xref: /openbmc/linux/net/tipc/socket.c (revision c1d3fb8a)
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, dnode = 0;
1368 	u32 type, inst;
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 		msg_set_type(hdr, TIPC_NAMED_MSG);
1422 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1423 		msg_set_nametype(hdr, type);
1424 		msg_set_nameinst(hdr, inst);
1425 		msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1426 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
1427 		msg_set_destnode(hdr, dnode);
1428 		msg_set_destport(hdr, dport);
1429 		if (unlikely(!dport && !dnode))
1430 			return -EHOSTUNREACH;
1431 	} else if (dest->addrtype == TIPC_ADDR_ID) {
1432 		dnode = dest->addr.id.node;
1433 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1434 		msg_set_lookup_scope(hdr, 0);
1435 		msg_set_destnode(hdr, dnode);
1436 		msg_set_destport(hdr, dest->addr.id.ref);
1437 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1438 	} else {
1439 		return -EINVAL;
1440 	}
1441 
1442 	/* Block or return if destination link is congested */
1443 	rc = tipc_wait_for_cond(sock, &timeout,
1444 				!tipc_dest_find(clinks, dnode, 0));
1445 	if (unlikely(rc))
1446 		return rc;
1447 
1448 	__skb_queue_head_init(&pkts);
1449 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
1450 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1451 	if (unlikely(rc != dlen))
1452 		return rc;
1453 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1454 		__skb_queue_purge(&pkts);
1455 		return -ENOMEM;
1456 	}
1457 
1458 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1459 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1460 	if (unlikely(rc == -ELINKCONG)) {
1461 		tipc_dest_push(clinks, dnode, 0);
1462 		tsk->cong_link_cnt++;
1463 		rc = 0;
1464 	}
1465 
1466 	if (unlikely(syn && !rc))
1467 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1468 
1469 	return rc ? rc : dlen;
1470 }
1471 
1472 /**
1473  * tipc_sendstream - send stream-oriented data
1474  * @sock: socket structure
1475  * @m: data to send
1476  * @dsz: total length of data to be transmitted
1477  *
1478  * Used for SOCK_STREAM data.
1479  *
1480  * Returns the number of bytes sent on success (or partial success),
1481  * or errno if no data sent
1482  */
1483 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1484 {
1485 	struct sock *sk = sock->sk;
1486 	int ret;
1487 
1488 	lock_sock(sk);
1489 	ret = __tipc_sendstream(sock, m, dsz);
1490 	release_sock(sk);
1491 
1492 	return ret;
1493 }
1494 
1495 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1496 {
1497 	struct sock *sk = sock->sk;
1498 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1499 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1500 	struct sk_buff_head *txq = &sk->sk_write_queue;
1501 	struct tipc_sock *tsk = tipc_sk(sk);
1502 	struct tipc_msg *hdr = &tsk->phdr;
1503 	struct net *net = sock_net(sk);
1504 	u32 dnode = tsk_peer_node(tsk);
1505 	int maxnagle = tsk->maxnagle;
1506 	int maxpkt = tsk->max_pkt;
1507 	int send, sent = 0;
1508 	int blocks, rc = 0;
1509 
1510 	if (unlikely(dlen > INT_MAX))
1511 		return -EMSGSIZE;
1512 
1513 	/* Handle implicit connection setup */
1514 	if (unlikely(dest)) {
1515 		rc = __tipc_sendmsg(sock, m, dlen);
1516 		if (dlen && dlen == rc) {
1517 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1518 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1519 		}
1520 		return rc;
1521 	}
1522 
1523 	do {
1524 		rc = tipc_wait_for_cond(sock, &timeout,
1525 					(!tsk->cong_link_cnt &&
1526 					 !tsk_conn_cong(tsk) &&
1527 					 tipc_sk_connected(sk)));
1528 		if (unlikely(rc))
1529 			break;
1530 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1531 		blocks = tsk->snd_backlog;
1532 		if (tsk->oneway++ >= 4 && send <= maxnagle) {
1533 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1534 			if (unlikely(rc < 0))
1535 				break;
1536 			blocks += rc;
1537 			if (blocks <= 64 && tsk->expect_ack) {
1538 				tsk->snd_backlog = blocks;
1539 				sent += send;
1540 				break;
1541 			}
1542 			tsk->expect_ack = true;
1543 		} else {
1544 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1545 			if (unlikely(rc != send))
1546 				break;
1547 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1548 		}
1549 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1550 					 TIPC_DUMP_SK_SNDQ, " ");
1551 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1552 		if (unlikely(rc == -ELINKCONG)) {
1553 			tsk->cong_link_cnt = 1;
1554 			rc = 0;
1555 		}
1556 		if (likely(!rc)) {
1557 			tsk->snt_unacked += blocks;
1558 			tsk->snd_backlog = 0;
1559 			sent += send;
1560 		}
1561 	} while (sent < dlen && !rc);
1562 
1563 	return sent ? sent : rc;
1564 }
1565 
1566 /**
1567  * tipc_send_packet - send a connection-oriented message
1568  * @sock: socket structure
1569  * @m: message to send
1570  * @dsz: length of data to be transmitted
1571  *
1572  * Used for SOCK_SEQPACKET messages.
1573  *
1574  * Returns the number of bytes sent on success, or errno otherwise
1575  */
1576 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1577 {
1578 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1579 		return -EMSGSIZE;
1580 
1581 	return tipc_sendstream(sock, m, dsz);
1582 }
1583 
1584 /* tipc_sk_finish_conn - complete the setup of a connection
1585  */
1586 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1587 				u32 peer_node)
1588 {
1589 	struct sock *sk = &tsk->sk;
1590 	struct net *net = sock_net(sk);
1591 	struct tipc_msg *msg = &tsk->phdr;
1592 
1593 	msg_set_syn(msg, 0);
1594 	msg_set_destnode(msg, peer_node);
1595 	msg_set_destport(msg, peer_port);
1596 	msg_set_type(msg, TIPC_CONN_MSG);
1597 	msg_set_lookup_scope(msg, 0);
1598 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1599 
1600 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1601 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1602 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1603 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1604 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1605 	tsk_set_nagle(tsk);
1606 	__skb_queue_purge(&sk->sk_write_queue);
1607 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1608 		return;
1609 
1610 	/* Fall back to message based flow control */
1611 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1612 	tsk->snd_win = FLOWCTL_MSG_WIN;
1613 }
1614 
1615 /**
1616  * tipc_sk_set_orig_addr - capture sender's address for received message
1617  * @m: descriptor for message info
1618  * @hdr: received message header
1619  *
1620  * Note: Address is not captured if not requested by receiver.
1621  */
1622 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1623 {
1624 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1625 	struct tipc_msg *hdr = buf_msg(skb);
1626 
1627 	if (!srcaddr)
1628 		return;
1629 
1630 	srcaddr->sock.family = AF_TIPC;
1631 	srcaddr->sock.addrtype = TIPC_ADDR_ID;
1632 	srcaddr->sock.scope = 0;
1633 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1634 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1635 	srcaddr->sock.addr.name.domain = 0;
1636 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1637 
1638 	if (!msg_in_group(hdr))
1639 		return;
1640 
1641 	/* Group message users may also want to know sending member's id */
1642 	srcaddr->member.family = AF_TIPC;
1643 	srcaddr->member.addrtype = TIPC_ADDR_NAME;
1644 	srcaddr->member.scope = 0;
1645 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1646 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1647 	srcaddr->member.addr.name.domain = 0;
1648 	m->msg_namelen = sizeof(*srcaddr);
1649 }
1650 
1651 /**
1652  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1653  * @m: descriptor for message info
1654  * @skb: received message buffer
1655  * @tsk: TIPC port associated with message
1656  *
1657  * Note: Ancillary data is not captured if not requested by receiver.
1658  *
1659  * Returns 0 if successful, otherwise errno
1660  */
1661 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1662 				 struct tipc_sock *tsk)
1663 {
1664 	struct tipc_msg *msg;
1665 	u32 anc_data[3];
1666 	u32 err;
1667 	u32 dest_type;
1668 	int has_name;
1669 	int res;
1670 
1671 	if (likely(m->msg_controllen == 0))
1672 		return 0;
1673 	msg = buf_msg(skb);
1674 
1675 	/* Optionally capture errored message object(s) */
1676 	err = msg ? msg_errcode(msg) : 0;
1677 	if (unlikely(err)) {
1678 		anc_data[0] = err;
1679 		anc_data[1] = msg_data_sz(msg);
1680 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1681 		if (res)
1682 			return res;
1683 		if (anc_data[1]) {
1684 			if (skb_linearize(skb))
1685 				return -ENOMEM;
1686 			msg = buf_msg(skb);
1687 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1688 				       msg_data(msg));
1689 			if (res)
1690 				return res;
1691 		}
1692 	}
1693 
1694 	/* Optionally capture message destination object */
1695 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1696 	switch (dest_type) {
1697 	case TIPC_NAMED_MSG:
1698 		has_name = 1;
1699 		anc_data[0] = msg_nametype(msg);
1700 		anc_data[1] = msg_namelower(msg);
1701 		anc_data[2] = msg_namelower(msg);
1702 		break;
1703 	case TIPC_MCAST_MSG:
1704 		has_name = 1;
1705 		anc_data[0] = msg_nametype(msg);
1706 		anc_data[1] = msg_namelower(msg);
1707 		anc_data[2] = msg_nameupper(msg);
1708 		break;
1709 	case TIPC_CONN_MSG:
1710 		has_name = (tsk->conn_type != 0);
1711 		anc_data[0] = tsk->conn_type;
1712 		anc_data[1] = tsk->conn_instance;
1713 		anc_data[2] = tsk->conn_instance;
1714 		break;
1715 	default:
1716 		has_name = 0;
1717 	}
1718 	if (has_name) {
1719 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1720 		if (res)
1721 			return res;
1722 	}
1723 
1724 	return 0;
1725 }
1726 
1727 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1728 {
1729 	struct sock *sk = &tsk->sk;
1730 	struct net *net = sock_net(sk);
1731 	struct sk_buff *skb = NULL;
1732 	struct tipc_msg *msg;
1733 	u32 peer_port = tsk_peer_port(tsk);
1734 	u32 dnode = tsk_peer_node(tsk);
1735 
1736 	if (!tipc_sk_connected(sk))
1737 		return;
1738 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1739 			      dnode, tsk_own_node(tsk), peer_port,
1740 			      tsk->portid, TIPC_OK);
1741 	if (!skb)
1742 		return;
1743 	msg = buf_msg(skb);
1744 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1745 	tsk->rcv_unacked = 0;
1746 
1747 	/* Adjust to and advertize the correct window limit */
1748 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1749 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1750 		msg_set_adv_win(msg, tsk->rcv_win);
1751 	}
1752 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1753 }
1754 
1755 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1756 {
1757 	struct sock *sk = sock->sk;
1758 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1759 	long timeo = *timeop;
1760 	int err = sock_error(sk);
1761 
1762 	if (err)
1763 		return err;
1764 
1765 	for (;;) {
1766 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1767 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1768 				err = -ENOTCONN;
1769 				break;
1770 			}
1771 			add_wait_queue(sk_sleep(sk), &wait);
1772 			release_sock(sk);
1773 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1774 			sched_annotate_sleep();
1775 			lock_sock(sk);
1776 			remove_wait_queue(sk_sleep(sk), &wait);
1777 		}
1778 		err = 0;
1779 		if (!skb_queue_empty(&sk->sk_receive_queue))
1780 			break;
1781 		err = -EAGAIN;
1782 		if (!timeo)
1783 			break;
1784 		err = sock_intr_errno(timeo);
1785 		if (signal_pending(current))
1786 			break;
1787 
1788 		err = sock_error(sk);
1789 		if (err)
1790 			break;
1791 	}
1792 	*timeop = timeo;
1793 	return err;
1794 }
1795 
1796 /**
1797  * tipc_recvmsg - receive packet-oriented message
1798  * @m: descriptor for message info
1799  * @buflen: length of user buffer area
1800  * @flags: receive flags
1801  *
1802  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1803  * If the complete message doesn't fit in user area, truncate it.
1804  *
1805  * Returns size of returned message data, errno otherwise
1806  */
1807 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1808 			size_t buflen,	int flags)
1809 {
1810 	struct sock *sk = sock->sk;
1811 	bool connected = !tipc_sk_type_connectionless(sk);
1812 	struct tipc_sock *tsk = tipc_sk(sk);
1813 	int rc, err, hlen, dlen, copy;
1814 	struct sk_buff_head xmitq;
1815 	struct tipc_msg *hdr;
1816 	struct sk_buff *skb;
1817 	bool grp_evt;
1818 	long timeout;
1819 
1820 	/* Catch invalid receive requests */
1821 	if (unlikely(!buflen))
1822 		return -EINVAL;
1823 
1824 	lock_sock(sk);
1825 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1826 		rc = -ENOTCONN;
1827 		goto exit;
1828 	}
1829 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1830 
1831 	/* Step rcv queue to first msg with data or error; wait if necessary */
1832 	do {
1833 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1834 		if (unlikely(rc))
1835 			goto exit;
1836 		skb = skb_peek(&sk->sk_receive_queue);
1837 		hdr = buf_msg(skb);
1838 		dlen = msg_data_sz(hdr);
1839 		hlen = msg_hdr_sz(hdr);
1840 		err = msg_errcode(hdr);
1841 		grp_evt = msg_is_grp_evt(hdr);
1842 		if (likely(dlen || err))
1843 			break;
1844 		tsk_advance_rx_queue(sk);
1845 	} while (1);
1846 
1847 	/* Collect msg meta data, including error code and rejected data */
1848 	tipc_sk_set_orig_addr(m, skb);
1849 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1850 	if (unlikely(rc))
1851 		goto exit;
1852 	hdr = buf_msg(skb);
1853 
1854 	/* Capture data if non-error msg, otherwise just set return value */
1855 	if (likely(!err)) {
1856 		copy = min_t(int, dlen, buflen);
1857 		if (unlikely(copy != dlen))
1858 			m->msg_flags |= MSG_TRUNC;
1859 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1860 	} else {
1861 		copy = 0;
1862 		rc = 0;
1863 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1864 			rc = -ECONNRESET;
1865 	}
1866 	if (unlikely(rc))
1867 		goto exit;
1868 
1869 	/* Mark message as group event if applicable */
1870 	if (unlikely(grp_evt)) {
1871 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1872 			m->msg_flags |= MSG_EOR;
1873 		m->msg_flags |= MSG_OOB;
1874 		copy = 0;
1875 	}
1876 
1877 	/* Caption of data or error code/rejected data was successful */
1878 	if (unlikely(flags & MSG_PEEK))
1879 		goto exit;
1880 
1881 	/* Send group flow control advertisement when applicable */
1882 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1883 		__skb_queue_head_init(&xmitq);
1884 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1885 					  msg_orignode(hdr), msg_origport(hdr),
1886 					  &xmitq);
1887 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1888 	}
1889 
1890 	tsk_advance_rx_queue(sk);
1891 
1892 	if (likely(!connected))
1893 		goto exit;
1894 
1895 	/* Send connection flow control advertisement when applicable */
1896 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1897 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1898 		tipc_sk_send_ack(tsk);
1899 exit:
1900 	release_sock(sk);
1901 	return rc ? rc : copy;
1902 }
1903 
1904 /**
1905  * tipc_recvstream - receive stream-oriented data
1906  * @m: descriptor for message info
1907  * @buflen: total size of user buffer area
1908  * @flags: receive flags
1909  *
1910  * Used for SOCK_STREAM messages only.  If not enough data is available
1911  * will optionally wait for more; never truncates data.
1912  *
1913  * Returns size of returned message data, errno otherwise
1914  */
1915 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1916 			   size_t buflen, int flags)
1917 {
1918 	struct sock *sk = sock->sk;
1919 	struct tipc_sock *tsk = tipc_sk(sk);
1920 	struct sk_buff *skb;
1921 	struct tipc_msg *hdr;
1922 	struct tipc_skb_cb *skb_cb;
1923 	bool peek = flags & MSG_PEEK;
1924 	int offset, required, copy, copied = 0;
1925 	int hlen, dlen, err, rc;
1926 	bool ack = false;
1927 	long timeout;
1928 
1929 	/* Catch invalid receive attempts */
1930 	if (unlikely(!buflen))
1931 		return -EINVAL;
1932 
1933 	lock_sock(sk);
1934 
1935 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
1936 		rc = -ENOTCONN;
1937 		goto exit;
1938 	}
1939 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1940 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1941 
1942 	do {
1943 		/* Look at first msg in receive queue; wait if necessary */
1944 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1945 		if (unlikely(rc))
1946 			break;
1947 		skb = skb_peek(&sk->sk_receive_queue);
1948 		skb_cb = TIPC_SKB_CB(skb);
1949 		hdr = buf_msg(skb);
1950 		dlen = msg_data_sz(hdr);
1951 		hlen = msg_hdr_sz(hdr);
1952 		err = msg_errcode(hdr);
1953 
1954 		/* Discard any empty non-errored (SYN-) message */
1955 		if (unlikely(!dlen && !err)) {
1956 			tsk_advance_rx_queue(sk);
1957 			continue;
1958 		}
1959 
1960 		/* Collect msg meta data, incl. error code and rejected data */
1961 		if (!copied) {
1962 			tipc_sk_set_orig_addr(m, skb);
1963 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
1964 			if (rc)
1965 				break;
1966 			hdr = buf_msg(skb);
1967 		}
1968 
1969 		/* Copy data if msg ok, otherwise return error/partial data */
1970 		if (likely(!err)) {
1971 			ack = msg_ack_required(hdr);
1972 			offset = skb_cb->bytes_read;
1973 			copy = min_t(int, dlen - offset, buflen - copied);
1974 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1975 			if (unlikely(rc))
1976 				break;
1977 			copied += copy;
1978 			offset += copy;
1979 			if (unlikely(offset < dlen)) {
1980 				if (!peek)
1981 					skb_cb->bytes_read = offset;
1982 				break;
1983 			}
1984 		} else {
1985 			rc = 0;
1986 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1987 				rc = -ECONNRESET;
1988 			if (copied || rc)
1989 				break;
1990 		}
1991 
1992 		if (unlikely(peek))
1993 			break;
1994 
1995 		tsk_advance_rx_queue(sk);
1996 
1997 		/* Send connection flow control advertisement when applicable */
1998 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1999 		if (ack || tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2000 			tipc_sk_send_ack(tsk);
2001 
2002 		/* Exit if all requested data or FIN/error received */
2003 		if (copied == buflen || err)
2004 			break;
2005 
2006 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2007 exit:
2008 	release_sock(sk);
2009 	return copied ? copied : rc;
2010 }
2011 
2012 /**
2013  * tipc_write_space - wake up thread if port congestion is released
2014  * @sk: socket
2015  */
2016 static void tipc_write_space(struct sock *sk)
2017 {
2018 	struct socket_wq *wq;
2019 
2020 	rcu_read_lock();
2021 	wq = rcu_dereference(sk->sk_wq);
2022 	if (skwq_has_sleeper(wq))
2023 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2024 						EPOLLWRNORM | EPOLLWRBAND);
2025 	rcu_read_unlock();
2026 }
2027 
2028 /**
2029  * tipc_data_ready - wake up threads to indicate messages have been received
2030  * @sk: socket
2031  * @len: the length of messages
2032  */
2033 static void tipc_data_ready(struct sock *sk)
2034 {
2035 	struct socket_wq *wq;
2036 
2037 	rcu_read_lock();
2038 	wq = rcu_dereference(sk->sk_wq);
2039 	if (skwq_has_sleeper(wq))
2040 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2041 						EPOLLRDNORM | EPOLLRDBAND);
2042 	rcu_read_unlock();
2043 }
2044 
2045 static void tipc_sock_destruct(struct sock *sk)
2046 {
2047 	__skb_queue_purge(&sk->sk_receive_queue);
2048 }
2049 
2050 static void tipc_sk_proto_rcv(struct sock *sk,
2051 			      struct sk_buff_head *inputq,
2052 			      struct sk_buff_head *xmitq)
2053 {
2054 	struct sk_buff *skb = __skb_dequeue(inputq);
2055 	struct tipc_sock *tsk = tipc_sk(sk);
2056 	struct tipc_msg *hdr = buf_msg(skb);
2057 	struct tipc_group *grp = tsk->group;
2058 	bool wakeup = false;
2059 
2060 	switch (msg_user(hdr)) {
2061 	case CONN_MANAGER:
2062 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2063 		return;
2064 	case SOCK_WAKEUP:
2065 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2066 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2067 		smp_wmb();
2068 		tsk->cong_link_cnt--;
2069 		wakeup = true;
2070 		tipc_sk_push_backlog(tsk);
2071 		break;
2072 	case GROUP_PROTOCOL:
2073 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2074 		break;
2075 	case TOP_SRV:
2076 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2077 				      hdr, inputq, xmitq);
2078 		break;
2079 	default:
2080 		break;
2081 	}
2082 
2083 	if (wakeup)
2084 		sk->sk_write_space(sk);
2085 
2086 	kfree_skb(skb);
2087 }
2088 
2089 /**
2090  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2091  * @tsk: TIPC socket
2092  * @skb: pointer to message buffer.
2093  * Returns true if message should be added to receive queue, false otherwise
2094  */
2095 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2096 {
2097 	struct sock *sk = &tsk->sk;
2098 	struct net *net = sock_net(sk);
2099 	struct tipc_msg *hdr = buf_msg(skb);
2100 	bool con_msg = msg_connected(hdr);
2101 	u32 pport = tsk_peer_port(tsk);
2102 	u32 pnode = tsk_peer_node(tsk);
2103 	u32 oport = msg_origport(hdr);
2104 	u32 onode = msg_orignode(hdr);
2105 	int err = msg_errcode(hdr);
2106 	unsigned long delay;
2107 
2108 	if (unlikely(msg_mcast(hdr)))
2109 		return false;
2110 	tsk->oneway = 0;
2111 
2112 	switch (sk->sk_state) {
2113 	case TIPC_CONNECTING:
2114 		/* Setup ACK */
2115 		if (likely(con_msg)) {
2116 			if (err)
2117 				break;
2118 			tipc_sk_finish_conn(tsk, oport, onode);
2119 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2120 			/* ACK+ message with data is added to receive queue */
2121 			if (msg_data_sz(hdr))
2122 				return true;
2123 			/* Empty ACK-, - wake up sleeping connect() and drop */
2124 			sk->sk_state_change(sk);
2125 			msg_set_dest_droppable(hdr, 1);
2126 			return false;
2127 		}
2128 		/* Ignore connectionless message if not from listening socket */
2129 		if (oport != pport || onode != pnode)
2130 			return false;
2131 
2132 		/* Rejected SYN */
2133 		if (err != TIPC_ERR_OVERLOAD)
2134 			break;
2135 
2136 		/* Prepare for new setup attempt if we have a SYN clone */
2137 		if (skb_queue_empty(&sk->sk_write_queue))
2138 			break;
2139 		get_random_bytes(&delay, 2);
2140 		delay %= (tsk->conn_timeout / 4);
2141 		delay = msecs_to_jiffies(delay + 100);
2142 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2143 		return false;
2144 	case TIPC_OPEN:
2145 	case TIPC_DISCONNECTING:
2146 		return false;
2147 	case TIPC_LISTEN:
2148 		/* Accept only SYN message */
2149 		if (!msg_is_syn(hdr) &&
2150 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2151 			return false;
2152 		if (!con_msg && !err)
2153 			return true;
2154 		return false;
2155 	case TIPC_ESTABLISHED:
2156 		if (!skb_queue_empty(&sk->sk_write_queue))
2157 			tipc_sk_push_backlog(tsk);
2158 		/* Accept only connection-based messages sent by peer */
2159 		if (likely(con_msg && !err && pport == oport && pnode == onode))
2160 			return true;
2161 		if (!tsk_peer_msg(tsk, hdr))
2162 			return false;
2163 		if (!err)
2164 			return true;
2165 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2166 		tipc_node_remove_conn(net, pnode, tsk->portid);
2167 		sk->sk_state_change(sk);
2168 		return true;
2169 	default:
2170 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2171 	}
2172 	/* Abort connection setup attempt */
2173 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2174 	sk->sk_err = ECONNREFUSED;
2175 	sk->sk_state_change(sk);
2176 	return true;
2177 }
2178 
2179 /**
2180  * rcvbuf_limit - get proper overload limit of socket receive queue
2181  * @sk: socket
2182  * @skb: message
2183  *
2184  * For connection oriented messages, irrespective of importance,
2185  * default queue limit is 2 MB.
2186  *
2187  * For connectionless messages, queue limits are based on message
2188  * importance as follows:
2189  *
2190  * TIPC_LOW_IMPORTANCE       (2 MB)
2191  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2192  * TIPC_HIGH_IMPORTANCE      (8 MB)
2193  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2194  *
2195  * Returns overload limit according to corresponding message importance
2196  */
2197 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2198 {
2199 	struct tipc_sock *tsk = tipc_sk(sk);
2200 	struct tipc_msg *hdr = buf_msg(skb);
2201 
2202 	if (unlikely(msg_in_group(hdr)))
2203 		return READ_ONCE(sk->sk_rcvbuf);
2204 
2205 	if (unlikely(!msg_connected(hdr)))
2206 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2207 
2208 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2209 		return READ_ONCE(sk->sk_rcvbuf);
2210 
2211 	return FLOWCTL_MSG_LIM;
2212 }
2213 
2214 /**
2215  * tipc_sk_filter_rcv - validate incoming message
2216  * @sk: socket
2217  * @skb: pointer to message.
2218  *
2219  * Enqueues message on receive queue if acceptable; optionally handles
2220  * disconnect indication for a connected socket.
2221  *
2222  * Called with socket lock already taken
2223  *
2224  */
2225 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2226 			       struct sk_buff_head *xmitq)
2227 {
2228 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2229 	struct tipc_sock *tsk = tipc_sk(sk);
2230 	struct tipc_group *grp = tsk->group;
2231 	struct tipc_msg *hdr = buf_msg(skb);
2232 	struct net *net = sock_net(sk);
2233 	struct sk_buff_head inputq;
2234 	int mtyp = msg_type(hdr);
2235 	int limit, err = TIPC_OK;
2236 
2237 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2238 	TIPC_SKB_CB(skb)->bytes_read = 0;
2239 	__skb_queue_head_init(&inputq);
2240 	__skb_queue_tail(&inputq, skb);
2241 
2242 	if (unlikely(!msg_isdata(hdr)))
2243 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2244 
2245 	if (unlikely(grp))
2246 		tipc_group_filter_msg(grp, &inputq, xmitq);
2247 
2248 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2249 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2250 
2251 	/* Validate and add to receive buffer if there is space */
2252 	while ((skb = __skb_dequeue(&inputq))) {
2253 		hdr = buf_msg(skb);
2254 		limit = rcvbuf_limit(sk, skb);
2255 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2256 		    (!sk_conn && msg_connected(hdr)) ||
2257 		    (!grp && msg_in_group(hdr)))
2258 			err = TIPC_ERR_NO_PORT;
2259 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2260 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2261 					   "err_overload2!");
2262 			atomic_inc(&sk->sk_drops);
2263 			err = TIPC_ERR_OVERLOAD;
2264 		}
2265 
2266 		if (unlikely(err)) {
2267 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2268 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2269 						      "@filter_rcv!");
2270 				__skb_queue_tail(xmitq, skb);
2271 			}
2272 			err = TIPC_OK;
2273 			continue;
2274 		}
2275 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2276 		skb_set_owner_r(skb, sk);
2277 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2278 					 "rcvq >90% allocated!");
2279 		sk->sk_data_ready(sk);
2280 	}
2281 }
2282 
2283 /**
2284  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2285  * @sk: socket
2286  * @skb: message
2287  *
2288  * Caller must hold socket lock
2289  */
2290 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2291 {
2292 	unsigned int before = sk_rmem_alloc_get(sk);
2293 	struct sk_buff_head xmitq;
2294 	unsigned int added;
2295 
2296 	__skb_queue_head_init(&xmitq);
2297 
2298 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2299 	added = sk_rmem_alloc_get(sk) - before;
2300 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2301 
2302 	/* Send pending response/rejected messages, if any */
2303 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2304 	return 0;
2305 }
2306 
2307 /**
2308  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2309  *                   inputq and try adding them to socket or backlog queue
2310  * @inputq: list of incoming buffers with potentially different destinations
2311  * @sk: socket where the buffers should be enqueued
2312  * @dport: port number for the socket
2313  *
2314  * Caller must hold socket lock
2315  */
2316 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2317 			    u32 dport, struct sk_buff_head *xmitq)
2318 {
2319 	unsigned long time_limit = jiffies + 2;
2320 	struct sk_buff *skb;
2321 	unsigned int lim;
2322 	atomic_t *dcnt;
2323 	u32 onode;
2324 
2325 	while (skb_queue_len(inputq)) {
2326 		if (unlikely(time_after_eq(jiffies, time_limit)))
2327 			return;
2328 
2329 		skb = tipc_skb_dequeue(inputq, dport);
2330 		if (unlikely(!skb))
2331 			return;
2332 
2333 		/* Add message directly to receive queue if possible */
2334 		if (!sock_owned_by_user(sk)) {
2335 			tipc_sk_filter_rcv(sk, skb, xmitq);
2336 			continue;
2337 		}
2338 
2339 		/* Try backlog, compensating for double-counted bytes */
2340 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2341 		if (!sk->sk_backlog.len)
2342 			atomic_set(dcnt, 0);
2343 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2344 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2345 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2346 						 "bklg & rcvq >90% allocated!");
2347 			continue;
2348 		}
2349 
2350 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2351 		/* Overload => reject message back to sender */
2352 		onode = tipc_own_addr(sock_net(sk));
2353 		atomic_inc(&sk->sk_drops);
2354 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2355 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2356 					      "@sk_enqueue!");
2357 			__skb_queue_tail(xmitq, skb);
2358 		}
2359 		break;
2360 	}
2361 }
2362 
2363 /**
2364  * tipc_sk_rcv - handle a chain of incoming buffers
2365  * @inputq: buffer list containing the buffers
2366  * Consumes all buffers in list until inputq is empty
2367  * Note: may be called in multiple threads referring to the same queue
2368  */
2369 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2370 {
2371 	struct sk_buff_head xmitq;
2372 	u32 dnode, dport = 0;
2373 	int err;
2374 	struct tipc_sock *tsk;
2375 	struct sock *sk;
2376 	struct sk_buff *skb;
2377 
2378 	__skb_queue_head_init(&xmitq);
2379 	while (skb_queue_len(inputq)) {
2380 		dport = tipc_skb_peek_port(inputq, dport);
2381 		tsk = tipc_sk_lookup(net, dport);
2382 
2383 		if (likely(tsk)) {
2384 			sk = &tsk->sk;
2385 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2386 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2387 				spin_unlock_bh(&sk->sk_lock.slock);
2388 			}
2389 			/* Send pending response/rejected messages, if any */
2390 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2391 			sock_put(sk);
2392 			continue;
2393 		}
2394 		/* No destination socket => dequeue skb if still there */
2395 		skb = tipc_skb_dequeue(inputq, dport);
2396 		if (!skb)
2397 			return;
2398 
2399 		/* Try secondary lookup if unresolved named message */
2400 		err = TIPC_ERR_NO_PORT;
2401 		if (tipc_msg_lookup_dest(net, skb, &err))
2402 			goto xmit;
2403 
2404 		/* Prepare for message rejection */
2405 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2406 			continue;
2407 
2408 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2409 xmit:
2410 		dnode = msg_destnode(buf_msg(skb));
2411 		tipc_node_xmit_skb(net, skb, dnode, dport);
2412 	}
2413 }
2414 
2415 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2416 {
2417 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2418 	struct sock *sk = sock->sk;
2419 	int done;
2420 
2421 	do {
2422 		int err = sock_error(sk);
2423 		if (err)
2424 			return err;
2425 		if (!*timeo_p)
2426 			return -ETIMEDOUT;
2427 		if (signal_pending(current))
2428 			return sock_intr_errno(*timeo_p);
2429 
2430 		add_wait_queue(sk_sleep(sk), &wait);
2431 		done = sk_wait_event(sk, timeo_p,
2432 				     sk->sk_state != TIPC_CONNECTING, &wait);
2433 		remove_wait_queue(sk_sleep(sk), &wait);
2434 	} while (!done);
2435 	return 0;
2436 }
2437 
2438 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2439 {
2440 	if (addr->family != AF_TIPC)
2441 		return false;
2442 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2443 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2444 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2445 		addr->addrtype == TIPC_SOCKET_ADDR);
2446 }
2447 
2448 /**
2449  * tipc_connect - establish a connection to another TIPC port
2450  * @sock: socket structure
2451  * @dest: socket address for destination port
2452  * @destlen: size of socket address data structure
2453  * @flags: file-related flags associated with socket
2454  *
2455  * Returns 0 on success, errno otherwise
2456  */
2457 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2458 			int destlen, int flags)
2459 {
2460 	struct sock *sk = sock->sk;
2461 	struct tipc_sock *tsk = tipc_sk(sk);
2462 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2463 	struct msghdr m = {NULL,};
2464 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2465 	int previous;
2466 	int res = 0;
2467 
2468 	if (destlen != sizeof(struct sockaddr_tipc))
2469 		return -EINVAL;
2470 
2471 	lock_sock(sk);
2472 
2473 	if (tsk->group) {
2474 		res = -EINVAL;
2475 		goto exit;
2476 	}
2477 
2478 	if (dst->family == AF_UNSPEC) {
2479 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2480 		if (!tipc_sk_type_connectionless(sk))
2481 			res = -EINVAL;
2482 		goto exit;
2483 	}
2484 	if (!tipc_sockaddr_is_sane(dst)) {
2485 		res = -EINVAL;
2486 		goto exit;
2487 	}
2488 	/* DGRAM/RDM connect(), just save the destaddr */
2489 	if (tipc_sk_type_connectionless(sk)) {
2490 		memcpy(&tsk->peer, dest, destlen);
2491 		goto exit;
2492 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2493 		res = -EINVAL;
2494 		goto exit;
2495 	}
2496 
2497 	previous = sk->sk_state;
2498 
2499 	switch (sk->sk_state) {
2500 	case TIPC_OPEN:
2501 		/* Send a 'SYN-' to destination */
2502 		m.msg_name = dest;
2503 		m.msg_namelen = destlen;
2504 
2505 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2506 		 * indicate send_msg() is never blocked.
2507 		 */
2508 		if (!timeout)
2509 			m.msg_flags = MSG_DONTWAIT;
2510 
2511 		res = __tipc_sendmsg(sock, &m, 0);
2512 		if ((res < 0) && (res != -EWOULDBLOCK))
2513 			goto exit;
2514 
2515 		/* Just entered TIPC_CONNECTING state; the only
2516 		 * difference is that return value in non-blocking
2517 		 * case is EINPROGRESS, rather than EALREADY.
2518 		 */
2519 		res = -EINPROGRESS;
2520 		/* fall through */
2521 	case TIPC_CONNECTING:
2522 		if (!timeout) {
2523 			if (previous == TIPC_CONNECTING)
2524 				res = -EALREADY;
2525 			goto exit;
2526 		}
2527 		timeout = msecs_to_jiffies(timeout);
2528 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2529 		res = tipc_wait_for_connect(sock, &timeout);
2530 		break;
2531 	case TIPC_ESTABLISHED:
2532 		res = -EISCONN;
2533 		break;
2534 	default:
2535 		res = -EINVAL;
2536 	}
2537 
2538 exit:
2539 	release_sock(sk);
2540 	return res;
2541 }
2542 
2543 /**
2544  * tipc_listen - allow socket to listen for incoming connections
2545  * @sock: socket structure
2546  * @len: (unused)
2547  *
2548  * Returns 0 on success, errno otherwise
2549  */
2550 static int tipc_listen(struct socket *sock, int len)
2551 {
2552 	struct sock *sk = sock->sk;
2553 	int res;
2554 
2555 	lock_sock(sk);
2556 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2557 	release_sock(sk);
2558 
2559 	return res;
2560 }
2561 
2562 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2563 {
2564 	struct sock *sk = sock->sk;
2565 	DEFINE_WAIT(wait);
2566 	int err;
2567 
2568 	/* True wake-one mechanism for incoming connections: only
2569 	 * one process gets woken up, not the 'whole herd'.
2570 	 * Since we do not 'race & poll' for established sockets
2571 	 * anymore, the common case will execute the loop only once.
2572 	*/
2573 	for (;;) {
2574 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2575 					  TASK_INTERRUPTIBLE);
2576 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2577 			release_sock(sk);
2578 			timeo = schedule_timeout(timeo);
2579 			lock_sock(sk);
2580 		}
2581 		err = 0;
2582 		if (!skb_queue_empty(&sk->sk_receive_queue))
2583 			break;
2584 		err = -EAGAIN;
2585 		if (!timeo)
2586 			break;
2587 		err = sock_intr_errno(timeo);
2588 		if (signal_pending(current))
2589 			break;
2590 	}
2591 	finish_wait(sk_sleep(sk), &wait);
2592 	return err;
2593 }
2594 
2595 /**
2596  * tipc_accept - wait for connection request
2597  * @sock: listening socket
2598  * @newsock: new socket that is to be connected
2599  * @flags: file-related flags associated with socket
2600  *
2601  * Returns 0 on success, errno otherwise
2602  */
2603 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2604 		       bool kern)
2605 {
2606 	struct sock *new_sk, *sk = sock->sk;
2607 	struct sk_buff *buf;
2608 	struct tipc_sock *new_tsock;
2609 	struct tipc_msg *msg;
2610 	long timeo;
2611 	int res;
2612 
2613 	lock_sock(sk);
2614 
2615 	if (sk->sk_state != TIPC_LISTEN) {
2616 		res = -EINVAL;
2617 		goto exit;
2618 	}
2619 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2620 	res = tipc_wait_for_accept(sock, timeo);
2621 	if (res)
2622 		goto exit;
2623 
2624 	buf = skb_peek(&sk->sk_receive_queue);
2625 
2626 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2627 	if (res)
2628 		goto exit;
2629 	security_sk_clone(sock->sk, new_sock->sk);
2630 
2631 	new_sk = new_sock->sk;
2632 	new_tsock = tipc_sk(new_sk);
2633 	msg = buf_msg(buf);
2634 
2635 	/* we lock on new_sk; but lockdep sees the lock on sk */
2636 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2637 
2638 	/*
2639 	 * Reject any stray messages received by new socket
2640 	 * before the socket lock was taken (very, very unlikely)
2641 	 */
2642 	tsk_rej_rx_queue(new_sk);
2643 
2644 	/* Connect new socket to it's peer */
2645 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2646 
2647 	tsk_set_importance(new_tsock, msg_importance(msg));
2648 	if (msg_named(msg)) {
2649 		new_tsock->conn_type = msg_nametype(msg);
2650 		new_tsock->conn_instance = msg_nameinst(msg);
2651 	}
2652 
2653 	/*
2654 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2655 	 * Respond to 'SYN+' by queuing it on new socket.
2656 	 */
2657 	if (!msg_data_sz(msg)) {
2658 		struct msghdr m = {NULL,};
2659 
2660 		tsk_advance_rx_queue(sk);
2661 		__tipc_sendstream(new_sock, &m, 0);
2662 	} else {
2663 		__skb_dequeue(&sk->sk_receive_queue);
2664 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2665 		skb_set_owner_r(buf, new_sk);
2666 	}
2667 	release_sock(new_sk);
2668 exit:
2669 	release_sock(sk);
2670 	return res;
2671 }
2672 
2673 /**
2674  * tipc_shutdown - shutdown socket connection
2675  * @sock: socket structure
2676  * @how: direction to close (must be SHUT_RDWR)
2677  *
2678  * Terminates connection (if necessary), then purges socket's receive queue.
2679  *
2680  * Returns 0 on success, errno otherwise
2681  */
2682 static int tipc_shutdown(struct socket *sock, int how)
2683 {
2684 	struct sock *sk = sock->sk;
2685 	int res;
2686 
2687 	if (how != SHUT_RDWR)
2688 		return -EINVAL;
2689 
2690 	lock_sock(sk);
2691 
2692 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2693 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2694 	sk->sk_shutdown = SEND_SHUTDOWN;
2695 
2696 	if (sk->sk_state == TIPC_DISCONNECTING) {
2697 		/* Discard any unreceived messages */
2698 		__skb_queue_purge(&sk->sk_receive_queue);
2699 
2700 		/* Wake up anyone sleeping in poll */
2701 		sk->sk_state_change(sk);
2702 		res = 0;
2703 	} else {
2704 		res = -ENOTCONN;
2705 	}
2706 
2707 	release_sock(sk);
2708 	return res;
2709 }
2710 
2711 static void tipc_sk_check_probing_state(struct sock *sk,
2712 					struct sk_buff_head *list)
2713 {
2714 	struct tipc_sock *tsk = tipc_sk(sk);
2715 	u32 pnode = tsk_peer_node(tsk);
2716 	u32 pport = tsk_peer_port(tsk);
2717 	u32 self = tsk_own_node(tsk);
2718 	u32 oport = tsk->portid;
2719 	struct sk_buff *skb;
2720 
2721 	if (tsk->probe_unacked) {
2722 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2723 		sk->sk_err = ECONNABORTED;
2724 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2725 		sk->sk_state_change(sk);
2726 		return;
2727 	}
2728 	/* Prepare new probe */
2729 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2730 			      pnode, self, pport, oport, TIPC_OK);
2731 	if (skb)
2732 		__skb_queue_tail(list, skb);
2733 	tsk->probe_unacked = true;
2734 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2735 }
2736 
2737 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2738 {
2739 	struct tipc_sock *tsk = tipc_sk(sk);
2740 
2741 	/* Try again later if dest link is congested */
2742 	if (tsk->cong_link_cnt) {
2743 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2744 		return;
2745 	}
2746 	/* Prepare SYN for retransmit */
2747 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2748 }
2749 
2750 static void tipc_sk_timeout(struct timer_list *t)
2751 {
2752 	struct sock *sk = from_timer(sk, t, sk_timer);
2753 	struct tipc_sock *tsk = tipc_sk(sk);
2754 	u32 pnode = tsk_peer_node(tsk);
2755 	struct sk_buff_head list;
2756 	int rc = 0;
2757 
2758 	__skb_queue_head_init(&list);
2759 	bh_lock_sock(sk);
2760 
2761 	/* Try again later if socket is busy */
2762 	if (sock_owned_by_user(sk)) {
2763 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2764 		bh_unlock_sock(sk);
2765 		sock_put(sk);
2766 		return;
2767 	}
2768 
2769 	if (sk->sk_state == TIPC_ESTABLISHED)
2770 		tipc_sk_check_probing_state(sk, &list);
2771 	else if (sk->sk_state == TIPC_CONNECTING)
2772 		tipc_sk_retry_connect(sk, &list);
2773 
2774 	bh_unlock_sock(sk);
2775 
2776 	if (!skb_queue_empty(&list))
2777 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2778 
2779 	/* SYN messages may cause link congestion */
2780 	if (rc == -ELINKCONG) {
2781 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2782 		tsk->cong_link_cnt = 1;
2783 	}
2784 	sock_put(sk);
2785 }
2786 
2787 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2788 			   struct tipc_name_seq const *seq)
2789 {
2790 	struct sock *sk = &tsk->sk;
2791 	struct net *net = sock_net(sk);
2792 	struct publication *publ;
2793 	u32 key;
2794 
2795 	if (scope != TIPC_NODE_SCOPE)
2796 		scope = TIPC_CLUSTER_SCOPE;
2797 
2798 	if (tipc_sk_connected(sk))
2799 		return -EINVAL;
2800 	key = tsk->portid + tsk->pub_count + 1;
2801 	if (key == tsk->portid)
2802 		return -EADDRINUSE;
2803 
2804 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2805 				    scope, tsk->portid, key);
2806 	if (unlikely(!publ))
2807 		return -EINVAL;
2808 
2809 	list_add(&publ->binding_sock, &tsk->publications);
2810 	tsk->pub_count++;
2811 	tsk->published = 1;
2812 	return 0;
2813 }
2814 
2815 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2816 			    struct tipc_name_seq const *seq)
2817 {
2818 	struct net *net = sock_net(&tsk->sk);
2819 	struct publication *publ;
2820 	struct publication *safe;
2821 	int rc = -EINVAL;
2822 
2823 	if (scope != TIPC_NODE_SCOPE)
2824 		scope = TIPC_CLUSTER_SCOPE;
2825 
2826 	list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2827 		if (seq) {
2828 			if (publ->scope != scope)
2829 				continue;
2830 			if (publ->type != seq->type)
2831 				continue;
2832 			if (publ->lower != seq->lower)
2833 				continue;
2834 			if (publ->upper != seq->upper)
2835 				break;
2836 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2837 					      publ->upper, publ->key);
2838 			rc = 0;
2839 			break;
2840 		}
2841 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2842 				      publ->upper, publ->key);
2843 		rc = 0;
2844 	}
2845 	if (list_empty(&tsk->publications))
2846 		tsk->published = 0;
2847 	return rc;
2848 }
2849 
2850 /* tipc_sk_reinit: set non-zero address in all existing sockets
2851  *                 when we go from standalone to network mode.
2852  */
2853 void tipc_sk_reinit(struct net *net)
2854 {
2855 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2856 	struct rhashtable_iter iter;
2857 	struct tipc_sock *tsk;
2858 	struct tipc_msg *msg;
2859 
2860 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2861 
2862 	do {
2863 		rhashtable_walk_start(&iter);
2864 
2865 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2866 			sock_hold(&tsk->sk);
2867 			rhashtable_walk_stop(&iter);
2868 			lock_sock(&tsk->sk);
2869 			msg = &tsk->phdr;
2870 			msg_set_prevnode(msg, tipc_own_addr(net));
2871 			msg_set_orignode(msg, tipc_own_addr(net));
2872 			release_sock(&tsk->sk);
2873 			rhashtable_walk_start(&iter);
2874 			sock_put(&tsk->sk);
2875 		}
2876 
2877 		rhashtable_walk_stop(&iter);
2878 	} while (tsk == ERR_PTR(-EAGAIN));
2879 
2880 	rhashtable_walk_exit(&iter);
2881 }
2882 
2883 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2884 {
2885 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2886 	struct tipc_sock *tsk;
2887 
2888 	rcu_read_lock();
2889 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2890 	if (tsk)
2891 		sock_hold(&tsk->sk);
2892 	rcu_read_unlock();
2893 
2894 	return tsk;
2895 }
2896 
2897 static int tipc_sk_insert(struct tipc_sock *tsk)
2898 {
2899 	struct sock *sk = &tsk->sk;
2900 	struct net *net = sock_net(sk);
2901 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2902 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2903 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2904 
2905 	while (remaining--) {
2906 		portid++;
2907 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2908 			portid = TIPC_MIN_PORT;
2909 		tsk->portid = portid;
2910 		sock_hold(&tsk->sk);
2911 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2912 						   tsk_rht_params))
2913 			return 0;
2914 		sock_put(&tsk->sk);
2915 	}
2916 
2917 	return -1;
2918 }
2919 
2920 static void tipc_sk_remove(struct tipc_sock *tsk)
2921 {
2922 	struct sock *sk = &tsk->sk;
2923 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2924 
2925 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2926 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2927 		__sock_put(sk);
2928 	}
2929 }
2930 
2931 static const struct rhashtable_params tsk_rht_params = {
2932 	.nelem_hint = 192,
2933 	.head_offset = offsetof(struct tipc_sock, node),
2934 	.key_offset = offsetof(struct tipc_sock, portid),
2935 	.key_len = sizeof(u32), /* portid */
2936 	.max_size = 1048576,
2937 	.min_size = 256,
2938 	.automatic_shrinking = true,
2939 };
2940 
2941 int tipc_sk_rht_init(struct net *net)
2942 {
2943 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2944 
2945 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2946 }
2947 
2948 void tipc_sk_rht_destroy(struct net *net)
2949 {
2950 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2951 
2952 	/* Wait for socket readers to complete */
2953 	synchronize_net();
2954 
2955 	rhashtable_destroy(&tn->sk_rht);
2956 }
2957 
2958 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2959 {
2960 	struct net *net = sock_net(&tsk->sk);
2961 	struct tipc_group *grp = tsk->group;
2962 	struct tipc_msg *hdr = &tsk->phdr;
2963 	struct tipc_name_seq seq;
2964 	int rc;
2965 
2966 	if (mreq->type < TIPC_RESERVED_TYPES)
2967 		return -EACCES;
2968 	if (mreq->scope > TIPC_NODE_SCOPE)
2969 		return -EINVAL;
2970 	if (grp)
2971 		return -EACCES;
2972 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2973 	if (!grp)
2974 		return -ENOMEM;
2975 	tsk->group = grp;
2976 	msg_set_lookup_scope(hdr, mreq->scope);
2977 	msg_set_nametype(hdr, mreq->type);
2978 	msg_set_dest_droppable(hdr, true);
2979 	seq.type = mreq->type;
2980 	seq.lower = mreq->instance;
2981 	seq.upper = seq.lower;
2982 	tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2983 	rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2984 	if (rc) {
2985 		tipc_group_delete(net, grp);
2986 		tsk->group = NULL;
2987 		return rc;
2988 	}
2989 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
2990 	tsk->mc_method.rcast = true;
2991 	tsk->mc_method.mandatory = true;
2992 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2993 	return rc;
2994 }
2995 
2996 static int tipc_sk_leave(struct tipc_sock *tsk)
2997 {
2998 	struct net *net = sock_net(&tsk->sk);
2999 	struct tipc_group *grp = tsk->group;
3000 	struct tipc_name_seq seq;
3001 	int scope;
3002 
3003 	if (!grp)
3004 		return -EINVAL;
3005 	tipc_group_self(grp, &seq, &scope);
3006 	tipc_group_delete(net, grp);
3007 	tsk->group = NULL;
3008 	tipc_sk_withdraw(tsk, scope, &seq);
3009 	return 0;
3010 }
3011 
3012 /**
3013  * tipc_setsockopt - set socket option
3014  * @sock: socket structure
3015  * @lvl: option level
3016  * @opt: option identifier
3017  * @ov: pointer to new option value
3018  * @ol: length of option value
3019  *
3020  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3021  * (to ease compatibility).
3022  *
3023  * Returns 0 on success, errno otherwise
3024  */
3025 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3026 			   char __user *ov, unsigned int ol)
3027 {
3028 	struct sock *sk = sock->sk;
3029 	struct tipc_sock *tsk = tipc_sk(sk);
3030 	struct tipc_group_req mreq;
3031 	u32 value = 0;
3032 	int res = 0;
3033 
3034 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3035 		return 0;
3036 	if (lvl != SOL_TIPC)
3037 		return -ENOPROTOOPT;
3038 
3039 	switch (opt) {
3040 	case TIPC_IMPORTANCE:
3041 	case TIPC_SRC_DROPPABLE:
3042 	case TIPC_DEST_DROPPABLE:
3043 	case TIPC_CONN_TIMEOUT:
3044 	case TIPC_NODELAY:
3045 		if (ol < sizeof(value))
3046 			return -EINVAL;
3047 		if (get_user(value, (u32 __user *)ov))
3048 			return -EFAULT;
3049 		break;
3050 	case TIPC_GROUP_JOIN:
3051 		if (ol < sizeof(mreq))
3052 			return -EINVAL;
3053 		if (copy_from_user(&mreq, ov, sizeof(mreq)))
3054 			return -EFAULT;
3055 		break;
3056 	default:
3057 		if (ov || ol)
3058 			return -EINVAL;
3059 	}
3060 
3061 	lock_sock(sk);
3062 
3063 	switch (opt) {
3064 	case TIPC_IMPORTANCE:
3065 		res = tsk_set_importance(tsk, value);
3066 		break;
3067 	case TIPC_SRC_DROPPABLE:
3068 		if (sock->type != SOCK_STREAM)
3069 			tsk_set_unreliable(tsk, value);
3070 		else
3071 			res = -ENOPROTOOPT;
3072 		break;
3073 	case TIPC_DEST_DROPPABLE:
3074 		tsk_set_unreturnable(tsk, value);
3075 		break;
3076 	case TIPC_CONN_TIMEOUT:
3077 		tipc_sk(sk)->conn_timeout = value;
3078 		break;
3079 	case TIPC_MCAST_BROADCAST:
3080 		tsk->mc_method.rcast = false;
3081 		tsk->mc_method.mandatory = true;
3082 		break;
3083 	case TIPC_MCAST_REPLICAST:
3084 		tsk->mc_method.rcast = true;
3085 		tsk->mc_method.mandatory = true;
3086 		break;
3087 	case TIPC_GROUP_JOIN:
3088 		res = tipc_sk_join(tsk, &mreq);
3089 		break;
3090 	case TIPC_GROUP_LEAVE:
3091 		res = tipc_sk_leave(tsk);
3092 		break;
3093 	case TIPC_NODELAY:
3094 		tsk->nodelay = !!value;
3095 		tsk_set_nagle(tsk);
3096 		break;
3097 	default:
3098 		res = -EINVAL;
3099 	}
3100 
3101 	release_sock(sk);
3102 
3103 	return res;
3104 }
3105 
3106 /**
3107  * tipc_getsockopt - get socket option
3108  * @sock: socket structure
3109  * @lvl: option level
3110  * @opt: option identifier
3111  * @ov: receptacle for option value
3112  * @ol: receptacle for length of option value
3113  *
3114  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3115  * (to ease compatibility).
3116  *
3117  * Returns 0 on success, errno otherwise
3118  */
3119 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3120 			   char __user *ov, int __user *ol)
3121 {
3122 	struct sock *sk = sock->sk;
3123 	struct tipc_sock *tsk = tipc_sk(sk);
3124 	struct tipc_name_seq seq;
3125 	int len, scope;
3126 	u32 value;
3127 	int res;
3128 
3129 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3130 		return put_user(0, ol);
3131 	if (lvl != SOL_TIPC)
3132 		return -ENOPROTOOPT;
3133 	res = get_user(len, ol);
3134 	if (res)
3135 		return res;
3136 
3137 	lock_sock(sk);
3138 
3139 	switch (opt) {
3140 	case TIPC_IMPORTANCE:
3141 		value = tsk_importance(tsk);
3142 		break;
3143 	case TIPC_SRC_DROPPABLE:
3144 		value = tsk_unreliable(tsk);
3145 		break;
3146 	case TIPC_DEST_DROPPABLE:
3147 		value = tsk_unreturnable(tsk);
3148 		break;
3149 	case TIPC_CONN_TIMEOUT:
3150 		value = tsk->conn_timeout;
3151 		/* no need to set "res", since already 0 at this point */
3152 		break;
3153 	case TIPC_NODE_RECVQ_DEPTH:
3154 		value = 0; /* was tipc_queue_size, now obsolete */
3155 		break;
3156 	case TIPC_SOCK_RECVQ_DEPTH:
3157 		value = skb_queue_len(&sk->sk_receive_queue);
3158 		break;
3159 	case TIPC_SOCK_RECVQ_USED:
3160 		value = sk_rmem_alloc_get(sk);
3161 		break;
3162 	case TIPC_GROUP_JOIN:
3163 		seq.type = 0;
3164 		if (tsk->group)
3165 			tipc_group_self(tsk->group, &seq, &scope);
3166 		value = seq.type;
3167 		break;
3168 	default:
3169 		res = -EINVAL;
3170 	}
3171 
3172 	release_sock(sk);
3173 
3174 	if (res)
3175 		return res;	/* "get" failed */
3176 
3177 	if (len < sizeof(value))
3178 		return -EINVAL;
3179 
3180 	if (copy_to_user(ov, &value, sizeof(value)))
3181 		return -EFAULT;
3182 
3183 	return put_user(sizeof(value), ol);
3184 }
3185 
3186 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3187 {
3188 	struct net *net = sock_net(sock->sk);
3189 	struct tipc_sioc_nodeid_req nr = {0};
3190 	struct tipc_sioc_ln_req lnr;
3191 	void __user *argp = (void __user *)arg;
3192 
3193 	switch (cmd) {
3194 	case SIOCGETLINKNAME:
3195 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3196 			return -EFAULT;
3197 		if (!tipc_node_get_linkname(net,
3198 					    lnr.bearer_id & 0xffff, lnr.peer,
3199 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3200 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3201 				return -EFAULT;
3202 			return 0;
3203 		}
3204 		return -EADDRNOTAVAIL;
3205 	case SIOCGETNODEID:
3206 		if (copy_from_user(&nr, argp, sizeof(nr)))
3207 			return -EFAULT;
3208 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3209 			return -EADDRNOTAVAIL;
3210 		if (copy_to_user(argp, &nr, sizeof(nr)))
3211 			return -EFAULT;
3212 		return 0;
3213 	default:
3214 		return -ENOIOCTLCMD;
3215 	}
3216 }
3217 
3218 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3219 {
3220 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3221 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3222 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3223 
3224 	tsk1->peer.family = AF_TIPC;
3225 	tsk1->peer.addrtype = TIPC_ADDR_ID;
3226 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3227 	tsk1->peer.addr.id.ref = tsk2->portid;
3228 	tsk1->peer.addr.id.node = onode;
3229 	tsk2->peer.family = AF_TIPC;
3230 	tsk2->peer.addrtype = TIPC_ADDR_ID;
3231 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3232 	tsk2->peer.addr.id.ref = tsk1->portid;
3233 	tsk2->peer.addr.id.node = onode;
3234 
3235 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3236 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3237 	return 0;
3238 }
3239 
3240 /* Protocol switches for the various types of TIPC sockets */
3241 
3242 static const struct proto_ops msg_ops = {
3243 	.owner		= THIS_MODULE,
3244 	.family		= AF_TIPC,
3245 	.release	= tipc_release,
3246 	.bind		= tipc_bind,
3247 	.connect	= tipc_connect,
3248 	.socketpair	= tipc_socketpair,
3249 	.accept		= sock_no_accept,
3250 	.getname	= tipc_getname,
3251 	.poll		= tipc_poll,
3252 	.ioctl		= tipc_ioctl,
3253 	.listen		= sock_no_listen,
3254 	.shutdown	= tipc_shutdown,
3255 	.setsockopt	= tipc_setsockopt,
3256 	.getsockopt	= tipc_getsockopt,
3257 	.sendmsg	= tipc_sendmsg,
3258 	.recvmsg	= tipc_recvmsg,
3259 	.mmap		= sock_no_mmap,
3260 	.sendpage	= sock_no_sendpage
3261 };
3262 
3263 static const struct proto_ops packet_ops = {
3264 	.owner		= THIS_MODULE,
3265 	.family		= AF_TIPC,
3266 	.release	= tipc_release,
3267 	.bind		= tipc_bind,
3268 	.connect	= tipc_connect,
3269 	.socketpair	= tipc_socketpair,
3270 	.accept		= tipc_accept,
3271 	.getname	= tipc_getname,
3272 	.poll		= tipc_poll,
3273 	.ioctl		= tipc_ioctl,
3274 	.listen		= tipc_listen,
3275 	.shutdown	= tipc_shutdown,
3276 	.setsockopt	= tipc_setsockopt,
3277 	.getsockopt	= tipc_getsockopt,
3278 	.sendmsg	= tipc_send_packet,
3279 	.recvmsg	= tipc_recvmsg,
3280 	.mmap		= sock_no_mmap,
3281 	.sendpage	= sock_no_sendpage
3282 };
3283 
3284 static const struct proto_ops stream_ops = {
3285 	.owner		= THIS_MODULE,
3286 	.family		= AF_TIPC,
3287 	.release	= tipc_release,
3288 	.bind		= tipc_bind,
3289 	.connect	= tipc_connect,
3290 	.socketpair	= tipc_socketpair,
3291 	.accept		= tipc_accept,
3292 	.getname	= tipc_getname,
3293 	.poll		= tipc_poll,
3294 	.ioctl		= tipc_ioctl,
3295 	.listen		= tipc_listen,
3296 	.shutdown	= tipc_shutdown,
3297 	.setsockopt	= tipc_setsockopt,
3298 	.getsockopt	= tipc_getsockopt,
3299 	.sendmsg	= tipc_sendstream,
3300 	.recvmsg	= tipc_recvstream,
3301 	.mmap		= sock_no_mmap,
3302 	.sendpage	= sock_no_sendpage
3303 };
3304 
3305 static const struct net_proto_family tipc_family_ops = {
3306 	.owner		= THIS_MODULE,
3307 	.family		= AF_TIPC,
3308 	.create		= tipc_sk_create
3309 };
3310 
3311 static struct proto tipc_proto = {
3312 	.name		= "TIPC",
3313 	.owner		= THIS_MODULE,
3314 	.obj_size	= sizeof(struct tipc_sock),
3315 	.sysctl_rmem	= sysctl_tipc_rmem
3316 };
3317 
3318 /**
3319  * tipc_socket_init - initialize TIPC socket interface
3320  *
3321  * Returns 0 on success, errno otherwise
3322  */
3323 int tipc_socket_init(void)
3324 {
3325 	int res;
3326 
3327 	res = proto_register(&tipc_proto, 1);
3328 	if (res) {
3329 		pr_err("Failed to register TIPC protocol type\n");
3330 		goto out;
3331 	}
3332 
3333 	res = sock_register(&tipc_family_ops);
3334 	if (res) {
3335 		pr_err("Failed to register TIPC socket type\n");
3336 		proto_unregister(&tipc_proto);
3337 		goto out;
3338 	}
3339  out:
3340 	return res;
3341 }
3342 
3343 /**
3344  * tipc_socket_stop - stop TIPC socket interface
3345  */
3346 void tipc_socket_stop(void)
3347 {
3348 	sock_unregister(tipc_family_ops.family);
3349 	proto_unregister(&tipc_proto);
3350 }
3351 
3352 /* Caller should hold socket lock for the passed tipc socket. */
3353 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3354 {
3355 	u32 peer_node;
3356 	u32 peer_port;
3357 	struct nlattr *nest;
3358 
3359 	peer_node = tsk_peer_node(tsk);
3360 	peer_port = tsk_peer_port(tsk);
3361 
3362 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3363 	if (!nest)
3364 		return -EMSGSIZE;
3365 
3366 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3367 		goto msg_full;
3368 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3369 		goto msg_full;
3370 
3371 	if (tsk->conn_type != 0) {
3372 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3373 			goto msg_full;
3374 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3375 			goto msg_full;
3376 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3377 			goto msg_full;
3378 	}
3379 	nla_nest_end(skb, nest);
3380 
3381 	return 0;
3382 
3383 msg_full:
3384 	nla_nest_cancel(skb, nest);
3385 
3386 	return -EMSGSIZE;
3387 }
3388 
3389 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3390 			  *tsk)
3391 {
3392 	struct net *net = sock_net(skb->sk);
3393 	struct sock *sk = &tsk->sk;
3394 
3395 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3396 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3397 		return -EMSGSIZE;
3398 
3399 	if (tipc_sk_connected(sk)) {
3400 		if (__tipc_nl_add_sk_con(skb, tsk))
3401 			return -EMSGSIZE;
3402 	} else if (!list_empty(&tsk->publications)) {
3403 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3404 			return -EMSGSIZE;
3405 	}
3406 	return 0;
3407 }
3408 
3409 /* Caller should hold socket lock for the passed tipc socket. */
3410 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3411 			    struct tipc_sock *tsk)
3412 {
3413 	struct nlattr *attrs;
3414 	void *hdr;
3415 
3416 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3417 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3418 	if (!hdr)
3419 		goto msg_cancel;
3420 
3421 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3422 	if (!attrs)
3423 		goto genlmsg_cancel;
3424 
3425 	if (__tipc_nl_add_sk_info(skb, tsk))
3426 		goto attr_msg_cancel;
3427 
3428 	nla_nest_end(skb, attrs);
3429 	genlmsg_end(skb, hdr);
3430 
3431 	return 0;
3432 
3433 attr_msg_cancel:
3434 	nla_nest_cancel(skb, attrs);
3435 genlmsg_cancel:
3436 	genlmsg_cancel(skb, hdr);
3437 msg_cancel:
3438 	return -EMSGSIZE;
3439 }
3440 
3441 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3442 		    int (*skb_handler)(struct sk_buff *skb,
3443 				       struct netlink_callback *cb,
3444 				       struct tipc_sock *tsk))
3445 {
3446 	struct rhashtable_iter *iter = (void *)cb->args[4];
3447 	struct tipc_sock *tsk;
3448 	int err;
3449 
3450 	rhashtable_walk_start(iter);
3451 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3452 		if (IS_ERR(tsk)) {
3453 			err = PTR_ERR(tsk);
3454 			if (err == -EAGAIN) {
3455 				err = 0;
3456 				continue;
3457 			}
3458 			break;
3459 		}
3460 
3461 		sock_hold(&tsk->sk);
3462 		rhashtable_walk_stop(iter);
3463 		lock_sock(&tsk->sk);
3464 		err = skb_handler(skb, cb, tsk);
3465 		if (err) {
3466 			release_sock(&tsk->sk);
3467 			sock_put(&tsk->sk);
3468 			goto out;
3469 		}
3470 		release_sock(&tsk->sk);
3471 		rhashtable_walk_start(iter);
3472 		sock_put(&tsk->sk);
3473 	}
3474 	rhashtable_walk_stop(iter);
3475 out:
3476 	return skb->len;
3477 }
3478 EXPORT_SYMBOL(tipc_nl_sk_walk);
3479 
3480 int tipc_dump_start(struct netlink_callback *cb)
3481 {
3482 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3483 }
3484 EXPORT_SYMBOL(tipc_dump_start);
3485 
3486 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3487 {
3488 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3489 	struct rhashtable_iter *iter = (void *)cb->args[4];
3490 	struct tipc_net *tn = tipc_net(net);
3491 
3492 	if (!iter) {
3493 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3494 		if (!iter)
3495 			return -ENOMEM;
3496 
3497 		cb->args[4] = (long)iter;
3498 	}
3499 
3500 	rhashtable_walk_enter(&tn->sk_rht, iter);
3501 	return 0;
3502 }
3503 
3504 int tipc_dump_done(struct netlink_callback *cb)
3505 {
3506 	struct rhashtable_iter *hti = (void *)cb->args[4];
3507 
3508 	rhashtable_walk_exit(hti);
3509 	kfree(hti);
3510 	return 0;
3511 }
3512 EXPORT_SYMBOL(tipc_dump_done);
3513 
3514 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3515 			   struct tipc_sock *tsk, u32 sk_filter_state,
3516 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3517 {
3518 	struct sock *sk = &tsk->sk;
3519 	struct nlattr *attrs;
3520 	struct nlattr *stat;
3521 
3522 	/*filter response w.r.t sk_state*/
3523 	if (!(sk_filter_state & (1 << sk->sk_state)))
3524 		return 0;
3525 
3526 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3527 	if (!attrs)
3528 		goto msg_cancel;
3529 
3530 	if (__tipc_nl_add_sk_info(skb, tsk))
3531 		goto attr_msg_cancel;
3532 
3533 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3534 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3535 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3536 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3537 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3538 					 sock_i_uid(sk))) ||
3539 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3540 			      tipc_diag_gen_cookie(sk),
3541 			      TIPC_NLA_SOCK_PAD))
3542 		goto attr_msg_cancel;
3543 
3544 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3545 	if (!stat)
3546 		goto attr_msg_cancel;
3547 
3548 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3549 			skb_queue_len(&sk->sk_receive_queue)) ||
3550 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3551 			skb_queue_len(&sk->sk_write_queue)) ||
3552 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3553 			atomic_read(&sk->sk_drops)))
3554 		goto stat_msg_cancel;
3555 
3556 	if (tsk->cong_link_cnt &&
3557 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3558 		goto stat_msg_cancel;
3559 
3560 	if (tsk_conn_cong(tsk) &&
3561 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3562 		goto stat_msg_cancel;
3563 
3564 	nla_nest_end(skb, stat);
3565 
3566 	if (tsk->group)
3567 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3568 			goto stat_msg_cancel;
3569 
3570 	nla_nest_end(skb, attrs);
3571 
3572 	return 0;
3573 
3574 stat_msg_cancel:
3575 	nla_nest_cancel(skb, stat);
3576 attr_msg_cancel:
3577 	nla_nest_cancel(skb, attrs);
3578 msg_cancel:
3579 	return -EMSGSIZE;
3580 }
3581 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3582 
3583 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3584 {
3585 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3586 }
3587 
3588 /* Caller should hold socket lock for the passed tipc socket. */
3589 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3590 				 struct netlink_callback *cb,
3591 				 struct publication *publ)
3592 {
3593 	void *hdr;
3594 	struct nlattr *attrs;
3595 
3596 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3597 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3598 	if (!hdr)
3599 		goto msg_cancel;
3600 
3601 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3602 	if (!attrs)
3603 		goto genlmsg_cancel;
3604 
3605 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3606 		goto attr_msg_cancel;
3607 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3608 		goto attr_msg_cancel;
3609 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3610 		goto attr_msg_cancel;
3611 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3612 		goto attr_msg_cancel;
3613 
3614 	nla_nest_end(skb, attrs);
3615 	genlmsg_end(skb, hdr);
3616 
3617 	return 0;
3618 
3619 attr_msg_cancel:
3620 	nla_nest_cancel(skb, attrs);
3621 genlmsg_cancel:
3622 	genlmsg_cancel(skb, hdr);
3623 msg_cancel:
3624 	return -EMSGSIZE;
3625 }
3626 
3627 /* Caller should hold socket lock for the passed tipc socket. */
3628 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3629 				  struct netlink_callback *cb,
3630 				  struct tipc_sock *tsk, u32 *last_publ)
3631 {
3632 	int err;
3633 	struct publication *p;
3634 
3635 	if (*last_publ) {
3636 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3637 			if (p->key == *last_publ)
3638 				break;
3639 		}
3640 		if (p->key != *last_publ) {
3641 			/* We never set seq or call nl_dump_check_consistent()
3642 			 * this means that setting prev_seq here will cause the
3643 			 * consistence check to fail in the netlink callback
3644 			 * handler. Resulting in the last NLMSG_DONE message
3645 			 * having the NLM_F_DUMP_INTR flag set.
3646 			 */
3647 			cb->prev_seq = 1;
3648 			*last_publ = 0;
3649 			return -EPIPE;
3650 		}
3651 	} else {
3652 		p = list_first_entry(&tsk->publications, struct publication,
3653 				     binding_sock);
3654 	}
3655 
3656 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3657 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3658 		if (err) {
3659 			*last_publ = p->key;
3660 			return err;
3661 		}
3662 	}
3663 	*last_publ = 0;
3664 
3665 	return 0;
3666 }
3667 
3668 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3669 {
3670 	int err;
3671 	u32 tsk_portid = cb->args[0];
3672 	u32 last_publ = cb->args[1];
3673 	u32 done = cb->args[2];
3674 	struct net *net = sock_net(skb->sk);
3675 	struct tipc_sock *tsk;
3676 
3677 	if (!tsk_portid) {
3678 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3679 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3680 
3681 		if (!attrs[TIPC_NLA_SOCK])
3682 			return -EINVAL;
3683 
3684 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3685 						  attrs[TIPC_NLA_SOCK],
3686 						  tipc_nl_sock_policy, NULL);
3687 		if (err)
3688 			return err;
3689 
3690 		if (!sock[TIPC_NLA_SOCK_REF])
3691 			return -EINVAL;
3692 
3693 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3694 	}
3695 
3696 	if (done)
3697 		return 0;
3698 
3699 	tsk = tipc_sk_lookup(net, tsk_portid);
3700 	if (!tsk)
3701 		return -EINVAL;
3702 
3703 	lock_sock(&tsk->sk);
3704 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3705 	if (!err)
3706 		done = 1;
3707 	release_sock(&tsk->sk);
3708 	sock_put(&tsk->sk);
3709 
3710 	cb->args[0] = tsk_portid;
3711 	cb->args[1] = last_publ;
3712 	cb->args[2] = done;
3713 
3714 	return skb->len;
3715 }
3716 
3717 /**
3718  * tipc_sk_filtering - check if a socket should be traced
3719  * @sk: the socket to be examined
3720  * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3721  *  (portid, sock type, name type, name lower, name upper)
3722  *
3723  * Returns true if the socket meets the socket tuple data
3724  * (value 0 = 'any') or when there is no tuple set (all = 0),
3725  * otherwise false
3726  */
3727 bool tipc_sk_filtering(struct sock *sk)
3728 {
3729 	struct tipc_sock *tsk;
3730 	struct publication *p;
3731 	u32 _port, _sktype, _type, _lower, _upper;
3732 	u32 type = 0, lower = 0, upper = 0;
3733 
3734 	if (!sk)
3735 		return true;
3736 
3737 	tsk = tipc_sk(sk);
3738 
3739 	_port = sysctl_tipc_sk_filter[0];
3740 	_sktype = sysctl_tipc_sk_filter[1];
3741 	_type = sysctl_tipc_sk_filter[2];
3742 	_lower = sysctl_tipc_sk_filter[3];
3743 	_upper = sysctl_tipc_sk_filter[4];
3744 
3745 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3746 		return true;
3747 
3748 	if (_port)
3749 		return (_port == tsk->portid);
3750 
3751 	if (_sktype && _sktype != sk->sk_type)
3752 		return false;
3753 
3754 	if (tsk->published) {
3755 		p = list_first_entry_or_null(&tsk->publications,
3756 					     struct publication, binding_sock);
3757 		if (p) {
3758 			type = p->type;
3759 			lower = p->lower;
3760 			upper = p->upper;
3761 		}
3762 	}
3763 
3764 	if (!tipc_sk_type_connectionless(sk)) {
3765 		type = tsk->conn_type;
3766 		lower = tsk->conn_instance;
3767 		upper = tsk->conn_instance;
3768 	}
3769 
3770 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3771 	    (_upper && _upper != upper))
3772 		return false;
3773 
3774 	return true;
3775 }
3776 
3777 u32 tipc_sock_get_portid(struct sock *sk)
3778 {
3779 	return (sk) ? (tipc_sk(sk))->portid : 0;
3780 }
3781 
3782 /**
3783  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3784  *			both the rcv and backlog queues are considered
3785  * @sk: tipc sk to be checked
3786  * @skb: tipc msg to be checked
3787  *
3788  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3789  */
3790 
3791 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3792 {
3793 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3794 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3795 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3796 
3797 	return (qsize > lim * 90 / 100);
3798 }
3799 
3800 /**
3801  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3802  *			only the rcv queue is considered
3803  * @sk: tipc sk to be checked
3804  * @skb: tipc msg to be checked
3805  *
3806  * Returns true if the socket rx queue allocation is > 90%, otherwise false
3807  */
3808 
3809 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3810 {
3811 	unsigned int lim = rcvbuf_limit(sk, skb);
3812 	unsigned int qsize = sk_rmem_alloc_get(sk);
3813 
3814 	return (qsize > lim * 90 / 100);
3815 }
3816 
3817 /**
3818  * tipc_sk_dump - dump TIPC socket
3819  * @sk: tipc sk to be dumped
3820  * @dqueues: bitmask to decide if any socket queue to be dumped?
3821  *           - TIPC_DUMP_NONE: don't dump socket queues
3822  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3823  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3824  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3825  *           - TIPC_DUMP_ALL: dump all the socket queues above
3826  * @buf: returned buffer of dump data in format
3827  */
3828 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3829 {
3830 	int i = 0;
3831 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3832 	struct tipc_sock *tsk;
3833 	struct publication *p;
3834 	bool tsk_connected;
3835 
3836 	if (!sk) {
3837 		i += scnprintf(buf, sz, "sk data: (null)\n");
3838 		return i;
3839 	}
3840 
3841 	tsk = tipc_sk(sk);
3842 	tsk_connected = !tipc_sk_type_connectionless(sk);
3843 
3844 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3845 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3846 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3847 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3848 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3849 	if (tsk_connected) {
3850 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3851 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3852 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3853 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3854 	}
3855 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3856 	if (tsk->published) {
3857 		p = list_first_entry_or_null(&tsk->publications,
3858 					     struct publication, binding_sock);
3859 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3860 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3861 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3862 	}
3863 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3864 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3865 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3866 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3867 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3868 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3869 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3870 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3871 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3872 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3873 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3874 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3875 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3876 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3877 
3878 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3879 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3880 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3881 	}
3882 
3883 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3884 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3885 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3886 	}
3887 
3888 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3889 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3890 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3891 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3892 			i += scnprintf(buf + i, sz - i, "  tail ");
3893 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3894 					   buf + i);
3895 		}
3896 	}
3897 
3898 	return i;
3899 }
3900