xref: /openbmc/linux/net/tipc/socket.c (revision 715f23b6)
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 #include "trace.h"
50 
51 #define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
52 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
53 #define TIPC_FWD_MSG		1
54 #define TIPC_MAX_PORT		0xffffffff
55 #define TIPC_MIN_PORT		1
56 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of of rcv window size */
57 
58 enum {
59 	TIPC_LISTEN = TCP_LISTEN,
60 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
61 	TIPC_OPEN = TCP_CLOSE,
62 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
63 	TIPC_CONNECTING = TCP_SYN_SENT,
64 };
65 
66 struct sockaddr_pair {
67 	struct sockaddr_tipc sock;
68 	struct sockaddr_tipc member;
69 };
70 
71 /**
72  * struct tipc_sock - TIPC socket structure
73  * @sk: socket - interacts with 'port' and with user via the socket API
74  * @conn_type: TIPC type used when connection was established
75  * @conn_instance: TIPC instance used when connection was established
76  * @published: non-zero if port has one or more associated names
77  * @max_pkt: maximum packet size "hint" used when building messages sent by port
78  * @maxnagle: maximum size of msg which can be subject to nagle
79  * @portid: unique port identity in TIPC socket hash table
80  * @phdr: preformatted message header used when sending messages
81  * #cong_links: list of congested links
82  * @publications: list of publications for port
83  * @blocking_link: address of the congested link we are currently sleeping on
84  * @pub_count: total # of publications port has made during its lifetime
85  * @conn_timeout: the time we can wait for an unresponded setup request
86  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87  * @cong_link_cnt: number of congested links
88  * @snt_unacked: # messages sent by socket, and not yet acked by peer
89  * @rcv_unacked: # messages read by user, but not yet acked back to peer
90  * @peer: 'connected' peer for dgram/rdm
91  * @node: hash table node
92  * @mc_method: cookie for use between socket and broadcast layer
93  * @rcu: rcu struct for tipc_sock
94  */
95 struct tipc_sock {
96 	struct sock sk;
97 	u32 conn_type;
98 	u32 conn_instance;
99 	int published;
100 	u32 max_pkt;
101 	u32 maxnagle;
102 	u32 portid;
103 	struct tipc_msg phdr;
104 	struct list_head cong_links;
105 	struct list_head publications;
106 	u32 pub_count;
107 	atomic_t dupl_rcvcnt;
108 	u16 conn_timeout;
109 	bool probe_unacked;
110 	u16 cong_link_cnt;
111 	u16 snt_unacked;
112 	u16 snd_win;
113 	u16 peer_caps;
114 	u16 rcv_unacked;
115 	u16 rcv_win;
116 	struct sockaddr_tipc peer;
117 	struct rhash_head node;
118 	struct tipc_mc_method mc_method;
119 	struct rcu_head rcu;
120 	struct tipc_group *group;
121 	u32 oneway;
122 	u16 snd_backlog;
123 	bool expect_ack;
124 	bool nodelay;
125 	bool group_is_open;
126 };
127 
128 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
129 static void tipc_data_ready(struct sock *sk);
130 static void tipc_write_space(struct sock *sk);
131 static void tipc_sock_destruct(struct sock *sk);
132 static int tipc_release(struct socket *sock);
133 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
134 		       bool kern);
135 static void tipc_sk_timeout(struct timer_list *t);
136 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
137 			   struct tipc_name_seq const *seq);
138 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
139 			    struct tipc_name_seq const *seq);
140 static int tipc_sk_leave(struct tipc_sock *tsk);
141 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
142 static int tipc_sk_insert(struct tipc_sock *tsk);
143 static void tipc_sk_remove(struct tipc_sock *tsk);
144 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
145 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
146 static void tipc_sk_push_backlog(struct tipc_sock *tsk);
147 
148 static const struct proto_ops packet_ops;
149 static const struct proto_ops stream_ops;
150 static const struct proto_ops msg_ops;
151 static struct proto tipc_proto;
152 static const struct rhashtable_params tsk_rht_params;
153 
154 static u32 tsk_own_node(struct tipc_sock *tsk)
155 {
156 	return msg_prevnode(&tsk->phdr);
157 }
158 
159 static u32 tsk_peer_node(struct tipc_sock *tsk)
160 {
161 	return msg_destnode(&tsk->phdr);
162 }
163 
164 static u32 tsk_peer_port(struct tipc_sock *tsk)
165 {
166 	return msg_destport(&tsk->phdr);
167 }
168 
169 static  bool tsk_unreliable(struct tipc_sock *tsk)
170 {
171 	return msg_src_droppable(&tsk->phdr) != 0;
172 }
173 
174 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
175 {
176 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
177 }
178 
179 static bool tsk_unreturnable(struct tipc_sock *tsk)
180 {
181 	return msg_dest_droppable(&tsk->phdr) != 0;
182 }
183 
184 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
185 {
186 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
187 }
188 
189 static int tsk_importance(struct tipc_sock *tsk)
190 {
191 	return msg_importance(&tsk->phdr);
192 }
193 
194 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
195 {
196 	if (imp > TIPC_CRITICAL_IMPORTANCE)
197 		return -EINVAL;
198 	msg_set_importance(&tsk->phdr, (u32)imp);
199 	return 0;
200 }
201 
202 static struct tipc_sock *tipc_sk(const struct sock *sk)
203 {
204 	return container_of(sk, struct tipc_sock, sk);
205 }
206 
207 static bool tsk_conn_cong(struct tipc_sock *tsk)
208 {
209 	return tsk->snt_unacked > tsk->snd_win;
210 }
211 
212 static u16 tsk_blocks(int len)
213 {
214 	return ((len / FLOWCTL_BLK_SZ) + 1);
215 }
216 
217 /* tsk_blocks(): translate a buffer size in bytes to number of
218  * advertisable blocks, taking into account the ratio truesize(len)/len
219  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
220  */
221 static u16 tsk_adv_blocks(int len)
222 {
223 	return len / FLOWCTL_BLK_SZ / 4;
224 }
225 
226 /* tsk_inc(): increment counter for sent or received data
227  * - If block based flow control is not supported by peer we
228  *   fall back to message based ditto, incrementing the counter
229  */
230 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
231 {
232 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
233 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
234 	return 1;
235 }
236 
237 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
238  */
239 static void tsk_set_nagle(struct tipc_sock *tsk)
240 {
241 	struct sock *sk = &tsk->sk;
242 
243 	tsk->maxnagle = 0;
244 	if (sk->sk_type != SOCK_STREAM)
245 		return;
246 	if (tsk->nodelay)
247 		return;
248 	if (!(tsk->peer_caps & TIPC_NAGLE))
249 		return;
250 	/* Limit node local buffer size to avoid receive queue overflow */
251 	if (tsk->max_pkt == MAX_MSG_SIZE)
252 		tsk->maxnagle = 1500;
253 	else
254 		tsk->maxnagle = tsk->max_pkt;
255 }
256 
257 /**
258  * tsk_advance_rx_queue - discard first buffer in socket receive queue
259  *
260  * Caller must hold socket lock
261  */
262 static void tsk_advance_rx_queue(struct sock *sk)
263 {
264 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
265 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
266 }
267 
268 /* tipc_sk_respond() : send response message back to sender
269  */
270 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
271 {
272 	u32 selector;
273 	u32 dnode;
274 	u32 onode = tipc_own_addr(sock_net(sk));
275 
276 	if (!tipc_msg_reverse(onode, &skb, err))
277 		return;
278 
279 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
280 	dnode = msg_destnode(buf_msg(skb));
281 	selector = msg_origport(buf_msg(skb));
282 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
283 }
284 
285 /**
286  * tsk_rej_rx_queue - reject all buffers in socket receive queue
287  *
288  * Caller must hold socket lock
289  */
290 static void tsk_rej_rx_queue(struct sock *sk)
291 {
292 	struct sk_buff *skb;
293 
294 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
295 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
296 }
297 
298 static bool tipc_sk_connected(struct sock *sk)
299 {
300 	return sk->sk_state == TIPC_ESTABLISHED;
301 }
302 
303 /* tipc_sk_type_connectionless - check if the socket is datagram socket
304  * @sk: socket
305  *
306  * Returns true if connection less, false otherwise
307  */
308 static bool tipc_sk_type_connectionless(struct sock *sk)
309 {
310 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
311 }
312 
313 /* tsk_peer_msg - verify if message was sent by connected port's peer
314  *
315  * Handles cases where the node's network address has changed from
316  * the default of <0.0.0> to its configured setting.
317  */
318 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
319 {
320 	struct sock *sk = &tsk->sk;
321 	u32 self = tipc_own_addr(sock_net(sk));
322 	u32 peer_port = tsk_peer_port(tsk);
323 	u32 orig_node, peer_node;
324 
325 	if (unlikely(!tipc_sk_connected(sk)))
326 		return false;
327 
328 	if (unlikely(msg_origport(msg) != peer_port))
329 		return false;
330 
331 	orig_node = msg_orignode(msg);
332 	peer_node = tsk_peer_node(tsk);
333 
334 	if (likely(orig_node == peer_node))
335 		return true;
336 
337 	if (!orig_node && peer_node == self)
338 		return true;
339 
340 	if (!peer_node && orig_node == self)
341 		return true;
342 
343 	return false;
344 }
345 
346 /* tipc_set_sk_state - set the sk_state of the socket
347  * @sk: socket
348  *
349  * Caller must hold socket lock
350  *
351  * Returns 0 on success, errno otherwise
352  */
353 static int tipc_set_sk_state(struct sock *sk, int state)
354 {
355 	int oldsk_state = sk->sk_state;
356 	int res = -EINVAL;
357 
358 	switch (state) {
359 	case TIPC_OPEN:
360 		res = 0;
361 		break;
362 	case TIPC_LISTEN:
363 	case TIPC_CONNECTING:
364 		if (oldsk_state == TIPC_OPEN)
365 			res = 0;
366 		break;
367 	case TIPC_ESTABLISHED:
368 		if (oldsk_state == TIPC_CONNECTING ||
369 		    oldsk_state == TIPC_OPEN)
370 			res = 0;
371 		break;
372 	case TIPC_DISCONNECTING:
373 		if (oldsk_state == TIPC_CONNECTING ||
374 		    oldsk_state == TIPC_ESTABLISHED)
375 			res = 0;
376 		break;
377 	}
378 
379 	if (!res)
380 		sk->sk_state = state;
381 
382 	return res;
383 }
384 
385 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
386 {
387 	struct sock *sk = sock->sk;
388 	int err = sock_error(sk);
389 	int typ = sock->type;
390 
391 	if (err)
392 		return err;
393 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
394 		if (sk->sk_state == TIPC_DISCONNECTING)
395 			return -EPIPE;
396 		else if (!tipc_sk_connected(sk))
397 			return -ENOTCONN;
398 	}
399 	if (!*timeout)
400 		return -EAGAIN;
401 	if (signal_pending(current))
402 		return sock_intr_errno(*timeout);
403 
404 	return 0;
405 }
406 
407 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
408 ({                                                                             \
409 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
410 	struct sock *sk_;						       \
411 	int rc_;							       \
412 									       \
413 	while ((rc_ = !(condition_))) {					       \
414 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
415 		smp_rmb();                                                     \
416 		sk_ = (sock_)->sk;					       \
417 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
418 		if (rc_)						       \
419 			break;						       \
420 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
421 		release_sock(sk_);					       \
422 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
423 		sched_annotate_sleep();				               \
424 		lock_sock(sk_);						       \
425 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
426 	}								       \
427 	rc_;								       \
428 })
429 
430 /**
431  * tipc_sk_create - create a TIPC socket
432  * @net: network namespace (must be default network)
433  * @sock: pre-allocated socket structure
434  * @protocol: protocol indicator (must be 0)
435  * @kern: caused by kernel or by userspace?
436  *
437  * This routine creates additional data structures used by the TIPC socket,
438  * initializes them, and links them together.
439  *
440  * Returns 0 on success, errno otherwise
441  */
442 static int tipc_sk_create(struct net *net, struct socket *sock,
443 			  int protocol, int kern)
444 {
445 	const struct proto_ops *ops;
446 	struct sock *sk;
447 	struct tipc_sock *tsk;
448 	struct tipc_msg *msg;
449 
450 	/* Validate arguments */
451 	if (unlikely(protocol != 0))
452 		return -EPROTONOSUPPORT;
453 
454 	switch (sock->type) {
455 	case SOCK_STREAM:
456 		ops = &stream_ops;
457 		break;
458 	case SOCK_SEQPACKET:
459 		ops = &packet_ops;
460 		break;
461 	case SOCK_DGRAM:
462 	case SOCK_RDM:
463 		ops = &msg_ops;
464 		break;
465 	default:
466 		return -EPROTOTYPE;
467 	}
468 
469 	/* Allocate socket's protocol area */
470 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
471 	if (sk == NULL)
472 		return -ENOMEM;
473 
474 	tsk = tipc_sk(sk);
475 	tsk->max_pkt = MAX_PKT_DEFAULT;
476 	tsk->maxnagle = 0;
477 	INIT_LIST_HEAD(&tsk->publications);
478 	INIT_LIST_HEAD(&tsk->cong_links);
479 	msg = &tsk->phdr;
480 
481 	/* Finish initializing socket data structures */
482 	sock->ops = ops;
483 	sock_init_data(sock, sk);
484 	tipc_set_sk_state(sk, TIPC_OPEN);
485 	if (tipc_sk_insert(tsk)) {
486 		pr_warn("Socket create failed; port number exhausted\n");
487 		return -EINVAL;
488 	}
489 
490 	/* Ensure tsk is visible before we read own_addr. */
491 	smp_mb();
492 
493 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
494 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
495 
496 	msg_set_origport(msg, tsk->portid);
497 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
498 	sk->sk_shutdown = 0;
499 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
500 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
501 	sk->sk_data_ready = tipc_data_ready;
502 	sk->sk_write_space = tipc_write_space;
503 	sk->sk_destruct = tipc_sock_destruct;
504 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
505 	tsk->group_is_open = true;
506 	atomic_set(&tsk->dupl_rcvcnt, 0);
507 
508 	/* Start out with safe limits until we receive an advertised window */
509 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
510 	tsk->rcv_win = tsk->snd_win;
511 
512 	if (tipc_sk_type_connectionless(sk)) {
513 		tsk_set_unreturnable(tsk, true);
514 		if (sock->type == SOCK_DGRAM)
515 			tsk_set_unreliable(tsk, true);
516 	}
517 	__skb_queue_head_init(&tsk->mc_method.deferredq);
518 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
519 	return 0;
520 }
521 
522 static void tipc_sk_callback(struct rcu_head *head)
523 {
524 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
525 
526 	sock_put(&tsk->sk);
527 }
528 
529 /* Caller should hold socket lock for the socket. */
530 static void __tipc_shutdown(struct socket *sock, int error)
531 {
532 	struct sock *sk = sock->sk;
533 	struct tipc_sock *tsk = tipc_sk(sk);
534 	struct net *net = sock_net(sk);
535 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
536 	u32 dnode = tsk_peer_node(tsk);
537 	struct sk_buff *skb;
538 
539 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
540 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
541 					    !tsk_conn_cong(tsk)));
542 
543 	/* Push out delayed messages if in Nagle mode */
544 	tipc_sk_push_backlog(tsk);
545 	/* Remove pending SYN */
546 	__skb_queue_purge(&sk->sk_write_queue);
547 
548 	/* Reject all unreceived messages, except on an active connection
549 	 * (which disconnects locally & sends a 'FIN+' to peer).
550 	 */
551 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
552 		if (TIPC_SKB_CB(skb)->bytes_read) {
553 			kfree_skb(skb);
554 			continue;
555 		}
556 		if (!tipc_sk_type_connectionless(sk) &&
557 		    sk->sk_state != TIPC_DISCONNECTING) {
558 			tipc_set_sk_state(sk, TIPC_DISCONNECTING);
559 			tipc_node_remove_conn(net, dnode, tsk->portid);
560 		}
561 		tipc_sk_respond(sk, skb, error);
562 	}
563 
564 	if (tipc_sk_type_connectionless(sk))
565 		return;
566 
567 	if (sk->sk_state != TIPC_DISCONNECTING) {
568 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
569 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
570 				      tsk_own_node(tsk), tsk_peer_port(tsk),
571 				      tsk->portid, error);
572 		if (skb)
573 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
574 		tipc_node_remove_conn(net, dnode, tsk->portid);
575 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
576 	}
577 }
578 
579 /**
580  * tipc_release - destroy a TIPC socket
581  * @sock: socket to destroy
582  *
583  * This routine cleans up any messages that are still queued on the socket.
584  * For DGRAM and RDM socket types, all queued messages are rejected.
585  * For SEQPACKET and STREAM socket types, the first message is rejected
586  * and any others are discarded.  (If the first message on a STREAM socket
587  * is partially-read, it is discarded and the next one is rejected instead.)
588  *
589  * NOTE: Rejected messages are not necessarily returned to the sender!  They
590  * are returned or discarded according to the "destination droppable" setting
591  * specified for the message by the sender.
592  *
593  * Returns 0 on success, errno otherwise
594  */
595 static int tipc_release(struct socket *sock)
596 {
597 	struct sock *sk = sock->sk;
598 	struct tipc_sock *tsk;
599 
600 	/*
601 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
602 	 * releases a pre-allocated child socket that was never used)
603 	 */
604 	if (sk == NULL)
605 		return 0;
606 
607 	tsk = tipc_sk(sk);
608 	lock_sock(sk);
609 
610 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
611 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
612 	sk->sk_shutdown = SHUTDOWN_MASK;
613 	tipc_sk_leave(tsk);
614 	tipc_sk_withdraw(tsk, 0, NULL);
615 	__skb_queue_purge(&tsk->mc_method.deferredq);
616 	sk_stop_timer(sk, &sk->sk_timer);
617 	tipc_sk_remove(tsk);
618 
619 	sock_orphan(sk);
620 	/* Reject any messages that accumulated in backlog queue */
621 	release_sock(sk);
622 	tipc_dest_list_purge(&tsk->cong_links);
623 	tsk->cong_link_cnt = 0;
624 	call_rcu(&tsk->rcu, tipc_sk_callback);
625 	sock->sk = NULL;
626 
627 	return 0;
628 }
629 
630 /**
631  * tipc_bind - associate or disassocate TIPC name(s) with a socket
632  * @sock: socket structure
633  * @uaddr: socket address describing name(s) and desired operation
634  * @uaddr_len: size of socket address data structure
635  *
636  * Name and name sequence binding is indicated using a positive scope value;
637  * a negative scope value unbinds the specified name.  Specifying no name
638  * (i.e. a socket address length of 0) unbinds all names from the socket.
639  *
640  * Returns 0 on success, errno otherwise
641  *
642  * NOTE: This routine doesn't need to take the socket lock since it doesn't
643  *       access any non-constant socket information.
644  */
645 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
646 		     int uaddr_len)
647 {
648 	struct sock *sk = sock->sk;
649 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
650 	struct tipc_sock *tsk = tipc_sk(sk);
651 	int res = -EINVAL;
652 
653 	lock_sock(sk);
654 	if (unlikely(!uaddr_len)) {
655 		res = tipc_sk_withdraw(tsk, 0, NULL);
656 		goto exit;
657 	}
658 	if (tsk->group) {
659 		res = -EACCES;
660 		goto exit;
661 	}
662 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
663 		res = -EINVAL;
664 		goto exit;
665 	}
666 	if (addr->family != AF_TIPC) {
667 		res = -EAFNOSUPPORT;
668 		goto exit;
669 	}
670 
671 	if (addr->addrtype == TIPC_ADDR_NAME)
672 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
673 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
674 		res = -EAFNOSUPPORT;
675 		goto exit;
676 	}
677 
678 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
679 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
680 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
681 		res = -EACCES;
682 		goto exit;
683 	}
684 
685 	res = (addr->scope >= 0) ?
686 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
687 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
688 exit:
689 	release_sock(sk);
690 	return res;
691 }
692 
693 /**
694  * tipc_getname - get port ID of socket or peer socket
695  * @sock: socket structure
696  * @uaddr: area for returned socket address
697  * @uaddr_len: area for returned length of socket address
698  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
699  *
700  * Returns 0 on success, errno otherwise
701  *
702  * NOTE: This routine doesn't need to take the socket lock since it only
703  *       accesses socket information that is unchanging (or which changes in
704  *       a completely predictable manner).
705  */
706 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
707 			int peer)
708 {
709 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
710 	struct sock *sk = sock->sk;
711 	struct tipc_sock *tsk = tipc_sk(sk);
712 
713 	memset(addr, 0, sizeof(*addr));
714 	if (peer) {
715 		if ((!tipc_sk_connected(sk)) &&
716 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
717 			return -ENOTCONN;
718 		addr->addr.id.ref = tsk_peer_port(tsk);
719 		addr->addr.id.node = tsk_peer_node(tsk);
720 	} else {
721 		addr->addr.id.ref = tsk->portid;
722 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
723 	}
724 
725 	addr->addrtype = TIPC_ADDR_ID;
726 	addr->family = AF_TIPC;
727 	addr->scope = 0;
728 	addr->addr.name.domain = 0;
729 
730 	return sizeof(*addr);
731 }
732 
733 /**
734  * tipc_poll - read and possibly block on pollmask
735  * @file: file structure associated with the socket
736  * @sock: socket for which to calculate the poll bits
737  * @wait: ???
738  *
739  * Returns pollmask value
740  *
741  * COMMENTARY:
742  * It appears that the usual socket locking mechanisms are not useful here
743  * since the pollmask info is potentially out-of-date the moment this routine
744  * exits.  TCP and other protocols seem to rely on higher level poll routines
745  * to handle any preventable race conditions, so TIPC will do the same ...
746  *
747  * IMPORTANT: The fact that a read or write operation is indicated does NOT
748  * imply that the operation will succeed, merely that it should be performed
749  * and will not block.
750  */
751 static __poll_t tipc_poll(struct file *file, struct socket *sock,
752 			      poll_table *wait)
753 {
754 	struct sock *sk = sock->sk;
755 	struct tipc_sock *tsk = tipc_sk(sk);
756 	__poll_t revents = 0;
757 
758 	sock_poll_wait(file, sock, wait);
759 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
760 
761 	if (sk->sk_shutdown & RCV_SHUTDOWN)
762 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
763 	if (sk->sk_shutdown == SHUTDOWN_MASK)
764 		revents |= EPOLLHUP;
765 
766 	switch (sk->sk_state) {
767 	case TIPC_ESTABLISHED:
768 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
769 			revents |= EPOLLOUT;
770 		/* fall through */
771 	case TIPC_LISTEN:
772 	case TIPC_CONNECTING:
773 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
774 			revents |= EPOLLIN | EPOLLRDNORM;
775 		break;
776 	case TIPC_OPEN:
777 		if (tsk->group_is_open && !tsk->cong_link_cnt)
778 			revents |= EPOLLOUT;
779 		if (!tipc_sk_type_connectionless(sk))
780 			break;
781 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
782 			break;
783 		revents |= EPOLLIN | EPOLLRDNORM;
784 		break;
785 	case TIPC_DISCONNECTING:
786 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
787 		break;
788 	}
789 	return revents;
790 }
791 
792 /**
793  * tipc_sendmcast - send multicast message
794  * @sock: socket structure
795  * @seq: destination address
796  * @msg: message to send
797  * @dlen: length of data to send
798  * @timeout: timeout to wait for wakeup
799  *
800  * Called from function tipc_sendmsg(), which has done all sanity checks
801  * Returns the number of bytes sent on success, or errno
802  */
803 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
804 			  struct msghdr *msg, size_t dlen, long timeout)
805 {
806 	struct sock *sk = sock->sk;
807 	struct tipc_sock *tsk = tipc_sk(sk);
808 	struct tipc_msg *hdr = &tsk->phdr;
809 	struct net *net = sock_net(sk);
810 	int mtu = tipc_bcast_get_mtu(net);
811 	struct tipc_mc_method *method = &tsk->mc_method;
812 	struct sk_buff_head pkts;
813 	struct tipc_nlist dsts;
814 	int rc;
815 
816 	if (tsk->group)
817 		return -EACCES;
818 
819 	/* Block or return if any destination link is congested */
820 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
821 	if (unlikely(rc))
822 		return rc;
823 
824 	/* Lookup destination nodes */
825 	tipc_nlist_init(&dsts, tipc_own_addr(net));
826 	tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
827 				      seq->upper, &dsts);
828 	if (!dsts.local && !dsts.remote)
829 		return -EHOSTUNREACH;
830 
831 	/* Build message header */
832 	msg_set_type(hdr, TIPC_MCAST_MSG);
833 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
834 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
835 	msg_set_destport(hdr, 0);
836 	msg_set_destnode(hdr, 0);
837 	msg_set_nametype(hdr, seq->type);
838 	msg_set_namelower(hdr, seq->lower);
839 	msg_set_nameupper(hdr, seq->upper);
840 
841 	/* Build message as chain of buffers */
842 	__skb_queue_head_init(&pkts);
843 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
844 
845 	/* Send message if build was successful */
846 	if (unlikely(rc == dlen)) {
847 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
848 					TIPC_DUMP_SK_SNDQ, " ");
849 		rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
850 				     &tsk->cong_link_cnt);
851 	}
852 
853 	tipc_nlist_purge(&dsts);
854 
855 	return rc ? rc : dlen;
856 }
857 
858 /**
859  * tipc_send_group_msg - send a message to a member in the group
860  * @net: network namespace
861  * @m: message to send
862  * @mb: group member
863  * @dnode: destination node
864  * @dport: destination port
865  * @dlen: total length of message data
866  */
867 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
868 			       struct msghdr *m, struct tipc_member *mb,
869 			       u32 dnode, u32 dport, int dlen)
870 {
871 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
872 	struct tipc_mc_method *method = &tsk->mc_method;
873 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
874 	struct tipc_msg *hdr = &tsk->phdr;
875 	struct sk_buff_head pkts;
876 	int mtu, rc;
877 
878 	/* Complete message header */
879 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
880 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
881 	msg_set_destport(hdr, dport);
882 	msg_set_destnode(hdr, dnode);
883 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
884 
885 	/* Build message as chain of buffers */
886 	__skb_queue_head_init(&pkts);
887 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
888 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
889 	if (unlikely(rc != dlen))
890 		return rc;
891 
892 	/* Send message */
893 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
894 	if (unlikely(rc == -ELINKCONG)) {
895 		tipc_dest_push(&tsk->cong_links, dnode, 0);
896 		tsk->cong_link_cnt++;
897 	}
898 
899 	/* Update send window */
900 	tipc_group_update_member(mb, blks);
901 
902 	/* A broadcast sent within next EXPIRE period must follow same path */
903 	method->rcast = true;
904 	method->mandatory = true;
905 	return dlen;
906 }
907 
908 /**
909  * tipc_send_group_unicast - send message to a member in the group
910  * @sock: socket structure
911  * @m: message to send
912  * @dlen: total length of message data
913  * @timeout: timeout to wait for wakeup
914  *
915  * Called from function tipc_sendmsg(), which has done all sanity checks
916  * Returns the number of bytes sent on success, or errno
917  */
918 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
919 				   int dlen, long timeout)
920 {
921 	struct sock *sk = sock->sk;
922 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
923 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
924 	struct tipc_sock *tsk = tipc_sk(sk);
925 	struct net *net = sock_net(sk);
926 	struct tipc_member *mb = NULL;
927 	u32 node, port;
928 	int rc;
929 
930 	node = dest->addr.id.node;
931 	port = dest->addr.id.ref;
932 	if (!port && !node)
933 		return -EHOSTUNREACH;
934 
935 	/* Block or return if destination link or member is congested */
936 	rc = tipc_wait_for_cond(sock, &timeout,
937 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
938 				tsk->group &&
939 				!tipc_group_cong(tsk->group, node, port, blks,
940 						 &mb));
941 	if (unlikely(rc))
942 		return rc;
943 
944 	if (unlikely(!mb))
945 		return -EHOSTUNREACH;
946 
947 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
948 
949 	return rc ? rc : dlen;
950 }
951 
952 /**
953  * tipc_send_group_anycast - send message to any member with given identity
954  * @sock: socket structure
955  * @m: message to send
956  * @dlen: total length of message data
957  * @timeout: timeout to wait for wakeup
958  *
959  * Called from function tipc_sendmsg(), which has done all sanity checks
960  * Returns the number of bytes sent on success, or errno
961  */
962 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
963 				   int dlen, long timeout)
964 {
965 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
966 	struct sock *sk = sock->sk;
967 	struct tipc_sock *tsk = tipc_sk(sk);
968 	struct list_head *cong_links = &tsk->cong_links;
969 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
970 	struct tipc_msg *hdr = &tsk->phdr;
971 	struct tipc_member *first = NULL;
972 	struct tipc_member *mbr = NULL;
973 	struct net *net = sock_net(sk);
974 	u32 node, port, exclude;
975 	struct list_head dsts;
976 	u32 type, inst, scope;
977 	int lookups = 0;
978 	int dstcnt, rc;
979 	bool cong;
980 
981 	INIT_LIST_HEAD(&dsts);
982 
983 	type = msg_nametype(hdr);
984 	inst = dest->addr.name.name.instance;
985 	scope = msg_lookup_scope(hdr);
986 
987 	while (++lookups < 4) {
988 		exclude = tipc_group_exclude(tsk->group);
989 
990 		first = NULL;
991 
992 		/* Look for a non-congested destination member, if any */
993 		while (1) {
994 			if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
995 						 &dstcnt, exclude, false))
996 				return -EHOSTUNREACH;
997 			tipc_dest_pop(&dsts, &node, &port);
998 			cong = tipc_group_cong(tsk->group, node, port, blks,
999 					       &mbr);
1000 			if (!cong)
1001 				break;
1002 			if (mbr == first)
1003 				break;
1004 			if (!first)
1005 				first = mbr;
1006 		}
1007 
1008 		/* Start over if destination was not in member list */
1009 		if (unlikely(!mbr))
1010 			continue;
1011 
1012 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1013 			break;
1014 
1015 		/* Block or return if destination link or member is congested */
1016 		rc = tipc_wait_for_cond(sock, &timeout,
1017 					!tipc_dest_find(cong_links, node, 0) &&
1018 					tsk->group &&
1019 					!tipc_group_cong(tsk->group, node, port,
1020 							 blks, &mbr));
1021 		if (unlikely(rc))
1022 			return rc;
1023 
1024 		/* Send, unless destination disappeared while waiting */
1025 		if (likely(mbr))
1026 			break;
1027 	}
1028 
1029 	if (unlikely(lookups >= 4))
1030 		return -EHOSTUNREACH;
1031 
1032 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1033 
1034 	return rc ? rc : dlen;
1035 }
1036 
1037 /**
1038  * tipc_send_group_bcast - send message to all members in communication group
1039  * @sk: socket structure
1040  * @m: message to send
1041  * @dlen: total length of message data
1042  * @timeout: timeout to wait for wakeup
1043  *
1044  * Called from function tipc_sendmsg(), which has done all sanity checks
1045  * Returns the number of bytes sent on success, or errno
1046  */
1047 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1048 				 int dlen, long timeout)
1049 {
1050 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1051 	struct sock *sk = sock->sk;
1052 	struct net *net = sock_net(sk);
1053 	struct tipc_sock *tsk = tipc_sk(sk);
1054 	struct tipc_nlist *dsts;
1055 	struct tipc_mc_method *method = &tsk->mc_method;
1056 	bool ack = method->mandatory && method->rcast;
1057 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1058 	struct tipc_msg *hdr = &tsk->phdr;
1059 	int mtu = tipc_bcast_get_mtu(net);
1060 	struct sk_buff_head pkts;
1061 	int rc = -EHOSTUNREACH;
1062 
1063 	/* Block or return if any destination link or member is congested */
1064 	rc = tipc_wait_for_cond(sock, &timeout,
1065 				!tsk->cong_link_cnt && tsk->group &&
1066 				!tipc_group_bc_cong(tsk->group, blks));
1067 	if (unlikely(rc))
1068 		return rc;
1069 
1070 	dsts = tipc_group_dests(tsk->group);
1071 	if (!dsts->local && !dsts->remote)
1072 		return -EHOSTUNREACH;
1073 
1074 	/* Complete message header */
1075 	if (dest) {
1076 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1077 		msg_set_nameinst(hdr, dest->addr.name.name.instance);
1078 	} else {
1079 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1080 		msg_set_nameinst(hdr, 0);
1081 	}
1082 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1083 	msg_set_destport(hdr, 0);
1084 	msg_set_destnode(hdr, 0);
1085 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1086 
1087 	/* Avoid getting stuck with repeated forced replicasts */
1088 	msg_set_grp_bc_ack_req(hdr, ack);
1089 
1090 	/* Build message as chain of buffers */
1091 	__skb_queue_head_init(&pkts);
1092 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1093 	if (unlikely(rc != dlen))
1094 		return rc;
1095 
1096 	/* Send message */
1097 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1098 	if (unlikely(rc))
1099 		return rc;
1100 
1101 	/* Update broadcast sequence number and send windows */
1102 	tipc_group_update_bc_members(tsk->group, blks, ack);
1103 
1104 	/* Broadcast link is now free to choose method for next broadcast */
1105 	method->mandatory = false;
1106 	method->expires = jiffies;
1107 
1108 	return dlen;
1109 }
1110 
1111 /**
1112  * tipc_send_group_mcast - send message to all members with given identity
1113  * @sock: socket structure
1114  * @m: message to send
1115  * @dlen: total length of message data
1116  * @timeout: timeout to wait for wakeup
1117  *
1118  * Called from function tipc_sendmsg(), which has done all sanity checks
1119  * Returns the number of bytes sent on success, or errno
1120  */
1121 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1122 				 int dlen, long timeout)
1123 {
1124 	struct sock *sk = sock->sk;
1125 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1126 	struct tipc_sock *tsk = tipc_sk(sk);
1127 	struct tipc_group *grp = tsk->group;
1128 	struct tipc_msg *hdr = &tsk->phdr;
1129 	struct net *net = sock_net(sk);
1130 	u32 type, inst, scope, exclude;
1131 	struct list_head dsts;
1132 	u32 dstcnt;
1133 
1134 	INIT_LIST_HEAD(&dsts);
1135 
1136 	type = msg_nametype(hdr);
1137 	inst = dest->addr.name.name.instance;
1138 	scope = msg_lookup_scope(hdr);
1139 	exclude = tipc_group_exclude(grp);
1140 
1141 	if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1142 				 &dstcnt, exclude, true))
1143 		return -EHOSTUNREACH;
1144 
1145 	if (dstcnt == 1) {
1146 		tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1147 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1148 	}
1149 
1150 	tipc_dest_list_purge(&dsts);
1151 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1152 }
1153 
1154 /**
1155  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1156  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1157  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1158  *
1159  * Multi-threaded: parallel calls with reference to same queues may occur
1160  */
1161 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1162 		       struct sk_buff_head *inputq)
1163 {
1164 	u32 self = tipc_own_addr(net);
1165 	u32 type, lower, upper, scope;
1166 	struct sk_buff *skb, *_skb;
1167 	u32 portid, onode;
1168 	struct sk_buff_head tmpq;
1169 	struct list_head dports;
1170 	struct tipc_msg *hdr;
1171 	int user, mtyp, hlen;
1172 	bool exact;
1173 
1174 	__skb_queue_head_init(&tmpq);
1175 	INIT_LIST_HEAD(&dports);
1176 
1177 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1178 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1179 		hdr = buf_msg(skb);
1180 		user = msg_user(hdr);
1181 		mtyp = msg_type(hdr);
1182 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1183 		onode = msg_orignode(hdr);
1184 		type = msg_nametype(hdr);
1185 
1186 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1187 			spin_lock_bh(&inputq->lock);
1188 			if (skb_peek(arrvq) == skb) {
1189 				__skb_dequeue(arrvq);
1190 				__skb_queue_tail(inputq, skb);
1191 			}
1192 			kfree_skb(skb);
1193 			spin_unlock_bh(&inputq->lock);
1194 			continue;
1195 		}
1196 
1197 		/* Group messages require exact scope match */
1198 		if (msg_in_group(hdr)) {
1199 			lower = 0;
1200 			upper = ~0;
1201 			scope = msg_lookup_scope(hdr);
1202 			exact = true;
1203 		} else {
1204 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1205 			if (onode == self)
1206 				scope = TIPC_NODE_SCOPE;
1207 			else
1208 				scope = TIPC_CLUSTER_SCOPE;
1209 			exact = false;
1210 			lower = msg_namelower(hdr);
1211 			upper = msg_nameupper(hdr);
1212 		}
1213 
1214 		/* Create destination port list: */
1215 		tipc_nametbl_mc_lookup(net, type, lower, upper,
1216 				       scope, exact, &dports);
1217 
1218 		/* Clone message per destination */
1219 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1220 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1221 			if (_skb) {
1222 				msg_set_destport(buf_msg(_skb), portid);
1223 				__skb_queue_tail(&tmpq, _skb);
1224 				continue;
1225 			}
1226 			pr_warn("Failed to clone mcast rcv buffer\n");
1227 		}
1228 		/* Append to inputq if not already done by other thread */
1229 		spin_lock_bh(&inputq->lock);
1230 		if (skb_peek(arrvq) == skb) {
1231 			skb_queue_splice_tail_init(&tmpq, inputq);
1232 			kfree_skb(__skb_dequeue(arrvq));
1233 		}
1234 		spin_unlock_bh(&inputq->lock);
1235 		__skb_queue_purge(&tmpq);
1236 		kfree_skb(skb);
1237 	}
1238 	tipc_sk_rcv(net, inputq);
1239 }
1240 
1241 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1242  *                         when socket is in Nagle mode
1243  */
1244 static void tipc_sk_push_backlog(struct tipc_sock *tsk)
1245 {
1246 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1247 	struct net *net = sock_net(&tsk->sk);
1248 	u32 dnode = tsk_peer_node(tsk);
1249 	struct sk_buff *skb = skb_peek(txq);
1250 	int rc;
1251 
1252 	if (!skb || tsk->cong_link_cnt)
1253 		return;
1254 
1255 	/* Do not send SYN again after congestion */
1256 	if (msg_is_syn(buf_msg(skb)))
1257 		return;
1258 
1259 	tsk->snt_unacked += tsk->snd_backlog;
1260 	tsk->snd_backlog = 0;
1261 	tsk->expect_ack = true;
1262 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1263 	if (rc == -ELINKCONG)
1264 		tsk->cong_link_cnt = 1;
1265 }
1266 
1267 /**
1268  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1269  * @tsk: receiving socket
1270  * @skb: pointer to message buffer.
1271  */
1272 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1273 				   struct sk_buff_head *inputq,
1274 				   struct sk_buff_head *xmitq)
1275 {
1276 	struct tipc_msg *hdr = buf_msg(skb);
1277 	u32 onode = tsk_own_node(tsk);
1278 	struct sock *sk = &tsk->sk;
1279 	int mtyp = msg_type(hdr);
1280 	bool was_cong;
1281 
1282 	/* Ignore if connection cannot be validated: */
1283 	if (!tsk_peer_msg(tsk, hdr)) {
1284 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1285 		goto exit;
1286 	}
1287 
1288 	if (unlikely(msg_errcode(hdr))) {
1289 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1290 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1291 				      tsk_peer_port(tsk));
1292 		sk->sk_state_change(sk);
1293 
1294 		/* State change is ignored if socket already awake,
1295 		 * - convert msg to abort msg and add to inqueue
1296 		 */
1297 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1298 		msg_set_type(hdr, TIPC_CONN_MSG);
1299 		msg_set_size(hdr, BASIC_H_SIZE);
1300 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1301 		__skb_queue_tail(inputq, skb);
1302 		return;
1303 	}
1304 
1305 	tsk->probe_unacked = false;
1306 
1307 	if (mtyp == CONN_PROBE) {
1308 		msg_set_type(hdr, CONN_PROBE_REPLY);
1309 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1310 			__skb_queue_tail(xmitq, skb);
1311 		return;
1312 	} else if (mtyp == CONN_ACK) {
1313 		was_cong = tsk_conn_cong(tsk);
1314 		tsk->expect_ack = false;
1315 		tipc_sk_push_backlog(tsk);
1316 		tsk->snt_unacked -= msg_conn_ack(hdr);
1317 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1318 			tsk->snd_win = msg_adv_win(hdr);
1319 		if (was_cong && !tsk_conn_cong(tsk))
1320 			sk->sk_write_space(sk);
1321 	} else if (mtyp != CONN_PROBE_REPLY) {
1322 		pr_warn("Received unknown CONN_PROTO msg\n");
1323 	}
1324 exit:
1325 	kfree_skb(skb);
1326 }
1327 
1328 /**
1329  * tipc_sendmsg - send message in connectionless manner
1330  * @sock: socket structure
1331  * @m: message to send
1332  * @dsz: amount of user data to be sent
1333  *
1334  * Message must have an destination specified explicitly.
1335  * Used for SOCK_RDM and SOCK_DGRAM messages,
1336  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1337  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1338  *
1339  * Returns the number of bytes sent on success, or errno otherwise
1340  */
1341 static int tipc_sendmsg(struct socket *sock,
1342 			struct msghdr *m, size_t dsz)
1343 {
1344 	struct sock *sk = sock->sk;
1345 	int ret;
1346 
1347 	lock_sock(sk);
1348 	ret = __tipc_sendmsg(sock, m, dsz);
1349 	release_sock(sk);
1350 
1351 	return ret;
1352 }
1353 
1354 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1355 {
1356 	struct sock *sk = sock->sk;
1357 	struct net *net = sock_net(sk);
1358 	struct tipc_sock *tsk = tipc_sk(sk);
1359 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1360 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1361 	struct list_head *clinks = &tsk->cong_links;
1362 	bool syn = !tipc_sk_type_connectionless(sk);
1363 	struct tipc_group *grp = tsk->group;
1364 	struct tipc_msg *hdr = &tsk->phdr;
1365 	struct tipc_name_seq *seq;
1366 	struct sk_buff_head pkts;
1367 	u32 dport = 0, dnode = 0;
1368 	u32 type = 0, inst = 0;
1369 	int mtu, rc;
1370 
1371 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1372 		return -EMSGSIZE;
1373 
1374 	if (likely(dest)) {
1375 		if (unlikely(m->msg_namelen < sizeof(*dest)))
1376 			return -EINVAL;
1377 		if (unlikely(dest->family != AF_TIPC))
1378 			return -EINVAL;
1379 	}
1380 
1381 	if (grp) {
1382 		if (!dest)
1383 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1384 		if (dest->addrtype == TIPC_ADDR_NAME)
1385 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1386 		if (dest->addrtype == TIPC_ADDR_ID)
1387 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1388 		if (dest->addrtype == TIPC_ADDR_MCAST)
1389 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1390 		return -EINVAL;
1391 	}
1392 
1393 	if (unlikely(!dest)) {
1394 		dest = &tsk->peer;
1395 		if (!syn && dest->family != AF_TIPC)
1396 			return -EDESTADDRREQ;
1397 	}
1398 
1399 	if (unlikely(syn)) {
1400 		if (sk->sk_state == TIPC_LISTEN)
1401 			return -EPIPE;
1402 		if (sk->sk_state != TIPC_OPEN)
1403 			return -EISCONN;
1404 		if (tsk->published)
1405 			return -EOPNOTSUPP;
1406 		if (dest->addrtype == TIPC_ADDR_NAME) {
1407 			tsk->conn_type = dest->addr.name.name.type;
1408 			tsk->conn_instance = dest->addr.name.name.instance;
1409 		}
1410 		msg_set_syn(hdr, 1);
1411 	}
1412 
1413 	seq = &dest->addr.nameseq;
1414 	if (dest->addrtype == TIPC_ADDR_MCAST)
1415 		return tipc_sendmcast(sock, seq, m, dlen, timeout);
1416 
1417 	if (dest->addrtype == TIPC_ADDR_NAME) {
1418 		type = dest->addr.name.name.type;
1419 		inst = dest->addr.name.name.instance;
1420 		dnode = dest->addr.name.domain;
1421 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1422 		if (unlikely(!dport && !dnode))
1423 			return -EHOSTUNREACH;
1424 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1425 		dnode = dest->addr.id.node;
1426 	} else {
1427 		return -EINVAL;
1428 	}
1429 
1430 	/* Block or return if destination link is congested */
1431 	rc = tipc_wait_for_cond(sock, &timeout,
1432 				!tipc_dest_find(clinks, dnode, 0));
1433 	if (unlikely(rc))
1434 		return rc;
1435 
1436 	if (dest->addrtype == TIPC_ADDR_NAME) {
1437 		msg_set_type(hdr, TIPC_NAMED_MSG);
1438 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1439 		msg_set_nametype(hdr, type);
1440 		msg_set_nameinst(hdr, inst);
1441 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1442 		msg_set_destnode(hdr, dnode);
1443 		msg_set_destport(hdr, dport);
1444 	} else { /* TIPC_ADDR_ID */
1445 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1446 		msg_set_lookup_scope(hdr, 0);
1447 		msg_set_destnode(hdr, dnode);
1448 		msg_set_destport(hdr, dest->addr.id.ref);
1449 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1450 	}
1451 
1452 	__skb_queue_head_init(&pkts);
1453 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
1454 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1455 	if (unlikely(rc != dlen))
1456 		return rc;
1457 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1458 		__skb_queue_purge(&pkts);
1459 		return -ENOMEM;
1460 	}
1461 
1462 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1463 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1464 	if (unlikely(rc == -ELINKCONG)) {
1465 		tipc_dest_push(clinks, dnode, 0);
1466 		tsk->cong_link_cnt++;
1467 		rc = 0;
1468 	}
1469 
1470 	if (unlikely(syn && !rc))
1471 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1472 
1473 	return rc ? rc : dlen;
1474 }
1475 
1476 /**
1477  * tipc_sendstream - send stream-oriented data
1478  * @sock: socket structure
1479  * @m: data to send
1480  * @dsz: total length of data to be transmitted
1481  *
1482  * Used for SOCK_STREAM data.
1483  *
1484  * Returns the number of bytes sent on success (or partial success),
1485  * or errno if no data sent
1486  */
1487 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1488 {
1489 	struct sock *sk = sock->sk;
1490 	int ret;
1491 
1492 	lock_sock(sk);
1493 	ret = __tipc_sendstream(sock, m, dsz);
1494 	release_sock(sk);
1495 
1496 	return ret;
1497 }
1498 
1499 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1500 {
1501 	struct sock *sk = sock->sk;
1502 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1503 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1504 	struct sk_buff_head *txq = &sk->sk_write_queue;
1505 	struct tipc_sock *tsk = tipc_sk(sk);
1506 	struct tipc_msg *hdr = &tsk->phdr;
1507 	struct net *net = sock_net(sk);
1508 	u32 dnode = tsk_peer_node(tsk);
1509 	int maxnagle = tsk->maxnagle;
1510 	int maxpkt = tsk->max_pkt;
1511 	int send, sent = 0;
1512 	int blocks, rc = 0;
1513 
1514 	if (unlikely(dlen > INT_MAX))
1515 		return -EMSGSIZE;
1516 
1517 	/* Handle implicit connection setup */
1518 	if (unlikely(dest)) {
1519 		rc = __tipc_sendmsg(sock, m, dlen);
1520 		if (dlen && dlen == rc) {
1521 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1522 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1523 		}
1524 		return rc;
1525 	}
1526 
1527 	do {
1528 		rc = tipc_wait_for_cond(sock, &timeout,
1529 					(!tsk->cong_link_cnt &&
1530 					 !tsk_conn_cong(tsk) &&
1531 					 tipc_sk_connected(sk)));
1532 		if (unlikely(rc))
1533 			break;
1534 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1535 		blocks = tsk->snd_backlog;
1536 		if (tsk->oneway++ >= 4 && send <= maxnagle) {
1537 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1538 			if (unlikely(rc < 0))
1539 				break;
1540 			blocks += rc;
1541 			if (blocks <= 64 && tsk->expect_ack) {
1542 				tsk->snd_backlog = blocks;
1543 				sent += send;
1544 				break;
1545 			}
1546 			tsk->expect_ack = true;
1547 		} else {
1548 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1549 			if (unlikely(rc != send))
1550 				break;
1551 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1552 		}
1553 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1554 					 TIPC_DUMP_SK_SNDQ, " ");
1555 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1556 		if (unlikely(rc == -ELINKCONG)) {
1557 			tsk->cong_link_cnt = 1;
1558 			rc = 0;
1559 		}
1560 		if (likely(!rc)) {
1561 			tsk->snt_unacked += blocks;
1562 			tsk->snd_backlog = 0;
1563 			sent += send;
1564 		}
1565 	} while (sent < dlen && !rc);
1566 
1567 	return sent ? sent : rc;
1568 }
1569 
1570 /**
1571  * tipc_send_packet - send a connection-oriented message
1572  * @sock: socket structure
1573  * @m: message to send
1574  * @dsz: length of data to be transmitted
1575  *
1576  * Used for SOCK_SEQPACKET messages.
1577  *
1578  * Returns the number of bytes sent on success, or errno otherwise
1579  */
1580 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1581 {
1582 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1583 		return -EMSGSIZE;
1584 
1585 	return tipc_sendstream(sock, m, dsz);
1586 }
1587 
1588 /* tipc_sk_finish_conn - complete the setup of a connection
1589  */
1590 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1591 				u32 peer_node)
1592 {
1593 	struct sock *sk = &tsk->sk;
1594 	struct net *net = sock_net(sk);
1595 	struct tipc_msg *msg = &tsk->phdr;
1596 
1597 	msg_set_syn(msg, 0);
1598 	msg_set_destnode(msg, peer_node);
1599 	msg_set_destport(msg, peer_port);
1600 	msg_set_type(msg, TIPC_CONN_MSG);
1601 	msg_set_lookup_scope(msg, 0);
1602 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1603 
1604 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1605 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1606 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1607 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1608 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1609 	tsk_set_nagle(tsk);
1610 	__skb_queue_purge(&sk->sk_write_queue);
1611 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1612 		return;
1613 
1614 	/* Fall back to message based flow control */
1615 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1616 	tsk->snd_win = FLOWCTL_MSG_WIN;
1617 }
1618 
1619 /**
1620  * tipc_sk_set_orig_addr - capture sender's address for received message
1621  * @m: descriptor for message info
1622  * @hdr: received message header
1623  *
1624  * Note: Address is not captured if not requested by receiver.
1625  */
1626 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1627 {
1628 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1629 	struct tipc_msg *hdr = buf_msg(skb);
1630 
1631 	if (!srcaddr)
1632 		return;
1633 
1634 	srcaddr->sock.family = AF_TIPC;
1635 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1636 	srcaddr->sock.scope = 0;
1637 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1638 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1639 	srcaddr->sock.addr.name.domain = 0;
1640 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1641 
1642 	if (!msg_in_group(hdr))
1643 		return;
1644 
1645 	/* Group message users may also want to know sending member's id */
1646 	srcaddr->member.family = AF_TIPC;
1647 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1648 	srcaddr->member.scope = 0;
1649 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1650 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1651 	srcaddr->member.addr.name.domain = 0;
1652 	m->msg_namelen = sizeof(*srcaddr);
1653 }
1654 
1655 /**
1656  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1657  * @m: descriptor for message info
1658  * @skb: received message buffer
1659  * @tsk: TIPC port associated with message
1660  *
1661  * Note: Ancillary data is not captured if not requested by receiver.
1662  *
1663  * Returns 0 if successful, otherwise errno
1664  */
1665 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1666 				 struct tipc_sock *tsk)
1667 {
1668 	struct tipc_msg *msg;
1669 	u32 anc_data[3];
1670 	u32 err;
1671 	u32 dest_type;
1672 	int has_name;
1673 	int res;
1674 
1675 	if (likely(m->msg_controllen == 0))
1676 		return 0;
1677 	msg = buf_msg(skb);
1678 
1679 	/* Optionally capture errored message object(s) */
1680 	err = msg ? msg_errcode(msg) : 0;
1681 	if (unlikely(err)) {
1682 		anc_data[0] = err;
1683 		anc_data[1] = msg_data_sz(msg);
1684 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1685 		if (res)
1686 			return res;
1687 		if (anc_data[1]) {
1688 			if (skb_linearize(skb))
1689 				return -ENOMEM;
1690 			msg = buf_msg(skb);
1691 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1692 				       msg_data(msg));
1693 			if (res)
1694 				return res;
1695 		}
1696 	}
1697 
1698 	/* Optionally capture message destination object */
1699 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1700 	switch (dest_type) {
1701 	case TIPC_NAMED_MSG:
1702 		has_name = 1;
1703 		anc_data[0] = msg_nametype(msg);
1704 		anc_data[1] = msg_namelower(msg);
1705 		anc_data[2] = msg_namelower(msg);
1706 		break;
1707 	case TIPC_MCAST_MSG:
1708 		has_name = 1;
1709 		anc_data[0] = msg_nametype(msg);
1710 		anc_data[1] = msg_namelower(msg);
1711 		anc_data[2] = msg_nameupper(msg);
1712 		break;
1713 	case TIPC_CONN_MSG:
1714 		has_name = (tsk->conn_type != 0);
1715 		anc_data[0] = tsk->conn_type;
1716 		anc_data[1] = tsk->conn_instance;
1717 		anc_data[2] = tsk->conn_instance;
1718 		break;
1719 	default:
1720 		has_name = 0;
1721 	}
1722 	if (has_name) {
1723 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1724 		if (res)
1725 			return res;
1726 	}
1727 
1728 	return 0;
1729 }
1730 
1731 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1732 {
1733 	struct sock *sk = &tsk->sk;
1734 	struct net *net = sock_net(sk);
1735 	struct sk_buff *skb = NULL;
1736 	struct tipc_msg *msg;
1737 	u32 peer_port = tsk_peer_port(tsk);
1738 	u32 dnode = tsk_peer_node(tsk);
1739 
1740 	if (!tipc_sk_connected(sk))
1741 		return;
1742 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1743 			      dnode, tsk_own_node(tsk), peer_port,
1744 			      tsk->portid, TIPC_OK);
1745 	if (!skb)
1746 		return;
1747 	msg = buf_msg(skb);
1748 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1749 	tsk->rcv_unacked = 0;
1750 
1751 	/* Adjust to and advertize the correct window limit */
1752 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1753 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1754 		msg_set_adv_win(msg, tsk->rcv_win);
1755 	}
1756 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1757 }
1758 
1759 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1760 {
1761 	struct sock *sk = sock->sk;
1762 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1763 	long timeo = *timeop;
1764 	int err = sock_error(sk);
1765 
1766 	if (err)
1767 		return err;
1768 
1769 	for (;;) {
1770 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1771 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1772 				err = -ENOTCONN;
1773 				break;
1774 			}
1775 			add_wait_queue(sk_sleep(sk), &wait);
1776 			release_sock(sk);
1777 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1778 			sched_annotate_sleep();
1779 			lock_sock(sk);
1780 			remove_wait_queue(sk_sleep(sk), &wait);
1781 		}
1782 		err = 0;
1783 		if (!skb_queue_empty(&sk->sk_receive_queue))
1784 			break;
1785 		err = -EAGAIN;
1786 		if (!timeo)
1787 			break;
1788 		err = sock_intr_errno(timeo);
1789 		if (signal_pending(current))
1790 			break;
1791 
1792 		err = sock_error(sk);
1793 		if (err)
1794 			break;
1795 	}
1796 	*timeop = timeo;
1797 	return err;
1798 }
1799 
1800 /**
1801  * tipc_recvmsg - receive packet-oriented message
1802  * @m: descriptor for message info
1803  * @buflen: length of user buffer area
1804  * @flags: receive flags
1805  *
1806  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1807  * If the complete message doesn't fit in user area, truncate it.
1808  *
1809  * Returns size of returned message data, errno otherwise
1810  */
1811 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1812 			size_t buflen,	int flags)
1813 {
1814 	struct sock *sk = sock->sk;
1815 	bool connected = !tipc_sk_type_connectionless(sk);
1816 	struct tipc_sock *tsk = tipc_sk(sk);
1817 	int rc, err, hlen, dlen, copy;
1818 	struct sk_buff_head xmitq;
1819 	struct tipc_msg *hdr;
1820 	struct sk_buff *skb;
1821 	bool grp_evt;
1822 	long timeout;
1823 
1824 	/* Catch invalid receive requests */
1825 	if (unlikely(!buflen))
1826 		return -EINVAL;
1827 
1828 	lock_sock(sk);
1829 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1830 		rc = -ENOTCONN;
1831 		goto exit;
1832 	}
1833 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1834 
1835 	/* Step rcv queue to first msg with data or error; wait if necessary */
1836 	do {
1837 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1838 		if (unlikely(rc))
1839 			goto exit;
1840 		skb = skb_peek(&sk->sk_receive_queue);
1841 		hdr = buf_msg(skb);
1842 		dlen = msg_data_sz(hdr);
1843 		hlen = msg_hdr_sz(hdr);
1844 		err = msg_errcode(hdr);
1845 		grp_evt = msg_is_grp_evt(hdr);
1846 		if (likely(dlen || err))
1847 			break;
1848 		tsk_advance_rx_queue(sk);
1849 	} while (1);
1850 
1851 	/* Collect msg meta data, including error code and rejected data */
1852 	tipc_sk_set_orig_addr(m, skb);
1853 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1854 	if (unlikely(rc))
1855 		goto exit;
1856 	hdr = buf_msg(skb);
1857 
1858 	/* Capture data if non-error msg, otherwise just set return value */
1859 	if (likely(!err)) {
1860 		copy = min_t(int, dlen, buflen);
1861 		if (unlikely(copy != dlen))
1862 			m->msg_flags |= MSG_TRUNC;
1863 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1864 	} else {
1865 		copy = 0;
1866 		rc = 0;
1867 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1868 			rc = -ECONNRESET;
1869 	}
1870 	if (unlikely(rc))
1871 		goto exit;
1872 
1873 	/* Mark message as group event if applicable */
1874 	if (unlikely(grp_evt)) {
1875 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1876 			m->msg_flags |= MSG_EOR;
1877 		m->msg_flags |= MSG_OOB;
1878 		copy = 0;
1879 	}
1880 
1881 	/* Caption of data or error code/rejected data was successful */
1882 	if (unlikely(flags & MSG_PEEK))
1883 		goto exit;
1884 
1885 	/* Send group flow control advertisement when applicable */
1886 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1887 		__skb_queue_head_init(&xmitq);
1888 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1889 					  msg_orignode(hdr), msg_origport(hdr),
1890 					  &xmitq);
1891 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1892 	}
1893 
1894 	tsk_advance_rx_queue(sk);
1895 
1896 	if (likely(!connected))
1897 		goto exit;
1898 
1899 	/* Send connection flow control advertisement when applicable */
1900 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1901 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1902 		tipc_sk_send_ack(tsk);
1903 exit:
1904 	release_sock(sk);
1905 	return rc ? rc : copy;
1906 }
1907 
1908 /**
1909  * tipc_recvstream - receive stream-oriented data
1910  * @m: descriptor for message info
1911  * @buflen: total size of user buffer area
1912  * @flags: receive flags
1913  *
1914  * Used for SOCK_STREAM messages only.  If not enough data is available
1915  * will optionally wait for more; never truncates data.
1916  *
1917  * Returns size of returned message data, errno otherwise
1918  */
1919 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1920 			   size_t buflen, int flags)
1921 {
1922 	struct sock *sk = sock->sk;
1923 	struct tipc_sock *tsk = tipc_sk(sk);
1924 	struct sk_buff *skb;
1925 	struct tipc_msg *hdr;
1926 	struct tipc_skb_cb *skb_cb;
1927 	bool peek = flags & MSG_PEEK;
1928 	int offset, required, copy, copied = 0;
1929 	int hlen, dlen, err, rc;
1930 	bool ack = false;
1931 	long timeout;
1932 
1933 	/* Catch invalid receive attempts */
1934 	if (unlikely(!buflen))
1935 		return -EINVAL;
1936 
1937 	lock_sock(sk);
1938 
1939 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1940 		rc = -ENOTCONN;
1941 		goto exit;
1942 	}
1943 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1944 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1945 
1946 	do {
1947 		/* Look at first msg in receive queue; wait if necessary */
1948 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1949 		if (unlikely(rc))
1950 			break;
1951 		skb = skb_peek(&sk->sk_receive_queue);
1952 		skb_cb = TIPC_SKB_CB(skb);
1953 		hdr = buf_msg(skb);
1954 		dlen = msg_data_sz(hdr);
1955 		hlen = msg_hdr_sz(hdr);
1956 		err = msg_errcode(hdr);
1957 
1958 		/* Discard any empty non-errored (SYN-) message */
1959 		if (unlikely(!dlen && !err)) {
1960 			tsk_advance_rx_queue(sk);
1961 			continue;
1962 		}
1963 
1964 		/* Collect msg meta data, incl. error code and rejected data */
1965 		if (!copied) {
1966 			tipc_sk_set_orig_addr(m, skb);
1967 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
1968 			if (rc)
1969 				break;
1970 			hdr = buf_msg(skb);
1971 		}
1972 
1973 		/* Copy data if msg ok, otherwise return error/partial data */
1974 		if (likely(!err)) {
1975 			ack = msg_ack_required(hdr);
1976 			offset = skb_cb->bytes_read;
1977 			copy = min_t(int, dlen - offset, buflen - copied);
1978 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1979 			if (unlikely(rc))
1980 				break;
1981 			copied += copy;
1982 			offset += copy;
1983 			if (unlikely(offset < dlen)) {
1984 				if (!peek)
1985 					skb_cb->bytes_read = offset;
1986 				break;
1987 			}
1988 		} else {
1989 			rc = 0;
1990 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1991 				rc = -ECONNRESET;
1992 			if (copied || rc)
1993 				break;
1994 		}
1995 
1996 		if (unlikely(peek))
1997 			break;
1998 
1999 		tsk_advance_rx_queue(sk);
2000 
2001 		/* Send connection flow control advertisement when applicable */
2002 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2003 		if (ack || tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2004 			tipc_sk_send_ack(tsk);
2005 
2006 		/* Exit if all requested data or FIN/error received */
2007 		if (copied == buflen || err)
2008 			break;
2009 
2010 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2011 exit:
2012 	release_sock(sk);
2013 	return copied ? copied : rc;
2014 }
2015 
2016 /**
2017  * tipc_write_space - wake up thread if port congestion is released
2018  * @sk: socket
2019  */
2020 static void tipc_write_space(struct sock *sk)
2021 {
2022 	struct socket_wq *wq;
2023 
2024 	rcu_read_lock();
2025 	wq = rcu_dereference(sk->sk_wq);
2026 	if (skwq_has_sleeper(wq))
2027 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2028 						EPOLLWRNORM | EPOLLWRBAND);
2029 	rcu_read_unlock();
2030 }
2031 
2032 /**
2033  * tipc_data_ready - wake up threads to indicate messages have been received
2034  * @sk: socket
2035  * @len: the length of messages
2036  */
2037 static void tipc_data_ready(struct sock *sk)
2038 {
2039 	struct socket_wq *wq;
2040 
2041 	rcu_read_lock();
2042 	wq = rcu_dereference(sk->sk_wq);
2043 	if (skwq_has_sleeper(wq))
2044 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2045 						EPOLLRDNORM | EPOLLRDBAND);
2046 	rcu_read_unlock();
2047 }
2048 
2049 static void tipc_sock_destruct(struct sock *sk)
2050 {
2051 	__skb_queue_purge(&sk->sk_receive_queue);
2052 }
2053 
2054 static void tipc_sk_proto_rcv(struct sock *sk,
2055 			      struct sk_buff_head *inputq,
2056 			      struct sk_buff_head *xmitq)
2057 {
2058 	struct sk_buff *skb = __skb_dequeue(inputq);
2059 	struct tipc_sock *tsk = tipc_sk(sk);
2060 	struct tipc_msg *hdr = buf_msg(skb);
2061 	struct tipc_group *grp = tsk->group;
2062 	bool wakeup = false;
2063 
2064 	switch (msg_user(hdr)) {
2065 	case CONN_MANAGER:
2066 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2067 		return;
2068 	case SOCK_WAKEUP:
2069 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2070 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2071 		smp_wmb();
2072 		tsk->cong_link_cnt--;
2073 		wakeup = true;
2074 		tipc_sk_push_backlog(tsk);
2075 		break;
2076 	case GROUP_PROTOCOL:
2077 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2078 		break;
2079 	case TOP_SRV:
2080 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2081 				      hdr, inputq, xmitq);
2082 		break;
2083 	default:
2084 		break;
2085 	}
2086 
2087 	if (wakeup)
2088 		sk->sk_write_space(sk);
2089 
2090 	kfree_skb(skb);
2091 }
2092 
2093 /**
2094  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2095  * @tsk: TIPC socket
2096  * @skb: pointer to message buffer.
2097  * Returns true if message should be added to receive queue, false otherwise
2098  */
2099 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2100 {
2101 	struct sock *sk = &tsk->sk;
2102 	struct net *net = sock_net(sk);
2103 	struct tipc_msg *hdr = buf_msg(skb);
2104 	bool con_msg = msg_connected(hdr);
2105 	u32 pport = tsk_peer_port(tsk);
2106 	u32 pnode = tsk_peer_node(tsk);
2107 	u32 oport = msg_origport(hdr);
2108 	u32 onode = msg_orignode(hdr);
2109 	int err = msg_errcode(hdr);
2110 	unsigned long delay;
2111 
2112 	if (unlikely(msg_mcast(hdr)))
2113 		return false;
2114 	tsk->oneway = 0;
2115 
2116 	switch (sk->sk_state) {
2117 	case TIPC_CONNECTING:
2118 		/* Setup ACK */
2119 		if (likely(con_msg)) {
2120 			if (err)
2121 				break;
2122 			tipc_sk_finish_conn(tsk, oport, onode);
2123 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2124 			/* ACK+ message with data is added to receive queue */
2125 			if (msg_data_sz(hdr))
2126 				return true;
2127 			/* Empty ACK-, - wake up sleeping connect() and drop */
2128 			sk->sk_state_change(sk);
2129 			msg_set_dest_droppable(hdr, 1);
2130 			return false;
2131 		}
2132 		/* Ignore connectionless message if not from listening socket */
2133 		if (oport != pport || onode != pnode)
2134 			return false;
2135 
2136 		/* Rejected SYN */
2137 		if (err != TIPC_ERR_OVERLOAD)
2138 			break;
2139 
2140 		/* Prepare for new setup attempt if we have a SYN clone */
2141 		if (skb_queue_empty(&sk->sk_write_queue))
2142 			break;
2143 		get_random_bytes(&delay, 2);
2144 		delay %= (tsk->conn_timeout / 4);
2145 		delay = msecs_to_jiffies(delay + 100);
2146 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2147 		return false;
2148 	case TIPC_OPEN:
2149 	case TIPC_DISCONNECTING:
2150 		return false;
2151 	case TIPC_LISTEN:
2152 		/* Accept only SYN message */
2153 		if (!msg_is_syn(hdr) &&
2154 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2155 			return false;
2156 		if (!con_msg && !err)
2157 			return true;
2158 		return false;
2159 	case TIPC_ESTABLISHED:
2160 		if (!skb_queue_empty(&sk->sk_write_queue))
2161 			tipc_sk_push_backlog(tsk);
2162 		/* Accept only connection-based messages sent by peer */
2163 		if (likely(con_msg && !err && pport == oport && pnode == onode))
2164 			return true;
2165 		if (!tsk_peer_msg(tsk, hdr))
2166 			return false;
2167 		if (!err)
2168 			return true;
2169 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2170 		tipc_node_remove_conn(net, pnode, tsk->portid);
2171 		sk->sk_state_change(sk);
2172 		return true;
2173 	default:
2174 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2175 	}
2176 	/* Abort connection setup attempt */
2177 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2178 	sk->sk_err = ECONNREFUSED;
2179 	sk->sk_state_change(sk);
2180 	return true;
2181 }
2182 
2183 /**
2184  * rcvbuf_limit - get proper overload limit of socket receive queue
2185  * @sk: socket
2186  * @skb: message
2187  *
2188  * For connection oriented messages, irrespective of importance,
2189  * default queue limit is 2 MB.
2190  *
2191  * For connectionless messages, queue limits are based on message
2192  * importance as follows:
2193  *
2194  * TIPC_LOW_IMPORTANCE       (2 MB)
2195  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2196  * TIPC_HIGH_IMPORTANCE      (8 MB)
2197  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2198  *
2199  * Returns overload limit according to corresponding message importance
2200  */
2201 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2202 {
2203 	struct tipc_sock *tsk = tipc_sk(sk);
2204 	struct tipc_msg *hdr = buf_msg(skb);
2205 
2206 	if (unlikely(msg_in_group(hdr)))
2207 		return READ_ONCE(sk->sk_rcvbuf);
2208 
2209 	if (unlikely(!msg_connected(hdr)))
2210 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2211 
2212 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2213 		return READ_ONCE(sk->sk_rcvbuf);
2214 
2215 	return FLOWCTL_MSG_LIM;
2216 }
2217 
2218 /**
2219  * tipc_sk_filter_rcv - validate incoming message
2220  * @sk: socket
2221  * @skb: pointer to message.
2222  *
2223  * Enqueues message on receive queue if acceptable; optionally handles
2224  * disconnect indication for a connected socket.
2225  *
2226  * Called with socket lock already taken
2227  *
2228  */
2229 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2230 			       struct sk_buff_head *xmitq)
2231 {
2232 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2233 	struct tipc_sock *tsk = tipc_sk(sk);
2234 	struct tipc_group *grp = tsk->group;
2235 	struct tipc_msg *hdr = buf_msg(skb);
2236 	struct net *net = sock_net(sk);
2237 	struct sk_buff_head inputq;
2238 	int mtyp = msg_type(hdr);
2239 	int limit, err = TIPC_OK;
2240 
2241 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2242 	TIPC_SKB_CB(skb)->bytes_read = 0;
2243 	__skb_queue_head_init(&inputq);
2244 	__skb_queue_tail(&inputq, skb);
2245 
2246 	if (unlikely(!msg_isdata(hdr)))
2247 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2248 
2249 	if (unlikely(grp))
2250 		tipc_group_filter_msg(grp, &inputq, xmitq);
2251 
2252 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2253 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2254 
2255 	/* Validate and add to receive buffer if there is space */
2256 	while ((skb = __skb_dequeue(&inputq))) {
2257 		hdr = buf_msg(skb);
2258 		limit = rcvbuf_limit(sk, skb);
2259 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2260 		    (!sk_conn && msg_connected(hdr)) ||
2261 		    (!grp && msg_in_group(hdr)))
2262 			err = TIPC_ERR_NO_PORT;
2263 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2264 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2265 					   "err_overload2!");
2266 			atomic_inc(&sk->sk_drops);
2267 			err = TIPC_ERR_OVERLOAD;
2268 		}
2269 
2270 		if (unlikely(err)) {
2271 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2272 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2273 						      "@filter_rcv!");
2274 				__skb_queue_tail(xmitq, skb);
2275 			}
2276 			err = TIPC_OK;
2277 			continue;
2278 		}
2279 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2280 		skb_set_owner_r(skb, sk);
2281 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2282 					 "rcvq >90% allocated!");
2283 		sk->sk_data_ready(sk);
2284 	}
2285 }
2286 
2287 /**
2288  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2289  * @sk: socket
2290  * @skb: message
2291  *
2292  * Caller must hold socket lock
2293  */
2294 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2295 {
2296 	unsigned int before = sk_rmem_alloc_get(sk);
2297 	struct sk_buff_head xmitq;
2298 	unsigned int added;
2299 
2300 	__skb_queue_head_init(&xmitq);
2301 
2302 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2303 	added = sk_rmem_alloc_get(sk) - before;
2304 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2305 
2306 	/* Send pending response/rejected messages, if any */
2307 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2308 	return 0;
2309 }
2310 
2311 /**
2312  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2313  *                   inputq and try adding them to socket or backlog queue
2314  * @inputq: list of incoming buffers with potentially different destinations
2315  * @sk: socket where the buffers should be enqueued
2316  * @dport: port number for the socket
2317  *
2318  * Caller must hold socket lock
2319  */
2320 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2321 			    u32 dport, struct sk_buff_head *xmitq)
2322 {
2323 	unsigned long time_limit = jiffies + 2;
2324 	struct sk_buff *skb;
2325 	unsigned int lim;
2326 	atomic_t *dcnt;
2327 	u32 onode;
2328 
2329 	while (skb_queue_len(inputq)) {
2330 		if (unlikely(time_after_eq(jiffies, time_limit)))
2331 			return;
2332 
2333 		skb = tipc_skb_dequeue(inputq, dport);
2334 		if (unlikely(!skb))
2335 			return;
2336 
2337 		/* Add message directly to receive queue if possible */
2338 		if (!sock_owned_by_user(sk)) {
2339 			tipc_sk_filter_rcv(sk, skb, xmitq);
2340 			continue;
2341 		}
2342 
2343 		/* Try backlog, compensating for double-counted bytes */
2344 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2345 		if (!sk->sk_backlog.len)
2346 			atomic_set(dcnt, 0);
2347 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2348 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2349 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2350 						 "bklg & rcvq >90% allocated!");
2351 			continue;
2352 		}
2353 
2354 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2355 		/* Overload => reject message back to sender */
2356 		onode = tipc_own_addr(sock_net(sk));
2357 		atomic_inc(&sk->sk_drops);
2358 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2359 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2360 					      "@sk_enqueue!");
2361 			__skb_queue_tail(xmitq, skb);
2362 		}
2363 		break;
2364 	}
2365 }
2366 
2367 /**
2368  * tipc_sk_rcv - handle a chain of incoming buffers
2369  * @inputq: buffer list containing the buffers
2370  * Consumes all buffers in list until inputq is empty
2371  * Note: may be called in multiple threads referring to the same queue
2372  */
2373 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2374 {
2375 	struct sk_buff_head xmitq;
2376 	u32 dnode, dport = 0;
2377 	int err;
2378 	struct tipc_sock *tsk;
2379 	struct sock *sk;
2380 	struct sk_buff *skb;
2381 
2382 	__skb_queue_head_init(&xmitq);
2383 	while (skb_queue_len(inputq)) {
2384 		dport = tipc_skb_peek_port(inputq, dport);
2385 		tsk = tipc_sk_lookup(net, dport);
2386 
2387 		if (likely(tsk)) {
2388 			sk = &tsk->sk;
2389 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2390 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2391 				spin_unlock_bh(&sk->sk_lock.slock);
2392 			}
2393 			/* Send pending response/rejected messages, if any */
2394 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2395 			sock_put(sk);
2396 			continue;
2397 		}
2398 		/* No destination socket => dequeue skb if still there */
2399 		skb = tipc_skb_dequeue(inputq, dport);
2400 		if (!skb)
2401 			return;
2402 
2403 		/* Try secondary lookup if unresolved named message */
2404 		err = TIPC_ERR_NO_PORT;
2405 		if (tipc_msg_lookup_dest(net, skb, &err))
2406 			goto xmit;
2407 
2408 		/* Prepare for message rejection */
2409 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2410 			continue;
2411 
2412 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2413 xmit:
2414 		dnode = msg_destnode(buf_msg(skb));
2415 		tipc_node_xmit_skb(net, skb, dnode, dport);
2416 	}
2417 }
2418 
2419 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2420 {
2421 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2422 	struct sock *sk = sock->sk;
2423 	int done;
2424 
2425 	do {
2426 		int err = sock_error(sk);
2427 		if (err)
2428 			return err;
2429 		if (!*timeo_p)
2430 			return -ETIMEDOUT;
2431 		if (signal_pending(current))
2432 			return sock_intr_errno(*timeo_p);
2433 
2434 		add_wait_queue(sk_sleep(sk), &wait);
2435 		done = sk_wait_event(sk, timeo_p,
2436 				     sk->sk_state != TIPC_CONNECTING, &wait);
2437 		remove_wait_queue(sk_sleep(sk), &wait);
2438 	} while (!done);
2439 	return 0;
2440 }
2441 
2442 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2443 {
2444 	if (addr->family != AF_TIPC)
2445 		return false;
2446 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2447 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2448 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2449 		addr->addrtype == TIPC_SOCKET_ADDR);
2450 }
2451 
2452 /**
2453  * tipc_connect - establish a connection to another TIPC port
2454  * @sock: socket structure
2455  * @dest: socket address for destination port
2456  * @destlen: size of socket address data structure
2457  * @flags: file-related flags associated with socket
2458  *
2459  * Returns 0 on success, errno otherwise
2460  */
2461 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2462 			int destlen, int flags)
2463 {
2464 	struct sock *sk = sock->sk;
2465 	struct tipc_sock *tsk = tipc_sk(sk);
2466 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2467 	struct msghdr m = {NULL,};
2468 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2469 	int previous;
2470 	int res = 0;
2471 
2472 	if (destlen != sizeof(struct sockaddr_tipc))
2473 		return -EINVAL;
2474 
2475 	lock_sock(sk);
2476 
2477 	if (tsk->group) {
2478 		res = -EINVAL;
2479 		goto exit;
2480 	}
2481 
2482 	if (dst->family == AF_UNSPEC) {
2483 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2484 		if (!tipc_sk_type_connectionless(sk))
2485 			res = -EINVAL;
2486 		goto exit;
2487 	}
2488 	if (!tipc_sockaddr_is_sane(dst)) {
2489 		res = -EINVAL;
2490 		goto exit;
2491 	}
2492 	/* DGRAM/RDM connect(), just save the destaddr */
2493 	if (tipc_sk_type_connectionless(sk)) {
2494 		memcpy(&tsk->peer, dest, destlen);
2495 		goto exit;
2496 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2497 		res = -EINVAL;
2498 		goto exit;
2499 	}
2500 
2501 	previous = sk->sk_state;
2502 
2503 	switch (sk->sk_state) {
2504 	case TIPC_OPEN:
2505 		/* Send a 'SYN-' to destination */
2506 		m.msg_name = dest;
2507 		m.msg_namelen = destlen;
2508 
2509 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2510 		 * indicate send_msg() is never blocked.
2511 		 */
2512 		if (!timeout)
2513 			m.msg_flags = MSG_DONTWAIT;
2514 
2515 		res = __tipc_sendmsg(sock, &m, 0);
2516 		if ((res < 0) && (res != -EWOULDBLOCK))
2517 			goto exit;
2518 
2519 		/* Just entered TIPC_CONNECTING state; the only
2520 		 * difference is that return value in non-blocking
2521 		 * case is EINPROGRESS, rather than EALREADY.
2522 		 */
2523 		res = -EINPROGRESS;
2524 		/* fall through */
2525 	case TIPC_CONNECTING:
2526 		if (!timeout) {
2527 			if (previous == TIPC_CONNECTING)
2528 				res = -EALREADY;
2529 			goto exit;
2530 		}
2531 		timeout = msecs_to_jiffies(timeout);
2532 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2533 		res = tipc_wait_for_connect(sock, &timeout);
2534 		break;
2535 	case TIPC_ESTABLISHED:
2536 		res = -EISCONN;
2537 		break;
2538 	default:
2539 		res = -EINVAL;
2540 	}
2541 
2542 exit:
2543 	release_sock(sk);
2544 	return res;
2545 }
2546 
2547 /**
2548  * tipc_listen - allow socket to listen for incoming connections
2549  * @sock: socket structure
2550  * @len: (unused)
2551  *
2552  * Returns 0 on success, errno otherwise
2553  */
2554 static int tipc_listen(struct socket *sock, int len)
2555 {
2556 	struct sock *sk = sock->sk;
2557 	int res;
2558 
2559 	lock_sock(sk);
2560 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2561 	release_sock(sk);
2562 
2563 	return res;
2564 }
2565 
2566 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2567 {
2568 	struct sock *sk = sock->sk;
2569 	DEFINE_WAIT(wait);
2570 	int err;
2571 
2572 	/* True wake-one mechanism for incoming connections: only
2573 	 * one process gets woken up, not the 'whole herd'.
2574 	 * Since we do not 'race & poll' for established sockets
2575 	 * anymore, the common case will execute the loop only once.
2576 	*/
2577 	for (;;) {
2578 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2579 					  TASK_INTERRUPTIBLE);
2580 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2581 			release_sock(sk);
2582 			timeo = schedule_timeout(timeo);
2583 			lock_sock(sk);
2584 		}
2585 		err = 0;
2586 		if (!skb_queue_empty(&sk->sk_receive_queue))
2587 			break;
2588 		err = -EAGAIN;
2589 		if (!timeo)
2590 			break;
2591 		err = sock_intr_errno(timeo);
2592 		if (signal_pending(current))
2593 			break;
2594 	}
2595 	finish_wait(sk_sleep(sk), &wait);
2596 	return err;
2597 }
2598 
2599 /**
2600  * tipc_accept - wait for connection request
2601  * @sock: listening socket
2602  * @newsock: new socket that is to be connected
2603  * @flags: file-related flags associated with socket
2604  *
2605  * Returns 0 on success, errno otherwise
2606  */
2607 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2608 		       bool kern)
2609 {
2610 	struct sock *new_sk, *sk = sock->sk;
2611 	struct sk_buff *buf;
2612 	struct tipc_sock *new_tsock;
2613 	struct tipc_msg *msg;
2614 	long timeo;
2615 	int res;
2616 
2617 	lock_sock(sk);
2618 
2619 	if (sk->sk_state != TIPC_LISTEN) {
2620 		res = -EINVAL;
2621 		goto exit;
2622 	}
2623 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2624 	res = tipc_wait_for_accept(sock, timeo);
2625 	if (res)
2626 		goto exit;
2627 
2628 	buf = skb_peek(&sk->sk_receive_queue);
2629 
2630 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2631 	if (res)
2632 		goto exit;
2633 	security_sk_clone(sock->sk, new_sock->sk);
2634 
2635 	new_sk = new_sock->sk;
2636 	new_tsock = tipc_sk(new_sk);
2637 	msg = buf_msg(buf);
2638 
2639 	/* we lock on new_sk; but lockdep sees the lock on sk */
2640 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2641 
2642 	/*
2643 	 * Reject any stray messages received by new socket
2644 	 * before the socket lock was taken (very, very unlikely)
2645 	 */
2646 	tsk_rej_rx_queue(new_sk);
2647 
2648 	/* Connect new socket to it's peer */
2649 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2650 
2651 	tsk_set_importance(new_tsock, msg_importance(msg));
2652 	if (msg_named(msg)) {
2653 		new_tsock->conn_type = msg_nametype(msg);
2654 		new_tsock->conn_instance = msg_nameinst(msg);
2655 	}
2656 
2657 	/*
2658 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2659 	 * Respond to 'SYN+' by queuing it on new socket.
2660 	 */
2661 	if (!msg_data_sz(msg)) {
2662 		struct msghdr m = {NULL,};
2663 
2664 		tsk_advance_rx_queue(sk);
2665 		__tipc_sendstream(new_sock, &m, 0);
2666 	} else {
2667 		__skb_dequeue(&sk->sk_receive_queue);
2668 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2669 		skb_set_owner_r(buf, new_sk);
2670 	}
2671 	release_sock(new_sk);
2672 exit:
2673 	release_sock(sk);
2674 	return res;
2675 }
2676 
2677 /**
2678  * tipc_shutdown - shutdown socket connection
2679  * @sock: socket structure
2680  * @how: direction to close (must be SHUT_RDWR)
2681  *
2682  * Terminates connection (if necessary), then purges socket's receive queue.
2683  *
2684  * Returns 0 on success, errno otherwise
2685  */
2686 static int tipc_shutdown(struct socket *sock, int how)
2687 {
2688 	struct sock *sk = sock->sk;
2689 	int res;
2690 
2691 	if (how != SHUT_RDWR)
2692 		return -EINVAL;
2693 
2694 	lock_sock(sk);
2695 
2696 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2697 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2698 	sk->sk_shutdown = SEND_SHUTDOWN;
2699 
2700 	if (sk->sk_state == TIPC_DISCONNECTING) {
2701 		/* Discard any unreceived messages */
2702 		__skb_queue_purge(&sk->sk_receive_queue);
2703 
2704 		/* Wake up anyone sleeping in poll */
2705 		sk->sk_state_change(sk);
2706 		res = 0;
2707 	} else {
2708 		res = -ENOTCONN;
2709 	}
2710 
2711 	release_sock(sk);
2712 	return res;
2713 }
2714 
2715 static void tipc_sk_check_probing_state(struct sock *sk,
2716 					struct sk_buff_head *list)
2717 {
2718 	struct tipc_sock *tsk = tipc_sk(sk);
2719 	u32 pnode = tsk_peer_node(tsk);
2720 	u32 pport = tsk_peer_port(tsk);
2721 	u32 self = tsk_own_node(tsk);
2722 	u32 oport = tsk->portid;
2723 	struct sk_buff *skb;
2724 
2725 	if (tsk->probe_unacked) {
2726 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2727 		sk->sk_err = ECONNABORTED;
2728 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2729 		sk->sk_state_change(sk);
2730 		return;
2731 	}
2732 	/* Prepare new probe */
2733 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2734 			      pnode, self, pport, oport, TIPC_OK);
2735 	if (skb)
2736 		__skb_queue_tail(list, skb);
2737 	tsk->probe_unacked = true;
2738 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2739 }
2740 
2741 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2742 {
2743 	struct tipc_sock *tsk = tipc_sk(sk);
2744 
2745 	/* Try again later if dest link is congested */
2746 	if (tsk->cong_link_cnt) {
2747 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2748 		return;
2749 	}
2750 	/* Prepare SYN for retransmit */
2751 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2752 }
2753 
2754 static void tipc_sk_timeout(struct timer_list *t)
2755 {
2756 	struct sock *sk = from_timer(sk, t, sk_timer);
2757 	struct tipc_sock *tsk = tipc_sk(sk);
2758 	u32 pnode = tsk_peer_node(tsk);
2759 	struct sk_buff_head list;
2760 	int rc = 0;
2761 
2762 	__skb_queue_head_init(&list);
2763 	bh_lock_sock(sk);
2764 
2765 	/* Try again later if socket is busy */
2766 	if (sock_owned_by_user(sk)) {
2767 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2768 		bh_unlock_sock(sk);
2769 		sock_put(sk);
2770 		return;
2771 	}
2772 
2773 	if (sk->sk_state == TIPC_ESTABLISHED)
2774 		tipc_sk_check_probing_state(sk, &list);
2775 	else if (sk->sk_state == TIPC_CONNECTING)
2776 		tipc_sk_retry_connect(sk, &list);
2777 
2778 	bh_unlock_sock(sk);
2779 
2780 	if (!skb_queue_empty(&list))
2781 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2782 
2783 	/* SYN messages may cause link congestion */
2784 	if (rc == -ELINKCONG) {
2785 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2786 		tsk->cong_link_cnt = 1;
2787 	}
2788 	sock_put(sk);
2789 }
2790 
2791 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2792 			   struct tipc_name_seq const *seq)
2793 {
2794 	struct sock *sk = &tsk->sk;
2795 	struct net *net = sock_net(sk);
2796 	struct publication *publ;
2797 	u32 key;
2798 
2799 	if (scope != TIPC_NODE_SCOPE)
2800 		scope = TIPC_CLUSTER_SCOPE;
2801 
2802 	if (tipc_sk_connected(sk))
2803 		return -EINVAL;
2804 	key = tsk->portid + tsk->pub_count + 1;
2805 	if (key == tsk->portid)
2806 		return -EADDRINUSE;
2807 
2808 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2809 				    scope, tsk->portid, key);
2810 	if (unlikely(!publ))
2811 		return -EINVAL;
2812 
2813 	list_add(&publ->binding_sock, &tsk->publications);
2814 	tsk->pub_count++;
2815 	tsk->published = 1;
2816 	return 0;
2817 }
2818 
2819 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2820 			    struct tipc_name_seq const *seq)
2821 {
2822 	struct net *net = sock_net(&tsk->sk);
2823 	struct publication *publ;
2824 	struct publication *safe;
2825 	int rc = -EINVAL;
2826 
2827 	if (scope != TIPC_NODE_SCOPE)
2828 		scope = TIPC_CLUSTER_SCOPE;
2829 
2830 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2831 		if (seq) {
2832 			if (publ->scope != scope)
2833 				continue;
2834 			if (publ->type != seq->type)
2835 				continue;
2836 			if (publ->lower != seq->lower)
2837 				continue;
2838 			if (publ->upper != seq->upper)
2839 				break;
2840 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2841 					      publ->upper, publ->key);
2842 			rc = 0;
2843 			break;
2844 		}
2845 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2846 				      publ->upper, publ->key);
2847 		rc = 0;
2848 	}
2849 	if (list_empty(&tsk->publications))
2850 		tsk->published = 0;
2851 	return rc;
2852 }
2853 
2854 /* tipc_sk_reinit: set non-zero address in all existing sockets
2855  *                 when we go from standalone to network mode.
2856  */
2857 void tipc_sk_reinit(struct net *net)
2858 {
2859 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2860 	struct rhashtable_iter iter;
2861 	struct tipc_sock *tsk;
2862 	struct tipc_msg *msg;
2863 
2864 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2865 
2866 	do {
2867 		rhashtable_walk_start(&iter);
2868 
2869 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2870 			sock_hold(&tsk->sk);
2871 			rhashtable_walk_stop(&iter);
2872 			lock_sock(&tsk->sk);
2873 			msg = &tsk->phdr;
2874 			msg_set_prevnode(msg, tipc_own_addr(net));
2875 			msg_set_orignode(msg, tipc_own_addr(net));
2876 			release_sock(&tsk->sk);
2877 			rhashtable_walk_start(&iter);
2878 			sock_put(&tsk->sk);
2879 		}
2880 
2881 		rhashtable_walk_stop(&iter);
2882 	} while (tsk == ERR_PTR(-EAGAIN));
2883 
2884 	rhashtable_walk_exit(&iter);
2885 }
2886 
2887 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2888 {
2889 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2890 	struct tipc_sock *tsk;
2891 
2892 	rcu_read_lock();
2893 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2894 	if (tsk)
2895 		sock_hold(&tsk->sk);
2896 	rcu_read_unlock();
2897 
2898 	return tsk;
2899 }
2900 
2901 static int tipc_sk_insert(struct tipc_sock *tsk)
2902 {
2903 	struct sock *sk = &tsk->sk;
2904 	struct net *net = sock_net(sk);
2905 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2906 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2907 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2908 
2909 	while (remaining--) {
2910 		portid++;
2911 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2912 			portid = TIPC_MIN_PORT;
2913 		tsk->portid = portid;
2914 		sock_hold(&tsk->sk);
2915 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2916 						   tsk_rht_params))
2917 			return 0;
2918 		sock_put(&tsk->sk);
2919 	}
2920 
2921 	return -1;
2922 }
2923 
2924 static void tipc_sk_remove(struct tipc_sock *tsk)
2925 {
2926 	struct sock *sk = &tsk->sk;
2927 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2928 
2929 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2930 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2931 		__sock_put(sk);
2932 	}
2933 }
2934 
2935 static const struct rhashtable_params tsk_rht_params = {
2936 	.nelem_hint = 192,
2937 	.head_offset = offsetof(struct tipc_sock, node),
2938 	.key_offset = offsetof(struct tipc_sock, portid),
2939 	.key_len = sizeof(u32), /* portid */
2940 	.max_size = 1048576,
2941 	.min_size = 256,
2942 	.automatic_shrinking = true,
2943 };
2944 
2945 int tipc_sk_rht_init(struct net *net)
2946 {
2947 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2948 
2949 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2950 }
2951 
2952 void tipc_sk_rht_destroy(struct net *net)
2953 {
2954 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2955 
2956 	/* Wait for socket readers to complete */
2957 	synchronize_net();
2958 
2959 	rhashtable_destroy(&tn->sk_rht);
2960 }
2961 
2962 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2963 {
2964 	struct net *net = sock_net(&tsk->sk);
2965 	struct tipc_group *grp = tsk->group;
2966 	struct tipc_msg *hdr = &tsk->phdr;
2967 	struct tipc_name_seq seq;
2968 	int rc;
2969 
2970 	if (mreq->type < TIPC_RESERVED_TYPES)
2971 		return -EACCES;
2972 	if (mreq->scope > TIPC_NODE_SCOPE)
2973 		return -EINVAL;
2974 	if (grp)
2975 		return -EACCES;
2976 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2977 	if (!grp)
2978 		return -ENOMEM;
2979 	tsk->group = grp;
2980 	msg_set_lookup_scope(hdr, mreq->scope);
2981 	msg_set_nametype(hdr, mreq->type);
2982 	msg_set_dest_droppable(hdr, true);
2983 	seq.type = mreq->type;
2984 	seq.lower = mreq->instance;
2985 	seq.upper = seq.lower;
2986 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2987 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2988 	if (rc) {
2989 		tipc_group_delete(net, grp);
2990 		tsk->group = NULL;
2991 		return rc;
2992 	}
2993 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2994 	tsk->mc_method.rcast = true;
2995 	tsk->mc_method.mandatory = true;
2996 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2997 	return rc;
2998 }
2999 
3000 static int tipc_sk_leave(struct tipc_sock *tsk)
3001 {
3002 	struct net *net = sock_net(&tsk->sk);
3003 	struct tipc_group *grp = tsk->group;
3004 	struct tipc_name_seq seq;
3005 	int scope;
3006 
3007 	if (!grp)
3008 		return -EINVAL;
3009 	tipc_group_self(grp, &seq, &scope);
3010 	tipc_group_delete(net, grp);
3011 	tsk->group = NULL;
3012 	tipc_sk_withdraw(tsk, scope, &seq);
3013 	return 0;
3014 }
3015 
3016 /**
3017  * tipc_setsockopt - set socket option
3018  * @sock: socket structure
3019  * @lvl: option level
3020  * @opt: option identifier
3021  * @ov: pointer to new option value
3022  * @ol: length of option value
3023  *
3024  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3025  * (to ease compatibility).
3026  *
3027  * Returns 0 on success, errno otherwise
3028  */
3029 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3030 			   char __user *ov, unsigned int ol)
3031 {
3032 	struct sock *sk = sock->sk;
3033 	struct tipc_sock *tsk = tipc_sk(sk);
3034 	struct tipc_group_req mreq;
3035 	u32 value = 0;
3036 	int res = 0;
3037 
3038 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3039 		return 0;
3040 	if (lvl != SOL_TIPC)
3041 		return -ENOPROTOOPT;
3042 
3043 	switch (opt) {
3044 	case TIPC_IMPORTANCE:
3045 	case TIPC_SRC_DROPPABLE:
3046 	case TIPC_DEST_DROPPABLE:
3047 	case TIPC_CONN_TIMEOUT:
3048 	case TIPC_NODELAY:
3049 		if (ol < sizeof(value))
3050 			return -EINVAL;
3051 		if (get_user(value, (u32 __user *)ov))
3052 			return -EFAULT;
3053 		break;
3054 	case TIPC_GROUP_JOIN:
3055 		if (ol < sizeof(mreq))
3056 			return -EINVAL;
3057 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
3058 			return -EFAULT;
3059 		break;
3060 	default:
3061 		if (ov || ol)
3062 			return -EINVAL;
3063 	}
3064 
3065 	lock_sock(sk);
3066 
3067 	switch (opt) {
3068 	case TIPC_IMPORTANCE:
3069 		res = tsk_set_importance(tsk, value);
3070 		break;
3071 	case TIPC_SRC_DROPPABLE:
3072 		if (sock->type != SOCK_STREAM)
3073 			tsk_set_unreliable(tsk, value);
3074 		else
3075 			res = -ENOPROTOOPT;
3076 		break;
3077 	case TIPC_DEST_DROPPABLE:
3078 		tsk_set_unreturnable(tsk, value);
3079 		break;
3080 	case TIPC_CONN_TIMEOUT:
3081 		tipc_sk(sk)->conn_timeout = value;
3082 		break;
3083 	case TIPC_MCAST_BROADCAST:
3084 		tsk->mc_method.rcast = false;
3085 		tsk->mc_method.mandatory = true;
3086 		break;
3087 	case TIPC_MCAST_REPLICAST:
3088 		tsk->mc_method.rcast = true;
3089 		tsk->mc_method.mandatory = true;
3090 		break;
3091 	case TIPC_GROUP_JOIN:
3092 		res = tipc_sk_join(tsk, &mreq);
3093 		break;
3094 	case TIPC_GROUP_LEAVE:
3095 		res = tipc_sk_leave(tsk);
3096 		break;
3097 	case TIPC_NODELAY:
3098 		tsk->nodelay = !!value;
3099 		tsk_set_nagle(tsk);
3100 		break;
3101 	default:
3102 		res = -EINVAL;
3103 	}
3104 
3105 	release_sock(sk);
3106 
3107 	return res;
3108 }
3109 
3110 /**
3111  * tipc_getsockopt - get socket option
3112  * @sock: socket structure
3113  * @lvl: option level
3114  * @opt: option identifier
3115  * @ov: receptacle for option value
3116  * @ol: receptacle for length of option value
3117  *
3118  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3119  * (to ease compatibility).
3120  *
3121  * Returns 0 on success, errno otherwise
3122  */
3123 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3124 			   char __user *ov, int __user *ol)
3125 {
3126 	struct sock *sk = sock->sk;
3127 	struct tipc_sock *tsk = tipc_sk(sk);
3128 	struct tipc_name_seq seq;
3129 	int len, scope;
3130 	u32 value;
3131 	int res;
3132 
3133 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3134 		return put_user(0, ol);
3135 	if (lvl != SOL_TIPC)
3136 		return -ENOPROTOOPT;
3137 	res = get_user(len, ol);
3138 	if (res)
3139 		return res;
3140 
3141 	lock_sock(sk);
3142 
3143 	switch (opt) {
3144 	case TIPC_IMPORTANCE:
3145 		value = tsk_importance(tsk);
3146 		break;
3147 	case TIPC_SRC_DROPPABLE:
3148 		value = tsk_unreliable(tsk);
3149 		break;
3150 	case TIPC_DEST_DROPPABLE:
3151 		value = tsk_unreturnable(tsk);
3152 		break;
3153 	case TIPC_CONN_TIMEOUT:
3154 		value = tsk->conn_timeout;
3155 		/* no need to set "res", since already 0 at this point */
3156 		break;
3157 	case TIPC_NODE_RECVQ_DEPTH:
3158 		value = 0; /* was tipc_queue_size, now obsolete */
3159 		break;
3160 	case TIPC_SOCK_RECVQ_DEPTH:
3161 		value = skb_queue_len(&sk->sk_receive_queue);
3162 		break;
3163 	case TIPC_SOCK_RECVQ_USED:
3164 		value = sk_rmem_alloc_get(sk);
3165 		break;
3166 	case TIPC_GROUP_JOIN:
3167 		seq.type = 0;
3168 		if (tsk->group)
3169 			tipc_group_self(tsk->group, &seq, &scope);
3170 		value = seq.type;
3171 		break;
3172 	default:
3173 		res = -EINVAL;
3174 	}
3175 
3176 	release_sock(sk);
3177 
3178 	if (res)
3179 		return res;	/* "get" failed */
3180 
3181 	if (len < sizeof(value))
3182 		return -EINVAL;
3183 
3184 	if (copy_to_user(ov, &value, sizeof(value)))
3185 		return -EFAULT;
3186 
3187 	return put_user(sizeof(value), ol);
3188 }
3189 
3190 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3191 {
3192 	struct net *net = sock_net(sock->sk);
3193 	struct tipc_sioc_nodeid_req nr = {0};
3194 	struct tipc_sioc_ln_req lnr;
3195 	void __user *argp = (void __user *)arg;
3196 
3197 	switch (cmd) {
3198 	case SIOCGETLINKNAME:
3199 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3200 			return -EFAULT;
3201 		if (!tipc_node_get_linkname(net,
3202 					    lnr.bearer_id & 0xffff, lnr.peer,
3203 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3204 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3205 				return -EFAULT;
3206 			return 0;
3207 		}
3208 		return -EADDRNOTAVAIL;
3209 	case SIOCGETNODEID:
3210 		if (copy_from_user(&nr, argp, sizeof(nr)))
3211 			return -EFAULT;
3212 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3213 			return -EADDRNOTAVAIL;
3214 		if (copy_to_user(argp, &nr, sizeof(nr)))
3215 			return -EFAULT;
3216 		return 0;
3217 	default:
3218 		return -ENOIOCTLCMD;
3219 	}
3220 }
3221 
3222 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3223 {
3224 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3225 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3226 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3227 
3228 	tsk1->peer.family = AF_TIPC;
3229 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3230 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3231 	tsk1->peer.addr.id.ref = tsk2->portid;
3232 	tsk1->peer.addr.id.node = onode;
3233 	tsk2->peer.family = AF_TIPC;
3234 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3235 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3236 	tsk2->peer.addr.id.ref = tsk1->portid;
3237 	tsk2->peer.addr.id.node = onode;
3238 
3239 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3240 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3241 	return 0;
3242 }
3243 
3244 /* Protocol switches for the various types of TIPC sockets */
3245 
3246 static const struct proto_ops msg_ops = {
3247 	.owner		= THIS_MODULE,
3248 	.family		= AF_TIPC,
3249 	.release	= tipc_release,
3250 	.bind		= tipc_bind,
3251 	.connect	= tipc_connect,
3252 	.socketpair	= tipc_socketpair,
3253 	.accept		= sock_no_accept,
3254 	.getname	= tipc_getname,
3255 	.poll		= tipc_poll,
3256 	.ioctl		= tipc_ioctl,
3257 	.listen		= sock_no_listen,
3258 	.shutdown	= tipc_shutdown,
3259 	.setsockopt	= tipc_setsockopt,
3260 	.getsockopt	= tipc_getsockopt,
3261 	.sendmsg	= tipc_sendmsg,
3262 	.recvmsg	= tipc_recvmsg,
3263 	.mmap		= sock_no_mmap,
3264 	.sendpage	= sock_no_sendpage
3265 };
3266 
3267 static const struct proto_ops packet_ops = {
3268 	.owner		= THIS_MODULE,
3269 	.family		= AF_TIPC,
3270 	.release	= tipc_release,
3271 	.bind		= tipc_bind,
3272 	.connect	= tipc_connect,
3273 	.socketpair	= tipc_socketpair,
3274 	.accept		= tipc_accept,
3275 	.getname	= tipc_getname,
3276 	.poll		= tipc_poll,
3277 	.ioctl		= tipc_ioctl,
3278 	.listen		= tipc_listen,
3279 	.shutdown	= tipc_shutdown,
3280 	.setsockopt	= tipc_setsockopt,
3281 	.getsockopt	= tipc_getsockopt,
3282 	.sendmsg	= tipc_send_packet,
3283 	.recvmsg	= tipc_recvmsg,
3284 	.mmap		= sock_no_mmap,
3285 	.sendpage	= sock_no_sendpage
3286 };
3287 
3288 static const struct proto_ops stream_ops = {
3289 	.owner		= THIS_MODULE,
3290 	.family		= AF_TIPC,
3291 	.release	= tipc_release,
3292 	.bind		= tipc_bind,
3293 	.connect	= tipc_connect,
3294 	.socketpair	= tipc_socketpair,
3295 	.accept		= tipc_accept,
3296 	.getname	= tipc_getname,
3297 	.poll		= tipc_poll,
3298 	.ioctl		= tipc_ioctl,
3299 	.listen		= tipc_listen,
3300 	.shutdown	= tipc_shutdown,
3301 	.setsockopt	= tipc_setsockopt,
3302 	.getsockopt	= tipc_getsockopt,
3303 	.sendmsg	= tipc_sendstream,
3304 	.recvmsg	= tipc_recvstream,
3305 	.mmap		= sock_no_mmap,
3306 	.sendpage	= sock_no_sendpage
3307 };
3308 
3309 static const struct net_proto_family tipc_family_ops = {
3310 	.owner		= THIS_MODULE,
3311 	.family		= AF_TIPC,
3312 	.create		= tipc_sk_create
3313 };
3314 
3315 static struct proto tipc_proto = {
3316 	.name		= "TIPC",
3317 	.owner		= THIS_MODULE,
3318 	.obj_size	= sizeof(struct tipc_sock),
3319 	.sysctl_rmem	= sysctl_tipc_rmem
3320 };
3321 
3322 /**
3323  * tipc_socket_init - initialize TIPC socket interface
3324  *
3325  * Returns 0 on success, errno otherwise
3326  */
3327 int tipc_socket_init(void)
3328 {
3329 	int res;
3330 
3331 	res = proto_register(&tipc_proto, 1);
3332 	if (res) {
3333 		pr_err("Failed to register TIPC protocol type\n");
3334 		goto out;
3335 	}
3336 
3337 	res = sock_register(&tipc_family_ops);
3338 	if (res) {
3339 		pr_err("Failed to register TIPC socket type\n");
3340 		proto_unregister(&tipc_proto);
3341 		goto out;
3342 	}
3343  out:
3344 	return res;
3345 }
3346 
3347 /**
3348  * tipc_socket_stop - stop TIPC socket interface
3349  */
3350 void tipc_socket_stop(void)
3351 {
3352 	sock_unregister(tipc_family_ops.family);
3353 	proto_unregister(&tipc_proto);
3354 }
3355 
3356 /* Caller should hold socket lock for the passed tipc socket. */
3357 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3358 {
3359 	u32 peer_node;
3360 	u32 peer_port;
3361 	struct nlattr *nest;
3362 
3363 	peer_node = tsk_peer_node(tsk);
3364 	peer_port = tsk_peer_port(tsk);
3365 
3366 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3367 	if (!nest)
3368 		return -EMSGSIZE;
3369 
3370 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3371 		goto msg_full;
3372 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3373 		goto msg_full;
3374 
3375 	if (tsk->conn_type != 0) {
3376 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3377 			goto msg_full;
3378 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3379 			goto msg_full;
3380 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3381 			goto msg_full;
3382 	}
3383 	nla_nest_end(skb, nest);
3384 
3385 	return 0;
3386 
3387 msg_full:
3388 	nla_nest_cancel(skb, nest);
3389 
3390 	return -EMSGSIZE;
3391 }
3392 
3393 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3394 			  *tsk)
3395 {
3396 	struct net *net = sock_net(skb->sk);
3397 	struct sock *sk = &tsk->sk;
3398 
3399 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3400 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3401 		return -EMSGSIZE;
3402 
3403 	if (tipc_sk_connected(sk)) {
3404 		if (__tipc_nl_add_sk_con(skb, tsk))
3405 			return -EMSGSIZE;
3406 	} else if (!list_empty(&tsk->publications)) {
3407 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3408 			return -EMSGSIZE;
3409 	}
3410 	return 0;
3411 }
3412 
3413 /* Caller should hold socket lock for the passed tipc socket. */
3414 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3415 			    struct tipc_sock *tsk)
3416 {
3417 	struct nlattr *attrs;
3418 	void *hdr;
3419 
3420 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3421 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3422 	if (!hdr)
3423 		goto msg_cancel;
3424 
3425 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3426 	if (!attrs)
3427 		goto genlmsg_cancel;
3428 
3429 	if (__tipc_nl_add_sk_info(skb, tsk))
3430 		goto attr_msg_cancel;
3431 
3432 	nla_nest_end(skb, attrs);
3433 	genlmsg_end(skb, hdr);
3434 
3435 	return 0;
3436 
3437 attr_msg_cancel:
3438 	nla_nest_cancel(skb, attrs);
3439 genlmsg_cancel:
3440 	genlmsg_cancel(skb, hdr);
3441 msg_cancel:
3442 	return -EMSGSIZE;
3443 }
3444 
3445 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3446 		    int (*skb_handler)(struct sk_buff *skb,
3447 				       struct netlink_callback *cb,
3448 				       struct tipc_sock *tsk))
3449 {
3450 	struct rhashtable_iter *iter = (void *)cb->args[4];
3451 	struct tipc_sock *tsk;
3452 	int err;
3453 
3454 	rhashtable_walk_start(iter);
3455 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3456 		if (IS_ERR(tsk)) {
3457 			err = PTR_ERR(tsk);
3458 			if (err == -EAGAIN) {
3459 				err = 0;
3460 				continue;
3461 			}
3462 			break;
3463 		}
3464 
3465 		sock_hold(&tsk->sk);
3466 		rhashtable_walk_stop(iter);
3467 		lock_sock(&tsk->sk);
3468 		err = skb_handler(skb, cb, tsk);
3469 		if (err) {
3470 			release_sock(&tsk->sk);
3471 			sock_put(&tsk->sk);
3472 			goto out;
3473 		}
3474 		release_sock(&tsk->sk);
3475 		rhashtable_walk_start(iter);
3476 		sock_put(&tsk->sk);
3477 	}
3478 	rhashtable_walk_stop(iter);
3479 out:
3480 	return skb->len;
3481 }
3482 EXPORT_SYMBOL(tipc_nl_sk_walk);
3483 
3484 int tipc_dump_start(struct netlink_callback *cb)
3485 {
3486 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3487 }
3488 EXPORT_SYMBOL(tipc_dump_start);
3489 
3490 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3491 {
3492 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3493 	struct rhashtable_iter *iter = (void *)cb->args[4];
3494 	struct tipc_net *tn = tipc_net(net);
3495 
3496 	if (!iter) {
3497 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3498 		if (!iter)
3499 			return -ENOMEM;
3500 
3501 		cb->args[4] = (long)iter;
3502 	}
3503 
3504 	rhashtable_walk_enter(&tn->sk_rht, iter);
3505 	return 0;
3506 }
3507 
3508 int tipc_dump_done(struct netlink_callback *cb)
3509 {
3510 	struct rhashtable_iter *hti = (void *)cb->args[4];
3511 
3512 	rhashtable_walk_exit(hti);
3513 	kfree(hti);
3514 	return 0;
3515 }
3516 EXPORT_SYMBOL(tipc_dump_done);
3517 
3518 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3519 			   struct tipc_sock *tsk, u32 sk_filter_state,
3520 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3521 {
3522 	struct sock *sk = &tsk->sk;
3523 	struct nlattr *attrs;
3524 	struct nlattr *stat;
3525 
3526 	/*filter response w.r.t sk_state*/
3527 	if (!(sk_filter_state & (1 << sk->sk_state)))
3528 		return 0;
3529 
3530 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3531 	if (!attrs)
3532 		goto msg_cancel;
3533 
3534 	if (__tipc_nl_add_sk_info(skb, tsk))
3535 		goto attr_msg_cancel;
3536 
3537 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3538 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3539 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3540 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3541 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3542 					 sock_i_uid(sk))) ||
3543 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3544 			      tipc_diag_gen_cookie(sk),
3545 			      TIPC_NLA_SOCK_PAD))
3546 		goto attr_msg_cancel;
3547 
3548 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3549 	if (!stat)
3550 		goto attr_msg_cancel;
3551 
3552 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3553 			skb_queue_len(&sk->sk_receive_queue)) ||
3554 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3555 			skb_queue_len(&sk->sk_write_queue)) ||
3556 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3557 			atomic_read(&sk->sk_drops)))
3558 		goto stat_msg_cancel;
3559 
3560 	if (tsk->cong_link_cnt &&
3561 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3562 		goto stat_msg_cancel;
3563 
3564 	if (tsk_conn_cong(tsk) &&
3565 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3566 		goto stat_msg_cancel;
3567 
3568 	nla_nest_end(skb, stat);
3569 
3570 	if (tsk->group)
3571 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3572 			goto stat_msg_cancel;
3573 
3574 	nla_nest_end(skb, attrs);
3575 
3576 	return 0;
3577 
3578 stat_msg_cancel:
3579 	nla_nest_cancel(skb, stat);
3580 attr_msg_cancel:
3581 	nla_nest_cancel(skb, attrs);
3582 msg_cancel:
3583 	return -EMSGSIZE;
3584 }
3585 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3586 
3587 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3588 {
3589 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3590 }
3591 
3592 /* Caller should hold socket lock for the passed tipc socket. */
3593 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3594 				 struct netlink_callback *cb,
3595 				 struct publication *publ)
3596 {
3597 	void *hdr;
3598 	struct nlattr *attrs;
3599 
3600 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3601 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3602 	if (!hdr)
3603 		goto msg_cancel;
3604 
3605 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3606 	if (!attrs)
3607 		goto genlmsg_cancel;
3608 
3609 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3610 		goto attr_msg_cancel;
3611 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3612 		goto attr_msg_cancel;
3613 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3614 		goto attr_msg_cancel;
3615 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3616 		goto attr_msg_cancel;
3617 
3618 	nla_nest_end(skb, attrs);
3619 	genlmsg_end(skb, hdr);
3620 
3621 	return 0;
3622 
3623 attr_msg_cancel:
3624 	nla_nest_cancel(skb, attrs);
3625 genlmsg_cancel:
3626 	genlmsg_cancel(skb, hdr);
3627 msg_cancel:
3628 	return -EMSGSIZE;
3629 }
3630 
3631 /* Caller should hold socket lock for the passed tipc socket. */
3632 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3633 				  struct netlink_callback *cb,
3634 				  struct tipc_sock *tsk, u32 *last_publ)
3635 {
3636 	int err;
3637 	struct publication *p;
3638 
3639 	if (*last_publ) {
3640 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3641 			if (p->key == *last_publ)
3642 				break;
3643 		}
3644 		if (p->key != *last_publ) {
3645 			/* We never set seq or call nl_dump_check_consistent()
3646 			 * this means that setting prev_seq here will cause the
3647 			 * consistence check to fail in the netlink callback
3648 			 * handler. Resulting in the last NLMSG_DONE message
3649 			 * having the NLM_F_DUMP_INTR flag set.
3650 			 */
3651 			cb->prev_seq = 1;
3652 			*last_publ = 0;
3653 			return -EPIPE;
3654 		}
3655 	} else {
3656 		p = list_first_entry(&tsk->publications, struct publication,
3657 				     binding_sock);
3658 	}
3659 
3660 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3661 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3662 		if (err) {
3663 			*last_publ = p->key;
3664 			return err;
3665 		}
3666 	}
3667 	*last_publ = 0;
3668 
3669 	return 0;
3670 }
3671 
3672 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3673 {
3674 	int err;
3675 	u32 tsk_portid = cb->args[0];
3676 	u32 last_publ = cb->args[1];
3677 	u32 done = cb->args[2];
3678 	struct net *net = sock_net(skb->sk);
3679 	struct tipc_sock *tsk;
3680 
3681 	if (!tsk_portid) {
3682 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3683 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3684 
3685 		if (!attrs[TIPC_NLA_SOCK])
3686 			return -EINVAL;
3687 
3688 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3689 						  attrs[TIPC_NLA_SOCK],
3690 						  tipc_nl_sock_policy, NULL);
3691 		if (err)
3692 			return err;
3693 
3694 		if (!sock[TIPC_NLA_SOCK_REF])
3695 			return -EINVAL;
3696 
3697 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3698 	}
3699 
3700 	if (done)
3701 		return 0;
3702 
3703 	tsk = tipc_sk_lookup(net, tsk_portid);
3704 	if (!tsk)
3705 		return -EINVAL;
3706 
3707 	lock_sock(&tsk->sk);
3708 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3709 	if (!err)
3710 		done = 1;
3711 	release_sock(&tsk->sk);
3712 	sock_put(&tsk->sk);
3713 
3714 	cb->args[0] = tsk_portid;
3715 	cb->args[1] = last_publ;
3716 	cb->args[2] = done;
3717 
3718 	return skb->len;
3719 }
3720 
3721 /**
3722  * tipc_sk_filtering - check if a socket should be traced
3723  * @sk: the socket to be examined
3724  * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3725  *  (portid, sock type, name type, name lower, name upper)
3726  *
3727  * Returns true if the socket meets the socket tuple data
3728  * (value 0 = 'any') or when there is no tuple set (all = 0),
3729  * otherwise false
3730  */
3731 bool tipc_sk_filtering(struct sock *sk)
3732 {
3733 	struct tipc_sock *tsk;
3734 	struct publication *p;
3735 	u32 _port, _sktype, _type, _lower, _upper;
3736 	u32 type = 0, lower = 0, upper = 0;
3737 
3738 	if (!sk)
3739 		return true;
3740 
3741 	tsk = tipc_sk(sk);
3742 
3743 	_port = sysctl_tipc_sk_filter[0];
3744 	_sktype = sysctl_tipc_sk_filter[1];
3745 	_type = sysctl_tipc_sk_filter[2];
3746 	_lower = sysctl_tipc_sk_filter[3];
3747 	_upper = sysctl_tipc_sk_filter[4];
3748 
3749 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3750 		return true;
3751 
3752 	if (_port)
3753 		return (_port == tsk->portid);
3754 
3755 	if (_sktype && _sktype != sk->sk_type)
3756 		return false;
3757 
3758 	if (tsk->published) {
3759 		p = list_first_entry_or_null(&tsk->publications,
3760 					     struct publication, binding_sock);
3761 		if (p) {
3762 			type = p->type;
3763 			lower = p->lower;
3764 			upper = p->upper;
3765 		}
3766 	}
3767 
3768 	if (!tipc_sk_type_connectionless(sk)) {
3769 		type = tsk->conn_type;
3770 		lower = tsk->conn_instance;
3771 		upper = tsk->conn_instance;
3772 	}
3773 
3774 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3775 	    (_upper && _upper != upper))
3776 		return false;
3777 
3778 	return true;
3779 }
3780 
3781 u32 tipc_sock_get_portid(struct sock *sk)
3782 {
3783 	return (sk) ? (tipc_sk(sk))->portid : 0;
3784 }
3785 
3786 /**
3787  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3788  *			both the rcv and backlog queues are considered
3789  * @sk: tipc sk to be checked
3790  * @skb: tipc msg to be checked
3791  *
3792  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3793  */
3794 
3795 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3796 {
3797 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3798 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3799 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3800 
3801 	return (qsize > lim * 90 / 100);
3802 }
3803 
3804 /**
3805  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3806  *			only the rcv queue is considered
3807  * @sk: tipc sk to be checked
3808  * @skb: tipc msg to be checked
3809  *
3810  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3811  */
3812 
3813 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3814 {
3815 	unsigned int lim = rcvbuf_limit(sk, skb);
3816 	unsigned int qsize = sk_rmem_alloc_get(sk);
3817 
3818 	return (qsize > lim * 90 / 100);
3819 }
3820 
3821 /**
3822  * tipc_sk_dump - dump TIPC socket
3823  * @sk: tipc sk to be dumped
3824  * @dqueues: bitmask to decide if any socket queue to be dumped?
3825  *           - TIPC_DUMP_NONE: don't dump socket queues
3826  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3827  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3828  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3829  *           - TIPC_DUMP_ALL: dump all the socket queues above
3830  * @buf: returned buffer of dump data in format
3831  */
3832 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3833 {
3834 	int i = 0;
3835 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3836 	struct tipc_sock *tsk;
3837 	struct publication *p;
3838 	bool tsk_connected;
3839 
3840 	if (!sk) {
3841 		i += scnprintf(buf, sz, "sk data: (null)\n");
3842 		return i;
3843 	}
3844 
3845 	tsk = tipc_sk(sk);
3846 	tsk_connected = !tipc_sk_type_connectionless(sk);
3847 
3848 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3849 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3850 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3851 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3852 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3853 	if (tsk_connected) {
3854 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3855 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3856 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3857 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3858 	}
3859 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3860 	if (tsk->published) {
3861 		p = list_first_entry_or_null(&tsk->publications,
3862 					     struct publication, binding_sock);
3863 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3864 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3865 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3866 	}
3867 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3868 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3869 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3870 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3871 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3872 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3873 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3874 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3875 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3876 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3877 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3878 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3879 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3880 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3881 
3882 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3883 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3884 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3885 	}
3886 
3887 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3888 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3889 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3890 	}
3891 
3892 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3893 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3894 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3895 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3896 			i += scnprintf(buf + i, sz - i, "  tail ");
3897 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3898 					   buf + i);
3899 		}
3900 	}
3901 
3902 	return i;
3903 }
3904