xref: /linux/net/tipc/socket.c (revision ab52c59103002b49f2455371e4b9c56ba3ef1781)
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2019, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6  * Copyright (c) 2020-2021, Red Hat Inc
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions are met:
11  *
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the names of the copyright holders nor the names of its
18  *    contributors may be used to endorse or promote products derived from
19  *    this software without specific prior written permission.
20  *
21  * Alternatively, this software may be distributed under the terms of the
22  * GNU General Public License ("GPL") version 2 as published by the Free
23  * Software Foundation.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <linux/rhashtable.h>
39 #include <linux/sched/signal.h>
40 #include <trace/events/sock.h>
41 
42 #include "core.h"
43 #include "name_table.h"
44 #include "node.h"
45 #include "link.h"
46 #include "name_distr.h"
47 #include "socket.h"
48 #include "bcast.h"
49 #include "netlink.h"
50 #include "group.h"
51 #include "trace.h"
52 
53 #define NAGLE_START_INIT	4
54 #define NAGLE_START_MAX		1024
55 #define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
56 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
57 #define TIPC_MAX_PORT		0xffffffff
58 #define TIPC_MIN_PORT		1
59 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of rcv window size */
60 
61 enum {
62 	TIPC_LISTEN = TCP_LISTEN,
63 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
64 	TIPC_OPEN = TCP_CLOSE,
65 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
66 	TIPC_CONNECTING = TCP_SYN_SENT,
67 };
68 
69 struct sockaddr_pair {
70 	struct sockaddr_tipc sock;
71 	struct sockaddr_tipc member;
72 };
73 
74 /**
75  * struct tipc_sock - TIPC socket structure
76  * @sk: socket - interacts with 'port' and with user via the socket API
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  * @pub_count: total # of publications port has made during its lifetime
84  * @conn_timeout: the time we can wait for an unresponded setup request
85  * @probe_unacked: probe has not received ack yet
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  * @snd_win: send window size
90  * @peer_caps: peer capabilities mask
91  * @rcv_unacked: # messages read by user, but not yet acked back to peer
92  * @rcv_win: receive window size
93  * @peer: 'connected' peer for dgram/rdm
94  * @node: hash table node
95  * @mc_method: cookie for use between socket and broadcast layer
96  * @rcu: rcu struct for tipc_sock
97  * @group: TIPC communications group
98  * @oneway: message count in one direction (FIXME)
99  * @nagle_start: current nagle value
100  * @snd_backlog: send backlog count
101  * @msg_acc: messages accepted; used in managing backlog and nagle
102  * @pkt_cnt: TIPC socket packet count
103  * @expect_ack: whether this TIPC socket is expecting an ack
104  * @nodelay: setsockopt() TIPC_NODELAY setting
105  * @group_is_open: TIPC socket group is fully open (FIXME)
106  * @published: true if port has one or more associated names
107  * @conn_addrtype: address type used when establishing connection
108  */
109 struct tipc_sock {
110 	struct sock sk;
111 	u32 max_pkt;
112 	u32 maxnagle;
113 	u32 portid;
114 	struct tipc_msg phdr;
115 	struct list_head cong_links;
116 	struct list_head publications;
117 	u32 pub_count;
118 	atomic_t dupl_rcvcnt;
119 	u16 conn_timeout;
120 	bool probe_unacked;
121 	u16 cong_link_cnt;
122 	u16 snt_unacked;
123 	u16 snd_win;
124 	u16 peer_caps;
125 	u16 rcv_unacked;
126 	u16 rcv_win;
127 	struct sockaddr_tipc peer;
128 	struct rhash_head node;
129 	struct tipc_mc_method mc_method;
130 	struct rcu_head rcu;
131 	struct tipc_group *group;
132 	u32 oneway;
133 	u32 nagle_start;
134 	u16 snd_backlog;
135 	u16 msg_acc;
136 	u16 pkt_cnt;
137 	bool expect_ack;
138 	bool nodelay;
139 	bool group_is_open;
140 	bool published;
141 	u8 conn_addrtype;
142 };
143 
144 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
145 static void tipc_data_ready(struct sock *sk);
146 static void tipc_write_space(struct sock *sk);
147 static void tipc_sock_destruct(struct sock *sk);
148 static int tipc_release(struct socket *sock);
149 static void tipc_sk_timeout(struct timer_list *t);
150 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
151 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
152 static int tipc_sk_leave(struct tipc_sock *tsk);
153 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
154 static int tipc_sk_insert(struct tipc_sock *tsk);
155 static void tipc_sk_remove(struct tipc_sock *tsk);
156 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
157 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
158 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
159 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
160 
161 static const struct proto_ops packet_ops;
162 static const struct proto_ops stream_ops;
163 static const struct proto_ops msg_ops;
164 static struct proto tipc_proto;
165 static const struct rhashtable_params tsk_rht_params;
166 
167 static u32 tsk_own_node(struct tipc_sock *tsk)
168 {
169 	return msg_prevnode(&tsk->phdr);
170 }
171 
172 static u32 tsk_peer_node(struct tipc_sock *tsk)
173 {
174 	return msg_destnode(&tsk->phdr);
175 }
176 
177 static u32 tsk_peer_port(struct tipc_sock *tsk)
178 {
179 	return msg_destport(&tsk->phdr);
180 }
181 
182 static  bool tsk_unreliable(struct tipc_sock *tsk)
183 {
184 	return msg_src_droppable(&tsk->phdr) != 0;
185 }
186 
187 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
188 {
189 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
190 }
191 
192 static bool tsk_unreturnable(struct tipc_sock *tsk)
193 {
194 	return msg_dest_droppable(&tsk->phdr) != 0;
195 }
196 
197 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
198 {
199 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
200 }
201 
202 static int tsk_importance(struct tipc_sock *tsk)
203 {
204 	return msg_importance(&tsk->phdr);
205 }
206 
207 static struct tipc_sock *tipc_sk(const struct sock *sk)
208 {
209 	return container_of(sk, struct tipc_sock, sk);
210 }
211 
212 int tsk_set_importance(struct sock *sk, int imp)
213 {
214 	if (imp > TIPC_CRITICAL_IMPORTANCE)
215 		return -EINVAL;
216 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
217 	return 0;
218 }
219 
220 static bool tsk_conn_cong(struct tipc_sock *tsk)
221 {
222 	return tsk->snt_unacked > tsk->snd_win;
223 }
224 
225 static u16 tsk_blocks(int len)
226 {
227 	return ((len / FLOWCTL_BLK_SZ) + 1);
228 }
229 
230 /* tsk_blocks(): translate a buffer size in bytes to number of
231  * advertisable blocks, taking into account the ratio truesize(len)/len
232  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
233  */
234 static u16 tsk_adv_blocks(int len)
235 {
236 	return len / FLOWCTL_BLK_SZ / 4;
237 }
238 
239 /* tsk_inc(): increment counter for sent or received data
240  * - If block based flow control is not supported by peer we
241  *   fall back to message based ditto, incrementing the counter
242  */
243 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
244 {
245 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
246 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
247 	return 1;
248 }
249 
250 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
251  */
252 static void tsk_set_nagle(struct tipc_sock *tsk)
253 {
254 	struct sock *sk = &tsk->sk;
255 
256 	tsk->maxnagle = 0;
257 	if (sk->sk_type != SOCK_STREAM)
258 		return;
259 	if (tsk->nodelay)
260 		return;
261 	if (!(tsk->peer_caps & TIPC_NAGLE))
262 		return;
263 	/* Limit node local buffer size to avoid receive queue overflow */
264 	if (tsk->max_pkt == MAX_MSG_SIZE)
265 		tsk->maxnagle = 1500;
266 	else
267 		tsk->maxnagle = tsk->max_pkt;
268 }
269 
270 /**
271  * tsk_advance_rx_queue - discard first buffer in socket receive queue
272  * @sk: network socket
273  *
274  * Caller must hold socket lock
275  */
276 static void tsk_advance_rx_queue(struct sock *sk)
277 {
278 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
279 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
280 }
281 
282 /* tipc_sk_respond() : send response message back to sender
283  */
284 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
285 {
286 	u32 selector;
287 	u32 dnode;
288 	u32 onode = tipc_own_addr(sock_net(sk));
289 
290 	if (!tipc_msg_reverse(onode, &skb, err))
291 		return;
292 
293 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
294 	dnode = msg_destnode(buf_msg(skb));
295 	selector = msg_origport(buf_msg(skb));
296 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
297 }
298 
299 /**
300  * tsk_rej_rx_queue - reject all buffers in socket receive queue
301  * @sk: network socket
302  * @error: response error code
303  *
304  * Caller must hold socket lock
305  */
306 static void tsk_rej_rx_queue(struct sock *sk, int error)
307 {
308 	struct sk_buff *skb;
309 
310 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
311 		tipc_sk_respond(sk, skb, error);
312 }
313 
314 static bool tipc_sk_connected(const struct sock *sk)
315 {
316 	return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED;
317 }
318 
319 /* tipc_sk_type_connectionless - check if the socket is datagram socket
320  * @sk: socket
321  *
322  * Returns true if connection less, false otherwise
323  */
324 static bool tipc_sk_type_connectionless(struct sock *sk)
325 {
326 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
327 }
328 
329 /* tsk_peer_msg - verify if message was sent by connected port's peer
330  *
331  * Handles cases where the node's network address has changed from
332  * the default of <0.0.0> to its configured setting.
333  */
334 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
335 {
336 	struct sock *sk = &tsk->sk;
337 	u32 self = tipc_own_addr(sock_net(sk));
338 	u32 peer_port = tsk_peer_port(tsk);
339 	u32 orig_node, peer_node;
340 
341 	if (unlikely(!tipc_sk_connected(sk)))
342 		return false;
343 
344 	if (unlikely(msg_origport(msg) != peer_port))
345 		return false;
346 
347 	orig_node = msg_orignode(msg);
348 	peer_node = tsk_peer_node(tsk);
349 
350 	if (likely(orig_node == peer_node))
351 		return true;
352 
353 	if (!orig_node && peer_node == self)
354 		return true;
355 
356 	if (!peer_node && orig_node == self)
357 		return true;
358 
359 	return false;
360 }
361 
362 /* tipc_set_sk_state - set the sk_state of the socket
363  * @sk: socket
364  *
365  * Caller must hold socket lock
366  *
367  * Returns 0 on success, errno otherwise
368  */
369 static int tipc_set_sk_state(struct sock *sk, int state)
370 {
371 	int oldsk_state = sk->sk_state;
372 	int res = -EINVAL;
373 
374 	switch (state) {
375 	case TIPC_OPEN:
376 		res = 0;
377 		break;
378 	case TIPC_LISTEN:
379 	case TIPC_CONNECTING:
380 		if (oldsk_state == TIPC_OPEN)
381 			res = 0;
382 		break;
383 	case TIPC_ESTABLISHED:
384 		if (oldsk_state == TIPC_CONNECTING ||
385 		    oldsk_state == TIPC_OPEN)
386 			res = 0;
387 		break;
388 	case TIPC_DISCONNECTING:
389 		if (oldsk_state == TIPC_CONNECTING ||
390 		    oldsk_state == TIPC_ESTABLISHED)
391 			res = 0;
392 		break;
393 	}
394 
395 	if (!res)
396 		sk->sk_state = state;
397 
398 	return res;
399 }
400 
401 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
402 {
403 	struct sock *sk = sock->sk;
404 	int err = sock_error(sk);
405 	int typ = sock->type;
406 
407 	if (err)
408 		return err;
409 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
410 		if (sk->sk_state == TIPC_DISCONNECTING)
411 			return -EPIPE;
412 		else if (!tipc_sk_connected(sk))
413 			return -ENOTCONN;
414 	}
415 	if (!*timeout)
416 		return -EAGAIN;
417 	if (signal_pending(current))
418 		return sock_intr_errno(*timeout);
419 
420 	return 0;
421 }
422 
423 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
424 ({                                                                             \
425 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
426 	struct sock *sk_;						       \
427 	int rc_;							       \
428 									       \
429 	while ((rc_ = !(condition_))) {					       \
430 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
431 		smp_rmb();                                                     \
432 		sk_ = (sock_)->sk;					       \
433 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
434 		if (rc_)						       \
435 			break;						       \
436 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
437 		release_sock(sk_);					       \
438 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
439 		sched_annotate_sleep();				               \
440 		lock_sock(sk_);						       \
441 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
442 	}								       \
443 	rc_;								       \
444 })
445 
446 /**
447  * tipc_sk_create - create a TIPC socket
448  * @net: network namespace (must be default network)
449  * @sock: pre-allocated socket structure
450  * @protocol: protocol indicator (must be 0)
451  * @kern: caused by kernel or by userspace?
452  *
453  * This routine creates additional data structures used by the TIPC socket,
454  * initializes them, and links them together.
455  *
456  * Return: 0 on success, errno otherwise
457  */
458 static int tipc_sk_create(struct net *net, struct socket *sock,
459 			  int protocol, int kern)
460 {
461 	const struct proto_ops *ops;
462 	struct sock *sk;
463 	struct tipc_sock *tsk;
464 	struct tipc_msg *msg;
465 
466 	/* Validate arguments */
467 	if (unlikely(protocol != 0))
468 		return -EPROTONOSUPPORT;
469 
470 	switch (sock->type) {
471 	case SOCK_STREAM:
472 		ops = &stream_ops;
473 		break;
474 	case SOCK_SEQPACKET:
475 		ops = &packet_ops;
476 		break;
477 	case SOCK_DGRAM:
478 	case SOCK_RDM:
479 		ops = &msg_ops;
480 		break;
481 	default:
482 		return -EPROTOTYPE;
483 	}
484 
485 	/* Allocate socket's protocol area */
486 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
487 	if (sk == NULL)
488 		return -ENOMEM;
489 
490 	tsk = tipc_sk(sk);
491 	tsk->max_pkt = MAX_PKT_DEFAULT;
492 	tsk->maxnagle = 0;
493 	tsk->nagle_start = NAGLE_START_INIT;
494 	INIT_LIST_HEAD(&tsk->publications);
495 	INIT_LIST_HEAD(&tsk->cong_links);
496 	msg = &tsk->phdr;
497 
498 	/* Finish initializing socket data structures */
499 	sock->ops = ops;
500 	sock_init_data(sock, sk);
501 	tipc_set_sk_state(sk, TIPC_OPEN);
502 	if (tipc_sk_insert(tsk)) {
503 		sk_free(sk);
504 		pr_warn("Socket create failed; port number exhausted\n");
505 		return -EINVAL;
506 	}
507 
508 	/* Ensure tsk is visible before we read own_addr. */
509 	smp_mb();
510 
511 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
512 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
513 
514 	msg_set_origport(msg, tsk->portid);
515 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
516 	sk->sk_shutdown = 0;
517 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
518 	sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
519 	sk->sk_data_ready = tipc_data_ready;
520 	sk->sk_write_space = tipc_write_space;
521 	sk->sk_destruct = tipc_sock_destruct;
522 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
523 	tsk->group_is_open = true;
524 	atomic_set(&tsk->dupl_rcvcnt, 0);
525 
526 	/* Start out with safe limits until we receive an advertised window */
527 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
528 	tsk->rcv_win = tsk->snd_win;
529 
530 	if (tipc_sk_type_connectionless(sk)) {
531 		tsk_set_unreturnable(tsk, true);
532 		if (sock->type == SOCK_DGRAM)
533 			tsk_set_unreliable(tsk, true);
534 	}
535 	__skb_queue_head_init(&tsk->mc_method.deferredq);
536 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
537 	return 0;
538 }
539 
540 static void tipc_sk_callback(struct rcu_head *head)
541 {
542 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
543 
544 	sock_put(&tsk->sk);
545 }
546 
547 /* Caller should hold socket lock for the socket. */
548 static void __tipc_shutdown(struct socket *sock, int error)
549 {
550 	struct sock *sk = sock->sk;
551 	struct tipc_sock *tsk = tipc_sk(sk);
552 	struct net *net = sock_net(sk);
553 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
554 	u32 dnode = tsk_peer_node(tsk);
555 	struct sk_buff *skb;
556 
557 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
558 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
559 					    !tsk_conn_cong(tsk)));
560 
561 	/* Push out delayed messages if in Nagle mode */
562 	tipc_sk_push_backlog(tsk, false);
563 	/* Remove pending SYN */
564 	__skb_queue_purge(&sk->sk_write_queue);
565 
566 	/* Remove partially received buffer if any */
567 	skb = skb_peek(&sk->sk_receive_queue);
568 	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
569 		__skb_unlink(skb, &sk->sk_receive_queue);
570 		kfree_skb(skb);
571 	}
572 
573 	/* Reject all unreceived messages if connectionless */
574 	if (tipc_sk_type_connectionless(sk)) {
575 		tsk_rej_rx_queue(sk, error);
576 		return;
577 	}
578 
579 	switch (sk->sk_state) {
580 	case TIPC_CONNECTING:
581 	case TIPC_ESTABLISHED:
582 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
583 		tipc_node_remove_conn(net, dnode, tsk->portid);
584 		/* Send a FIN+/- to its peer */
585 		skb = __skb_dequeue(&sk->sk_receive_queue);
586 		if (skb) {
587 			__skb_queue_purge(&sk->sk_receive_queue);
588 			tipc_sk_respond(sk, skb, error);
589 			break;
590 		}
591 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
592 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
593 				      tsk_own_node(tsk), tsk_peer_port(tsk),
594 				      tsk->portid, error);
595 		if (skb)
596 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
597 		break;
598 	case TIPC_LISTEN:
599 		/* Reject all SYN messages */
600 		tsk_rej_rx_queue(sk, error);
601 		break;
602 	default:
603 		__skb_queue_purge(&sk->sk_receive_queue);
604 		break;
605 	}
606 }
607 
608 /**
609  * tipc_release - destroy a TIPC socket
610  * @sock: socket to destroy
611  *
612  * This routine cleans up any messages that are still queued on the socket.
613  * For DGRAM and RDM socket types, all queued messages are rejected.
614  * For SEQPACKET and STREAM socket types, the first message is rejected
615  * and any others are discarded.  (If the first message on a STREAM socket
616  * is partially-read, it is discarded and the next one is rejected instead.)
617  *
618  * NOTE: Rejected messages are not necessarily returned to the sender!  They
619  * are returned or discarded according to the "destination droppable" setting
620  * specified for the message by the sender.
621  *
622  * Return: 0 on success, errno otherwise
623  */
624 static int tipc_release(struct socket *sock)
625 {
626 	struct sock *sk = sock->sk;
627 	struct tipc_sock *tsk;
628 
629 	/*
630 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
631 	 * releases a pre-allocated child socket that was never used)
632 	 */
633 	if (sk == NULL)
634 		return 0;
635 
636 	tsk = tipc_sk(sk);
637 	lock_sock(sk);
638 
639 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
640 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
641 	sk->sk_shutdown = SHUTDOWN_MASK;
642 	tipc_sk_leave(tsk);
643 	tipc_sk_withdraw(tsk, NULL);
644 	__skb_queue_purge(&tsk->mc_method.deferredq);
645 	sk_stop_timer(sk, &sk->sk_timer);
646 	tipc_sk_remove(tsk);
647 
648 	sock_orphan(sk);
649 	/* Reject any messages that accumulated in backlog queue */
650 	release_sock(sk);
651 	tipc_dest_list_purge(&tsk->cong_links);
652 	tsk->cong_link_cnt = 0;
653 	call_rcu(&tsk->rcu, tipc_sk_callback);
654 	sock->sk = NULL;
655 
656 	return 0;
657 }
658 
659 /**
660  * __tipc_bind - associate or disassocate TIPC name(s) with a socket
661  * @sock: socket structure
662  * @skaddr: socket address describing name(s) and desired operation
663  * @alen: size of socket address data structure
664  *
665  * Name and name sequence binding are indicated using a positive scope value;
666  * a negative scope value unbinds the specified name.  Specifying no name
667  * (i.e. a socket address length of 0) unbinds all names from the socket.
668  *
669  * Return: 0 on success, errno otherwise
670  *
671  * NOTE: This routine doesn't need to take the socket lock since it doesn't
672  *       access any non-constant socket information.
673  */
674 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
675 {
676 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
677 	struct tipc_sock *tsk = tipc_sk(sock->sk);
678 	bool unbind = false;
679 
680 	if (unlikely(!alen))
681 		return tipc_sk_withdraw(tsk, NULL);
682 
683 	if (ua->addrtype == TIPC_SERVICE_ADDR) {
684 		ua->addrtype = TIPC_SERVICE_RANGE;
685 		ua->sr.upper = ua->sr.lower;
686 	}
687 	if (ua->scope < 0) {
688 		unbind = true;
689 		ua->scope = -ua->scope;
690 	}
691 	/* Users may still use deprecated TIPC_ZONE_SCOPE */
692 	if (ua->scope != TIPC_NODE_SCOPE)
693 		ua->scope = TIPC_CLUSTER_SCOPE;
694 
695 	if (tsk->group)
696 		return -EACCES;
697 
698 	if (unbind)
699 		return tipc_sk_withdraw(tsk, ua);
700 	return tipc_sk_publish(tsk, ua);
701 }
702 
703 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
704 {
705 	int res;
706 
707 	lock_sock(sock->sk);
708 	res = __tipc_bind(sock, skaddr, alen);
709 	release_sock(sock->sk);
710 	return res;
711 }
712 
713 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
714 {
715 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
716 	u32 atype = ua->addrtype;
717 
718 	if (alen) {
719 		if (!tipc_uaddr_valid(ua, alen))
720 			return -EINVAL;
721 		if (atype == TIPC_SOCKET_ADDR)
722 			return -EAFNOSUPPORT;
723 		if (ua->sr.type < TIPC_RESERVED_TYPES) {
724 			pr_warn_once("Can't bind to reserved service type %u\n",
725 				     ua->sr.type);
726 			return -EACCES;
727 		}
728 	}
729 	return tipc_sk_bind(sock, skaddr, alen);
730 }
731 
732 /**
733  * tipc_getname - get port ID of socket or peer socket
734  * @sock: socket structure
735  * @uaddr: area for returned socket address
736  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
737  *
738  * Return: 0 on success, errno otherwise
739  *
740  * NOTE: This routine doesn't need to take the socket lock since it only
741  *       accesses socket information that is unchanging (or which changes in
742  *       a completely predictable manner).
743  */
744 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
745 			int peer)
746 {
747 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
748 	struct sock *sk = sock->sk;
749 	struct tipc_sock *tsk = tipc_sk(sk);
750 
751 	memset(addr, 0, sizeof(*addr));
752 	if (peer) {
753 		if ((!tipc_sk_connected(sk)) &&
754 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
755 			return -ENOTCONN;
756 		addr->addr.id.ref = tsk_peer_port(tsk);
757 		addr->addr.id.node = tsk_peer_node(tsk);
758 	} else {
759 		addr->addr.id.ref = tsk->portid;
760 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
761 	}
762 
763 	addr->addrtype = TIPC_SOCKET_ADDR;
764 	addr->family = AF_TIPC;
765 	addr->scope = 0;
766 	addr->addr.name.domain = 0;
767 
768 	return sizeof(*addr);
769 }
770 
771 /**
772  * tipc_poll - read and possibly block on pollmask
773  * @file: file structure associated with the socket
774  * @sock: socket for which to calculate the poll bits
775  * @wait: ???
776  *
777  * Return: pollmask value
778  *
779  * COMMENTARY:
780  * It appears that the usual socket locking mechanisms are not useful here
781  * since the pollmask info is potentially out-of-date the moment this routine
782  * exits.  TCP and other protocols seem to rely on higher level poll routines
783  * to handle any preventable race conditions, so TIPC will do the same ...
784  *
785  * IMPORTANT: The fact that a read or write operation is indicated does NOT
786  * imply that the operation will succeed, merely that it should be performed
787  * and will not block.
788  */
789 static __poll_t tipc_poll(struct file *file, struct socket *sock,
790 			      poll_table *wait)
791 {
792 	struct sock *sk = sock->sk;
793 	struct tipc_sock *tsk = tipc_sk(sk);
794 	__poll_t revents = 0;
795 
796 	sock_poll_wait(file, sock, wait);
797 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
798 
799 	if (sk->sk_shutdown & RCV_SHUTDOWN)
800 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
801 	if (sk->sk_shutdown == SHUTDOWN_MASK)
802 		revents |= EPOLLHUP;
803 
804 	switch (sk->sk_state) {
805 	case TIPC_ESTABLISHED:
806 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
807 			revents |= EPOLLOUT;
808 		fallthrough;
809 	case TIPC_LISTEN:
810 	case TIPC_CONNECTING:
811 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
812 			revents |= EPOLLIN | EPOLLRDNORM;
813 		break;
814 	case TIPC_OPEN:
815 		if (tsk->group_is_open && !tsk->cong_link_cnt)
816 			revents |= EPOLLOUT;
817 		if (!tipc_sk_type_connectionless(sk))
818 			break;
819 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
820 			break;
821 		revents |= EPOLLIN | EPOLLRDNORM;
822 		break;
823 	case TIPC_DISCONNECTING:
824 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
825 		break;
826 	}
827 	return revents;
828 }
829 
830 /**
831  * tipc_sendmcast - send multicast message
832  * @sock: socket structure
833  * @ua: destination address struct
834  * @msg: message to send
835  * @dlen: length of data to send
836  * @timeout: timeout to wait for wakeup
837  *
838  * Called from function tipc_sendmsg(), which has done all sanity checks
839  * Return: the number of bytes sent on success, or errno
840  */
841 static int tipc_sendmcast(struct  socket *sock, struct tipc_uaddr *ua,
842 			  struct msghdr *msg, size_t dlen, long timeout)
843 {
844 	struct sock *sk = sock->sk;
845 	struct tipc_sock *tsk = tipc_sk(sk);
846 	struct tipc_msg *hdr = &tsk->phdr;
847 	struct net *net = sock_net(sk);
848 	int mtu = tipc_bcast_get_mtu(net);
849 	struct sk_buff_head pkts;
850 	struct tipc_nlist dsts;
851 	int rc;
852 
853 	if (tsk->group)
854 		return -EACCES;
855 
856 	/* Block or return if any destination link is congested */
857 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
858 	if (unlikely(rc))
859 		return rc;
860 
861 	/* Lookup destination nodes */
862 	tipc_nlist_init(&dsts, tipc_own_addr(net));
863 	tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
864 	if (!dsts.local && !dsts.remote)
865 		return -EHOSTUNREACH;
866 
867 	/* Build message header */
868 	msg_set_type(hdr, TIPC_MCAST_MSG);
869 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
870 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
871 	msg_set_destport(hdr, 0);
872 	msg_set_destnode(hdr, 0);
873 	msg_set_nametype(hdr, ua->sr.type);
874 	msg_set_namelower(hdr, ua->sr.lower);
875 	msg_set_nameupper(hdr, ua->sr.upper);
876 
877 	/* Build message as chain of buffers */
878 	__skb_queue_head_init(&pkts);
879 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
880 
881 	/* Send message if build was successful */
882 	if (unlikely(rc == dlen)) {
883 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
884 					TIPC_DUMP_SK_SNDQ, " ");
885 		rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
886 				     &tsk->cong_link_cnt);
887 	}
888 
889 	tipc_nlist_purge(&dsts);
890 
891 	return rc ? rc : dlen;
892 }
893 
894 /**
895  * tipc_send_group_msg - send a message to a member in the group
896  * @net: network namespace
897  * @tsk: tipc socket
898  * @m: message to send
899  * @mb: group member
900  * @dnode: destination node
901  * @dport: destination port
902  * @dlen: total length of message data
903  */
904 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
905 			       struct msghdr *m, struct tipc_member *mb,
906 			       u32 dnode, u32 dport, int dlen)
907 {
908 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
909 	struct tipc_mc_method *method = &tsk->mc_method;
910 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
911 	struct tipc_msg *hdr = &tsk->phdr;
912 	struct sk_buff_head pkts;
913 	int mtu, rc;
914 
915 	/* Complete message header */
916 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
917 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
918 	msg_set_destport(hdr, dport);
919 	msg_set_destnode(hdr, dnode);
920 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
921 
922 	/* Build message as chain of buffers */
923 	__skb_queue_head_init(&pkts);
924 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
925 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
926 	if (unlikely(rc != dlen))
927 		return rc;
928 
929 	/* Send message */
930 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
931 	if (unlikely(rc == -ELINKCONG)) {
932 		tipc_dest_push(&tsk->cong_links, dnode, 0);
933 		tsk->cong_link_cnt++;
934 	}
935 
936 	/* Update send window */
937 	tipc_group_update_member(mb, blks);
938 
939 	/* A broadcast sent within next EXPIRE period must follow same path */
940 	method->rcast = true;
941 	method->mandatory = true;
942 	return dlen;
943 }
944 
945 /**
946  * tipc_send_group_unicast - send message to a member in the group
947  * @sock: socket structure
948  * @m: message to send
949  * @dlen: total length of message data
950  * @timeout: timeout to wait for wakeup
951  *
952  * Called from function tipc_sendmsg(), which has done all sanity checks
953  * Return: the number of bytes sent on success, or errno
954  */
955 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
956 				   int dlen, long timeout)
957 {
958 	struct sock *sk = sock->sk;
959 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
960 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
961 	struct tipc_sock *tsk = tipc_sk(sk);
962 	struct net *net = sock_net(sk);
963 	struct tipc_member *mb = NULL;
964 	u32 node, port;
965 	int rc;
966 
967 	node = ua->sk.node;
968 	port = ua->sk.ref;
969 	if (!port && !node)
970 		return -EHOSTUNREACH;
971 
972 	/* Block or return if destination link or member is congested */
973 	rc = tipc_wait_for_cond(sock, &timeout,
974 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
975 				tsk->group &&
976 				!tipc_group_cong(tsk->group, node, port, blks,
977 						 &mb));
978 	if (unlikely(rc))
979 		return rc;
980 
981 	if (unlikely(!mb))
982 		return -EHOSTUNREACH;
983 
984 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
985 
986 	return rc ? rc : dlen;
987 }
988 
989 /**
990  * tipc_send_group_anycast - send message to any member with given identity
991  * @sock: socket structure
992  * @m: message to send
993  * @dlen: total length of message data
994  * @timeout: timeout to wait for wakeup
995  *
996  * Called from function tipc_sendmsg(), which has done all sanity checks
997  * Return: the number of bytes sent on success, or errno
998  */
999 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1000 				   int dlen, long timeout)
1001 {
1002 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1003 	struct sock *sk = sock->sk;
1004 	struct tipc_sock *tsk = tipc_sk(sk);
1005 	struct list_head *cong_links = &tsk->cong_links;
1006 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1007 	struct tipc_msg *hdr = &tsk->phdr;
1008 	struct tipc_member *first = NULL;
1009 	struct tipc_member *mbr = NULL;
1010 	struct net *net = sock_net(sk);
1011 	u32 node, port, exclude;
1012 	struct list_head dsts;
1013 	int lookups = 0;
1014 	int dstcnt, rc;
1015 	bool cong;
1016 
1017 	INIT_LIST_HEAD(&dsts);
1018 	ua->sa.type = msg_nametype(hdr);
1019 	ua->scope = msg_lookup_scope(hdr);
1020 
1021 	while (++lookups < 4) {
1022 		exclude = tipc_group_exclude(tsk->group);
1023 
1024 		first = NULL;
1025 
1026 		/* Look for a non-congested destination member, if any */
1027 		while (1) {
1028 			if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1029 						       exclude, false))
1030 				return -EHOSTUNREACH;
1031 			tipc_dest_pop(&dsts, &node, &port);
1032 			cong = tipc_group_cong(tsk->group, node, port, blks,
1033 					       &mbr);
1034 			if (!cong)
1035 				break;
1036 			if (mbr == first)
1037 				break;
1038 			if (!first)
1039 				first = mbr;
1040 		}
1041 
1042 		/* Start over if destination was not in member list */
1043 		if (unlikely(!mbr))
1044 			continue;
1045 
1046 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1047 			break;
1048 
1049 		/* Block or return if destination link or member is congested */
1050 		rc = tipc_wait_for_cond(sock, &timeout,
1051 					!tipc_dest_find(cong_links, node, 0) &&
1052 					tsk->group &&
1053 					!tipc_group_cong(tsk->group, node, port,
1054 							 blks, &mbr));
1055 		if (unlikely(rc))
1056 			return rc;
1057 
1058 		/* Send, unless destination disappeared while waiting */
1059 		if (likely(mbr))
1060 			break;
1061 	}
1062 
1063 	if (unlikely(lookups >= 4))
1064 		return -EHOSTUNREACH;
1065 
1066 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1067 
1068 	return rc ? rc : dlen;
1069 }
1070 
1071 /**
1072  * tipc_send_group_bcast - send message to all members in communication group
1073  * @sock: socket structure
1074  * @m: message to send
1075  * @dlen: total length of message data
1076  * @timeout: timeout to wait for wakeup
1077  *
1078  * Called from function tipc_sendmsg(), which has done all sanity checks
1079  * Return: the number of bytes sent on success, or errno
1080  */
1081 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1082 				 int dlen, long timeout)
1083 {
1084 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1085 	struct sock *sk = sock->sk;
1086 	struct net *net = sock_net(sk);
1087 	struct tipc_sock *tsk = tipc_sk(sk);
1088 	struct tipc_nlist *dsts;
1089 	struct tipc_mc_method *method = &tsk->mc_method;
1090 	bool ack = method->mandatory && method->rcast;
1091 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1092 	struct tipc_msg *hdr = &tsk->phdr;
1093 	int mtu = tipc_bcast_get_mtu(net);
1094 	struct sk_buff_head pkts;
1095 	int rc = -EHOSTUNREACH;
1096 
1097 	/* Block or return if any destination link or member is congested */
1098 	rc = tipc_wait_for_cond(sock, &timeout,
1099 				!tsk->cong_link_cnt && tsk->group &&
1100 				!tipc_group_bc_cong(tsk->group, blks));
1101 	if (unlikely(rc))
1102 		return rc;
1103 
1104 	dsts = tipc_group_dests(tsk->group);
1105 	if (!dsts->local && !dsts->remote)
1106 		return -EHOSTUNREACH;
1107 
1108 	/* Complete message header */
1109 	if (ua) {
1110 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1111 		msg_set_nameinst(hdr, ua->sa.instance);
1112 	} else {
1113 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1114 		msg_set_nameinst(hdr, 0);
1115 	}
1116 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1117 	msg_set_destport(hdr, 0);
1118 	msg_set_destnode(hdr, 0);
1119 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1120 
1121 	/* Avoid getting stuck with repeated forced replicasts */
1122 	msg_set_grp_bc_ack_req(hdr, ack);
1123 
1124 	/* Build message as chain of buffers */
1125 	__skb_queue_head_init(&pkts);
1126 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1127 	if (unlikely(rc != dlen))
1128 		return rc;
1129 
1130 	/* Send message */
1131 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1132 	if (unlikely(rc))
1133 		return rc;
1134 
1135 	/* Update broadcast sequence number and send windows */
1136 	tipc_group_update_bc_members(tsk->group, blks, ack);
1137 
1138 	/* Broadcast link is now free to choose method for next broadcast */
1139 	method->mandatory = false;
1140 	method->expires = jiffies;
1141 
1142 	return dlen;
1143 }
1144 
1145 /**
1146  * tipc_send_group_mcast - send message to all members with given identity
1147  * @sock: socket structure
1148  * @m: message to send
1149  * @dlen: total length of message data
1150  * @timeout: timeout to wait for wakeup
1151  *
1152  * Called from function tipc_sendmsg(), which has done all sanity checks
1153  * Return: the number of bytes sent on success, or errno
1154  */
1155 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1156 				 int dlen, long timeout)
1157 {
1158 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1159 	struct sock *sk = sock->sk;
1160 	struct tipc_sock *tsk = tipc_sk(sk);
1161 	struct tipc_group *grp = tsk->group;
1162 	struct tipc_msg *hdr = &tsk->phdr;
1163 	struct net *net = sock_net(sk);
1164 	struct list_head dsts;
1165 	u32 dstcnt, exclude;
1166 
1167 	INIT_LIST_HEAD(&dsts);
1168 	ua->sa.type = msg_nametype(hdr);
1169 	ua->scope = msg_lookup_scope(hdr);
1170 	exclude = tipc_group_exclude(grp);
1171 
1172 	if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1173 		return -EHOSTUNREACH;
1174 
1175 	if (dstcnt == 1) {
1176 		tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1177 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1178 	}
1179 
1180 	tipc_dest_list_purge(&dsts);
1181 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1182 }
1183 
1184 /**
1185  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1186  * @net: the associated network namespace
1187  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1188  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1189  *
1190  * Multi-threaded: parallel calls with reference to same queues may occur
1191  */
1192 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1193 		       struct sk_buff_head *inputq)
1194 {
1195 	u32 self = tipc_own_addr(net);
1196 	struct sk_buff *skb, *_skb;
1197 	u32 portid, onode;
1198 	struct sk_buff_head tmpq;
1199 	struct list_head dports;
1200 	struct tipc_msg *hdr;
1201 	struct tipc_uaddr ua;
1202 	int user, mtyp, hlen;
1203 
1204 	__skb_queue_head_init(&tmpq);
1205 	INIT_LIST_HEAD(&dports);
1206 	ua.addrtype = TIPC_SERVICE_RANGE;
1207 
1208 	/* tipc_skb_peek() increments the head skb's reference counter */
1209 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1210 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1211 		hdr = buf_msg(skb);
1212 		user = msg_user(hdr);
1213 		mtyp = msg_type(hdr);
1214 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1215 		onode = msg_orignode(hdr);
1216 		ua.sr.type = msg_nametype(hdr);
1217 		ua.sr.lower = msg_namelower(hdr);
1218 		ua.sr.upper = msg_nameupper(hdr);
1219 		if (onode == self)
1220 			ua.scope = TIPC_ANY_SCOPE;
1221 		else
1222 			ua.scope = TIPC_CLUSTER_SCOPE;
1223 
1224 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1225 			spin_lock_bh(&inputq->lock);
1226 			if (skb_peek(arrvq) == skb) {
1227 				__skb_dequeue(arrvq);
1228 				__skb_queue_tail(inputq, skb);
1229 			}
1230 			kfree_skb(skb);
1231 			spin_unlock_bh(&inputq->lock);
1232 			continue;
1233 		}
1234 
1235 		/* Group messages require exact scope match */
1236 		if (msg_in_group(hdr)) {
1237 			ua.sr.lower = 0;
1238 			ua.sr.upper = ~0;
1239 			ua.scope = msg_lookup_scope(hdr);
1240 		}
1241 
1242 		/* Create destination port list: */
1243 		tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1244 
1245 		/* Clone message per destination */
1246 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1247 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1248 			if (_skb) {
1249 				msg_set_destport(buf_msg(_skb), portid);
1250 				__skb_queue_tail(&tmpq, _skb);
1251 				continue;
1252 			}
1253 			pr_warn("Failed to clone mcast rcv buffer\n");
1254 		}
1255 		/* Append clones to inputq only if skb is still head of arrvq */
1256 		spin_lock_bh(&inputq->lock);
1257 		if (skb_peek(arrvq) == skb) {
1258 			skb_queue_splice_tail_init(&tmpq, inputq);
1259 			/* Decrement the skb's refcnt */
1260 			kfree_skb(__skb_dequeue(arrvq));
1261 		}
1262 		spin_unlock_bh(&inputq->lock);
1263 		__skb_queue_purge(&tmpq);
1264 		kfree_skb(skb);
1265 	}
1266 	tipc_sk_rcv(net, inputq);
1267 }
1268 
1269 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1270  *                         when socket is in Nagle mode
1271  */
1272 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1273 {
1274 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1275 	struct sk_buff *skb = skb_peek_tail(txq);
1276 	struct net *net = sock_net(&tsk->sk);
1277 	u32 dnode = tsk_peer_node(tsk);
1278 	int rc;
1279 
1280 	if (nagle_ack) {
1281 		tsk->pkt_cnt += skb_queue_len(txq);
1282 		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1283 			tsk->oneway = 0;
1284 			if (tsk->nagle_start < NAGLE_START_MAX)
1285 				tsk->nagle_start *= 2;
1286 			tsk->expect_ack = false;
1287 			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1288 				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1289 				 tsk->nagle_start);
1290 		} else {
1291 			tsk->nagle_start = NAGLE_START_INIT;
1292 			if (skb) {
1293 				msg_set_ack_required(buf_msg(skb));
1294 				tsk->expect_ack = true;
1295 			} else {
1296 				tsk->expect_ack = false;
1297 			}
1298 		}
1299 		tsk->msg_acc = 0;
1300 		tsk->pkt_cnt = 0;
1301 	}
1302 
1303 	if (!skb || tsk->cong_link_cnt)
1304 		return;
1305 
1306 	/* Do not send SYN again after congestion */
1307 	if (msg_is_syn(buf_msg(skb)))
1308 		return;
1309 
1310 	if (tsk->msg_acc)
1311 		tsk->pkt_cnt += skb_queue_len(txq);
1312 	tsk->snt_unacked += tsk->snd_backlog;
1313 	tsk->snd_backlog = 0;
1314 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1315 	if (rc == -ELINKCONG)
1316 		tsk->cong_link_cnt = 1;
1317 }
1318 
1319 /**
1320  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1321  * @tsk: receiving socket
1322  * @skb: pointer to message buffer.
1323  * @inputq: buffer list containing the buffers
1324  * @xmitq: output message area
1325  */
1326 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1327 				   struct sk_buff_head *inputq,
1328 				   struct sk_buff_head *xmitq)
1329 {
1330 	struct tipc_msg *hdr = buf_msg(skb);
1331 	u32 onode = tsk_own_node(tsk);
1332 	struct sock *sk = &tsk->sk;
1333 	int mtyp = msg_type(hdr);
1334 	bool was_cong;
1335 
1336 	/* Ignore if connection cannot be validated: */
1337 	if (!tsk_peer_msg(tsk, hdr)) {
1338 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1339 		goto exit;
1340 	}
1341 
1342 	if (unlikely(msg_errcode(hdr))) {
1343 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1344 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1345 				      tsk_peer_port(tsk));
1346 		sk->sk_state_change(sk);
1347 
1348 		/* State change is ignored if socket already awake,
1349 		 * - convert msg to abort msg and add to inqueue
1350 		 */
1351 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1352 		msg_set_type(hdr, TIPC_CONN_MSG);
1353 		msg_set_size(hdr, BASIC_H_SIZE);
1354 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1355 		__skb_queue_tail(inputq, skb);
1356 		return;
1357 	}
1358 
1359 	tsk->probe_unacked = false;
1360 
1361 	if (mtyp == CONN_PROBE) {
1362 		msg_set_type(hdr, CONN_PROBE_REPLY);
1363 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1364 			__skb_queue_tail(xmitq, skb);
1365 		return;
1366 	} else if (mtyp == CONN_ACK) {
1367 		was_cong = tsk_conn_cong(tsk);
1368 		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1369 		tsk->snt_unacked -= msg_conn_ack(hdr);
1370 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1371 			tsk->snd_win = msg_adv_win(hdr);
1372 		if (was_cong && !tsk_conn_cong(tsk))
1373 			sk->sk_write_space(sk);
1374 	} else if (mtyp != CONN_PROBE_REPLY) {
1375 		pr_warn("Received unknown CONN_PROTO msg\n");
1376 	}
1377 exit:
1378 	kfree_skb(skb);
1379 }
1380 
1381 /**
1382  * tipc_sendmsg - send message in connectionless manner
1383  * @sock: socket structure
1384  * @m: message to send
1385  * @dsz: amount of user data to be sent
1386  *
1387  * Message must have an destination specified explicitly.
1388  * Used for SOCK_RDM and SOCK_DGRAM messages,
1389  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1390  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1391  *
1392  * Return: the number of bytes sent on success, or errno otherwise
1393  */
1394 static int tipc_sendmsg(struct socket *sock,
1395 			struct msghdr *m, size_t dsz)
1396 {
1397 	struct sock *sk = sock->sk;
1398 	int ret;
1399 
1400 	lock_sock(sk);
1401 	ret = __tipc_sendmsg(sock, m, dsz);
1402 	release_sock(sk);
1403 
1404 	return ret;
1405 }
1406 
1407 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1408 {
1409 	struct sock *sk = sock->sk;
1410 	struct net *net = sock_net(sk);
1411 	struct tipc_sock *tsk = tipc_sk(sk);
1412 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1413 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1414 	struct list_head *clinks = &tsk->cong_links;
1415 	bool syn = !tipc_sk_type_connectionless(sk);
1416 	struct tipc_group *grp = tsk->group;
1417 	struct tipc_msg *hdr = &tsk->phdr;
1418 	struct tipc_socket_addr skaddr;
1419 	struct sk_buff_head pkts;
1420 	int atype, mtu, rc;
1421 
1422 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1423 		return -EMSGSIZE;
1424 
1425 	if (ua) {
1426 		if (!tipc_uaddr_valid(ua, m->msg_namelen))
1427 			return -EINVAL;
1428 		atype = ua->addrtype;
1429 	}
1430 
1431 	/* If socket belongs to a communication group follow other paths */
1432 	if (grp) {
1433 		if (!ua)
1434 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1435 		if (atype == TIPC_SERVICE_ADDR)
1436 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1437 		if (atype == TIPC_SOCKET_ADDR)
1438 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1439 		if (atype == TIPC_SERVICE_RANGE)
1440 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1441 		return -EINVAL;
1442 	}
1443 
1444 	if (!ua) {
1445 		ua = (struct tipc_uaddr *)&tsk->peer;
1446 		if (!syn && ua->family != AF_TIPC)
1447 			return -EDESTADDRREQ;
1448 		atype = ua->addrtype;
1449 	}
1450 
1451 	if (unlikely(syn)) {
1452 		if (sk->sk_state == TIPC_LISTEN)
1453 			return -EPIPE;
1454 		if (sk->sk_state != TIPC_OPEN)
1455 			return -EISCONN;
1456 		if (tsk->published)
1457 			return -EOPNOTSUPP;
1458 		if (atype == TIPC_SERVICE_ADDR)
1459 			tsk->conn_addrtype = atype;
1460 		msg_set_syn(hdr, 1);
1461 	}
1462 
1463 	memset(&skaddr, 0, sizeof(skaddr));
1464 
1465 	/* Determine destination */
1466 	if (atype == TIPC_SERVICE_RANGE) {
1467 		return tipc_sendmcast(sock, ua, m, dlen, timeout);
1468 	} else if (atype == TIPC_SERVICE_ADDR) {
1469 		skaddr.node = ua->lookup_node;
1470 		ua->scope = tipc_node2scope(skaddr.node);
1471 		if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1472 			return -EHOSTUNREACH;
1473 	} else if (atype == TIPC_SOCKET_ADDR) {
1474 		skaddr = ua->sk;
1475 	} else {
1476 		return -EINVAL;
1477 	}
1478 
1479 	/* Block or return if destination link is congested */
1480 	rc = tipc_wait_for_cond(sock, &timeout,
1481 				!tipc_dest_find(clinks, skaddr.node, 0));
1482 	if (unlikely(rc))
1483 		return rc;
1484 
1485 	/* Finally build message header */
1486 	msg_set_destnode(hdr, skaddr.node);
1487 	msg_set_destport(hdr, skaddr.ref);
1488 	if (atype == TIPC_SERVICE_ADDR) {
1489 		msg_set_type(hdr, TIPC_NAMED_MSG);
1490 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1491 		msg_set_nametype(hdr, ua->sa.type);
1492 		msg_set_nameinst(hdr, ua->sa.instance);
1493 		msg_set_lookup_scope(hdr, ua->scope);
1494 	} else { /* TIPC_SOCKET_ADDR */
1495 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1496 		msg_set_lookup_scope(hdr, 0);
1497 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1498 	}
1499 
1500 	/* Add message body */
1501 	__skb_queue_head_init(&pkts);
1502 	mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1503 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1504 	if (unlikely(rc != dlen))
1505 		return rc;
1506 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1507 		__skb_queue_purge(&pkts);
1508 		return -ENOMEM;
1509 	}
1510 
1511 	/* Send message */
1512 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1513 	rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1514 	if (unlikely(rc == -ELINKCONG)) {
1515 		tipc_dest_push(clinks, skaddr.node, 0);
1516 		tsk->cong_link_cnt++;
1517 		rc = 0;
1518 	}
1519 
1520 	if (unlikely(syn && !rc)) {
1521 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1522 		if (dlen && timeout) {
1523 			timeout = msecs_to_jiffies(timeout);
1524 			tipc_wait_for_connect(sock, &timeout);
1525 		}
1526 	}
1527 
1528 	return rc ? rc : dlen;
1529 }
1530 
1531 /**
1532  * tipc_sendstream - send stream-oriented data
1533  * @sock: socket structure
1534  * @m: data to send
1535  * @dsz: total length of data to be transmitted
1536  *
1537  * Used for SOCK_STREAM data.
1538  *
1539  * Return: the number of bytes sent on success (or partial success),
1540  * or errno if no data sent
1541  */
1542 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1543 {
1544 	struct sock *sk = sock->sk;
1545 	int ret;
1546 
1547 	lock_sock(sk);
1548 	ret = __tipc_sendstream(sock, m, dsz);
1549 	release_sock(sk);
1550 
1551 	return ret;
1552 }
1553 
1554 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1555 {
1556 	struct sock *sk = sock->sk;
1557 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1558 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1559 	struct sk_buff_head *txq = &sk->sk_write_queue;
1560 	struct tipc_sock *tsk = tipc_sk(sk);
1561 	struct tipc_msg *hdr = &tsk->phdr;
1562 	struct net *net = sock_net(sk);
1563 	struct sk_buff *skb;
1564 	u32 dnode = tsk_peer_node(tsk);
1565 	int maxnagle = tsk->maxnagle;
1566 	int maxpkt = tsk->max_pkt;
1567 	int send, sent = 0;
1568 	int blocks, rc = 0;
1569 
1570 	if (unlikely(dlen > INT_MAX))
1571 		return -EMSGSIZE;
1572 
1573 	/* Handle implicit connection setup */
1574 	if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1575 		rc = __tipc_sendmsg(sock, m, dlen);
1576 		if (dlen && dlen == rc) {
1577 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1578 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1579 		}
1580 		return rc;
1581 	}
1582 
1583 	do {
1584 		rc = tipc_wait_for_cond(sock, &timeout,
1585 					(!tsk->cong_link_cnt &&
1586 					 !tsk_conn_cong(tsk) &&
1587 					 tipc_sk_connected(sk)));
1588 		if (unlikely(rc))
1589 			break;
1590 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1591 		blocks = tsk->snd_backlog;
1592 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1593 		    send <= maxnagle) {
1594 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1595 			if (unlikely(rc < 0))
1596 				break;
1597 			blocks += rc;
1598 			tsk->msg_acc++;
1599 			if (blocks <= 64 && tsk->expect_ack) {
1600 				tsk->snd_backlog = blocks;
1601 				sent += send;
1602 				break;
1603 			} else if (blocks > 64) {
1604 				tsk->pkt_cnt += skb_queue_len(txq);
1605 			} else {
1606 				skb = skb_peek_tail(txq);
1607 				if (skb) {
1608 					msg_set_ack_required(buf_msg(skb));
1609 					tsk->expect_ack = true;
1610 				} else {
1611 					tsk->expect_ack = false;
1612 				}
1613 				tsk->msg_acc = 0;
1614 				tsk->pkt_cnt = 0;
1615 			}
1616 		} else {
1617 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1618 			if (unlikely(rc != send))
1619 				break;
1620 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1621 		}
1622 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1623 					 TIPC_DUMP_SK_SNDQ, " ");
1624 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1625 		if (unlikely(rc == -ELINKCONG)) {
1626 			tsk->cong_link_cnt = 1;
1627 			rc = 0;
1628 		}
1629 		if (likely(!rc)) {
1630 			tsk->snt_unacked += blocks;
1631 			tsk->snd_backlog = 0;
1632 			sent += send;
1633 		}
1634 	} while (sent < dlen && !rc);
1635 
1636 	return sent ? sent : rc;
1637 }
1638 
1639 /**
1640  * tipc_send_packet - send a connection-oriented message
1641  * @sock: socket structure
1642  * @m: message to send
1643  * @dsz: length of data to be transmitted
1644  *
1645  * Used for SOCK_SEQPACKET messages.
1646  *
1647  * Return: the number of bytes sent on success, or errno otherwise
1648  */
1649 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1650 {
1651 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1652 		return -EMSGSIZE;
1653 
1654 	return tipc_sendstream(sock, m, dsz);
1655 }
1656 
1657 /* tipc_sk_finish_conn - complete the setup of a connection
1658  */
1659 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1660 				u32 peer_node)
1661 {
1662 	struct sock *sk = &tsk->sk;
1663 	struct net *net = sock_net(sk);
1664 	struct tipc_msg *msg = &tsk->phdr;
1665 
1666 	msg_set_syn(msg, 0);
1667 	msg_set_destnode(msg, peer_node);
1668 	msg_set_destport(msg, peer_port);
1669 	msg_set_type(msg, TIPC_CONN_MSG);
1670 	msg_set_lookup_scope(msg, 0);
1671 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1672 
1673 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1674 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1675 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1676 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1677 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1678 	tsk_set_nagle(tsk);
1679 	__skb_queue_purge(&sk->sk_write_queue);
1680 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1681 		return;
1682 
1683 	/* Fall back to message based flow control */
1684 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1685 	tsk->snd_win = FLOWCTL_MSG_WIN;
1686 }
1687 
1688 /**
1689  * tipc_sk_set_orig_addr - capture sender's address for received message
1690  * @m: descriptor for message info
1691  * @skb: received message
1692  *
1693  * Note: Address is not captured if not requested by receiver.
1694  */
1695 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1696 {
1697 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1698 	struct tipc_msg *hdr = buf_msg(skb);
1699 
1700 	if (!srcaddr)
1701 		return;
1702 
1703 	srcaddr->sock.family = AF_TIPC;
1704 	srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1705 	srcaddr->sock.scope = 0;
1706 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1707 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1708 	srcaddr->sock.addr.name.domain = 0;
1709 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1710 
1711 	if (!msg_in_group(hdr))
1712 		return;
1713 
1714 	/* Group message users may also want to know sending member's id */
1715 	srcaddr->member.family = AF_TIPC;
1716 	srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1717 	srcaddr->member.scope = 0;
1718 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1719 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1720 	srcaddr->member.addr.name.domain = 0;
1721 	m->msg_namelen = sizeof(*srcaddr);
1722 }
1723 
1724 /**
1725  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1726  * @m: descriptor for message info
1727  * @skb: received message buffer
1728  * @tsk: TIPC port associated with message
1729  *
1730  * Note: Ancillary data is not captured if not requested by receiver.
1731  *
1732  * Return: 0 if successful, otherwise errno
1733  */
1734 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1735 				 struct tipc_sock *tsk)
1736 {
1737 	struct tipc_msg *hdr;
1738 	u32 data[3] = {0,};
1739 	bool has_addr;
1740 	int dlen, rc;
1741 
1742 	if (likely(m->msg_controllen == 0))
1743 		return 0;
1744 
1745 	hdr = buf_msg(skb);
1746 	dlen = msg_data_sz(hdr);
1747 
1748 	/* Capture errored message object, if any */
1749 	if (msg_errcode(hdr)) {
1750 		if (skb_linearize(skb))
1751 			return -ENOMEM;
1752 		hdr = buf_msg(skb);
1753 		data[0] = msg_errcode(hdr);
1754 		data[1] = dlen;
1755 		rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1756 		if (rc || !dlen)
1757 			return rc;
1758 		rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1759 		if (rc)
1760 			return rc;
1761 	}
1762 
1763 	/* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1764 	switch (msg_type(hdr)) {
1765 	case TIPC_NAMED_MSG:
1766 		has_addr = true;
1767 		data[0] = msg_nametype(hdr);
1768 		data[1] = msg_namelower(hdr);
1769 		data[2] = data[1];
1770 		break;
1771 	case TIPC_MCAST_MSG:
1772 		has_addr = true;
1773 		data[0] = msg_nametype(hdr);
1774 		data[1] = msg_namelower(hdr);
1775 		data[2] = msg_nameupper(hdr);
1776 		break;
1777 	case TIPC_CONN_MSG:
1778 		has_addr = !!tsk->conn_addrtype;
1779 		data[0] = msg_nametype(&tsk->phdr);
1780 		data[1] = msg_nameinst(&tsk->phdr);
1781 		data[2] = data[1];
1782 		break;
1783 	default:
1784 		has_addr = false;
1785 	}
1786 	if (!has_addr)
1787 		return 0;
1788 	return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1789 }
1790 
1791 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1792 {
1793 	struct sock *sk = &tsk->sk;
1794 	struct sk_buff *skb = NULL;
1795 	struct tipc_msg *msg;
1796 	u32 peer_port = tsk_peer_port(tsk);
1797 	u32 dnode = tsk_peer_node(tsk);
1798 
1799 	if (!tipc_sk_connected(sk))
1800 		return NULL;
1801 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1802 			      dnode, tsk_own_node(tsk), peer_port,
1803 			      tsk->portid, TIPC_OK);
1804 	if (!skb)
1805 		return NULL;
1806 	msg = buf_msg(skb);
1807 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1808 	tsk->rcv_unacked = 0;
1809 
1810 	/* Adjust to and advertize the correct window limit */
1811 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1812 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1813 		msg_set_adv_win(msg, tsk->rcv_win);
1814 	}
1815 	return skb;
1816 }
1817 
1818 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1819 {
1820 	struct sk_buff *skb;
1821 
1822 	skb = tipc_sk_build_ack(tsk);
1823 	if (!skb)
1824 		return;
1825 
1826 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1827 			   msg_link_selector(buf_msg(skb)));
1828 }
1829 
1830 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1831 {
1832 	struct sock *sk = sock->sk;
1833 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1834 	long timeo = *timeop;
1835 	int err = sock_error(sk);
1836 
1837 	if (err)
1838 		return err;
1839 
1840 	for (;;) {
1841 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1842 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1843 				err = -ENOTCONN;
1844 				break;
1845 			}
1846 			add_wait_queue(sk_sleep(sk), &wait);
1847 			release_sock(sk);
1848 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1849 			sched_annotate_sleep();
1850 			lock_sock(sk);
1851 			remove_wait_queue(sk_sleep(sk), &wait);
1852 		}
1853 		err = 0;
1854 		if (!skb_queue_empty(&sk->sk_receive_queue))
1855 			break;
1856 		err = -EAGAIN;
1857 		if (!timeo)
1858 			break;
1859 		err = sock_intr_errno(timeo);
1860 		if (signal_pending(current))
1861 			break;
1862 
1863 		err = sock_error(sk);
1864 		if (err)
1865 			break;
1866 	}
1867 	*timeop = timeo;
1868 	return err;
1869 }
1870 
1871 /**
1872  * tipc_recvmsg - receive packet-oriented message
1873  * @sock: network socket
1874  * @m: descriptor for message info
1875  * @buflen: length of user buffer area
1876  * @flags: receive flags
1877  *
1878  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1879  * If the complete message doesn't fit in user area, truncate it.
1880  *
1881  * Return: size of returned message data, errno otherwise
1882  */
1883 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1884 			size_t buflen,	int flags)
1885 {
1886 	struct sock *sk = sock->sk;
1887 	bool connected = !tipc_sk_type_connectionless(sk);
1888 	struct tipc_sock *tsk = tipc_sk(sk);
1889 	int rc, err, hlen, dlen, copy;
1890 	struct tipc_skb_cb *skb_cb;
1891 	struct sk_buff_head xmitq;
1892 	struct tipc_msg *hdr;
1893 	struct sk_buff *skb;
1894 	bool grp_evt;
1895 	long timeout;
1896 
1897 	/* Catch invalid receive requests */
1898 	if (unlikely(!buflen))
1899 		return -EINVAL;
1900 
1901 	lock_sock(sk);
1902 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1903 		rc = -ENOTCONN;
1904 		goto exit;
1905 	}
1906 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1907 
1908 	/* Step rcv queue to first msg with data or error; wait if necessary */
1909 	do {
1910 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1911 		if (unlikely(rc))
1912 			goto exit;
1913 		skb = skb_peek(&sk->sk_receive_queue);
1914 		skb_cb = TIPC_SKB_CB(skb);
1915 		hdr = buf_msg(skb);
1916 		dlen = msg_data_sz(hdr);
1917 		hlen = msg_hdr_sz(hdr);
1918 		err = msg_errcode(hdr);
1919 		grp_evt = msg_is_grp_evt(hdr);
1920 		if (likely(dlen || err))
1921 			break;
1922 		tsk_advance_rx_queue(sk);
1923 	} while (1);
1924 
1925 	/* Collect msg meta data, including error code and rejected data */
1926 	tipc_sk_set_orig_addr(m, skb);
1927 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1928 	if (unlikely(rc))
1929 		goto exit;
1930 	hdr = buf_msg(skb);
1931 
1932 	/* Capture data if non-error msg, otherwise just set return value */
1933 	if (likely(!err)) {
1934 		int offset = skb_cb->bytes_read;
1935 
1936 		copy = min_t(int, dlen - offset, buflen);
1937 		rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1938 		if (unlikely(rc))
1939 			goto exit;
1940 		if (unlikely(offset + copy < dlen)) {
1941 			if (flags & MSG_EOR) {
1942 				if (!(flags & MSG_PEEK))
1943 					skb_cb->bytes_read = offset + copy;
1944 			} else {
1945 				m->msg_flags |= MSG_TRUNC;
1946 				skb_cb->bytes_read = 0;
1947 			}
1948 		} else {
1949 			if (flags & MSG_EOR)
1950 				m->msg_flags |= MSG_EOR;
1951 			skb_cb->bytes_read = 0;
1952 		}
1953 	} else {
1954 		copy = 0;
1955 		rc = 0;
1956 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1957 			rc = -ECONNRESET;
1958 			goto exit;
1959 		}
1960 	}
1961 
1962 	/* Mark message as group event if applicable */
1963 	if (unlikely(grp_evt)) {
1964 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1965 			m->msg_flags |= MSG_EOR;
1966 		m->msg_flags |= MSG_OOB;
1967 		copy = 0;
1968 	}
1969 
1970 	/* Caption of data or error code/rejected data was successful */
1971 	if (unlikely(flags & MSG_PEEK))
1972 		goto exit;
1973 
1974 	/* Send group flow control advertisement when applicable */
1975 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1976 		__skb_queue_head_init(&xmitq);
1977 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1978 					  msg_orignode(hdr), msg_origport(hdr),
1979 					  &xmitq);
1980 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1981 	}
1982 
1983 	if (skb_cb->bytes_read)
1984 		goto exit;
1985 
1986 	tsk_advance_rx_queue(sk);
1987 
1988 	if (likely(!connected))
1989 		goto exit;
1990 
1991 	/* Send connection flow control advertisement when applicable */
1992 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1993 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1994 		tipc_sk_send_ack(tsk);
1995 exit:
1996 	release_sock(sk);
1997 	return rc ? rc : copy;
1998 }
1999 
2000 /**
2001  * tipc_recvstream - receive stream-oriented data
2002  * @sock: network socket
2003  * @m: descriptor for message info
2004  * @buflen: total size of user buffer area
2005  * @flags: receive flags
2006  *
2007  * Used for SOCK_STREAM messages only.  If not enough data is available
2008  * will optionally wait for more; never truncates data.
2009  *
2010  * Return: size of returned message data, errno otherwise
2011  */
2012 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2013 			   size_t buflen, int flags)
2014 {
2015 	struct sock *sk = sock->sk;
2016 	struct tipc_sock *tsk = tipc_sk(sk);
2017 	struct sk_buff *skb;
2018 	struct tipc_msg *hdr;
2019 	struct tipc_skb_cb *skb_cb;
2020 	bool peek = flags & MSG_PEEK;
2021 	int offset, required, copy, copied = 0;
2022 	int hlen, dlen, err, rc;
2023 	long timeout;
2024 
2025 	/* Catch invalid receive attempts */
2026 	if (unlikely(!buflen))
2027 		return -EINVAL;
2028 
2029 	lock_sock(sk);
2030 
2031 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2032 		rc = -ENOTCONN;
2033 		goto exit;
2034 	}
2035 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2036 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2037 
2038 	do {
2039 		/* Look at first msg in receive queue; wait if necessary */
2040 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2041 		if (unlikely(rc))
2042 			break;
2043 		skb = skb_peek(&sk->sk_receive_queue);
2044 		skb_cb = TIPC_SKB_CB(skb);
2045 		hdr = buf_msg(skb);
2046 		dlen = msg_data_sz(hdr);
2047 		hlen = msg_hdr_sz(hdr);
2048 		err = msg_errcode(hdr);
2049 
2050 		/* Discard any empty non-errored (SYN-) message */
2051 		if (unlikely(!dlen && !err)) {
2052 			tsk_advance_rx_queue(sk);
2053 			continue;
2054 		}
2055 
2056 		/* Collect msg meta data, incl. error code and rejected data */
2057 		if (!copied) {
2058 			tipc_sk_set_orig_addr(m, skb);
2059 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2060 			if (rc)
2061 				break;
2062 			hdr = buf_msg(skb);
2063 		}
2064 
2065 		/* Copy data if msg ok, otherwise return error/partial data */
2066 		if (likely(!err)) {
2067 			offset = skb_cb->bytes_read;
2068 			copy = min_t(int, dlen - offset, buflen - copied);
2069 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2070 			if (unlikely(rc))
2071 				break;
2072 			copied += copy;
2073 			offset += copy;
2074 			if (unlikely(offset < dlen)) {
2075 				if (!peek)
2076 					skb_cb->bytes_read = offset;
2077 				break;
2078 			}
2079 		} else {
2080 			rc = 0;
2081 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2082 				rc = -ECONNRESET;
2083 			if (copied || rc)
2084 				break;
2085 		}
2086 
2087 		if (unlikely(peek))
2088 			break;
2089 
2090 		tsk_advance_rx_queue(sk);
2091 
2092 		/* Send connection flow control advertisement when applicable */
2093 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2094 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2095 			tipc_sk_send_ack(tsk);
2096 
2097 		/* Exit if all requested data or FIN/error received */
2098 		if (copied == buflen || err)
2099 			break;
2100 
2101 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2102 exit:
2103 	release_sock(sk);
2104 	return copied ? copied : rc;
2105 }
2106 
2107 /**
2108  * tipc_write_space - wake up thread if port congestion is released
2109  * @sk: socket
2110  */
2111 static void tipc_write_space(struct sock *sk)
2112 {
2113 	struct socket_wq *wq;
2114 
2115 	rcu_read_lock();
2116 	wq = rcu_dereference(sk->sk_wq);
2117 	if (skwq_has_sleeper(wq))
2118 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2119 						EPOLLWRNORM | EPOLLWRBAND);
2120 	rcu_read_unlock();
2121 }
2122 
2123 /**
2124  * tipc_data_ready - wake up threads to indicate messages have been received
2125  * @sk: socket
2126  */
2127 static void tipc_data_ready(struct sock *sk)
2128 {
2129 	struct socket_wq *wq;
2130 
2131 	trace_sk_data_ready(sk);
2132 
2133 	rcu_read_lock();
2134 	wq = rcu_dereference(sk->sk_wq);
2135 	if (skwq_has_sleeper(wq))
2136 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2137 						EPOLLRDNORM | EPOLLRDBAND);
2138 	rcu_read_unlock();
2139 }
2140 
2141 static void tipc_sock_destruct(struct sock *sk)
2142 {
2143 	__skb_queue_purge(&sk->sk_receive_queue);
2144 }
2145 
2146 static void tipc_sk_proto_rcv(struct sock *sk,
2147 			      struct sk_buff_head *inputq,
2148 			      struct sk_buff_head *xmitq)
2149 {
2150 	struct sk_buff *skb = __skb_dequeue(inputq);
2151 	struct tipc_sock *tsk = tipc_sk(sk);
2152 	struct tipc_msg *hdr = buf_msg(skb);
2153 	struct tipc_group *grp = tsk->group;
2154 	bool wakeup = false;
2155 
2156 	switch (msg_user(hdr)) {
2157 	case CONN_MANAGER:
2158 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2159 		return;
2160 	case SOCK_WAKEUP:
2161 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2162 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2163 		smp_wmb();
2164 		tsk->cong_link_cnt--;
2165 		wakeup = true;
2166 		tipc_sk_push_backlog(tsk, false);
2167 		break;
2168 	case GROUP_PROTOCOL:
2169 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2170 		break;
2171 	case TOP_SRV:
2172 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2173 				      hdr, inputq, xmitq);
2174 		break;
2175 	default:
2176 		break;
2177 	}
2178 
2179 	if (wakeup)
2180 		sk->sk_write_space(sk);
2181 
2182 	kfree_skb(skb);
2183 }
2184 
2185 /**
2186  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2187  * @tsk: TIPC socket
2188  * @skb: pointer to message buffer.
2189  * @xmitq: for Nagle ACK if any
2190  * Return: true if message should be added to receive queue, false otherwise
2191  */
2192 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2193 				   struct sk_buff_head *xmitq)
2194 {
2195 	struct sock *sk = &tsk->sk;
2196 	struct net *net = sock_net(sk);
2197 	struct tipc_msg *hdr = buf_msg(skb);
2198 	bool con_msg = msg_connected(hdr);
2199 	u32 pport = tsk_peer_port(tsk);
2200 	u32 pnode = tsk_peer_node(tsk);
2201 	u32 oport = msg_origport(hdr);
2202 	u32 onode = msg_orignode(hdr);
2203 	int err = msg_errcode(hdr);
2204 	unsigned long delay;
2205 
2206 	if (unlikely(msg_mcast(hdr)))
2207 		return false;
2208 	tsk->oneway = 0;
2209 
2210 	switch (sk->sk_state) {
2211 	case TIPC_CONNECTING:
2212 		/* Setup ACK */
2213 		if (likely(con_msg)) {
2214 			if (err)
2215 				break;
2216 			tipc_sk_finish_conn(tsk, oport, onode);
2217 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2218 			/* ACK+ message with data is added to receive queue */
2219 			if (msg_data_sz(hdr))
2220 				return true;
2221 			/* Empty ACK-, - wake up sleeping connect() and drop */
2222 			sk->sk_state_change(sk);
2223 			msg_set_dest_droppable(hdr, 1);
2224 			return false;
2225 		}
2226 		/* Ignore connectionless message if not from listening socket */
2227 		if (oport != pport || onode != pnode)
2228 			return false;
2229 
2230 		/* Rejected SYN */
2231 		if (err != TIPC_ERR_OVERLOAD)
2232 			break;
2233 
2234 		/* Prepare for new setup attempt if we have a SYN clone */
2235 		if (skb_queue_empty(&sk->sk_write_queue))
2236 			break;
2237 		get_random_bytes(&delay, 2);
2238 		delay %= (tsk->conn_timeout / 4);
2239 		delay = msecs_to_jiffies(delay + 100);
2240 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2241 		return false;
2242 	case TIPC_OPEN:
2243 	case TIPC_DISCONNECTING:
2244 		return false;
2245 	case TIPC_LISTEN:
2246 		/* Accept only SYN message */
2247 		if (!msg_is_syn(hdr) &&
2248 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2249 			return false;
2250 		if (!con_msg && !err)
2251 			return true;
2252 		return false;
2253 	case TIPC_ESTABLISHED:
2254 		if (!skb_queue_empty(&sk->sk_write_queue))
2255 			tipc_sk_push_backlog(tsk, false);
2256 		/* Accept only connection-based messages sent by peer */
2257 		if (likely(con_msg && !err && pport == oport &&
2258 			   pnode == onode)) {
2259 			if (msg_ack_required(hdr)) {
2260 				struct sk_buff *skb;
2261 
2262 				skb = tipc_sk_build_ack(tsk);
2263 				if (skb) {
2264 					msg_set_nagle_ack(buf_msg(skb));
2265 					__skb_queue_tail(xmitq, skb);
2266 				}
2267 			}
2268 			return true;
2269 		}
2270 		if (!tsk_peer_msg(tsk, hdr))
2271 			return false;
2272 		if (!err)
2273 			return true;
2274 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2275 		tipc_node_remove_conn(net, pnode, tsk->portid);
2276 		sk->sk_state_change(sk);
2277 		return true;
2278 	default:
2279 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2280 	}
2281 	/* Abort connection setup attempt */
2282 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2283 	sk->sk_err = ECONNREFUSED;
2284 	sk->sk_state_change(sk);
2285 	return true;
2286 }
2287 
2288 /**
2289  * rcvbuf_limit - get proper overload limit of socket receive queue
2290  * @sk: socket
2291  * @skb: message
2292  *
2293  * For connection oriented messages, irrespective of importance,
2294  * default queue limit is 2 MB.
2295  *
2296  * For connectionless messages, queue limits are based on message
2297  * importance as follows:
2298  *
2299  * TIPC_LOW_IMPORTANCE       (2 MB)
2300  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2301  * TIPC_HIGH_IMPORTANCE      (8 MB)
2302  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2303  *
2304  * Return: overload limit according to corresponding message importance
2305  */
2306 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2307 {
2308 	struct tipc_sock *tsk = tipc_sk(sk);
2309 	struct tipc_msg *hdr = buf_msg(skb);
2310 
2311 	if (unlikely(msg_in_group(hdr)))
2312 		return READ_ONCE(sk->sk_rcvbuf);
2313 
2314 	if (unlikely(!msg_connected(hdr)))
2315 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2316 
2317 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2318 		return READ_ONCE(sk->sk_rcvbuf);
2319 
2320 	return FLOWCTL_MSG_LIM;
2321 }
2322 
2323 /**
2324  * tipc_sk_filter_rcv - validate incoming message
2325  * @sk: socket
2326  * @skb: pointer to message.
2327  * @xmitq: output message area (FIXME)
2328  *
2329  * Enqueues message on receive queue if acceptable; optionally handles
2330  * disconnect indication for a connected socket.
2331  *
2332  * Called with socket lock already taken
2333  */
2334 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2335 			       struct sk_buff_head *xmitq)
2336 {
2337 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2338 	struct tipc_sock *tsk = tipc_sk(sk);
2339 	struct tipc_group *grp = tsk->group;
2340 	struct tipc_msg *hdr = buf_msg(skb);
2341 	struct net *net = sock_net(sk);
2342 	struct sk_buff_head inputq;
2343 	int mtyp = msg_type(hdr);
2344 	int limit, err = TIPC_OK;
2345 
2346 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2347 	TIPC_SKB_CB(skb)->bytes_read = 0;
2348 	__skb_queue_head_init(&inputq);
2349 	__skb_queue_tail(&inputq, skb);
2350 
2351 	if (unlikely(!msg_isdata(hdr)))
2352 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2353 
2354 	if (unlikely(grp))
2355 		tipc_group_filter_msg(grp, &inputq, xmitq);
2356 
2357 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2358 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2359 
2360 	/* Validate and add to receive buffer if there is space */
2361 	while ((skb = __skb_dequeue(&inputq))) {
2362 		hdr = buf_msg(skb);
2363 		limit = rcvbuf_limit(sk, skb);
2364 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2365 		    (!sk_conn && msg_connected(hdr)) ||
2366 		    (!grp && msg_in_group(hdr)))
2367 			err = TIPC_ERR_NO_PORT;
2368 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2369 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2370 					   "err_overload2!");
2371 			atomic_inc(&sk->sk_drops);
2372 			err = TIPC_ERR_OVERLOAD;
2373 		}
2374 
2375 		if (unlikely(err)) {
2376 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2377 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2378 						      "@filter_rcv!");
2379 				__skb_queue_tail(xmitq, skb);
2380 			}
2381 			err = TIPC_OK;
2382 			continue;
2383 		}
2384 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2385 		skb_set_owner_r(skb, sk);
2386 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2387 					 "rcvq >90% allocated!");
2388 		sk->sk_data_ready(sk);
2389 	}
2390 }
2391 
2392 /**
2393  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2394  * @sk: socket
2395  * @skb: message
2396  *
2397  * Caller must hold socket lock
2398  */
2399 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2400 {
2401 	unsigned int before = sk_rmem_alloc_get(sk);
2402 	struct sk_buff_head xmitq;
2403 	unsigned int added;
2404 
2405 	__skb_queue_head_init(&xmitq);
2406 
2407 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2408 	added = sk_rmem_alloc_get(sk) - before;
2409 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2410 
2411 	/* Send pending response/rejected messages, if any */
2412 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2413 	return 0;
2414 }
2415 
2416 /**
2417  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2418  *                   inputq and try adding them to socket or backlog queue
2419  * @inputq: list of incoming buffers with potentially different destinations
2420  * @sk: socket where the buffers should be enqueued
2421  * @dport: port number for the socket
2422  * @xmitq: output queue
2423  *
2424  * Caller must hold socket lock
2425  */
2426 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2427 			    u32 dport, struct sk_buff_head *xmitq)
2428 {
2429 	unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2430 	struct sk_buff *skb;
2431 	unsigned int lim;
2432 	atomic_t *dcnt;
2433 	u32 onode;
2434 
2435 	while (skb_queue_len(inputq)) {
2436 		if (unlikely(time_after_eq(jiffies, time_limit)))
2437 			return;
2438 
2439 		skb = tipc_skb_dequeue(inputq, dport);
2440 		if (unlikely(!skb))
2441 			return;
2442 
2443 		/* Add message directly to receive queue if possible */
2444 		if (!sock_owned_by_user(sk)) {
2445 			tipc_sk_filter_rcv(sk, skb, xmitq);
2446 			continue;
2447 		}
2448 
2449 		/* Try backlog, compensating for double-counted bytes */
2450 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2451 		if (!sk->sk_backlog.len)
2452 			atomic_set(dcnt, 0);
2453 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2454 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2455 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2456 						 "bklg & rcvq >90% allocated!");
2457 			continue;
2458 		}
2459 
2460 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2461 		/* Overload => reject message back to sender */
2462 		onode = tipc_own_addr(sock_net(sk));
2463 		atomic_inc(&sk->sk_drops);
2464 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2465 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2466 					      "@sk_enqueue!");
2467 			__skb_queue_tail(xmitq, skb);
2468 		}
2469 		break;
2470 	}
2471 }
2472 
2473 /**
2474  * tipc_sk_rcv - handle a chain of incoming buffers
2475  * @net: the associated network namespace
2476  * @inputq: buffer list containing the buffers
2477  * Consumes all buffers in list until inputq is empty
2478  * Note: may be called in multiple threads referring to the same queue
2479  */
2480 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2481 {
2482 	struct sk_buff_head xmitq;
2483 	u32 dnode, dport = 0;
2484 	int err;
2485 	struct tipc_sock *tsk;
2486 	struct sock *sk;
2487 	struct sk_buff *skb;
2488 
2489 	__skb_queue_head_init(&xmitq);
2490 	while (skb_queue_len(inputq)) {
2491 		dport = tipc_skb_peek_port(inputq, dport);
2492 		tsk = tipc_sk_lookup(net, dport);
2493 
2494 		if (likely(tsk)) {
2495 			sk = &tsk->sk;
2496 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2497 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2498 				spin_unlock_bh(&sk->sk_lock.slock);
2499 			}
2500 			/* Send pending response/rejected messages, if any */
2501 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2502 			sock_put(sk);
2503 			continue;
2504 		}
2505 		/* No destination socket => dequeue skb if still there */
2506 		skb = tipc_skb_dequeue(inputq, dport);
2507 		if (!skb)
2508 			return;
2509 
2510 		/* Try secondary lookup if unresolved named message */
2511 		err = TIPC_ERR_NO_PORT;
2512 		if (tipc_msg_lookup_dest(net, skb, &err))
2513 			goto xmit;
2514 
2515 		/* Prepare for message rejection */
2516 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2517 			continue;
2518 
2519 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2520 xmit:
2521 		dnode = msg_destnode(buf_msg(skb));
2522 		tipc_node_xmit_skb(net, skb, dnode, dport);
2523 	}
2524 }
2525 
2526 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2527 {
2528 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2529 	struct sock *sk = sock->sk;
2530 	int done;
2531 
2532 	do {
2533 		int err = sock_error(sk);
2534 		if (err)
2535 			return err;
2536 		if (!*timeo_p)
2537 			return -ETIMEDOUT;
2538 		if (signal_pending(current))
2539 			return sock_intr_errno(*timeo_p);
2540 		if (sk->sk_state == TIPC_DISCONNECTING)
2541 			break;
2542 
2543 		add_wait_queue(sk_sleep(sk), &wait);
2544 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2545 				     &wait);
2546 		remove_wait_queue(sk_sleep(sk), &wait);
2547 	} while (!done);
2548 	return 0;
2549 }
2550 
2551 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2552 {
2553 	if (addr->family != AF_TIPC)
2554 		return false;
2555 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2556 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2557 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2558 		addr->addrtype == TIPC_SOCKET_ADDR);
2559 }
2560 
2561 /**
2562  * tipc_connect - establish a connection to another TIPC port
2563  * @sock: socket structure
2564  * @dest: socket address for destination port
2565  * @destlen: size of socket address data structure
2566  * @flags: file-related flags associated with socket
2567  *
2568  * Return: 0 on success, errno otherwise
2569  */
2570 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2571 			int destlen, int flags)
2572 {
2573 	struct sock *sk = sock->sk;
2574 	struct tipc_sock *tsk = tipc_sk(sk);
2575 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2576 	struct msghdr m = {NULL,};
2577 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2578 	int previous;
2579 	int res = 0;
2580 
2581 	if (destlen != sizeof(struct sockaddr_tipc))
2582 		return -EINVAL;
2583 
2584 	lock_sock(sk);
2585 
2586 	if (tsk->group) {
2587 		res = -EINVAL;
2588 		goto exit;
2589 	}
2590 
2591 	if (dst->family == AF_UNSPEC) {
2592 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2593 		if (!tipc_sk_type_connectionless(sk))
2594 			res = -EINVAL;
2595 		goto exit;
2596 	}
2597 	if (!tipc_sockaddr_is_sane(dst)) {
2598 		res = -EINVAL;
2599 		goto exit;
2600 	}
2601 	/* DGRAM/RDM connect(), just save the destaddr */
2602 	if (tipc_sk_type_connectionless(sk)) {
2603 		memcpy(&tsk->peer, dest, destlen);
2604 		goto exit;
2605 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2606 		res = -EINVAL;
2607 		goto exit;
2608 	}
2609 
2610 	previous = sk->sk_state;
2611 
2612 	switch (sk->sk_state) {
2613 	case TIPC_OPEN:
2614 		/* Send a 'SYN-' to destination */
2615 		m.msg_name = dest;
2616 		m.msg_namelen = destlen;
2617 		iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2618 
2619 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2620 		 * indicate send_msg() is never blocked.
2621 		 */
2622 		if (!timeout)
2623 			m.msg_flags = MSG_DONTWAIT;
2624 
2625 		res = __tipc_sendmsg(sock, &m, 0);
2626 		if ((res < 0) && (res != -EWOULDBLOCK))
2627 			goto exit;
2628 
2629 		/* Just entered TIPC_CONNECTING state; the only
2630 		 * difference is that return value in non-blocking
2631 		 * case is EINPROGRESS, rather than EALREADY.
2632 		 */
2633 		res = -EINPROGRESS;
2634 		fallthrough;
2635 	case TIPC_CONNECTING:
2636 		if (!timeout) {
2637 			if (previous == TIPC_CONNECTING)
2638 				res = -EALREADY;
2639 			goto exit;
2640 		}
2641 		timeout = msecs_to_jiffies(timeout);
2642 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2643 		res = tipc_wait_for_connect(sock, &timeout);
2644 		break;
2645 	case TIPC_ESTABLISHED:
2646 		res = -EISCONN;
2647 		break;
2648 	default:
2649 		res = -EINVAL;
2650 	}
2651 
2652 exit:
2653 	release_sock(sk);
2654 	return res;
2655 }
2656 
2657 /**
2658  * tipc_listen - allow socket to listen for incoming connections
2659  * @sock: socket structure
2660  * @len: (unused)
2661  *
2662  * Return: 0 on success, errno otherwise
2663  */
2664 static int tipc_listen(struct socket *sock, int len)
2665 {
2666 	struct sock *sk = sock->sk;
2667 	int res;
2668 
2669 	lock_sock(sk);
2670 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2671 	release_sock(sk);
2672 
2673 	return res;
2674 }
2675 
2676 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2677 {
2678 	struct sock *sk = sock->sk;
2679 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2680 	int err;
2681 
2682 	/* True wake-one mechanism for incoming connections: only
2683 	 * one process gets woken up, not the 'whole herd'.
2684 	 * Since we do not 'race & poll' for established sockets
2685 	 * anymore, the common case will execute the loop only once.
2686 	*/
2687 	for (;;) {
2688 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2689 			add_wait_queue(sk_sleep(sk), &wait);
2690 			release_sock(sk);
2691 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2692 			lock_sock(sk);
2693 			remove_wait_queue(sk_sleep(sk), &wait);
2694 		}
2695 		err = 0;
2696 		if (!skb_queue_empty(&sk->sk_receive_queue))
2697 			break;
2698 		err = -EAGAIN;
2699 		if (!timeo)
2700 			break;
2701 		err = sock_intr_errno(timeo);
2702 		if (signal_pending(current))
2703 			break;
2704 	}
2705 	return err;
2706 }
2707 
2708 /**
2709  * tipc_accept - wait for connection request
2710  * @sock: listening socket
2711  * @new_sock: new socket that is to be connected
2712  * @arg: arguments for accept
2713  *
2714  * Return: 0 on success, errno otherwise
2715  */
2716 static int tipc_accept(struct socket *sock, struct socket *new_sock,
2717 		       struct proto_accept_arg *arg)
2718 {
2719 	struct sock *new_sk, *sk = sock->sk;
2720 	struct tipc_sock *new_tsock;
2721 	struct msghdr m = {NULL,};
2722 	struct tipc_msg *msg;
2723 	struct sk_buff *buf;
2724 	long timeo;
2725 	int res;
2726 
2727 	lock_sock(sk);
2728 
2729 	if (sk->sk_state != TIPC_LISTEN) {
2730 		res = -EINVAL;
2731 		goto exit;
2732 	}
2733 	timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK);
2734 	res = tipc_wait_for_accept(sock, timeo);
2735 	if (res)
2736 		goto exit;
2737 
2738 	buf = skb_peek(&sk->sk_receive_queue);
2739 
2740 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, arg->kern);
2741 	if (res)
2742 		goto exit;
2743 	security_sk_clone(sock->sk, new_sock->sk);
2744 
2745 	new_sk = new_sock->sk;
2746 	new_tsock = tipc_sk(new_sk);
2747 	msg = buf_msg(buf);
2748 
2749 	/* we lock on new_sk; but lockdep sees the lock on sk */
2750 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2751 
2752 	/*
2753 	 * Reject any stray messages received by new socket
2754 	 * before the socket lock was taken (very, very unlikely)
2755 	 */
2756 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2757 
2758 	/* Connect new socket to it's peer */
2759 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2760 
2761 	tsk_set_importance(new_sk, msg_importance(msg));
2762 	if (msg_named(msg)) {
2763 		new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2764 		msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2765 		msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2766 	}
2767 
2768 	/*
2769 	 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2770 	 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2771 	 */
2772 	if (!msg_data_sz(msg)) {
2773 		tsk_advance_rx_queue(sk);
2774 	} else {
2775 		__skb_dequeue(&sk->sk_receive_queue);
2776 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2777 		skb_set_owner_r(buf, new_sk);
2778 	}
2779 	iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2780 	__tipc_sendstream(new_sock, &m, 0);
2781 	release_sock(new_sk);
2782 exit:
2783 	release_sock(sk);
2784 	return res;
2785 }
2786 
2787 /**
2788  * tipc_shutdown - shutdown socket connection
2789  * @sock: socket structure
2790  * @how: direction to close (must be SHUT_RDWR)
2791  *
2792  * Terminates connection (if necessary), then purges socket's receive queue.
2793  *
2794  * Return: 0 on success, errno otherwise
2795  */
2796 static int tipc_shutdown(struct socket *sock, int how)
2797 {
2798 	struct sock *sk = sock->sk;
2799 	int res;
2800 
2801 	if (how != SHUT_RDWR)
2802 		return -EINVAL;
2803 
2804 	lock_sock(sk);
2805 
2806 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2807 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2808 	sk->sk_shutdown = SHUTDOWN_MASK;
2809 
2810 	if (sk->sk_state == TIPC_DISCONNECTING) {
2811 		/* Discard any unreceived messages */
2812 		__skb_queue_purge(&sk->sk_receive_queue);
2813 
2814 		res = 0;
2815 	} else {
2816 		res = -ENOTCONN;
2817 	}
2818 	/* Wake up anyone sleeping in poll. */
2819 	sk->sk_state_change(sk);
2820 
2821 	release_sock(sk);
2822 	return res;
2823 }
2824 
2825 static void tipc_sk_check_probing_state(struct sock *sk,
2826 					struct sk_buff_head *list)
2827 {
2828 	struct tipc_sock *tsk = tipc_sk(sk);
2829 	u32 pnode = tsk_peer_node(tsk);
2830 	u32 pport = tsk_peer_port(tsk);
2831 	u32 self = tsk_own_node(tsk);
2832 	u32 oport = tsk->portid;
2833 	struct sk_buff *skb;
2834 
2835 	if (tsk->probe_unacked) {
2836 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2837 		sk->sk_err = ECONNABORTED;
2838 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2839 		sk->sk_state_change(sk);
2840 		return;
2841 	}
2842 	/* Prepare new probe */
2843 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2844 			      pnode, self, pport, oport, TIPC_OK);
2845 	if (skb)
2846 		__skb_queue_tail(list, skb);
2847 	tsk->probe_unacked = true;
2848 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2849 }
2850 
2851 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2852 {
2853 	struct tipc_sock *tsk = tipc_sk(sk);
2854 
2855 	/* Try again later if dest link is congested */
2856 	if (tsk->cong_link_cnt) {
2857 		sk_reset_timer(sk, &sk->sk_timer,
2858 			       jiffies + msecs_to_jiffies(100));
2859 		return;
2860 	}
2861 	/* Prepare SYN for retransmit */
2862 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2863 }
2864 
2865 static void tipc_sk_timeout(struct timer_list *t)
2866 {
2867 	struct sock *sk = from_timer(sk, t, sk_timer);
2868 	struct tipc_sock *tsk = tipc_sk(sk);
2869 	u32 pnode = tsk_peer_node(tsk);
2870 	struct sk_buff_head list;
2871 	int rc = 0;
2872 
2873 	__skb_queue_head_init(&list);
2874 	bh_lock_sock(sk);
2875 
2876 	/* Try again later if socket is busy */
2877 	if (sock_owned_by_user(sk)) {
2878 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2879 		bh_unlock_sock(sk);
2880 		sock_put(sk);
2881 		return;
2882 	}
2883 
2884 	if (sk->sk_state == TIPC_ESTABLISHED)
2885 		tipc_sk_check_probing_state(sk, &list);
2886 	else if (sk->sk_state == TIPC_CONNECTING)
2887 		tipc_sk_retry_connect(sk, &list);
2888 
2889 	bh_unlock_sock(sk);
2890 
2891 	if (!skb_queue_empty(&list))
2892 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2893 
2894 	/* SYN messages may cause link congestion */
2895 	if (rc == -ELINKCONG) {
2896 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2897 		tsk->cong_link_cnt = 1;
2898 	}
2899 	sock_put(sk);
2900 }
2901 
2902 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2903 {
2904 	struct sock *sk = &tsk->sk;
2905 	struct net *net = sock_net(sk);
2906 	struct tipc_socket_addr skaddr;
2907 	struct publication *p;
2908 	u32 key;
2909 
2910 	if (tipc_sk_connected(sk))
2911 		return -EINVAL;
2912 	key = tsk->portid + tsk->pub_count + 1;
2913 	if (key == tsk->portid)
2914 		return -EADDRINUSE;
2915 	skaddr.ref = tsk->portid;
2916 	skaddr.node = tipc_own_addr(net);
2917 	p = tipc_nametbl_publish(net, ua, &skaddr, key);
2918 	if (unlikely(!p))
2919 		return -EINVAL;
2920 
2921 	list_add(&p->binding_sock, &tsk->publications);
2922 	tsk->pub_count++;
2923 	tsk->published = true;
2924 	return 0;
2925 }
2926 
2927 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2928 {
2929 	struct net *net = sock_net(&tsk->sk);
2930 	struct publication *safe, *p;
2931 	struct tipc_uaddr _ua;
2932 	int rc = -EINVAL;
2933 
2934 	list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2935 		if (!ua) {
2936 			tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2937 				   p->sr.type, p->sr.lower, p->sr.upper);
2938 			tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2939 			continue;
2940 		}
2941 		/* Unbind specific publication */
2942 		if (p->scope != ua->scope)
2943 			continue;
2944 		if (p->sr.type != ua->sr.type)
2945 			continue;
2946 		if (p->sr.lower != ua->sr.lower)
2947 			continue;
2948 		if (p->sr.upper != ua->sr.upper)
2949 			break;
2950 		tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2951 		rc = 0;
2952 		break;
2953 	}
2954 	if (list_empty(&tsk->publications)) {
2955 		tsk->published = 0;
2956 		rc = 0;
2957 	}
2958 	return rc;
2959 }
2960 
2961 /* tipc_sk_reinit: set non-zero address in all existing sockets
2962  *                 when we go from standalone to network mode.
2963  */
2964 void tipc_sk_reinit(struct net *net)
2965 {
2966 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2967 	struct rhashtable_iter iter;
2968 	struct tipc_sock *tsk;
2969 	struct tipc_msg *msg;
2970 
2971 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2972 
2973 	do {
2974 		rhashtable_walk_start(&iter);
2975 
2976 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2977 			sock_hold(&tsk->sk);
2978 			rhashtable_walk_stop(&iter);
2979 			lock_sock(&tsk->sk);
2980 			msg = &tsk->phdr;
2981 			msg_set_prevnode(msg, tipc_own_addr(net));
2982 			msg_set_orignode(msg, tipc_own_addr(net));
2983 			release_sock(&tsk->sk);
2984 			rhashtable_walk_start(&iter);
2985 			sock_put(&tsk->sk);
2986 		}
2987 
2988 		rhashtable_walk_stop(&iter);
2989 	} while (tsk == ERR_PTR(-EAGAIN));
2990 
2991 	rhashtable_walk_exit(&iter);
2992 }
2993 
2994 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2995 {
2996 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2997 	struct tipc_sock *tsk;
2998 
2999 	rcu_read_lock();
3000 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
3001 	if (tsk)
3002 		sock_hold(&tsk->sk);
3003 	rcu_read_unlock();
3004 
3005 	return tsk;
3006 }
3007 
3008 static int tipc_sk_insert(struct tipc_sock *tsk)
3009 {
3010 	struct sock *sk = &tsk->sk;
3011 	struct net *net = sock_net(sk);
3012 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3013 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3014 	u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT;
3015 
3016 	while (remaining--) {
3017 		portid++;
3018 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3019 			portid = TIPC_MIN_PORT;
3020 		tsk->portid = portid;
3021 		sock_hold(&tsk->sk);
3022 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3023 						   tsk_rht_params))
3024 			return 0;
3025 		sock_put(&tsk->sk);
3026 	}
3027 
3028 	return -1;
3029 }
3030 
3031 static void tipc_sk_remove(struct tipc_sock *tsk)
3032 {
3033 	struct sock *sk = &tsk->sk;
3034 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3035 
3036 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3037 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3038 		__sock_put(sk);
3039 	}
3040 }
3041 
3042 static const struct rhashtable_params tsk_rht_params = {
3043 	.nelem_hint = 192,
3044 	.head_offset = offsetof(struct tipc_sock, node),
3045 	.key_offset = offsetof(struct tipc_sock, portid),
3046 	.key_len = sizeof(u32), /* portid */
3047 	.max_size = 1048576,
3048 	.min_size = 256,
3049 	.automatic_shrinking = true,
3050 };
3051 
3052 int tipc_sk_rht_init(struct net *net)
3053 {
3054 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3055 
3056 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3057 }
3058 
3059 void tipc_sk_rht_destroy(struct net *net)
3060 {
3061 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3062 
3063 	/* Wait for socket readers to complete */
3064 	synchronize_net();
3065 
3066 	rhashtable_destroy(&tn->sk_rht);
3067 }
3068 
3069 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3070 {
3071 	struct net *net = sock_net(&tsk->sk);
3072 	struct tipc_group *grp = tsk->group;
3073 	struct tipc_msg *hdr = &tsk->phdr;
3074 	struct tipc_uaddr ua;
3075 	int rc;
3076 
3077 	if (mreq->type < TIPC_RESERVED_TYPES)
3078 		return -EACCES;
3079 	if (mreq->scope > TIPC_NODE_SCOPE)
3080 		return -EINVAL;
3081 	if (mreq->scope != TIPC_NODE_SCOPE)
3082 		mreq->scope = TIPC_CLUSTER_SCOPE;
3083 	if (grp)
3084 		return -EACCES;
3085 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3086 	if (!grp)
3087 		return -ENOMEM;
3088 	tsk->group = grp;
3089 	msg_set_lookup_scope(hdr, mreq->scope);
3090 	msg_set_nametype(hdr, mreq->type);
3091 	msg_set_dest_droppable(hdr, true);
3092 	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3093 		   mreq->type, mreq->instance, mreq->instance);
3094 	tipc_nametbl_build_group(net, grp, &ua);
3095 	rc = tipc_sk_publish(tsk, &ua);
3096 	if (rc) {
3097 		tipc_group_delete(net, grp);
3098 		tsk->group = NULL;
3099 		return rc;
3100 	}
3101 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3102 	tsk->mc_method.rcast = true;
3103 	tsk->mc_method.mandatory = true;
3104 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3105 	return rc;
3106 }
3107 
3108 static int tipc_sk_leave(struct tipc_sock *tsk)
3109 {
3110 	struct net *net = sock_net(&tsk->sk);
3111 	struct tipc_group *grp = tsk->group;
3112 	struct tipc_uaddr ua;
3113 	int scope;
3114 
3115 	if (!grp)
3116 		return -EINVAL;
3117 	ua.addrtype = TIPC_SERVICE_RANGE;
3118 	tipc_group_self(grp, &ua.sr, &scope);
3119 	ua.scope = scope;
3120 	tipc_group_delete(net, grp);
3121 	tsk->group = NULL;
3122 	tipc_sk_withdraw(tsk, &ua);
3123 	return 0;
3124 }
3125 
3126 /**
3127  * tipc_setsockopt - set socket option
3128  * @sock: socket structure
3129  * @lvl: option level
3130  * @opt: option identifier
3131  * @ov: pointer to new option value
3132  * @ol: length of option value
3133  *
3134  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3135  * (to ease compatibility).
3136  *
3137  * Return: 0 on success, errno otherwise
3138  */
3139 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3140 			   sockptr_t ov, unsigned int ol)
3141 {
3142 	struct sock *sk = sock->sk;
3143 	struct tipc_sock *tsk = tipc_sk(sk);
3144 	struct tipc_group_req mreq;
3145 	u32 value = 0;
3146 	int res = 0;
3147 
3148 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3149 		return 0;
3150 	if (lvl != SOL_TIPC)
3151 		return -ENOPROTOOPT;
3152 
3153 	switch (opt) {
3154 	case TIPC_IMPORTANCE:
3155 	case TIPC_SRC_DROPPABLE:
3156 	case TIPC_DEST_DROPPABLE:
3157 	case TIPC_CONN_TIMEOUT:
3158 	case TIPC_NODELAY:
3159 		if (ol < sizeof(value))
3160 			return -EINVAL;
3161 		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3162 			return -EFAULT;
3163 		break;
3164 	case TIPC_GROUP_JOIN:
3165 		if (ol < sizeof(mreq))
3166 			return -EINVAL;
3167 		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3168 			return -EFAULT;
3169 		break;
3170 	default:
3171 		if (!sockptr_is_null(ov) || ol)
3172 			return -EINVAL;
3173 	}
3174 
3175 	lock_sock(sk);
3176 
3177 	switch (opt) {
3178 	case TIPC_IMPORTANCE:
3179 		res = tsk_set_importance(sk, value);
3180 		break;
3181 	case TIPC_SRC_DROPPABLE:
3182 		if (sock->type != SOCK_STREAM)
3183 			tsk_set_unreliable(tsk, value);
3184 		else
3185 			res = -ENOPROTOOPT;
3186 		break;
3187 	case TIPC_DEST_DROPPABLE:
3188 		tsk_set_unreturnable(tsk, value);
3189 		break;
3190 	case TIPC_CONN_TIMEOUT:
3191 		tipc_sk(sk)->conn_timeout = value;
3192 		break;
3193 	case TIPC_MCAST_BROADCAST:
3194 		tsk->mc_method.rcast = false;
3195 		tsk->mc_method.mandatory = true;
3196 		break;
3197 	case TIPC_MCAST_REPLICAST:
3198 		tsk->mc_method.rcast = true;
3199 		tsk->mc_method.mandatory = true;
3200 		break;
3201 	case TIPC_GROUP_JOIN:
3202 		res = tipc_sk_join(tsk, &mreq);
3203 		break;
3204 	case TIPC_GROUP_LEAVE:
3205 		res = tipc_sk_leave(tsk);
3206 		break;
3207 	case TIPC_NODELAY:
3208 		tsk->nodelay = !!value;
3209 		tsk_set_nagle(tsk);
3210 		break;
3211 	default:
3212 		res = -EINVAL;
3213 	}
3214 
3215 	release_sock(sk);
3216 
3217 	return res;
3218 }
3219 
3220 /**
3221  * tipc_getsockopt - get socket option
3222  * @sock: socket structure
3223  * @lvl: option level
3224  * @opt: option identifier
3225  * @ov: receptacle for option value
3226  * @ol: receptacle for length of option value
3227  *
3228  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3229  * (to ease compatibility).
3230  *
3231  * Return: 0 on success, errno otherwise
3232  */
3233 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3234 			   char __user *ov, int __user *ol)
3235 {
3236 	struct sock *sk = sock->sk;
3237 	struct tipc_sock *tsk = tipc_sk(sk);
3238 	struct tipc_service_range seq;
3239 	int len, scope;
3240 	u32 value;
3241 	int res;
3242 
3243 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3244 		return put_user(0, ol);
3245 	if (lvl != SOL_TIPC)
3246 		return -ENOPROTOOPT;
3247 	res = get_user(len, ol);
3248 	if (res)
3249 		return res;
3250 
3251 	lock_sock(sk);
3252 
3253 	switch (opt) {
3254 	case TIPC_IMPORTANCE:
3255 		value = tsk_importance(tsk);
3256 		break;
3257 	case TIPC_SRC_DROPPABLE:
3258 		value = tsk_unreliable(tsk);
3259 		break;
3260 	case TIPC_DEST_DROPPABLE:
3261 		value = tsk_unreturnable(tsk);
3262 		break;
3263 	case TIPC_CONN_TIMEOUT:
3264 		value = tsk->conn_timeout;
3265 		/* no need to set "res", since already 0 at this point */
3266 		break;
3267 	case TIPC_NODE_RECVQ_DEPTH:
3268 		value = 0; /* was tipc_queue_size, now obsolete */
3269 		break;
3270 	case TIPC_SOCK_RECVQ_DEPTH:
3271 		value = skb_queue_len(&sk->sk_receive_queue);
3272 		break;
3273 	case TIPC_SOCK_RECVQ_USED:
3274 		value = sk_rmem_alloc_get(sk);
3275 		break;
3276 	case TIPC_GROUP_JOIN:
3277 		seq.type = 0;
3278 		if (tsk->group)
3279 			tipc_group_self(tsk->group, &seq, &scope);
3280 		value = seq.type;
3281 		break;
3282 	default:
3283 		res = -EINVAL;
3284 	}
3285 
3286 	release_sock(sk);
3287 
3288 	if (res)
3289 		return res;	/* "get" failed */
3290 
3291 	if (len < sizeof(value))
3292 		return -EINVAL;
3293 
3294 	if (copy_to_user(ov, &value, sizeof(value)))
3295 		return -EFAULT;
3296 
3297 	return put_user(sizeof(value), ol);
3298 }
3299 
3300 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3301 {
3302 	struct net *net = sock_net(sock->sk);
3303 	struct tipc_sioc_nodeid_req nr = {0};
3304 	struct tipc_sioc_ln_req lnr;
3305 	void __user *argp = (void __user *)arg;
3306 
3307 	switch (cmd) {
3308 	case SIOCGETLINKNAME:
3309 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3310 			return -EFAULT;
3311 		if (!tipc_node_get_linkname(net,
3312 					    lnr.bearer_id & 0xffff, lnr.peer,
3313 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3314 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3315 				return -EFAULT;
3316 			return 0;
3317 		}
3318 		return -EADDRNOTAVAIL;
3319 	case SIOCGETNODEID:
3320 		if (copy_from_user(&nr, argp, sizeof(nr)))
3321 			return -EFAULT;
3322 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3323 			return -EADDRNOTAVAIL;
3324 		if (copy_to_user(argp, &nr, sizeof(nr)))
3325 			return -EFAULT;
3326 		return 0;
3327 	default:
3328 		return -ENOIOCTLCMD;
3329 	}
3330 }
3331 
3332 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3333 {
3334 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3335 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3336 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3337 
3338 	tsk1->peer.family = AF_TIPC;
3339 	tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3340 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3341 	tsk1->peer.addr.id.ref = tsk2->portid;
3342 	tsk1->peer.addr.id.node = onode;
3343 	tsk2->peer.family = AF_TIPC;
3344 	tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3345 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3346 	tsk2->peer.addr.id.ref = tsk1->portid;
3347 	tsk2->peer.addr.id.node = onode;
3348 
3349 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3350 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3351 	return 0;
3352 }
3353 
3354 /* Protocol switches for the various types of TIPC sockets */
3355 
3356 static const struct proto_ops msg_ops = {
3357 	.owner		= THIS_MODULE,
3358 	.family		= AF_TIPC,
3359 	.release	= tipc_release,
3360 	.bind		= tipc_bind,
3361 	.connect	= tipc_connect,
3362 	.socketpair	= tipc_socketpair,
3363 	.accept		= sock_no_accept,
3364 	.getname	= tipc_getname,
3365 	.poll		= tipc_poll,
3366 	.ioctl		= tipc_ioctl,
3367 	.listen		= sock_no_listen,
3368 	.shutdown	= tipc_shutdown,
3369 	.setsockopt	= tipc_setsockopt,
3370 	.getsockopt	= tipc_getsockopt,
3371 	.sendmsg	= tipc_sendmsg,
3372 	.recvmsg	= tipc_recvmsg,
3373 	.mmap		= sock_no_mmap,
3374 };
3375 
3376 static const struct proto_ops packet_ops = {
3377 	.owner		= THIS_MODULE,
3378 	.family		= AF_TIPC,
3379 	.release	= tipc_release,
3380 	.bind		= tipc_bind,
3381 	.connect	= tipc_connect,
3382 	.socketpair	= tipc_socketpair,
3383 	.accept		= tipc_accept,
3384 	.getname	= tipc_getname,
3385 	.poll		= tipc_poll,
3386 	.ioctl		= tipc_ioctl,
3387 	.listen		= tipc_listen,
3388 	.shutdown	= tipc_shutdown,
3389 	.setsockopt	= tipc_setsockopt,
3390 	.getsockopt	= tipc_getsockopt,
3391 	.sendmsg	= tipc_send_packet,
3392 	.recvmsg	= tipc_recvmsg,
3393 	.mmap		= sock_no_mmap,
3394 };
3395 
3396 static const struct proto_ops stream_ops = {
3397 	.owner		= THIS_MODULE,
3398 	.family		= AF_TIPC,
3399 	.release	= tipc_release,
3400 	.bind		= tipc_bind,
3401 	.connect	= tipc_connect,
3402 	.socketpair	= tipc_socketpair,
3403 	.accept		= tipc_accept,
3404 	.getname	= tipc_getname,
3405 	.poll		= tipc_poll,
3406 	.ioctl		= tipc_ioctl,
3407 	.listen		= tipc_listen,
3408 	.shutdown	= tipc_shutdown,
3409 	.setsockopt	= tipc_setsockopt,
3410 	.getsockopt	= tipc_getsockopt,
3411 	.sendmsg	= tipc_sendstream,
3412 	.recvmsg	= tipc_recvstream,
3413 	.mmap		= sock_no_mmap,
3414 };
3415 
3416 static const struct net_proto_family tipc_family_ops = {
3417 	.owner		= THIS_MODULE,
3418 	.family		= AF_TIPC,
3419 	.create		= tipc_sk_create
3420 };
3421 
3422 static struct proto tipc_proto = {
3423 	.name		= "TIPC",
3424 	.owner		= THIS_MODULE,
3425 	.obj_size	= sizeof(struct tipc_sock),
3426 	.sysctl_rmem	= sysctl_tipc_rmem
3427 };
3428 
3429 /**
3430  * tipc_socket_init - initialize TIPC socket interface
3431  *
3432  * Return: 0 on success, errno otherwise
3433  */
3434 int tipc_socket_init(void)
3435 {
3436 	int res;
3437 
3438 	res = proto_register(&tipc_proto, 1);
3439 	if (res) {
3440 		pr_err("Failed to register TIPC protocol type\n");
3441 		goto out;
3442 	}
3443 
3444 	res = sock_register(&tipc_family_ops);
3445 	if (res) {
3446 		pr_err("Failed to register TIPC socket type\n");
3447 		proto_unregister(&tipc_proto);
3448 		goto out;
3449 	}
3450  out:
3451 	return res;
3452 }
3453 
3454 /**
3455  * tipc_socket_stop - stop TIPC socket interface
3456  */
3457 void tipc_socket_stop(void)
3458 {
3459 	sock_unregister(tipc_family_ops.family);
3460 	proto_unregister(&tipc_proto);
3461 }
3462 
3463 /* Caller should hold socket lock for the passed tipc socket. */
3464 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3465 {
3466 	u32 peer_node, peer_port;
3467 	u32 conn_type, conn_instance;
3468 	struct nlattr *nest;
3469 
3470 	peer_node = tsk_peer_node(tsk);
3471 	peer_port = tsk_peer_port(tsk);
3472 	conn_type = msg_nametype(&tsk->phdr);
3473 	conn_instance = msg_nameinst(&tsk->phdr);
3474 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3475 	if (!nest)
3476 		return -EMSGSIZE;
3477 
3478 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3479 		goto msg_full;
3480 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3481 		goto msg_full;
3482 
3483 	if (tsk->conn_addrtype != 0) {
3484 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3485 			goto msg_full;
3486 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3487 			goto msg_full;
3488 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3489 			goto msg_full;
3490 	}
3491 	nla_nest_end(skb, nest);
3492 
3493 	return 0;
3494 
3495 msg_full:
3496 	nla_nest_cancel(skb, nest);
3497 
3498 	return -EMSGSIZE;
3499 }
3500 
3501 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3502 			  *tsk)
3503 {
3504 	struct net *net = sock_net(skb->sk);
3505 	struct sock *sk = &tsk->sk;
3506 
3507 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3508 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3509 		return -EMSGSIZE;
3510 
3511 	if (tipc_sk_connected(sk)) {
3512 		if (__tipc_nl_add_sk_con(skb, tsk))
3513 			return -EMSGSIZE;
3514 	} else if (!list_empty(&tsk->publications)) {
3515 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3516 			return -EMSGSIZE;
3517 	}
3518 	return 0;
3519 }
3520 
3521 /* Caller should hold socket lock for the passed tipc socket. */
3522 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3523 			    struct tipc_sock *tsk)
3524 {
3525 	struct nlattr *attrs;
3526 	void *hdr;
3527 
3528 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3529 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3530 	if (!hdr)
3531 		goto msg_cancel;
3532 
3533 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3534 	if (!attrs)
3535 		goto genlmsg_cancel;
3536 
3537 	if (__tipc_nl_add_sk_info(skb, tsk))
3538 		goto attr_msg_cancel;
3539 
3540 	nla_nest_end(skb, attrs);
3541 	genlmsg_end(skb, hdr);
3542 
3543 	return 0;
3544 
3545 attr_msg_cancel:
3546 	nla_nest_cancel(skb, attrs);
3547 genlmsg_cancel:
3548 	genlmsg_cancel(skb, hdr);
3549 msg_cancel:
3550 	return -EMSGSIZE;
3551 }
3552 
3553 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3554 		    int (*skb_handler)(struct sk_buff *skb,
3555 				       struct netlink_callback *cb,
3556 				       struct tipc_sock *tsk))
3557 {
3558 	struct rhashtable_iter *iter = (void *)cb->args[4];
3559 	struct tipc_sock *tsk;
3560 	int err;
3561 
3562 	rhashtable_walk_start(iter);
3563 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3564 		if (IS_ERR(tsk)) {
3565 			if (PTR_ERR(tsk) == -EAGAIN)
3566 				continue;
3567 			break;
3568 		}
3569 
3570 		sock_hold(&tsk->sk);
3571 		rhashtable_walk_stop(iter);
3572 		lock_sock(&tsk->sk);
3573 		err = skb_handler(skb, cb, tsk);
3574 		if (err) {
3575 			release_sock(&tsk->sk);
3576 			sock_put(&tsk->sk);
3577 			goto out;
3578 		}
3579 		release_sock(&tsk->sk);
3580 		rhashtable_walk_start(iter);
3581 		sock_put(&tsk->sk);
3582 	}
3583 	rhashtable_walk_stop(iter);
3584 out:
3585 	return skb->len;
3586 }
3587 EXPORT_SYMBOL(tipc_nl_sk_walk);
3588 
3589 int tipc_dump_start(struct netlink_callback *cb)
3590 {
3591 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3592 }
3593 EXPORT_SYMBOL(tipc_dump_start);
3594 
3595 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3596 {
3597 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3598 	struct rhashtable_iter *iter = (void *)cb->args[4];
3599 	struct tipc_net *tn = tipc_net(net);
3600 
3601 	if (!iter) {
3602 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3603 		if (!iter)
3604 			return -ENOMEM;
3605 
3606 		cb->args[4] = (long)iter;
3607 	}
3608 
3609 	rhashtable_walk_enter(&tn->sk_rht, iter);
3610 	return 0;
3611 }
3612 
3613 int tipc_dump_done(struct netlink_callback *cb)
3614 {
3615 	struct rhashtable_iter *hti = (void *)cb->args[4];
3616 
3617 	rhashtable_walk_exit(hti);
3618 	kfree(hti);
3619 	return 0;
3620 }
3621 EXPORT_SYMBOL(tipc_dump_done);
3622 
3623 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3624 			   struct tipc_sock *tsk, u32 sk_filter_state,
3625 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3626 {
3627 	struct sock *sk = &tsk->sk;
3628 	struct nlattr *attrs;
3629 	struct nlattr *stat;
3630 
3631 	/*filter response w.r.t sk_state*/
3632 	if (!(sk_filter_state & (1 << sk->sk_state)))
3633 		return 0;
3634 
3635 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3636 	if (!attrs)
3637 		goto msg_cancel;
3638 
3639 	if (__tipc_nl_add_sk_info(skb, tsk))
3640 		goto attr_msg_cancel;
3641 
3642 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3643 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3644 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3645 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3646 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3647 					 sock_i_uid(sk))) ||
3648 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3649 			      tipc_diag_gen_cookie(sk),
3650 			      TIPC_NLA_SOCK_PAD))
3651 		goto attr_msg_cancel;
3652 
3653 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3654 	if (!stat)
3655 		goto attr_msg_cancel;
3656 
3657 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3658 			skb_queue_len(&sk->sk_receive_queue)) ||
3659 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3660 			skb_queue_len(&sk->sk_write_queue)) ||
3661 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3662 			atomic_read(&sk->sk_drops)))
3663 		goto stat_msg_cancel;
3664 
3665 	if (tsk->cong_link_cnt &&
3666 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3667 		goto stat_msg_cancel;
3668 
3669 	if (tsk_conn_cong(tsk) &&
3670 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3671 		goto stat_msg_cancel;
3672 
3673 	nla_nest_end(skb, stat);
3674 
3675 	if (tsk->group)
3676 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3677 			goto stat_msg_cancel;
3678 
3679 	nla_nest_end(skb, attrs);
3680 
3681 	return 0;
3682 
3683 stat_msg_cancel:
3684 	nla_nest_cancel(skb, stat);
3685 attr_msg_cancel:
3686 	nla_nest_cancel(skb, attrs);
3687 msg_cancel:
3688 	return -EMSGSIZE;
3689 }
3690 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3691 
3692 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3693 {
3694 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3695 }
3696 
3697 /* Caller should hold socket lock for the passed tipc socket. */
3698 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3699 				 struct netlink_callback *cb,
3700 				 struct publication *publ)
3701 {
3702 	void *hdr;
3703 	struct nlattr *attrs;
3704 
3705 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3706 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3707 	if (!hdr)
3708 		goto msg_cancel;
3709 
3710 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3711 	if (!attrs)
3712 		goto genlmsg_cancel;
3713 
3714 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3715 		goto attr_msg_cancel;
3716 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3717 		goto attr_msg_cancel;
3718 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3719 		goto attr_msg_cancel;
3720 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3721 		goto attr_msg_cancel;
3722 
3723 	nla_nest_end(skb, attrs);
3724 	genlmsg_end(skb, hdr);
3725 
3726 	return 0;
3727 
3728 attr_msg_cancel:
3729 	nla_nest_cancel(skb, attrs);
3730 genlmsg_cancel:
3731 	genlmsg_cancel(skb, hdr);
3732 msg_cancel:
3733 	return -EMSGSIZE;
3734 }
3735 
3736 /* Caller should hold socket lock for the passed tipc socket. */
3737 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3738 				  struct netlink_callback *cb,
3739 				  struct tipc_sock *tsk, u32 *last_publ)
3740 {
3741 	int err;
3742 	struct publication *p;
3743 
3744 	if (*last_publ) {
3745 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3746 			if (p->key == *last_publ)
3747 				break;
3748 		}
3749 		if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3750 			/* We never set seq or call nl_dump_check_consistent()
3751 			 * this means that setting prev_seq here will cause the
3752 			 * consistence check to fail in the netlink callback
3753 			 * handler. Resulting in the last NLMSG_DONE message
3754 			 * having the NLM_F_DUMP_INTR flag set.
3755 			 */
3756 			cb->prev_seq = 1;
3757 			*last_publ = 0;
3758 			return -EPIPE;
3759 		}
3760 	} else {
3761 		p = list_first_entry(&tsk->publications, struct publication,
3762 				     binding_sock);
3763 	}
3764 
3765 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3766 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3767 		if (err) {
3768 			*last_publ = p->key;
3769 			return err;
3770 		}
3771 	}
3772 	*last_publ = 0;
3773 
3774 	return 0;
3775 }
3776 
3777 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3778 {
3779 	int err;
3780 	u32 tsk_portid = cb->args[0];
3781 	u32 last_publ = cb->args[1];
3782 	u32 done = cb->args[2];
3783 	struct net *net = sock_net(skb->sk);
3784 	struct tipc_sock *tsk;
3785 
3786 	if (!tsk_portid) {
3787 		struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs;
3788 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3789 
3790 		if (!attrs[TIPC_NLA_SOCK])
3791 			return -EINVAL;
3792 
3793 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3794 						  attrs[TIPC_NLA_SOCK],
3795 						  tipc_nl_sock_policy, NULL);
3796 		if (err)
3797 			return err;
3798 
3799 		if (!sock[TIPC_NLA_SOCK_REF])
3800 			return -EINVAL;
3801 
3802 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3803 	}
3804 
3805 	if (done)
3806 		return 0;
3807 
3808 	tsk = tipc_sk_lookup(net, tsk_portid);
3809 	if (!tsk)
3810 		return -EINVAL;
3811 
3812 	lock_sock(&tsk->sk);
3813 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3814 	if (!err)
3815 		done = 1;
3816 	release_sock(&tsk->sk);
3817 	sock_put(&tsk->sk);
3818 
3819 	cb->args[0] = tsk_portid;
3820 	cb->args[1] = last_publ;
3821 	cb->args[2] = done;
3822 
3823 	return skb->len;
3824 }
3825 
3826 /**
3827  * tipc_sk_filtering - check if a socket should be traced
3828  * @sk: the socket to be examined
3829  *
3830  * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3831  * (portid, sock type, name type, name lower, name upper)
3832  *
3833  * Return: true if the socket meets the socket tuple data
3834  * (value 0 = 'any') or when there is no tuple set (all = 0),
3835  * otherwise false
3836  */
3837 bool tipc_sk_filtering(struct sock *sk)
3838 {
3839 	struct tipc_sock *tsk;
3840 	struct publication *p;
3841 	u32 _port, _sktype, _type, _lower, _upper;
3842 	u32 type = 0, lower = 0, upper = 0;
3843 
3844 	if (!sk)
3845 		return true;
3846 
3847 	tsk = tipc_sk(sk);
3848 
3849 	_port = sysctl_tipc_sk_filter[0];
3850 	_sktype = sysctl_tipc_sk_filter[1];
3851 	_type = sysctl_tipc_sk_filter[2];
3852 	_lower = sysctl_tipc_sk_filter[3];
3853 	_upper = sysctl_tipc_sk_filter[4];
3854 
3855 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3856 		return true;
3857 
3858 	if (_port)
3859 		return (_port == tsk->portid);
3860 
3861 	if (_sktype && _sktype != sk->sk_type)
3862 		return false;
3863 
3864 	if (tsk->published) {
3865 		p = list_first_entry_or_null(&tsk->publications,
3866 					     struct publication, binding_sock);
3867 		if (p) {
3868 			type = p->sr.type;
3869 			lower = p->sr.lower;
3870 			upper = p->sr.upper;
3871 		}
3872 	}
3873 
3874 	if (!tipc_sk_type_connectionless(sk)) {
3875 		type = msg_nametype(&tsk->phdr);
3876 		lower = msg_nameinst(&tsk->phdr);
3877 		upper = lower;
3878 	}
3879 
3880 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3881 	    (_upper && _upper != upper))
3882 		return false;
3883 
3884 	return true;
3885 }
3886 
3887 u32 tipc_sock_get_portid(struct sock *sk)
3888 {
3889 	return (sk) ? (tipc_sk(sk))->portid : 0;
3890 }
3891 
3892 /**
3893  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3894  *			both the rcv and backlog queues are considered
3895  * @sk: tipc sk to be checked
3896  * @skb: tipc msg to be checked
3897  *
3898  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3899  */
3900 
3901 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3902 {
3903 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3904 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3905 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3906 
3907 	return (qsize > lim * 90 / 100);
3908 }
3909 
3910 /**
3911  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3912  *			only the rcv queue is considered
3913  * @sk: tipc sk to be checked
3914  * @skb: tipc msg to be checked
3915  *
3916  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3917  */
3918 
3919 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3920 {
3921 	unsigned int lim = rcvbuf_limit(sk, skb);
3922 	unsigned int qsize = sk_rmem_alloc_get(sk);
3923 
3924 	return (qsize > lim * 90 / 100);
3925 }
3926 
3927 /**
3928  * tipc_sk_dump - dump TIPC socket
3929  * @sk: tipc sk to be dumped
3930  * @dqueues: bitmask to decide if any socket queue to be dumped?
3931  *           - TIPC_DUMP_NONE: don't dump socket queues
3932  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3933  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3934  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3935  *           - TIPC_DUMP_ALL: dump all the socket queues above
3936  * @buf: returned buffer of dump data in format
3937  */
3938 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3939 {
3940 	int i = 0;
3941 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3942 	u32 conn_type, conn_instance;
3943 	struct tipc_sock *tsk;
3944 	struct publication *p;
3945 	bool tsk_connected;
3946 
3947 	if (!sk) {
3948 		i += scnprintf(buf, sz, "sk data: (null)\n");
3949 		return i;
3950 	}
3951 
3952 	tsk = tipc_sk(sk);
3953 	tsk_connected = !tipc_sk_type_connectionless(sk);
3954 
3955 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3956 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3957 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3958 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3959 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3960 	if (tsk_connected) {
3961 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3962 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3963 		conn_type = msg_nametype(&tsk->phdr);
3964 		conn_instance = msg_nameinst(&tsk->phdr);
3965 		i += scnprintf(buf + i, sz - i, " %u", conn_type);
3966 		i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3967 	}
3968 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3969 	if (tsk->published) {
3970 		p = list_first_entry_or_null(&tsk->publications,
3971 					     struct publication, binding_sock);
3972 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3973 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3974 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3975 	}
3976 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3977 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3978 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3979 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3980 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3981 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3982 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3983 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3984 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3985 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3986 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3987 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3988 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3989 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3990 
3991 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3992 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3993 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3994 	}
3995 
3996 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3997 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3998 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3999 	}
4000 
4001 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4002 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
4003 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4004 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4005 			i += scnprintf(buf + i, sz - i, "  tail ");
4006 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
4007 					   buf + i);
4008 		}
4009 	}
4010 
4011 	return i;
4012 }
4013