xref: /openbmc/linux/net/tipc/socket.c (revision 01a6e126)
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 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
293 	u32 peer_port = tsk_peer_port(tsk);
294 	u32 orig_node;
295 	u32 peer_node;
296 
297 	if (unlikely(!tipc_sk_connected(sk)))
298 		return false;
299 
300 	if (unlikely(msg_origport(msg) != peer_port))
301 		return false;
302 
303 	orig_node = msg_orignode(msg);
304 	peer_node = tsk_peer_node(tsk);
305 
306 	if (likely(orig_node == peer_node))
307 		return true;
308 
309 	if (!orig_node && (peer_node == tn->own_addr))
310 		return true;
311 
312 	if (!peer_node && (orig_node == tn->own_addr))
313 		return true;
314 
315 	return false;
316 }
317 
318 /* tipc_set_sk_state - set the sk_state of the socket
319  * @sk: socket
320  *
321  * Caller must hold socket lock
322  *
323  * Returns 0 on success, errno otherwise
324  */
325 static int tipc_set_sk_state(struct sock *sk, int state)
326 {
327 	int oldsk_state = sk->sk_state;
328 	int res = -EINVAL;
329 
330 	switch (state) {
331 	case TIPC_OPEN:
332 		res = 0;
333 		break;
334 	case TIPC_LISTEN:
335 	case TIPC_CONNECTING:
336 		if (oldsk_state == TIPC_OPEN)
337 			res = 0;
338 		break;
339 	case TIPC_ESTABLISHED:
340 		if (oldsk_state == TIPC_CONNECTING ||
341 		    oldsk_state == TIPC_OPEN)
342 			res = 0;
343 		break;
344 	case TIPC_DISCONNECTING:
345 		if (oldsk_state == TIPC_CONNECTING ||
346 		    oldsk_state == TIPC_ESTABLISHED)
347 			res = 0;
348 		break;
349 	}
350 
351 	if (!res)
352 		sk->sk_state = state;
353 
354 	return res;
355 }
356 
357 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
358 {
359 	struct sock *sk = sock->sk;
360 	int err = sock_error(sk);
361 	int typ = sock->type;
362 
363 	if (err)
364 		return err;
365 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
366 		if (sk->sk_state == TIPC_DISCONNECTING)
367 			return -EPIPE;
368 		else if (!tipc_sk_connected(sk))
369 			return -ENOTCONN;
370 	}
371 	if (!*timeout)
372 		return -EAGAIN;
373 	if (signal_pending(current))
374 		return sock_intr_errno(*timeout);
375 
376 	return 0;
377 }
378 
379 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
380 ({                                                                             \
381 	struct sock *sk_;						       \
382 	int rc_;							       \
383 									       \
384 	while ((rc_ = !(condition_))) {					       \
385 		DEFINE_WAIT_FUNC(wait_, woken_wake_function);	               \
386 		sk_ = (sock_)->sk;					       \
387 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
388 		if (rc_)						       \
389 			break;						       \
390 		prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE);    \
391 		release_sock(sk_);					       \
392 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
393 		sched_annotate_sleep();				               \
394 		lock_sock(sk_);						       \
395 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
396 	}								       \
397 	rc_;								       \
398 })
399 
400 /**
401  * tipc_sk_create - create a TIPC socket
402  * @net: network namespace (must be default network)
403  * @sock: pre-allocated socket structure
404  * @protocol: protocol indicator (must be 0)
405  * @kern: caused by kernel or by userspace?
406  *
407  * This routine creates additional data structures used by the TIPC socket,
408  * initializes them, and links them together.
409  *
410  * Returns 0 on success, errno otherwise
411  */
412 static int tipc_sk_create(struct net *net, struct socket *sock,
413 			  int protocol, int kern)
414 {
415 	struct tipc_net *tn;
416 	const struct proto_ops *ops;
417 	struct sock *sk;
418 	struct tipc_sock *tsk;
419 	struct tipc_msg *msg;
420 
421 	/* Validate arguments */
422 	if (unlikely(protocol != 0))
423 		return -EPROTONOSUPPORT;
424 
425 	switch (sock->type) {
426 	case SOCK_STREAM:
427 		ops = &stream_ops;
428 		break;
429 	case SOCK_SEQPACKET:
430 		ops = &packet_ops;
431 		break;
432 	case SOCK_DGRAM:
433 	case SOCK_RDM:
434 		ops = &msg_ops;
435 		break;
436 	default:
437 		return -EPROTOTYPE;
438 	}
439 
440 	/* Allocate socket's protocol area */
441 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
442 	if (sk == NULL)
443 		return -ENOMEM;
444 
445 	tsk = tipc_sk(sk);
446 	tsk->max_pkt = MAX_PKT_DEFAULT;
447 	INIT_LIST_HEAD(&tsk->publications);
448 	INIT_LIST_HEAD(&tsk->cong_links);
449 	msg = &tsk->phdr;
450 	tn = net_generic(sock_net(sk), tipc_net_id);
451 
452 	/* Finish initializing socket data structures */
453 	sock->ops = ops;
454 	sock_init_data(sock, sk);
455 	tipc_set_sk_state(sk, TIPC_OPEN);
456 	if (tipc_sk_insert(tsk)) {
457 		pr_warn("Socket create failed; port number exhausted\n");
458 		return -EINVAL;
459 	}
460 
461 	/* Ensure tsk is visible before we read own_addr. */
462 	smp_mb();
463 
464 	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
465 		      NAMED_H_SIZE, 0);
466 
467 	msg_set_origport(msg, tsk->portid);
468 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
469 	sk->sk_shutdown = 0;
470 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
471 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
472 	sk->sk_data_ready = tipc_data_ready;
473 	sk->sk_write_space = tipc_write_space;
474 	sk->sk_destruct = tipc_sock_destruct;
475 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
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 *uaddr_len, 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 	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
674 
675 	memset(addr, 0, sizeof(*addr));
676 	if (peer) {
677 		if ((!tipc_sk_connected(sk)) &&
678 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
679 			return -ENOTCONN;
680 		addr->addr.id.ref = tsk_peer_port(tsk);
681 		addr->addr.id.node = tsk_peer_node(tsk);
682 	} else {
683 		addr->addr.id.ref = tsk->portid;
684 		addr->addr.id.node = tn->own_addr;
685 	}
686 
687 	*uaddr_len = sizeof(*addr);
688 	addr->addrtype = TIPC_ADDR_ID;
689 	addr->family = AF_TIPC;
690 	addr->scope = 0;
691 	addr->addr.name.domain = 0;
692 
693 	return 0;
694 }
695 
696 /**
697  * tipc_poll - read and possibly block on pollmask
698  * @file: file structure associated with the socket
699  * @sock: socket for which to calculate the poll bits
700  * @wait: ???
701  *
702  * Returns pollmask value
703  *
704  * COMMENTARY:
705  * It appears that the usual socket locking mechanisms are not useful here
706  * since the pollmask info is potentially out-of-date the moment this routine
707  * exits.  TCP and other protocols seem to rely on higher level poll routines
708  * to handle any preventable race conditions, so TIPC will do the same ...
709  *
710  * IMPORTANT: The fact that a read or write operation is indicated does NOT
711  * imply that the operation will succeed, merely that it should be performed
712  * and will not block.
713  */
714 static __poll_t tipc_poll(struct file *file, struct socket *sock,
715 			      poll_table *wait)
716 {
717 	struct sock *sk = sock->sk;
718 	struct tipc_sock *tsk = tipc_sk(sk);
719 	__poll_t revents = 0;
720 
721 	sock_poll_wait(file, sk_sleep(sk), wait);
722 
723 	if (sk->sk_shutdown & RCV_SHUTDOWN)
724 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
725 	if (sk->sk_shutdown == SHUTDOWN_MASK)
726 		revents |= EPOLLHUP;
727 
728 	switch (sk->sk_state) {
729 	case TIPC_ESTABLISHED:
730 	case TIPC_CONNECTING:
731 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
732 			revents |= EPOLLOUT;
733 		/* fall thru' */
734 	case TIPC_LISTEN:
735 		if (!skb_queue_empty(&sk->sk_receive_queue))
736 			revents |= EPOLLIN | EPOLLRDNORM;
737 		break;
738 	case TIPC_OPEN:
739 		if (tsk->group_is_open && !tsk->cong_link_cnt)
740 			revents |= EPOLLOUT;
741 		if (!tipc_sk_type_connectionless(sk))
742 			break;
743 		if (skb_queue_empty(&sk->sk_receive_queue))
744 			break;
745 		revents |= EPOLLIN | EPOLLRDNORM;
746 		break;
747 	case TIPC_DISCONNECTING:
748 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
749 		break;
750 	}
751 	return revents;
752 }
753 
754 /**
755  * tipc_sendmcast - send multicast message
756  * @sock: socket structure
757  * @seq: destination address
758  * @msg: message to send
759  * @dlen: length of data to send
760  * @timeout: timeout to wait for wakeup
761  *
762  * Called from function tipc_sendmsg(), which has done all sanity checks
763  * Returns the number of bytes sent on success, or errno
764  */
765 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
766 			  struct msghdr *msg, size_t dlen, long timeout)
767 {
768 	struct sock *sk = sock->sk;
769 	struct tipc_sock *tsk = tipc_sk(sk);
770 	struct tipc_msg *hdr = &tsk->phdr;
771 	struct net *net = sock_net(sk);
772 	int mtu = tipc_bcast_get_mtu(net);
773 	struct tipc_mc_method *method = &tsk->mc_method;
774 	struct sk_buff_head pkts;
775 	struct tipc_nlist dsts;
776 	int rc;
777 
778 	if (tsk->group)
779 		return -EACCES;
780 
781 	/* Block or return if any destination link is congested */
782 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
783 	if (unlikely(rc))
784 		return rc;
785 
786 	/* Lookup destination nodes */
787 	tipc_nlist_init(&dsts, tipc_own_addr(net));
788 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
789 				      seq->upper, &dsts);
790 	if (!dsts.local && !dsts.remote)
791 		return -EHOSTUNREACH;
792 
793 	/* Build message header */
794 	msg_set_type(hdr, TIPC_MCAST_MSG);
795 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
796 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
797 	msg_set_destport(hdr, 0);
798 	msg_set_destnode(hdr, 0);
799 	msg_set_nametype(hdr, seq->type);
800 	msg_set_namelower(hdr, seq->lower);
801 	msg_set_nameupper(hdr, seq->upper);
802 
803 	/* Build message as chain of buffers */
804 	skb_queue_head_init(&pkts);
805 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
806 
807 	/* Send message if build was successful */
808 	if (unlikely(rc == dlen))
809 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
810 				     &tsk->cong_link_cnt);
811 
812 	tipc_nlist_purge(&dsts);
813 
814 	return rc ? rc : dlen;
815 }
816 
817 /**
818  * tipc_send_group_msg - send a message to a member in the group
819  * @net: network namespace
820  * @m: message to send
821  * @mb: group member
822  * @dnode: destination node
823  * @dport: destination port
824  * @dlen: total length of message data
825  */
826 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
827 			       struct msghdr *m, struct tipc_member *mb,
828 			       u32 dnode, u32 dport, int dlen)
829 {
830 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
831 	struct tipc_mc_method *method = &tsk->mc_method;
832 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
833 	struct tipc_msg *hdr = &tsk->phdr;
834 	struct sk_buff_head pkts;
835 	int mtu, rc;
836 
837 	/* Complete message header */
838 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
839 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
840 	msg_set_destport(hdr, dport);
841 	msg_set_destnode(hdr, dnode);
842 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
843 
844 	/* Build message as chain of buffers */
845 	skb_queue_head_init(&pkts);
846 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
847 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
848 	if (unlikely(rc != dlen))
849 		return rc;
850 
851 	/* Send message */
852 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
853 	if (unlikely(rc == -ELINKCONG)) {
854 		tipc_dest_push(&tsk->cong_links, dnode, 0);
855 		tsk->cong_link_cnt++;
856 	}
857 
858 	/* Update send window */
859 	tipc_group_update_member(mb, blks);
860 
861 	/* A broadcast sent within next EXPIRE period must follow same path */
862 	method->rcast = true;
863 	method->mandatory = true;
864 	return dlen;
865 }
866 
867 /**
868  * tipc_send_group_unicast - send message to a member in the group
869  * @sock: socket structure
870  * @m: message to send
871  * @dlen: total length of message data
872  * @timeout: timeout to wait for wakeup
873  *
874  * Called from function tipc_sendmsg(), which has done all sanity checks
875  * Returns the number of bytes sent on success, or errno
876  */
877 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
878 				   int dlen, long timeout)
879 {
880 	struct sock *sk = sock->sk;
881 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
882 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
883 	struct tipc_sock *tsk = tipc_sk(sk);
884 	struct tipc_group *grp = tsk->group;
885 	struct net *net = sock_net(sk);
886 	struct tipc_member *mb = NULL;
887 	u32 node, port;
888 	int rc;
889 
890 	node = dest->addr.id.node;
891 	port = dest->addr.id.ref;
892 	if (!port && !node)
893 		return -EHOSTUNREACH;
894 
895 	/* Block or return if destination link or member is congested */
896 	rc = tipc_wait_for_cond(sock, &timeout,
897 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
898 				!tipc_group_cong(grp, node, port, blks, &mb));
899 	if (unlikely(rc))
900 		return rc;
901 
902 	if (unlikely(!mb))
903 		return -EHOSTUNREACH;
904 
905 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
906 
907 	return rc ? rc : dlen;
908 }
909 
910 /**
911  * tipc_send_group_anycast - send message to any member with given identity
912  * @sock: socket structure
913  * @m: message to send
914  * @dlen: total length of message data
915  * @timeout: timeout to wait for wakeup
916  *
917  * Called from function tipc_sendmsg(), which has done all sanity checks
918  * Returns the number of bytes sent on success, or errno
919  */
920 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
921 				   int dlen, long timeout)
922 {
923 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
924 	struct sock *sk = sock->sk;
925 	struct tipc_sock *tsk = tipc_sk(sk);
926 	struct list_head *cong_links = &tsk->cong_links;
927 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
928 	struct tipc_group *grp = tsk->group;
929 	struct tipc_msg *hdr = &tsk->phdr;
930 	struct tipc_member *first = NULL;
931 	struct tipc_member *mbr = NULL;
932 	struct net *net = sock_net(sk);
933 	u32 node, port, exclude;
934 	struct list_head dsts;
935 	u32 type, inst, scope;
936 	int lookups = 0;
937 	int dstcnt, rc;
938 	bool cong;
939 
940 	INIT_LIST_HEAD(&dsts);
941 
942 	type = msg_nametype(hdr);
943 	inst = dest->addr.name.name.instance;
944 	scope = msg_lookup_scope(hdr);
945 	exclude = tipc_group_exclude(grp);
946 
947 	while (++lookups < 4) {
948 		first = NULL;
949 
950 		/* Look for a non-congested destination member, if any */
951 		while (1) {
952 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
953 						 &dstcnt, exclude, false))
954 				return -EHOSTUNREACH;
955 			tipc_dest_pop(&dsts, &node, &port);
956 			cong = tipc_group_cong(grp, node, port, blks, &mbr);
957 			if (!cong)
958 				break;
959 			if (mbr == first)
960 				break;
961 			if (!first)
962 				first = mbr;
963 		}
964 
965 		/* Start over if destination was not in member list */
966 		if (unlikely(!mbr))
967 			continue;
968 
969 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
970 			break;
971 
972 		/* Block or return if destination link or member is congested */
973 		rc = tipc_wait_for_cond(sock, &timeout,
974 					!tipc_dest_find(cong_links, node, 0) &&
975 					!tipc_group_cong(grp, node, port,
976 							 blks, &mbr));
977 		if (unlikely(rc))
978 			return rc;
979 
980 		/* Send, unless destination disappeared while waiting */
981 		if (likely(mbr))
982 			break;
983 	}
984 
985 	if (unlikely(lookups >= 4))
986 		return -EHOSTUNREACH;
987 
988 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
989 
990 	return rc ? rc : dlen;
991 }
992 
993 /**
994  * tipc_send_group_bcast - send message to all members in communication group
995  * @sk: socket structure
996  * @m: message to send
997  * @dlen: total length of message data
998  * @timeout: timeout to wait for wakeup
999  *
1000  * Called from function tipc_sendmsg(), which has done all sanity checks
1001  * Returns the number of bytes sent on success, or errno
1002  */
1003 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1004 				 int dlen, long timeout)
1005 {
1006 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1007 	struct sock *sk = sock->sk;
1008 	struct net *net = sock_net(sk);
1009 	struct tipc_sock *tsk = tipc_sk(sk);
1010 	struct tipc_group *grp = tsk->group;
1011 	struct tipc_nlist *dsts = tipc_group_dests(grp);
1012 	struct tipc_mc_method *method = &tsk->mc_method;
1013 	bool ack = method->mandatory && method->rcast;
1014 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1015 	struct tipc_msg *hdr = &tsk->phdr;
1016 	int mtu = tipc_bcast_get_mtu(net);
1017 	struct sk_buff_head pkts;
1018 	int rc = -EHOSTUNREACH;
1019 
1020 	if (!dsts->local && !dsts->remote)
1021 		return -EHOSTUNREACH;
1022 
1023 	/* Block or return if any destination link or member is congested */
1024 	rc = tipc_wait_for_cond(sock, &timeout,	!tsk->cong_link_cnt &&
1025 				!tipc_group_bc_cong(grp, blks));
1026 	if (unlikely(rc))
1027 		return rc;
1028 
1029 	/* Complete message header */
1030 	if (dest) {
1031 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1032 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1033 	} else {
1034 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1035 		msg_set_nameinst(hdr, 0);
1036 	}
1037 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1038 	msg_set_destport(hdr, 0);
1039 	msg_set_destnode(hdr, 0);
1040 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1041 
1042 	/* Avoid getting stuck with repeated forced replicasts */
1043 	msg_set_grp_bc_ack_req(hdr, ack);
1044 
1045 	/* Build message as chain of buffers */
1046 	skb_queue_head_init(&pkts);
1047 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1048 	if (unlikely(rc != dlen))
1049 		return rc;
1050 
1051 	/* Send message */
1052 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1053 	if (unlikely(rc))
1054 		return rc;
1055 
1056 	/* Update broadcast sequence number and send windows */
1057 	tipc_group_update_bc_members(tsk->group, blks, ack);
1058 
1059 	/* Broadcast link is now free to choose method for next broadcast */
1060 	method->mandatory = false;
1061 	method->expires = jiffies;
1062 
1063 	return dlen;
1064 }
1065 
1066 /**
1067  * tipc_send_group_mcast - send message to all members with given identity
1068  * @sock: socket structure
1069  * @m: message to send
1070  * @dlen: total length of message data
1071  * @timeout: timeout to wait for wakeup
1072  *
1073  * Called from function tipc_sendmsg(), which has done all sanity checks
1074  * Returns the number of bytes sent on success, or errno
1075  */
1076 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1077 				 int dlen, long timeout)
1078 {
1079 	struct sock *sk = sock->sk;
1080 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1081 	struct tipc_sock *tsk = tipc_sk(sk);
1082 	struct tipc_group *grp = tsk->group;
1083 	struct tipc_msg *hdr = &tsk->phdr;
1084 	struct net *net = sock_net(sk);
1085 	u32 type, inst, scope, exclude;
1086 	struct list_head dsts;
1087 	u32 dstcnt;
1088 
1089 	INIT_LIST_HEAD(&dsts);
1090 
1091 	type = msg_nametype(hdr);
1092 	inst = dest->addr.name.name.instance;
1093 	scope = msg_lookup_scope(hdr);
1094 	exclude = tipc_group_exclude(grp);
1095 
1096 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1097 				 &dstcnt, exclude, true))
1098 		return -EHOSTUNREACH;
1099 
1100 	if (dstcnt == 1) {
1101 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1102 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1103 	}
1104 
1105 	tipc_dest_list_purge(&dsts);
1106 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1107 }
1108 
1109 /**
1110  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1111  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1112  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1113  *
1114  * Multi-threaded: parallel calls with reference to same queues may occur
1115  */
1116 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1117 		       struct sk_buff_head *inputq)
1118 {
1119 	u32 self = tipc_own_addr(net);
1120 	u32 type, lower, upper, scope;
1121 	struct sk_buff *skb, *_skb;
1122 	u32 portid, oport, onode;
1123 	struct sk_buff_head tmpq;
1124 	struct list_head dports;
1125 	struct tipc_msg *hdr;
1126 	int user, mtyp, hlen;
1127 	bool exact;
1128 
1129 	__skb_queue_head_init(&tmpq);
1130 	INIT_LIST_HEAD(&dports);
1131 
1132 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1133 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1134 		hdr = buf_msg(skb);
1135 		user = msg_user(hdr);
1136 		mtyp = msg_type(hdr);
1137 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1138 		oport = msg_origport(hdr);
1139 		onode = msg_orignode(hdr);
1140 		type = msg_nametype(hdr);
1141 
1142 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1143 			spin_lock_bh(&inputq->lock);
1144 			if (skb_peek(arrvq) == skb) {
1145 				__skb_dequeue(arrvq);
1146 				__skb_queue_tail(inputq, skb);
1147 			}
1148 			kfree_skb(skb);
1149 			spin_unlock_bh(&inputq->lock);
1150 			continue;
1151 		}
1152 
1153 		/* Group messages require exact scope match */
1154 		if (msg_in_group(hdr)) {
1155 			lower = 0;
1156 			upper = ~0;
1157 			scope = msg_lookup_scope(hdr);
1158 			exact = true;
1159 		} else {
1160 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1161 			if (onode == self)
1162 				scope = TIPC_NODE_SCOPE;
1163 			else
1164 				scope = TIPC_CLUSTER_SCOPE;
1165 			exact = false;
1166 			lower = msg_namelower(hdr);
1167 			upper = msg_nameupper(hdr);
1168 		}
1169 
1170 		/* Create destination port list: */
1171 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1172 				       scope, exact, &dports);
1173 
1174 		/* Clone message per destination */
1175 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1176 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1177 			if (_skb) {
1178 				msg_set_destport(buf_msg(_skb), portid);
1179 				__skb_queue_tail(&tmpq, _skb);
1180 				continue;
1181 			}
1182 			pr_warn("Failed to clone mcast rcv buffer\n");
1183 		}
1184 		/* Append to inputq if not already done by other thread */
1185 		spin_lock_bh(&inputq->lock);
1186 		if (skb_peek(arrvq) == skb) {
1187 			skb_queue_splice_tail_init(&tmpq, inputq);
1188 			kfree_skb(__skb_dequeue(arrvq));
1189 		}
1190 		spin_unlock_bh(&inputq->lock);
1191 		__skb_queue_purge(&tmpq);
1192 		kfree_skb(skb);
1193 	}
1194 	tipc_sk_rcv(net, inputq);
1195 }
1196 
1197 /**
1198  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1199  * @tsk: receiving socket
1200  * @skb: pointer to message buffer.
1201  */
1202 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1203 				   struct sk_buff_head *xmitq)
1204 {
1205 	struct tipc_msg *hdr = buf_msg(skb);
1206 	u32 onode = tsk_own_node(tsk);
1207 	struct sock *sk = &tsk->sk;
1208 	int mtyp = msg_type(hdr);
1209 	bool conn_cong;
1210 
1211 	/* Ignore if connection cannot be validated: */
1212 	if (!tsk_peer_msg(tsk, hdr))
1213 		goto exit;
1214 
1215 	if (unlikely(msg_errcode(hdr))) {
1216 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1217 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1218 				      tsk_peer_port(tsk));
1219 		sk->sk_state_change(sk);
1220 		goto exit;
1221 	}
1222 
1223 	tsk->probe_unacked = false;
1224 
1225 	if (mtyp == CONN_PROBE) {
1226 		msg_set_type(hdr, CONN_PROBE_REPLY);
1227 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1228 			__skb_queue_tail(xmitq, skb);
1229 		return;
1230 	} else if (mtyp == CONN_ACK) {
1231 		conn_cong = tsk_conn_cong(tsk);
1232 		tsk->snt_unacked -= msg_conn_ack(hdr);
1233 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1234 			tsk->snd_win = msg_adv_win(hdr);
1235 		if (conn_cong)
1236 			sk->sk_write_space(sk);
1237 	} else if (mtyp != CONN_PROBE_REPLY) {
1238 		pr_warn("Received unknown CONN_PROTO msg\n");
1239 	}
1240 exit:
1241 	kfree_skb(skb);
1242 }
1243 
1244 /**
1245  * tipc_sendmsg - send message in connectionless manner
1246  * @sock: socket structure
1247  * @m: message to send
1248  * @dsz: amount of user data to be sent
1249  *
1250  * Message must have an destination specified explicitly.
1251  * Used for SOCK_RDM and SOCK_DGRAM messages,
1252  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1253  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1254  *
1255  * Returns the number of bytes sent on success, or errno otherwise
1256  */
1257 static int tipc_sendmsg(struct socket *sock,
1258 			struct msghdr *m, size_t dsz)
1259 {
1260 	struct sock *sk = sock->sk;
1261 	int ret;
1262 
1263 	lock_sock(sk);
1264 	ret = __tipc_sendmsg(sock, m, dsz);
1265 	release_sock(sk);
1266 
1267 	return ret;
1268 }
1269 
1270 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1271 {
1272 	struct sock *sk = sock->sk;
1273 	struct net *net = sock_net(sk);
1274 	struct tipc_sock *tsk = tipc_sk(sk);
1275 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1276 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1277 	struct list_head *clinks = &tsk->cong_links;
1278 	bool syn = !tipc_sk_type_connectionless(sk);
1279 	struct tipc_group *grp = tsk->group;
1280 	struct tipc_msg *hdr = &tsk->phdr;
1281 	struct tipc_name_seq *seq;
1282 	struct sk_buff_head pkts;
1283 	u32 type, inst, domain;
1284 	u32 dnode, dport;
1285 	int mtu, rc;
1286 
1287 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1288 		return -EMSGSIZE;
1289 
1290 	if (likely(dest)) {
1291 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1292 			return -EINVAL;
1293 		if (unlikely(dest->family != AF_TIPC))
1294 			return -EINVAL;
1295 	}
1296 
1297 	if (grp) {
1298 		if (!dest)
1299 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1300 		if (dest->addrtype == TIPC_ADDR_NAME)
1301 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1302 		if (dest->addrtype == TIPC_ADDR_ID)
1303 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1304 		if (dest->addrtype == TIPC_ADDR_MCAST)
1305 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1306 		return -EINVAL;
1307 	}
1308 
1309 	if (unlikely(!dest)) {
1310 		dest = &tsk->peer;
1311 		if (!syn || dest->family != AF_TIPC)
1312 			return -EDESTADDRREQ;
1313 	}
1314 
1315 	if (unlikely(syn)) {
1316 		if (sk->sk_state == TIPC_LISTEN)
1317 			return -EPIPE;
1318 		if (sk->sk_state != TIPC_OPEN)
1319 			return -EISCONN;
1320 		if (tsk->published)
1321 			return -EOPNOTSUPP;
1322 		if (dest->addrtype == TIPC_ADDR_NAME) {
1323 			tsk->conn_type = dest->addr.name.name.type;
1324 			tsk->conn_instance = dest->addr.name.name.instance;
1325 		}
1326 	}
1327 
1328 	seq = &dest->addr.nameseq;
1329 	if (dest->addrtype == TIPC_ADDR_MCAST)
1330 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1331 
1332 	if (dest->addrtype == TIPC_ADDR_NAME) {
1333 		type = dest->addr.name.name.type;
1334 		inst = dest->addr.name.name.instance;
1335 		domain = dest->addr.name.domain;
1336 		dnode = domain;
1337 		msg_set_type(hdr, TIPC_NAMED_MSG);
1338 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1339 		msg_set_nametype(hdr, type);
1340 		msg_set_nameinst(hdr, inst);
1341 		msg_set_lookup_scope(hdr, tipc_addr_scope(domain));
1342 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1343 		msg_set_destnode(hdr, dnode);
1344 		msg_set_destport(hdr, dport);
1345 		if (unlikely(!dport && !dnode))
1346 			return -EHOSTUNREACH;
1347 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1348 		dnode = dest->addr.id.node;
1349 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1350 		msg_set_lookup_scope(hdr, 0);
1351 		msg_set_destnode(hdr, dnode);
1352 		msg_set_destport(hdr, dest->addr.id.ref);
1353 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1354 	}
1355 
1356 	/* Block or return if destination link is congested */
1357 	rc = tipc_wait_for_cond(sock, &timeout,
1358 				!tipc_dest_find(clinks, dnode, 0));
1359 	if (unlikely(rc))
1360 		return rc;
1361 
1362 	skb_queue_head_init(&pkts);
1363 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1364 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1365 	if (unlikely(rc != dlen))
1366 		return rc;
1367 
1368 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1369 	if (unlikely(rc == -ELINKCONG)) {
1370 		tipc_dest_push(clinks, dnode, 0);
1371 		tsk->cong_link_cnt++;
1372 		rc = 0;
1373 	}
1374 
1375 	if (unlikely(syn && !rc))
1376 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1377 
1378 	return rc ? rc : dlen;
1379 }
1380 
1381 /**
1382  * tipc_sendstream - send stream-oriented data
1383  * @sock: socket structure
1384  * @m: data to send
1385  * @dsz: total length of data to be transmitted
1386  *
1387  * Used for SOCK_STREAM data.
1388  *
1389  * Returns the number of bytes sent on success (or partial success),
1390  * or errno if no data sent
1391  */
1392 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1393 {
1394 	struct sock *sk = sock->sk;
1395 	int ret;
1396 
1397 	lock_sock(sk);
1398 	ret = __tipc_sendstream(sock, m, dsz);
1399 	release_sock(sk);
1400 
1401 	return ret;
1402 }
1403 
1404 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1405 {
1406 	struct sock *sk = sock->sk;
1407 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1408 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1409 	struct tipc_sock *tsk = tipc_sk(sk);
1410 	struct tipc_msg *hdr = &tsk->phdr;
1411 	struct net *net = sock_net(sk);
1412 	struct sk_buff_head pkts;
1413 	u32 dnode = tsk_peer_node(tsk);
1414 	int send, sent = 0;
1415 	int rc = 0;
1416 
1417 	skb_queue_head_init(&pkts);
1418 
1419 	if (unlikely(dlen > INT_MAX))
1420 		return -EMSGSIZE;
1421 
1422 	/* Handle implicit connection setup */
1423 	if (unlikely(dest)) {
1424 		rc = __tipc_sendmsg(sock, m, dlen);
1425 		if (dlen && (dlen == rc))
1426 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1427 		return rc;
1428 	}
1429 
1430 	do {
1431 		rc = tipc_wait_for_cond(sock, &timeout,
1432 					(!tsk->cong_link_cnt &&
1433 					 !tsk_conn_cong(tsk) &&
1434 					 tipc_sk_connected(sk)));
1435 		if (unlikely(rc))
1436 			break;
1437 
1438 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1439 		rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1440 		if (unlikely(rc != send))
1441 			break;
1442 
1443 		rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1444 		if (unlikely(rc == -ELINKCONG)) {
1445 			tsk->cong_link_cnt = 1;
1446 			rc = 0;
1447 		}
1448 		if (likely(!rc)) {
1449 			tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1450 			sent += send;
1451 		}
1452 	} while (sent < dlen && !rc);
1453 
1454 	return sent ? sent : rc;
1455 }
1456 
1457 /**
1458  * tipc_send_packet - send a connection-oriented message
1459  * @sock: socket structure
1460  * @m: message to send
1461  * @dsz: length of data to be transmitted
1462  *
1463  * Used for SOCK_SEQPACKET messages.
1464  *
1465  * Returns the number of bytes sent on success, or errno otherwise
1466  */
1467 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1468 {
1469 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1470 		return -EMSGSIZE;
1471 
1472 	return tipc_sendstream(sock, m, dsz);
1473 }
1474 
1475 /* tipc_sk_finish_conn - complete the setup of a connection
1476  */
1477 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1478 				u32 peer_node)
1479 {
1480 	struct sock *sk = &tsk->sk;
1481 	struct net *net = sock_net(sk);
1482 	struct tipc_msg *msg = &tsk->phdr;
1483 
1484 	msg_set_destnode(msg, peer_node);
1485 	msg_set_destport(msg, peer_port);
1486 	msg_set_type(msg, TIPC_CONN_MSG);
1487 	msg_set_lookup_scope(msg, 0);
1488 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1489 
1490 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1491 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1492 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1493 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1494 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1495 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1496 		return;
1497 
1498 	/* Fall back to message based flow control */
1499 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1500 	tsk->snd_win = FLOWCTL_MSG_WIN;
1501 }
1502 
1503 /**
1504  * tipc_sk_set_orig_addr - capture sender's address for received message
1505  * @m: descriptor for message info
1506  * @hdr: received message header
1507  *
1508  * Note: Address is not captured if not requested by receiver.
1509  */
1510 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1511 {
1512 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1513 	struct tipc_msg *hdr = buf_msg(skb);
1514 
1515 	if (!srcaddr)
1516 		return;
1517 
1518 	srcaddr->sock.family = AF_TIPC;
1519 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1520 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1521 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1522 	srcaddr->sock.addr.name.domain = 0;
1523 	srcaddr->sock.scope = 0;
1524 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1525 
1526 	if (!msg_in_group(hdr))
1527 		return;
1528 
1529 	/* Group message users may also want to know sending member's id */
1530 	srcaddr->member.family = AF_TIPC;
1531 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1532 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1533 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1534 	srcaddr->member.addr.name.domain = 0;
1535 	m->msg_namelen = sizeof(*srcaddr);
1536 }
1537 
1538 /**
1539  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1540  * @m: descriptor for message info
1541  * @msg: received message header
1542  * @tsk: TIPC port associated with message
1543  *
1544  * Note: Ancillary data is not captured if not requested by receiver.
1545  *
1546  * Returns 0 if successful, otherwise errno
1547  */
1548 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1549 				 struct tipc_sock *tsk)
1550 {
1551 	u32 anc_data[3];
1552 	u32 err;
1553 	u32 dest_type;
1554 	int has_name;
1555 	int res;
1556 
1557 	if (likely(m->msg_controllen == 0))
1558 		return 0;
1559 
1560 	/* Optionally capture errored message object(s) */
1561 	err = msg ? msg_errcode(msg) : 0;
1562 	if (unlikely(err)) {
1563 		anc_data[0] = err;
1564 		anc_data[1] = msg_data_sz(msg);
1565 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1566 		if (res)
1567 			return res;
1568 		if (anc_data[1]) {
1569 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1570 				       msg_data(msg));
1571 			if (res)
1572 				return res;
1573 		}
1574 	}
1575 
1576 	/* Optionally capture message destination object */
1577 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1578 	switch (dest_type) {
1579 	case TIPC_NAMED_MSG:
1580 		has_name = 1;
1581 		anc_data[0] = msg_nametype(msg);
1582 		anc_data[1] = msg_namelower(msg);
1583 		anc_data[2] = msg_namelower(msg);
1584 		break;
1585 	case TIPC_MCAST_MSG:
1586 		has_name = 1;
1587 		anc_data[0] = msg_nametype(msg);
1588 		anc_data[1] = msg_namelower(msg);
1589 		anc_data[2] = msg_nameupper(msg);
1590 		break;
1591 	case TIPC_CONN_MSG:
1592 		has_name = (tsk->conn_type != 0);
1593 		anc_data[0] = tsk->conn_type;
1594 		anc_data[1] = tsk->conn_instance;
1595 		anc_data[2] = tsk->conn_instance;
1596 		break;
1597 	default:
1598 		has_name = 0;
1599 	}
1600 	if (has_name) {
1601 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1602 		if (res)
1603 			return res;
1604 	}
1605 
1606 	return 0;
1607 }
1608 
1609 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1610 {
1611 	struct sock *sk = &tsk->sk;
1612 	struct net *net = sock_net(sk);
1613 	struct sk_buff *skb = NULL;
1614 	struct tipc_msg *msg;
1615 	u32 peer_port = tsk_peer_port(tsk);
1616 	u32 dnode = tsk_peer_node(tsk);
1617 
1618 	if (!tipc_sk_connected(sk))
1619 		return;
1620 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1621 			      dnode, tsk_own_node(tsk), peer_port,
1622 			      tsk->portid, TIPC_OK);
1623 	if (!skb)
1624 		return;
1625 	msg = buf_msg(skb);
1626 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1627 	tsk->rcv_unacked = 0;
1628 
1629 	/* Adjust to and advertize the correct window limit */
1630 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1631 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1632 		msg_set_adv_win(msg, tsk->rcv_win);
1633 	}
1634 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1635 }
1636 
1637 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1638 {
1639 	struct sock *sk = sock->sk;
1640 	DEFINE_WAIT(wait);
1641 	long timeo = *timeop;
1642 	int err = sock_error(sk);
1643 
1644 	if (err)
1645 		return err;
1646 
1647 	for (;;) {
1648 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1649 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1650 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1651 				err = -ENOTCONN;
1652 				break;
1653 			}
1654 			release_sock(sk);
1655 			timeo = schedule_timeout(timeo);
1656 			lock_sock(sk);
1657 		}
1658 		err = 0;
1659 		if (!skb_queue_empty(&sk->sk_receive_queue))
1660 			break;
1661 		err = -EAGAIN;
1662 		if (!timeo)
1663 			break;
1664 		err = sock_intr_errno(timeo);
1665 		if (signal_pending(current))
1666 			break;
1667 
1668 		err = sock_error(sk);
1669 		if (err)
1670 			break;
1671 	}
1672 	finish_wait(sk_sleep(sk), &wait);
1673 	*timeop = timeo;
1674 	return err;
1675 }
1676 
1677 /**
1678  * tipc_recvmsg - receive packet-oriented message
1679  * @m: descriptor for message info
1680  * @buflen: length of user buffer area
1681  * @flags: receive flags
1682  *
1683  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1684  * If the complete message doesn't fit in user area, truncate it.
1685  *
1686  * Returns size of returned message data, errno otherwise
1687  */
1688 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1689 			size_t buflen,	int flags)
1690 {
1691 	struct sock *sk = sock->sk;
1692 	bool connected = !tipc_sk_type_connectionless(sk);
1693 	struct tipc_sock *tsk = tipc_sk(sk);
1694 	int rc, err, hlen, dlen, copy;
1695 	struct sk_buff_head xmitq;
1696 	struct tipc_msg *hdr;
1697 	struct sk_buff *skb;
1698 	bool grp_evt;
1699 	long timeout;
1700 
1701 	/* Catch invalid receive requests */
1702 	if (unlikely(!buflen))
1703 		return -EINVAL;
1704 
1705 	lock_sock(sk);
1706 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1707 		rc = -ENOTCONN;
1708 		goto exit;
1709 	}
1710 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1711 
1712 	/* Step rcv queue to first msg with data or error; wait if necessary */
1713 	do {
1714 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1715 		if (unlikely(rc))
1716 			goto exit;
1717 		skb = skb_peek(&sk->sk_receive_queue);
1718 		hdr = buf_msg(skb);
1719 		dlen = msg_data_sz(hdr);
1720 		hlen = msg_hdr_sz(hdr);
1721 		err = msg_errcode(hdr);
1722 		grp_evt = msg_is_grp_evt(hdr);
1723 		if (likely(dlen || err))
1724 			break;
1725 		tsk_advance_rx_queue(sk);
1726 	} while (1);
1727 
1728 	/* Collect msg meta data, including error code and rejected data */
1729 	tipc_sk_set_orig_addr(m, skb);
1730 	rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1731 	if (unlikely(rc))
1732 		goto exit;
1733 
1734 	/* Capture data if non-error msg, otherwise just set return value */
1735 	if (likely(!err)) {
1736 		copy = min_t(int, dlen, buflen);
1737 		if (unlikely(copy != dlen))
1738 			m->msg_flags |= MSG_TRUNC;
1739 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1740 	} else {
1741 		copy = 0;
1742 		rc = 0;
1743 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1744 			rc = -ECONNRESET;
1745 	}
1746 	if (unlikely(rc))
1747 		goto exit;
1748 
1749 	/* Mark message as group event if applicable */
1750 	if (unlikely(grp_evt)) {
1751 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1752 			m->msg_flags |= MSG_EOR;
1753 		m->msg_flags |= MSG_OOB;
1754 		copy = 0;
1755 	}
1756 
1757 	/* Caption of data or error code/rejected data was successful */
1758 	if (unlikely(flags & MSG_PEEK))
1759 		goto exit;
1760 
1761 	/* Send group flow control advertisement when applicable */
1762 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1763 		skb_queue_head_init(&xmitq);
1764 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1765 					  msg_orignode(hdr), msg_origport(hdr),
1766 					  &xmitq);
1767 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1768 	}
1769 
1770 	tsk_advance_rx_queue(sk);
1771 
1772 	if (likely(!connected))
1773 		goto exit;
1774 
1775 	/* Send connection flow control advertisement when applicable */
1776 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1777 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1778 		tipc_sk_send_ack(tsk);
1779 exit:
1780 	release_sock(sk);
1781 	return rc ? rc : copy;
1782 }
1783 
1784 /**
1785  * tipc_recvstream - receive stream-oriented data
1786  * @m: descriptor for message info
1787  * @buflen: total size of user buffer area
1788  * @flags: receive flags
1789  *
1790  * Used for SOCK_STREAM messages only.  If not enough data is available
1791  * will optionally wait for more; never truncates data.
1792  *
1793  * Returns size of returned message data, errno otherwise
1794  */
1795 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1796 			   size_t buflen, int flags)
1797 {
1798 	struct sock *sk = sock->sk;
1799 	struct tipc_sock *tsk = tipc_sk(sk);
1800 	struct sk_buff *skb;
1801 	struct tipc_msg *hdr;
1802 	struct tipc_skb_cb *skb_cb;
1803 	bool peek = flags & MSG_PEEK;
1804 	int offset, required, copy, copied = 0;
1805 	int hlen, dlen, err, rc;
1806 	long timeout;
1807 
1808 	/* Catch invalid receive attempts */
1809 	if (unlikely(!buflen))
1810 		return -EINVAL;
1811 
1812 	lock_sock(sk);
1813 
1814 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1815 		rc = -ENOTCONN;
1816 		goto exit;
1817 	}
1818 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1819 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1820 
1821 	do {
1822 		/* Look at first msg in receive queue; wait if necessary */
1823 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1824 		if (unlikely(rc))
1825 			break;
1826 		skb = skb_peek(&sk->sk_receive_queue);
1827 		skb_cb = TIPC_SKB_CB(skb);
1828 		hdr = buf_msg(skb);
1829 		dlen = msg_data_sz(hdr);
1830 		hlen = msg_hdr_sz(hdr);
1831 		err = msg_errcode(hdr);
1832 
1833 		/* Discard any empty non-errored (SYN-) message */
1834 		if (unlikely(!dlen && !err)) {
1835 			tsk_advance_rx_queue(sk);
1836 			continue;
1837 		}
1838 
1839 		/* Collect msg meta data, incl. error code and rejected data */
1840 		if (!copied) {
1841 			tipc_sk_set_orig_addr(m, skb);
1842 			rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1843 			if (rc)
1844 				break;
1845 		}
1846 
1847 		/* Copy data if msg ok, otherwise return error/partial data */
1848 		if (likely(!err)) {
1849 			offset = skb_cb->bytes_read;
1850 			copy = min_t(int, dlen - offset, buflen - copied);
1851 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1852 			if (unlikely(rc))
1853 				break;
1854 			copied += copy;
1855 			offset += copy;
1856 			if (unlikely(offset < dlen)) {
1857 				if (!peek)
1858 					skb_cb->bytes_read = offset;
1859 				break;
1860 			}
1861 		} else {
1862 			rc = 0;
1863 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1864 				rc = -ECONNRESET;
1865 			if (copied || rc)
1866 				break;
1867 		}
1868 
1869 		if (unlikely(peek))
1870 			break;
1871 
1872 		tsk_advance_rx_queue(sk);
1873 
1874 		/* Send connection flow control advertisement when applicable */
1875 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1876 		if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1877 			tipc_sk_send_ack(tsk);
1878 
1879 		/* Exit if all requested data or FIN/error received */
1880 		if (copied == buflen || err)
1881 			break;
1882 
1883 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1884 exit:
1885 	release_sock(sk);
1886 	return copied ? copied : rc;
1887 }
1888 
1889 /**
1890  * tipc_write_space - wake up thread if port congestion is released
1891  * @sk: socket
1892  */
1893 static void tipc_write_space(struct sock *sk)
1894 {
1895 	struct socket_wq *wq;
1896 
1897 	rcu_read_lock();
1898 	wq = rcu_dereference(sk->sk_wq);
1899 	if (skwq_has_sleeper(wq))
1900 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1901 						EPOLLWRNORM | EPOLLWRBAND);
1902 	rcu_read_unlock();
1903 }
1904 
1905 /**
1906  * tipc_data_ready - wake up threads to indicate messages have been received
1907  * @sk: socket
1908  * @len: the length of messages
1909  */
1910 static void tipc_data_ready(struct sock *sk)
1911 {
1912 	struct socket_wq *wq;
1913 
1914 	rcu_read_lock();
1915 	wq = rcu_dereference(sk->sk_wq);
1916 	if (skwq_has_sleeper(wq))
1917 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1918 						EPOLLRDNORM | EPOLLRDBAND);
1919 	rcu_read_unlock();
1920 }
1921 
1922 static void tipc_sock_destruct(struct sock *sk)
1923 {
1924 	__skb_queue_purge(&sk->sk_receive_queue);
1925 }
1926 
1927 static void tipc_sk_proto_rcv(struct sock *sk,
1928 			      struct sk_buff_head *inputq,
1929 			      struct sk_buff_head *xmitq)
1930 {
1931 	struct sk_buff *skb = __skb_dequeue(inputq);
1932 	struct tipc_sock *tsk = tipc_sk(sk);
1933 	struct tipc_msg *hdr = buf_msg(skb);
1934 	struct tipc_group *grp = tsk->group;
1935 	bool wakeup = false;
1936 
1937 	switch (msg_user(hdr)) {
1938 	case CONN_MANAGER:
1939 		tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1940 		return;
1941 	case SOCK_WAKEUP:
1942 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1943 		tsk->cong_link_cnt--;
1944 		wakeup = true;
1945 		break;
1946 	case GROUP_PROTOCOL:
1947 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1948 		break;
1949 	case TOP_SRV:
1950 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1951 				      hdr, inputq, xmitq);
1952 		break;
1953 	default:
1954 		break;
1955 	}
1956 
1957 	if (wakeup)
1958 		sk->sk_write_space(sk);
1959 
1960 	kfree_skb(skb);
1961 }
1962 
1963 /**
1964  * tipc_filter_connect - Handle incoming message for a connection-based socket
1965  * @tsk: TIPC socket
1966  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1967  *
1968  * Returns true if everything ok, false otherwise
1969  */
1970 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1971 {
1972 	struct sock *sk = &tsk->sk;
1973 	struct net *net = sock_net(sk);
1974 	struct tipc_msg *hdr = buf_msg(skb);
1975 	u32 pport = msg_origport(hdr);
1976 	u32 pnode = msg_orignode(hdr);
1977 
1978 	if (unlikely(msg_mcast(hdr)))
1979 		return false;
1980 
1981 	switch (sk->sk_state) {
1982 	case TIPC_CONNECTING:
1983 		/* Accept only ACK or NACK message */
1984 		if (unlikely(!msg_connected(hdr))) {
1985 			if (pport != tsk_peer_port(tsk) ||
1986 			    pnode != tsk_peer_node(tsk))
1987 				return false;
1988 
1989 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1990 			sk->sk_err = ECONNREFUSED;
1991 			sk->sk_state_change(sk);
1992 			return true;
1993 		}
1994 
1995 		if (unlikely(msg_errcode(hdr))) {
1996 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1997 			sk->sk_err = ECONNREFUSED;
1998 			sk->sk_state_change(sk);
1999 			return true;
2000 		}
2001 
2002 		if (unlikely(!msg_isdata(hdr))) {
2003 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2004 			sk->sk_err = EINVAL;
2005 			sk->sk_state_change(sk);
2006 			return true;
2007 		}
2008 
2009 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2010 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
2011 
2012 		/* If 'ACK+' message, add to socket receive queue */
2013 		if (msg_data_sz(hdr))
2014 			return true;
2015 
2016 		/* If empty 'ACK-' message, wake up sleeping connect() */
2017 		sk->sk_data_ready(sk);
2018 
2019 		/* 'ACK-' message is neither accepted nor rejected: */
2020 		msg_set_dest_droppable(hdr, 1);
2021 		return false;
2022 
2023 	case TIPC_OPEN:
2024 	case TIPC_DISCONNECTING:
2025 		break;
2026 	case TIPC_LISTEN:
2027 		/* Accept only SYN message */
2028 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2029 			return true;
2030 		break;
2031 	case TIPC_ESTABLISHED:
2032 		/* Accept only connection-based messages sent by peer */
2033 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
2034 			return false;
2035 
2036 		if (unlikely(msg_errcode(hdr))) {
2037 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2038 			/* Let timer expire on it's own */
2039 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
2040 					      tsk->portid);
2041 			sk->sk_state_change(sk);
2042 		}
2043 		return true;
2044 	default:
2045 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2046 	}
2047 
2048 	return false;
2049 }
2050 
2051 /**
2052  * rcvbuf_limit - get proper overload limit of socket receive queue
2053  * @sk: socket
2054  * @skb: message
2055  *
2056  * For connection oriented messages, irrespective of importance,
2057  * default queue limit is 2 MB.
2058  *
2059  * For connectionless messages, queue limits are based on message
2060  * importance as follows:
2061  *
2062  * TIPC_LOW_IMPORTANCE       (2 MB)
2063  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2064  * TIPC_HIGH_IMPORTANCE      (8 MB)
2065  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2066  *
2067  * Returns overload limit according to corresponding message importance
2068  */
2069 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2070 {
2071 	struct tipc_sock *tsk = tipc_sk(sk);
2072 	struct tipc_msg *hdr = buf_msg(skb);
2073 
2074 	if (unlikely(msg_in_group(hdr)))
2075 		return sk->sk_rcvbuf;
2076 
2077 	if (unlikely(!msg_connected(hdr)))
2078 		return sk->sk_rcvbuf << msg_importance(hdr);
2079 
2080 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2081 		return sk->sk_rcvbuf;
2082 
2083 	return FLOWCTL_MSG_LIM;
2084 }
2085 
2086 /**
2087  * tipc_sk_filter_rcv - validate incoming message
2088  * @sk: socket
2089  * @skb: pointer to message.
2090  *
2091  * Enqueues message on receive queue if acceptable; optionally handles
2092  * disconnect indication for a connected socket.
2093  *
2094  * Called with socket lock already taken
2095  *
2096  */
2097 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2098 			       struct sk_buff_head *xmitq)
2099 {
2100 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2101 	struct tipc_sock *tsk = tipc_sk(sk);
2102 	struct tipc_group *grp = tsk->group;
2103 	struct tipc_msg *hdr = buf_msg(skb);
2104 	struct net *net = sock_net(sk);
2105 	struct sk_buff_head inputq;
2106 	int limit, err = TIPC_OK;
2107 
2108 	TIPC_SKB_CB(skb)->bytes_read = 0;
2109 	__skb_queue_head_init(&inputq);
2110 	__skb_queue_tail(&inputq, skb);
2111 
2112 	if (unlikely(!msg_isdata(hdr)))
2113 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2114 
2115 	if (unlikely(grp))
2116 		tipc_group_filter_msg(grp, &inputq, xmitq);
2117 
2118 	/* Validate and add to receive buffer if there is space */
2119 	while ((skb = __skb_dequeue(&inputq))) {
2120 		hdr = buf_msg(skb);
2121 		limit = rcvbuf_limit(sk, skb);
2122 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2123 		    (!sk_conn && msg_connected(hdr)) ||
2124 		    (!grp && msg_in_group(hdr)))
2125 			err = TIPC_ERR_NO_PORT;
2126 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit)
2127 			err = TIPC_ERR_OVERLOAD;
2128 
2129 		if (unlikely(err)) {
2130 			tipc_skb_reject(net, err, skb, xmitq);
2131 			err = TIPC_OK;
2132 			continue;
2133 		}
2134 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2135 		skb_set_owner_r(skb, sk);
2136 		sk->sk_data_ready(sk);
2137 	}
2138 }
2139 
2140 /**
2141  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2142  * @sk: socket
2143  * @skb: message
2144  *
2145  * Caller must hold socket lock
2146  */
2147 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2148 {
2149 	unsigned int before = sk_rmem_alloc_get(sk);
2150 	struct sk_buff_head xmitq;
2151 	unsigned int added;
2152 
2153 	__skb_queue_head_init(&xmitq);
2154 
2155 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2156 	added = sk_rmem_alloc_get(sk) - before;
2157 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2158 
2159 	/* Send pending response/rejected messages, if any */
2160 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2161 	return 0;
2162 }
2163 
2164 /**
2165  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2166  *                   inputq and try adding them to socket or backlog queue
2167  * @inputq: list of incoming buffers with potentially different destinations
2168  * @sk: socket where the buffers should be enqueued
2169  * @dport: port number for the socket
2170  *
2171  * Caller must hold socket lock
2172  */
2173 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2174 			    u32 dport, struct sk_buff_head *xmitq)
2175 {
2176 	unsigned long time_limit = jiffies + 2;
2177 	struct sk_buff *skb;
2178 	unsigned int lim;
2179 	atomic_t *dcnt;
2180 	u32 onode;
2181 
2182 	while (skb_queue_len(inputq)) {
2183 		if (unlikely(time_after_eq(jiffies, time_limit)))
2184 			return;
2185 
2186 		skb = tipc_skb_dequeue(inputq, dport);
2187 		if (unlikely(!skb))
2188 			return;
2189 
2190 		/* Add message directly to receive queue if possible */
2191 		if (!sock_owned_by_user(sk)) {
2192 			tipc_sk_filter_rcv(sk, skb, xmitq);
2193 			continue;
2194 		}
2195 
2196 		/* Try backlog, compensating for double-counted bytes */
2197 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2198 		if (!sk->sk_backlog.len)
2199 			atomic_set(dcnt, 0);
2200 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2201 		if (likely(!sk_add_backlog(sk, skb, lim)))
2202 			continue;
2203 
2204 		/* Overload => reject message back to sender */
2205 		onode = tipc_own_addr(sock_net(sk));
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 (tipc_sk_connected(sk))
2596 		return -EINVAL;
2597 	key = tsk->portid + tsk->pub_count + 1;
2598 	if (key == tsk->portid)
2599 		return -EADDRINUSE;
2600 
2601 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2602 				    scope, tsk->portid, key);
2603 	if (unlikely(!publ))
2604 		return -EINVAL;
2605 
2606 	list_add(&publ->pport_list, &tsk->publications);
2607 	tsk->pub_count++;
2608 	tsk->published = 1;
2609 	return 0;
2610 }
2611 
2612 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2613 			    struct tipc_name_seq const *seq)
2614 {
2615 	struct net *net = sock_net(&tsk->sk);
2616 	struct publication *publ;
2617 	struct publication *safe;
2618 	int rc = -EINVAL;
2619 
2620 	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2621 		if (seq) {
2622 			if (publ->scope != scope)
2623 				continue;
2624 			if (publ->type != seq->type)
2625 				continue;
2626 			if (publ->lower != seq->lower)
2627 				continue;
2628 			if (publ->upper != seq->upper)
2629 				break;
2630 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2631 					      publ->ref, publ->key);
2632 			rc = 0;
2633 			break;
2634 		}
2635 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2636 				      publ->ref, publ->key);
2637 		rc = 0;
2638 	}
2639 	if (list_empty(&tsk->publications))
2640 		tsk->published = 0;
2641 	return rc;
2642 }
2643 
2644 /* tipc_sk_reinit: set non-zero address in all existing sockets
2645  *                 when we go from standalone to network mode.
2646  */
2647 void tipc_sk_reinit(struct net *net)
2648 {
2649 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2650 	struct rhashtable_iter iter;
2651 	struct tipc_sock *tsk;
2652 	struct tipc_msg *msg;
2653 
2654 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2655 
2656 	do {
2657 		rhashtable_walk_start(&iter);
2658 
2659 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2660 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2661 			msg = &tsk->phdr;
2662 			msg_set_prevnode(msg, tn->own_addr);
2663 			msg_set_orignode(msg, tn->own_addr);
2664 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2665 		}
2666 
2667 		rhashtable_walk_stop(&iter);
2668 	} while (tsk == ERR_PTR(-EAGAIN));
2669 }
2670 
2671 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2672 {
2673 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2674 	struct tipc_sock *tsk;
2675 
2676 	rcu_read_lock();
2677 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2678 	if (tsk)
2679 		sock_hold(&tsk->sk);
2680 	rcu_read_unlock();
2681 
2682 	return tsk;
2683 }
2684 
2685 static int tipc_sk_insert(struct tipc_sock *tsk)
2686 {
2687 	struct sock *sk = &tsk->sk;
2688 	struct net *net = sock_net(sk);
2689 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2690 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2691 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2692 
2693 	while (remaining--) {
2694 		portid++;
2695 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2696 			portid = TIPC_MIN_PORT;
2697 		tsk->portid = portid;
2698 		sock_hold(&tsk->sk);
2699 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2700 						   tsk_rht_params))
2701 			return 0;
2702 		sock_put(&tsk->sk);
2703 	}
2704 
2705 	return -1;
2706 }
2707 
2708 static void tipc_sk_remove(struct tipc_sock *tsk)
2709 {
2710 	struct sock *sk = &tsk->sk;
2711 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2712 
2713 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2714 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2715 		__sock_put(sk);
2716 	}
2717 }
2718 
2719 static const struct rhashtable_params tsk_rht_params = {
2720 	.nelem_hint = 192,
2721 	.head_offset = offsetof(struct tipc_sock, node),
2722 	.key_offset = offsetof(struct tipc_sock, portid),
2723 	.key_len = sizeof(u32), /* portid */
2724 	.max_size = 1048576,
2725 	.min_size = 256,
2726 	.automatic_shrinking = true,
2727 };
2728 
2729 int tipc_sk_rht_init(struct net *net)
2730 {
2731 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2732 
2733 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2734 }
2735 
2736 void tipc_sk_rht_destroy(struct net *net)
2737 {
2738 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2739 
2740 	/* Wait for socket readers to complete */
2741 	synchronize_net();
2742 
2743 	rhashtable_destroy(&tn->sk_rht);
2744 }
2745 
2746 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2747 {
2748 	struct net *net = sock_net(&tsk->sk);
2749 	struct tipc_group *grp = tsk->group;
2750 	struct tipc_msg *hdr = &tsk->phdr;
2751 	struct tipc_name_seq seq;
2752 	int rc;
2753 
2754 	if (mreq->type < TIPC_RESERVED_TYPES)
2755 		return -EACCES;
2756 	if (mreq->scope > TIPC_NODE_SCOPE)
2757 		return -EINVAL;
2758 	if (grp)
2759 		return -EACCES;
2760 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2761 	if (!grp)
2762 		return -ENOMEM;
2763 	tsk->group = grp;
2764 	msg_set_lookup_scope(hdr, mreq->scope);
2765 	msg_set_nametype(hdr, mreq->type);
2766 	msg_set_dest_droppable(hdr, true);
2767 	seq.type = mreq->type;
2768 	seq.lower = mreq->instance;
2769 	seq.upper = seq.lower;
2770 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2771 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2772 	if (rc) {
2773 		tipc_group_delete(net, grp);
2774 		tsk->group = NULL;
2775 		return rc;
2776 	}
2777 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2778 	tsk->mc_method.rcast = true;
2779 	tsk->mc_method.mandatory = true;
2780 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2781 	return rc;
2782 }
2783 
2784 static int tipc_sk_leave(struct tipc_sock *tsk)
2785 {
2786 	struct net *net = sock_net(&tsk->sk);
2787 	struct tipc_group *grp = tsk->group;
2788 	struct tipc_name_seq seq;
2789 	int scope;
2790 
2791 	if (!grp)
2792 		return -EINVAL;
2793 	tipc_group_self(grp, &seq, &scope);
2794 	tipc_group_delete(net, grp);
2795 	tsk->group = NULL;
2796 	tipc_sk_withdraw(tsk, scope, &seq);
2797 	return 0;
2798 }
2799 
2800 /**
2801  * tipc_setsockopt - set socket option
2802  * @sock: socket structure
2803  * @lvl: option level
2804  * @opt: option identifier
2805  * @ov: pointer to new option value
2806  * @ol: length of option value
2807  *
2808  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2809  * (to ease compatibility).
2810  *
2811  * Returns 0 on success, errno otherwise
2812  */
2813 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2814 			   char __user *ov, unsigned int ol)
2815 {
2816 	struct sock *sk = sock->sk;
2817 	struct tipc_sock *tsk = tipc_sk(sk);
2818 	struct tipc_group_req mreq;
2819 	u32 value = 0;
2820 	int res = 0;
2821 
2822 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2823 		return 0;
2824 	if (lvl != SOL_TIPC)
2825 		return -ENOPROTOOPT;
2826 
2827 	switch (opt) {
2828 	case TIPC_IMPORTANCE:
2829 	case TIPC_SRC_DROPPABLE:
2830 	case TIPC_DEST_DROPPABLE:
2831 	case TIPC_CONN_TIMEOUT:
2832 		if (ol < sizeof(value))
2833 			return -EINVAL;
2834 		if (get_user(value, (u32 __user *)ov))
2835 			return -EFAULT;
2836 		break;
2837 	case TIPC_GROUP_JOIN:
2838 		if (ol < sizeof(mreq))
2839 			return -EINVAL;
2840 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
2841 			return -EFAULT;
2842 		break;
2843 	default:
2844 		if (ov || ol)
2845 			return -EINVAL;
2846 	}
2847 
2848 	lock_sock(sk);
2849 
2850 	switch (opt) {
2851 	case TIPC_IMPORTANCE:
2852 		res = tsk_set_importance(tsk, value);
2853 		break;
2854 	case TIPC_SRC_DROPPABLE:
2855 		if (sock->type != SOCK_STREAM)
2856 			tsk_set_unreliable(tsk, value);
2857 		else
2858 			res = -ENOPROTOOPT;
2859 		break;
2860 	case TIPC_DEST_DROPPABLE:
2861 		tsk_set_unreturnable(tsk, value);
2862 		break;
2863 	case TIPC_CONN_TIMEOUT:
2864 		tipc_sk(sk)->conn_timeout = value;
2865 		break;
2866 	case TIPC_MCAST_BROADCAST:
2867 		tsk->mc_method.rcast = false;
2868 		tsk->mc_method.mandatory = true;
2869 		break;
2870 	case TIPC_MCAST_REPLICAST:
2871 		tsk->mc_method.rcast = true;
2872 		tsk->mc_method.mandatory = true;
2873 		break;
2874 	case TIPC_GROUP_JOIN:
2875 		res = tipc_sk_join(tsk, &mreq);
2876 		break;
2877 	case TIPC_GROUP_LEAVE:
2878 		res = tipc_sk_leave(tsk);
2879 		break;
2880 	default:
2881 		res = -EINVAL;
2882 	}
2883 
2884 	release_sock(sk);
2885 
2886 	return res;
2887 }
2888 
2889 /**
2890  * tipc_getsockopt - get socket option
2891  * @sock: socket structure
2892  * @lvl: option level
2893  * @opt: option identifier
2894  * @ov: receptacle for option value
2895  * @ol: receptacle for length of option value
2896  *
2897  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2898  * (to ease compatibility).
2899  *
2900  * Returns 0 on success, errno otherwise
2901  */
2902 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2903 			   char __user *ov, int __user *ol)
2904 {
2905 	struct sock *sk = sock->sk;
2906 	struct tipc_sock *tsk = tipc_sk(sk);
2907 	struct tipc_name_seq seq;
2908 	int len, scope;
2909 	u32 value;
2910 	int res;
2911 
2912 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2913 		return put_user(0, ol);
2914 	if (lvl != SOL_TIPC)
2915 		return -ENOPROTOOPT;
2916 	res = get_user(len, ol);
2917 	if (res)
2918 		return res;
2919 
2920 	lock_sock(sk);
2921 
2922 	switch (opt) {
2923 	case TIPC_IMPORTANCE:
2924 		value = tsk_importance(tsk);
2925 		break;
2926 	case TIPC_SRC_DROPPABLE:
2927 		value = tsk_unreliable(tsk);
2928 		break;
2929 	case TIPC_DEST_DROPPABLE:
2930 		value = tsk_unreturnable(tsk);
2931 		break;
2932 	case TIPC_CONN_TIMEOUT:
2933 		value = tsk->conn_timeout;
2934 		/* no need to set "res", since already 0 at this point */
2935 		break;
2936 	case TIPC_NODE_RECVQ_DEPTH:
2937 		value = 0; /* was tipc_queue_size, now obsolete */
2938 		break;
2939 	case TIPC_SOCK_RECVQ_DEPTH:
2940 		value = skb_queue_len(&sk->sk_receive_queue);
2941 		break;
2942 	case TIPC_GROUP_JOIN:
2943 		seq.type = 0;
2944 		if (tsk->group)
2945 			tipc_group_self(tsk->group, &seq, &scope);
2946 		value = seq.type;
2947 		break;
2948 	default:
2949 		res = -EINVAL;
2950 	}
2951 
2952 	release_sock(sk);
2953 
2954 	if (res)
2955 		return res;	/* "get" failed */
2956 
2957 	if (len < sizeof(value))
2958 		return -EINVAL;
2959 
2960 	if (copy_to_user(ov, &value, sizeof(value)))
2961 		return -EFAULT;
2962 
2963 	return put_user(sizeof(value), ol);
2964 }
2965 
2966 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2967 {
2968 	struct sock *sk = sock->sk;
2969 	struct tipc_sioc_ln_req lnr;
2970 	void __user *argp = (void __user *)arg;
2971 
2972 	switch (cmd) {
2973 	case SIOCGETLINKNAME:
2974 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2975 			return -EFAULT;
2976 		if (!tipc_node_get_linkname(sock_net(sk),
2977 					    lnr.bearer_id & 0xffff, lnr.peer,
2978 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2979 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2980 				return -EFAULT;
2981 			return 0;
2982 		}
2983 		return -EADDRNOTAVAIL;
2984 	default:
2985 		return -ENOIOCTLCMD;
2986 	}
2987 }
2988 
2989 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
2990 {
2991 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
2992 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
2993 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
2994 
2995 	tsk1->peer.family = AF_TIPC;
2996 	tsk1->peer.addrtype = TIPC_ADDR_ID;
2997 	tsk1->peer.scope = TIPC_NODE_SCOPE;
2998 	tsk1->peer.addr.id.ref = tsk2->portid;
2999 	tsk1->peer.addr.id.node = onode;
3000 	tsk2->peer.family = AF_TIPC;
3001 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3002 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3003 	tsk2->peer.addr.id.ref = tsk1->portid;
3004 	tsk2->peer.addr.id.node = onode;
3005 
3006 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3007 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3008 	return 0;
3009 }
3010 
3011 /* Protocol switches for the various types of TIPC sockets */
3012 
3013 static const struct proto_ops msg_ops = {
3014 	.owner		= THIS_MODULE,
3015 	.family		= AF_TIPC,
3016 	.release	= tipc_release,
3017 	.bind		= tipc_bind,
3018 	.connect	= tipc_connect,
3019 	.socketpair	= tipc_socketpair,
3020 	.accept		= sock_no_accept,
3021 	.getname	= tipc_getname,
3022 	.poll		= tipc_poll,
3023 	.ioctl		= tipc_ioctl,
3024 	.listen		= sock_no_listen,
3025 	.shutdown	= tipc_shutdown,
3026 	.setsockopt	= tipc_setsockopt,
3027 	.getsockopt	= tipc_getsockopt,
3028 	.sendmsg	= tipc_sendmsg,
3029 	.recvmsg	= tipc_recvmsg,
3030 	.mmap		= sock_no_mmap,
3031 	.sendpage	= sock_no_sendpage
3032 };
3033 
3034 static const struct proto_ops packet_ops = {
3035 	.owner		= THIS_MODULE,
3036 	.family		= AF_TIPC,
3037 	.release	= tipc_release,
3038 	.bind		= tipc_bind,
3039 	.connect	= tipc_connect,
3040 	.socketpair	= tipc_socketpair,
3041 	.accept		= tipc_accept,
3042 	.getname	= tipc_getname,
3043 	.poll		= tipc_poll,
3044 	.ioctl		= tipc_ioctl,
3045 	.listen		= tipc_listen,
3046 	.shutdown	= tipc_shutdown,
3047 	.setsockopt	= tipc_setsockopt,
3048 	.getsockopt	= tipc_getsockopt,
3049 	.sendmsg	= tipc_send_packet,
3050 	.recvmsg	= tipc_recvmsg,
3051 	.mmap		= sock_no_mmap,
3052 	.sendpage	= sock_no_sendpage
3053 };
3054 
3055 static const struct proto_ops stream_ops = {
3056 	.owner		= THIS_MODULE,
3057 	.family		= AF_TIPC,
3058 	.release	= tipc_release,
3059 	.bind		= tipc_bind,
3060 	.connect	= tipc_connect,
3061 	.socketpair	= tipc_socketpair,
3062 	.accept		= tipc_accept,
3063 	.getname	= tipc_getname,
3064 	.poll		= tipc_poll,
3065 	.ioctl		= tipc_ioctl,
3066 	.listen		= tipc_listen,
3067 	.shutdown	= tipc_shutdown,
3068 	.setsockopt	= tipc_setsockopt,
3069 	.getsockopt	= tipc_getsockopt,
3070 	.sendmsg	= tipc_sendstream,
3071 	.recvmsg	= tipc_recvstream,
3072 	.mmap		= sock_no_mmap,
3073 	.sendpage	= sock_no_sendpage
3074 };
3075 
3076 static const struct net_proto_family tipc_family_ops = {
3077 	.owner		= THIS_MODULE,
3078 	.family		= AF_TIPC,
3079 	.create		= tipc_sk_create
3080 };
3081 
3082 static struct proto tipc_proto = {
3083 	.name		= "TIPC",
3084 	.owner		= THIS_MODULE,
3085 	.obj_size	= sizeof(struct tipc_sock),
3086 	.sysctl_rmem	= sysctl_tipc_rmem
3087 };
3088 
3089 /**
3090  * tipc_socket_init - initialize TIPC socket interface
3091  *
3092  * Returns 0 on success, errno otherwise
3093  */
3094 int tipc_socket_init(void)
3095 {
3096 	int res;
3097 
3098 	res = proto_register(&tipc_proto, 1);
3099 	if (res) {
3100 		pr_err("Failed to register TIPC protocol type\n");
3101 		goto out;
3102 	}
3103 
3104 	res = sock_register(&tipc_family_ops);
3105 	if (res) {
3106 		pr_err("Failed to register TIPC socket type\n");
3107 		proto_unregister(&tipc_proto);
3108 		goto out;
3109 	}
3110  out:
3111 	return res;
3112 }
3113 
3114 /**
3115  * tipc_socket_stop - stop TIPC socket interface
3116  */
3117 void tipc_socket_stop(void)
3118 {
3119 	sock_unregister(tipc_family_ops.family);
3120 	proto_unregister(&tipc_proto);
3121 }
3122 
3123 /* Caller should hold socket lock for the passed tipc socket. */
3124 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3125 {
3126 	u32 peer_node;
3127 	u32 peer_port;
3128 	struct nlattr *nest;
3129 
3130 	peer_node = tsk_peer_node(tsk);
3131 	peer_port = tsk_peer_port(tsk);
3132 
3133 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3134 
3135 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3136 		goto msg_full;
3137 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3138 		goto msg_full;
3139 
3140 	if (tsk->conn_type != 0) {
3141 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3142 			goto msg_full;
3143 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3144 			goto msg_full;
3145 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3146 			goto msg_full;
3147 	}
3148 	nla_nest_end(skb, nest);
3149 
3150 	return 0;
3151 
3152 msg_full:
3153 	nla_nest_cancel(skb, nest);
3154 
3155 	return -EMSGSIZE;
3156 }
3157 
3158 /* Caller should hold socket lock for the passed tipc socket. */
3159 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3160 			    struct tipc_sock *tsk)
3161 {
3162 	int err;
3163 	void *hdr;
3164 	struct nlattr *attrs;
3165 	struct net *net = sock_net(skb->sk);
3166 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3167 	struct sock *sk = &tsk->sk;
3168 
3169 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3170 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3171 	if (!hdr)
3172 		goto msg_cancel;
3173 
3174 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3175 	if (!attrs)
3176 		goto genlmsg_cancel;
3177 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
3178 		goto attr_msg_cancel;
3179 	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
3180 		goto attr_msg_cancel;
3181 
3182 	if (tipc_sk_connected(sk)) {
3183 		err = __tipc_nl_add_sk_con(skb, tsk);
3184 		if (err)
3185 			goto attr_msg_cancel;
3186 	} else if (!list_empty(&tsk->publications)) {
3187 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3188 			goto attr_msg_cancel;
3189 	}
3190 	nla_nest_end(skb, attrs);
3191 	genlmsg_end(skb, hdr);
3192 
3193 	return 0;
3194 
3195 attr_msg_cancel:
3196 	nla_nest_cancel(skb, attrs);
3197 genlmsg_cancel:
3198 	genlmsg_cancel(skb, hdr);
3199 msg_cancel:
3200 	return -EMSGSIZE;
3201 }
3202 
3203 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3204 {
3205 	int err;
3206 	struct tipc_sock *tsk;
3207 	const struct bucket_table *tbl;
3208 	struct rhash_head *pos;
3209 	struct net *net = sock_net(skb->sk);
3210 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3211 	u32 tbl_id = cb->args[0];
3212 	u32 prev_portid = cb->args[1];
3213 
3214 	rcu_read_lock();
3215 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3216 	for (; tbl_id < tbl->size; tbl_id++) {
3217 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3218 			spin_lock_bh(&tsk->sk.sk_lock.slock);
3219 			if (prev_portid && prev_portid != tsk->portid) {
3220 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3221 				continue;
3222 			}
3223 
3224 			err = __tipc_nl_add_sk(skb, cb, tsk);
3225 			if (err) {
3226 				prev_portid = tsk->portid;
3227 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
3228 				goto out;
3229 			}
3230 			prev_portid = 0;
3231 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
3232 		}
3233 	}
3234 out:
3235 	rcu_read_unlock();
3236 	cb->args[0] = tbl_id;
3237 	cb->args[1] = prev_portid;
3238 
3239 	return skb->len;
3240 }
3241 
3242 /* Caller should hold socket lock for the passed tipc socket. */
3243 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3244 				 struct netlink_callback *cb,
3245 				 struct publication *publ)
3246 {
3247 	void *hdr;
3248 	struct nlattr *attrs;
3249 
3250 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3251 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3252 	if (!hdr)
3253 		goto msg_cancel;
3254 
3255 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3256 	if (!attrs)
3257 		goto genlmsg_cancel;
3258 
3259 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3260 		goto attr_msg_cancel;
3261 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3262 		goto attr_msg_cancel;
3263 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3264 		goto attr_msg_cancel;
3265 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3266 		goto attr_msg_cancel;
3267 
3268 	nla_nest_end(skb, attrs);
3269 	genlmsg_end(skb, hdr);
3270 
3271 	return 0;
3272 
3273 attr_msg_cancel:
3274 	nla_nest_cancel(skb, attrs);
3275 genlmsg_cancel:
3276 	genlmsg_cancel(skb, hdr);
3277 msg_cancel:
3278 	return -EMSGSIZE;
3279 }
3280 
3281 /* Caller should hold socket lock for the passed tipc socket. */
3282 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3283 				  struct netlink_callback *cb,
3284 				  struct tipc_sock *tsk, u32 *last_publ)
3285 {
3286 	int err;
3287 	struct publication *p;
3288 
3289 	if (*last_publ) {
3290 		list_for_each_entry(p, &tsk->publications, pport_list) {
3291 			if (p->key == *last_publ)
3292 				break;
3293 		}
3294 		if (p->key != *last_publ) {
3295 			/* We never set seq or call nl_dump_check_consistent()
3296 			 * this means that setting prev_seq here will cause the
3297 			 * consistence check to fail in the netlink callback
3298 			 * handler. Resulting in the last NLMSG_DONE message
3299 			 * having the NLM_F_DUMP_INTR flag set.
3300 			 */
3301 			cb->prev_seq = 1;
3302 			*last_publ = 0;
3303 			return -EPIPE;
3304 		}
3305 	} else {
3306 		p = list_first_entry(&tsk->publications, struct publication,
3307 				     pport_list);
3308 	}
3309 
3310 	list_for_each_entry_from(p, &tsk->publications, pport_list) {
3311 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3312 		if (err) {
3313 			*last_publ = p->key;
3314 			return err;
3315 		}
3316 	}
3317 	*last_publ = 0;
3318 
3319 	return 0;
3320 }
3321 
3322 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3323 {
3324 	int err;
3325 	u32 tsk_portid = cb->args[0];
3326 	u32 last_publ = cb->args[1];
3327 	u32 done = cb->args[2];
3328 	struct net *net = sock_net(skb->sk);
3329 	struct tipc_sock *tsk;
3330 
3331 	if (!tsk_portid) {
3332 		struct nlattr **attrs;
3333 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3334 
3335 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
3336 		if (err)
3337 			return err;
3338 
3339 		if (!attrs[TIPC_NLA_SOCK])
3340 			return -EINVAL;
3341 
3342 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3343 				       attrs[TIPC_NLA_SOCK],
3344 				       tipc_nl_sock_policy, NULL);
3345 		if (err)
3346 			return err;
3347 
3348 		if (!sock[TIPC_NLA_SOCK_REF])
3349 			return -EINVAL;
3350 
3351 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3352 	}
3353 
3354 	if (done)
3355 		return 0;
3356 
3357 	tsk = tipc_sk_lookup(net, tsk_portid);
3358 	if (!tsk)
3359 		return -EINVAL;
3360 
3361 	lock_sock(&tsk->sk);
3362 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3363 	if (!err)
3364 		done = 1;
3365 	release_sock(&tsk->sk);
3366 	sock_put(&tsk->sk);
3367 
3368 	cb->args[0] = tsk_portid;
3369 	cb->args[1] = last_publ;
3370 	cb->args[2] = done;
3371 
3372 	return skb->len;
3373 }
3374