol < sizeof(value))
2839 			return -EINVAL;
2840 		if (get_user(value, (u32 __user *)ov))
2841 			return -EFAULT;
2842 		break;
2843 	case TIPC_GROUP_JOIN:
2844 		if (ol < sizeof(mreq))
2845 			return -EINVAL;
2846 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2847 			return -EFAULT;
2848 		break;
2849 	default:
2850 		if (ov || ol)
2851 			return -EINVAL;
2852 	}
2853 
2854 	lock_sock(sk);
2855 
2856 	switch (opt) {
2857 	case TIPC_IMPORTANCE:
2858 		res = tsk_set_importance(tsk, value);
2859 		break;
2860 	case TIPC_SRC_DROPPABLE:
2861 		if (sock->type != SOCK_STREAM)
2862 			tsk_set_unreliable(tsk, value);
2863 		else
2864 			res = -ENOPROTOOPT;
2865 		break;
2866 	case TIPC_DEST_DROPPABLE:
2867 		tsk_set_unreturnable(tsk, value);
2868 		break;
2869 	case TIPC_CONN_TIMEOUT:
2870 		tipc_sk(sk)->conn_timeout = value;
2871 		break;
2872 	case TIPC_MCAST_BROADCAST:
2873 		tsk->mc_method.rcast = false;
2874 		tsk->mc_method.mandatory = true;
2875 		break;
2876 	case TIPC_MCAST_REPLICAST:
2877 		tsk->mc_method.rcast = true;
2878 		tsk->mc_method.mandatory = true;
2879 		break;
2880 	case TIPC_GROUP_JOIN:
2881 		res = tipc_sk_join(tsk, &mreq);
2882 		break;
2883 	case TIPC_GROUP_LEAVE:
2884 		res = tipc_sk_leave(tsk);
2885 		break;
2886 	default:
2887 		res = -EINVAL;
2888 	}
2889 
2890 	release_sock(sk);
2891 
2892 	return res;
2893 }
2894 
2895 /**
2896  * tipc_getsockopt - get socket option
2897  * @sock: socket structure
2898  * @lvl: option level
2899  * @opt: option identifier
2900  * @ov: receptacle for option value
2901  * @ol: receptacle for length of option value
2902  *
2903  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2904  * (to ease compatibility).
2905  *
2906  * Returns 0 on success, errno otherwise
2907  */
2908 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2909 			   char __user *ov, int __user *ol)
2910 {
2911 	struct sock *sk = sock->sk;
2912 	struct tipc_sock *tsk = tipc_sk(sk);
2913 	struct tipc_name_seq seq;
2914 	int len, scope;
2915 	u32 value;
2916 	int res;
2917 
2918 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2919 		return put_user(0, ol);
2920 	if (lvl != SOL_TIPC)
2921 		return -ENOPROTOOPT;
2922 	res = get_user(len, ol);
2923 	if (res)
2924 		return res;
2925 
2926 	lock_sock(sk);
2927 
2928 	switch (opt) {
2929 	case TIPC_IMPORTANCE:
2930 		value = tsk_importance(tsk);
2931 		break;
2932 	case TIPC_SRC_DROPPABLE:
2933 		value = tsk_unreliable(tsk);
2934 		break;
2935 	case TIPC_DEST_DROPPABLE:
2936 		value = tsk_unreturnable(tsk);
2937 		break;
2938 	case TIPC_CONN_TIMEOUT:
2939 		value = tsk->conn_timeout;
2940 		/* no need to set "res", since already 0 at this point */
2941 		break;
2942 	case TIPC_NODE_RECVQ_DEPTH:
2943 		value = 0; /* was tipc_queue_size, now obsolete */
2944 		break;
2945 	case TIPC_SOCK_RECVQ_DEPTH:
2946 		value = skb_queue_len(&sk->sk_receive_queue);
2947 		break;
2948 	case TIPC_GROUP_JOIN:
2949 		seq.type = 0;
2950 		if (tsk->group)
2951 			tipc_group_self(tsk->group, &seq, &scope);
2952 		value = seq.type;
2953 		break;
2954 	default:
2955 		res = -EINVAL;
2956 	}
2957 
2958 	release_sock(sk);
2959 
2960 	if (res)
2961 		return res;	/* "get" failed */
2962 
2963 	if (len < sizeof(value))
2964 		return -EINVAL;
2965 
2966 	if (copy_to_user(ov, &value, sizeof(value)))
2967 		return -EFAULT;
2968 
2969 	return put_user(sizeof(value), ol);
2970 }
2971 
2972 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2973 {
2974 	struct sock *sk = sock->sk;
2975 	struct tipc_sioc_ln_req lnr;
2976 	void __user *argp = (void __user *)arg;
2977 
2978 	switch (cmd) {
2979 	case SIOCGETLINKNAME:
2980 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2981 			return -EFAULT;
2982 		if (!tipc_node_get_linkname(sock_net(sk),
2983 					    lnr.bearer_id & 0xffff, lnr.peer,
2984 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2985 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2986 				return -EFAULT;
2987 			return 0;
2988 		}
2989 		return -EADDRNOTAVAIL;
2990 	default:
2991 		return -ENOIOCTLCMD;
2992 	}
2993 }
2994 
2995 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
2996 {
2997 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
2998 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
2999 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3000 
3001 	tsk1->peer.family = AF_TIPC;
3002 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3003 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3004 	tsk1->peer.addr.id.ref = tsk2->portid;
3005 	tsk1->peer.addr.id.node = onode;
3006 	tsk2->peer.family = AF_TIPC;
3007 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3008 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3009 	tsk2->peer.addr.id.ref = tsk1->portid;
3010 	tsk2->peer.addr.id.node = onode;
3011 
3012 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3013 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3014 	return 0;
3015 }
3016 
3017 /* Protocol switches for the various types of TIPC sockets */
3018 
3019 static const struct proto_ops msg_ops = {
3020 	.owner		= THIS_MODULE,
3021 	.family		= AF_TIPC,
3022 	.release	= tipc_release,
3023 	.bind		= tipc_bind,
3024 	.connect	= tipc_connect,
3025 	.socketpair	= tipc_socketpair,
3026 	.accept		= sock_no_accept,
3027 	.getname	= tipc_getname,
3028 	.poll		= tipc_poll,
3029 	.ioctl		= tipc_ioctl,
3030 	.listen		= sock_no_listen,
3031 	.shutdown	= tipc_shutdown,
3032 	.setsockopt	= tipc_setsockopt,
3033 	.getsockopt	= tipc_getsockopt,
3034 	.sendmsg	= tipc_sendmsg,
3035 	.recvmsg	= tipc_recvmsg,
3036 	.mmap		= sock_no_mmap,
3037 	.sendpage	= sock_no_sendpage
3038 };
3039 
3040 static const struct proto_ops packet_ops = {
3041 	.owner		= THIS_MODULE,
3042 	.family		= AF_TIPC,
3043 	.release	= tipc_release,
3044 	.bind		= tipc_bind,
3045 	.connect	= tipc_connect,
3046 	.socketpair	= tipc_socketpair,
3047 	.accept		= tipc_accept,
3048 	.getname	= tipc_getname,
3049 	.poll		= tipc_poll,
3050 	.ioctl		= tipc_ioctl,
3051 	.listen		= tipc_listen,
3052 	.shutdown	= tipc_shutdown,
3053 	.setsockopt	= tipc_setsockopt,
3054 	.getsockopt	= tipc_getsockopt,
3055 	.sendmsg	= tipc_send_packet,
3056 	.recvmsg	= tipc_recvmsg,
3057 	.mmap		= sock_no_mmap,
3058 	.sendpage	= sock_no_sendpage
3059 };
3060 
3061 static const struct proto_ops stream_ops = {
3062 	.owner		= THIS_MODULE,
3063 	.family		= AF_TIPC,
3064 	.release	= tipc_release,
3065 	.bind		= tipc_bind,
3066 	.connect	= tipc_connect,
3067 	.socketpair	= tipc_socketpair,
3068 	.accept		= tipc_accept,
3069 	.getname	= tipc_getname,
3070 	.poll		= tipc_poll,
3071 	.ioctl		= tipc_ioctl,
3072 	.listen		= tipc_listen,
3073 	.shutdown	= tipc_shutdown,
3074 	.setsockopt	= tipc_setsockopt,
3075 	.getsockopt	= tipc_getsockopt,
3076 	.sendmsg	= tipc_sendstream,
3077 	.recvmsg	= tipc_recvstream,
3078 	.mmap		= sock_no_mmap,
3079 	.sendpage	= sock_no_sendpage
3080 };
3081 
3082 static const struct net_proto_family tipc_family_ops = {
3083 	.owner		= THIS_MODULE,
3084 	.family		= AF_TIPC,
3085 	.create		= tipc_sk_create
3086 };
3087 
3088 static struct proto tipc_proto = {
3089 	.name		= "TIPC",
3090 	.owner		= THIS_MODULE,
3091 	.obj_size	= sizeof(struct tipc_sock),
3092 	.sysctl_rmem	= sysctl_tipc_rmem
3093 };
3094 
3095 /**
3096  * tipc_socket_init - initialize TIPC socket interface
3097  *
3098  * Returns 0 on success, errno otherwise
3099  */
3100 int tipc_socket_init(void)
3101 {
3102 	int res;
3103 
3104 	res = proto_register(&tipc_proto, 1);
3105 	if (res) {
3106 		pr_err("Failed to register TIPC protocol type\n");
3107 		goto out;
3108 	}
3109 
3110 	res = sock_register(&tipc_family_ops);
3111 	if (res) {
3112 		pr_err("Failed to register TIPC socket type\n");
3113 		proto_unregister(&tipc_proto);
3114 		goto out;
3115 	}
3116  out:
3117 	return res;
3118 }
3119 
3120 /**
3121  * tipc_socket_stop - stop TIPC socket interface
3122  */
3123 void tipc_socket_stop(void)
3124 {
3125 	sock_unregister(tipc_family_ops.family);
3126 	proto_unregister(&tipc_proto);
3127 }
3128 
3129 /* Caller should hold socket lock for the passed tipc socket. */
3130 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3131 {
3132 	u32 peer_node;
3133 	u32 peer_port;
3134 	struct nlattr *nest;
3135 
3136 	peer_node = tsk_peer_node(tsk);
3137 	peer_port = tsk_peer_port(tsk);
3138 
3139 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3140 
3141 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3142 		goto msg_full;
3143 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3144 		goto msg_full;
3145 
3146 	if (tsk->conn_type != 0) {
3147 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3148 			goto msg_full;
3149 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3150 			goto msg_full;
3151 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3152 			goto msg_full;
3153 	}
3154 	nla_nest_end(skb, nest);
3155 
3156 	return 0;
3157 
3158 msg_full:
3159 	nla_nest_cancel(skb, nest);
3160 
3161 	return -EMSGSIZE;
3162 }
3163 
3164 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3165 			  *tsk)
3166 {
3167 	struct net *net = sock_net(skb->sk);
3168 	struct sock *sk = &tsk->sk;
3169 
3170 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3171 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3172 		return -EMSGSIZE;
3173 
3174 	if (tipc_sk_connected(sk)) {
3175 		if (__tipc_nl_add_sk_con(skb, tsk))
3176 			return -EMSGSIZE;
3177 	} else if (!list_empty(&tsk->publications)) {
3178 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3179 			return -EMSGSIZE;
3180 	}
3181 	return 0;
3182 }
3183 
3184 /* Caller should hold socket lock for the passed tipc socket. */
3185 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3186 			    struct tipc_sock *tsk)
3187 {
3188 	struct nlattr *attrs;
3189 	void *hdr;
3190 
3191 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3192 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3193 	if (!hdr)
3194 		goto msg_cancel;
3195 
3196 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3197 	if (!attrs)
3198 		goto genlmsg_cancel;
3199 
3200 	if (__tipc_nl_add_sk_info(skb, tsk))
3201 		goto attr_msg_cancel;
3202 
3203 	nla_nest_end(skb, attrs);
3204 	genlmsg_end(skb, hdr);
3205 
3206 	return 0;
3207 
3208 attr_msg_cancel:
3209 	nla_nest_cancel(skb, attrs);
3210 genlmsg_cancel:
3211 	genlmsg_cancel(skb, hdr);
3212 msg_cancel:
3213 	return -EMSGSIZE;
3214 }
3215 
3216 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3217 		    int (*skb_handler)(struct sk_buff *skb,
3218 				       struct netlink_callback *cb,
3219 				       struct tipc_sock *tsk))
3220 {
3221 	struct net *net = sock_net(skb->sk);
3222 	struct tipc_net *tn = tipc_net(net);
3223 	const struct bucket_table *tbl;
3224 	u32 prev_portid = cb->args[1];
3225 	u32 tbl_id = cb->args[0];
3226 	struct rhash_head *pos;
3227 	struct tipc_sock *tsk;
3228 	int err;
3229 
3230 	rcu_read_lock();
3231 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3232 	for (; tbl_id < tbl->size; tbl_id++) {
3233 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3234 			spin_lock_bh(&tsk->sk.sk_lock.slock);
3235 			if (prev_portid && prev_portid != tsk->portid) {
3236 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3237 				continue;
3238 			}
3239 
3240 			err = skb_handler(skb, cb, tsk);
3241 			if (err) {
3242 				prev_portid = tsk->portid;
3243 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3244 				goto out;
3245 			}
3246 
3247 			prev_portid = 0;
3248 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
3249 		}
3250 	}
3251 out:
3252 	rcu_read_unlock();
3253 	cb->args[0] = tbl_id;
3254 	cb->args[1] = prev_portid;
3255 
3256 	return skb->len;
3257 }
3258 EXPORT_SYMBOL(tipc_nl_sk_walk);
3259 
3260 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3261 			   struct tipc_sock *tsk, u32 sk_filter_state,
3262 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3263 {
3264 	struct sock *sk = &tsk->sk;
3265 	struct nlattr *attrs;
3266 	struct nlattr *stat;
3267 
3268 	/*filter response w.r.t sk_state*/
3269 	if (!(sk_filter_state & (1 << sk->sk_state)))
3270 		return 0;
3271 
3272 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3273 	if (!attrs)
3274 		goto msg_cancel;
3275 
3276 	if (__tipc_nl_add_sk_info(skb, tsk))
3277 		goto attr_msg_cancel;
3278 
3279 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3280 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3281 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3282 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3283 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3284 					 sock_i_uid(sk))) ||
3285 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3286 			      tipc_diag_gen_cookie(sk),
3287 			      TIPC_NLA_SOCK_PAD))
3288 		goto attr_msg_cancel;
3289 
3290 	stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3291 	if (!stat)
3292 		goto attr_msg_cancel;
3293 
3294 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3295 			skb_queue_len(&sk->sk_receive_queue)) ||
3296 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3297 			skb_queue_len(&sk->sk_write_queue)) ||
3298 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3299 			atomic_read(&sk->sk_drops)))
3300 		goto stat_msg_cancel;
3301 
3302 	if (tsk->cong_link_cnt &&
3303 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3304 		goto stat_msg_cancel;
3305 
3306 	if (tsk_conn_cong(tsk) &&
3307 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3308 		goto stat_msg_cancel;
3309 
3310 	nla_nest_end(skb, stat);
3311 	nla_nest_end(skb, attrs);
3312 
3313 	return 0;
3314 
3315 stat_msg_cancel:
3316 	nla_nest_cancel(skb, stat);
3317 attr_msg_cancel:
3318 	nla_nest_cancel(skb, attrs);
3319 msg_cancel:
3320 	return -EMSGSIZE;
3321 }
3322 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3323 
3324 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3325 {
3326 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3327 }
3328 
3329 /* Caller should hold socket lock for the passed tipc socket. */
3330 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3331 				 struct netlink_callback *cb,
3332 				 struct publication *publ)
3333 {
3334 	void *hdr;
3335 	struct nlattr *attrs;
3336 
3337 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3338 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3339 	if (!hdr)
3340 		goto msg_cancel;
3341 
3342 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3343 	if (!attrs)
3344 		goto genlmsg_cancel;
3345 
3346 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3347 		goto attr_msg_cancel;
3348 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3349 		goto attr_msg_cancel;
3350 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3351 		goto attr_msg_cancel;
3352 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3353 		goto attr_msg_cancel;
3354 
3355 	nla_nest_end(skb, attrs);
3356 	genlmsg_end(skb, hdr);
3357 
3358 	return 0;
3359 
3360 attr_msg_cancel:
3361 	nla_nest_cancel(skb, attrs);
3362 genlmsg_cancel:
3363 	genlmsg_cancel(skb, hdr);
3364 msg_cancel:
3365 	return -EMSGSIZE;
3366 }
3367 
3368 /* Caller should hold socket lock for the passed tipc socket. */
3369 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3370 				  struct netlink_callback *cb,
3371 				  struct tipc_sock *tsk, u32 *last_publ)
3372 {
3373 	int err;
3374 	struct publication *p;
3375 
3376 	if (*last_publ) {
3377 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3378 			if (p->key == *last_publ)
3379 				break;
3380 		}
3381 		if (p->key != *last_publ) {
3382 			/* We never set seq or call nl_dump_check_consistent()
3383 			 * this means that setting prev_seq here will cause the
3384 			 * consistence check to fail in the netlink callback
3385 			 * handler. Resulting in the last NLMSG_DONE message
3386 			 * having the NLM_F_DUMP_INTR flag set.
3387 			 */
3388 			cb->prev_seq = 1;
3389 			*last_publ = 0;
3390 			return -EPIPE;
3391 		}
3392 	} else {
3393 		p = list_first_entry(&tsk->publications, struct publication,
3394 				     binding_sock);
3395 	}
3396 
3397 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3398 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3399 		if (err) {
3400 			*last_publ = p->key;
3401 			return err;
3402 		}
3403 	}
3404 	*last_publ = 0;
3405 
3406 	return 0;
3407 }
3408 
3409 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3410 {
3411 	int err;
3412 	u32 tsk_portid = cb->args[0];
3413 	u32 last_publ = cb->args[1];
3414 	u32 done = cb->args[2];
3415 	struct net *net = sock_net(skb->sk);
3416 	struct tipc_sock *tsk;
3417 
3418 	if (!tsk_portid) {
3419 		struct nlattr **attrs;
3420 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3421 
3422 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3423 		if (err)
3424 			return err;
3425 
3426 		if (!attrs[TIPC_NLA_SOCK])
3427 			return -EINVAL;
3428 
3429 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3430 				       attrs[TIPC_NLA_SOCK],
3431 				       tipc_nl_sock_policy, NULL);
3432 		if (err)
3433 			return err;
3434 
3435 		if (!sock[TIPC_NLA_SOCK_REF])
3436 			return -EINVAL;
3437 
3438 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3439 	}
3440 
3441 	if (done)
3442 		return 0;
3443 
3444 	tsk = tipc_sk_lookup(net, tsk_portid);
3445 	if (!tsk)
3446 		return -EINVAL;
3447 
3448 	lock_sock(&tsk->sk);
3449 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3450 	if (!err)
3451 		done = 1;
3452 	release_sock(&tsk->sk);
3453 	sock_put(&tsk->sk);
3454 
3455 	cb->args[0] = tsk_portid;
3456 	cb->args[1] = last_publ;
3457 	cb->args[2] = done;
3458 
3459 	return skb->len;
3460 }
3461