xref: /linux/net/tipc/socket.c (revision 49545357bf7e134a4012d4652c2df5f78f4485af)
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 
41 #include "core.h"
42 #include "name_table.h"
43 #include "node.h"
44 #include "link.h"
45 #include "name_distr.h"
46 #include "socket.h"
47 #include "bcast.h"
48 #include "netlink.h"
49 #include "group.h"
50 #include "trace.h"
51 
52 #define NAGLE_START_INIT	4
53 #define NAGLE_START_MAX		1024
54 #define CONN_TIMEOUT_DEFAULT    8000    /* default connect timeout = 8s */
55 #define CONN_PROBING_INTV	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
56 #define TIPC_MAX_PORT		0xffffffff
57 #define TIPC_MIN_PORT		1
58 #define TIPC_ACK_RATE		4       /* ACK at 1/4 of rcv window size */
59 
60 enum {
61 	TIPC_LISTEN = TCP_LISTEN,
62 	TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 	TIPC_OPEN = TCP_CLOSE,
64 	TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 	TIPC_CONNECTING = TCP_SYN_SENT,
66 };
67 
68 struct sockaddr_pair {
69 	struct sockaddr_tipc sock;
70 	struct sockaddr_tipc member;
71 };
72 
73 /**
74  * struct tipc_sock - TIPC socket structure
75  * @sk: socket - interacts with 'port' and with user via the socket API
76  * @max_pkt: maximum packet size "hint" used when building messages sent by port
77  * @maxnagle: maximum size of msg which can be subject to nagle
78  * @portid: unique port identity in TIPC socket hash table
79  * @phdr: preformatted message header used when sending messages
80  * @cong_links: list of congested links
81  * @publications: list of publications for port
82  * @blocking_link: address of the congested link we are currently sleeping on
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 int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
150 		       bool kern);
151 static void tipc_sk_timeout(struct timer_list *t);
152 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
153 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
154 static int tipc_sk_leave(struct tipc_sock *tsk);
155 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
156 static int tipc_sk_insert(struct tipc_sock *tsk);
157 static void tipc_sk_remove(struct tipc_sock *tsk);
158 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
159 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
160 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
161 
162 static const struct proto_ops packet_ops;
163 static const struct proto_ops stream_ops;
164 static const struct proto_ops msg_ops;
165 static struct proto tipc_proto;
166 static const struct rhashtable_params tsk_rht_params;
167 
168 static u32 tsk_own_node(struct tipc_sock *tsk)
169 {
170 	return msg_prevnode(&tsk->phdr);
171 }
172 
173 static u32 tsk_peer_node(struct tipc_sock *tsk)
174 {
175 	return msg_destnode(&tsk->phdr);
176 }
177 
178 static u32 tsk_peer_port(struct tipc_sock *tsk)
179 {
180 	return msg_destport(&tsk->phdr);
181 }
182 
183 static  bool tsk_unreliable(struct tipc_sock *tsk)
184 {
185 	return msg_src_droppable(&tsk->phdr) != 0;
186 }
187 
188 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
189 {
190 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
191 }
192 
193 static bool tsk_unreturnable(struct tipc_sock *tsk)
194 {
195 	return msg_dest_droppable(&tsk->phdr) != 0;
196 }
197 
198 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
199 {
200 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
201 }
202 
203 static int tsk_importance(struct tipc_sock *tsk)
204 {
205 	return msg_importance(&tsk->phdr);
206 }
207 
208 static struct tipc_sock *tipc_sk(const struct sock *sk)
209 {
210 	return container_of(sk, struct tipc_sock, sk);
211 }
212 
213 int tsk_set_importance(struct sock *sk, int imp)
214 {
215 	if (imp > TIPC_CRITICAL_IMPORTANCE)
216 		return -EINVAL;
217 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
218 	return 0;
219 }
220 
221 static bool tsk_conn_cong(struct tipc_sock *tsk)
222 {
223 	return tsk->snt_unacked > tsk->snd_win;
224 }
225 
226 static u16 tsk_blocks(int len)
227 {
228 	return ((len / FLOWCTL_BLK_SZ) + 1);
229 }
230 
231 /* tsk_blocks(): translate a buffer size in bytes to number of
232  * advertisable blocks, taking into account the ratio truesize(len)/len
233  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
234  */
235 static u16 tsk_adv_blocks(int len)
236 {
237 	return len / FLOWCTL_BLK_SZ / 4;
238 }
239 
240 /* tsk_inc(): increment counter for sent or received data
241  * - If block based flow control is not supported by peer we
242  *   fall back to message based ditto, incrementing the counter
243  */
244 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
245 {
246 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
247 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
248 	return 1;
249 }
250 
251 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
252  */
253 static void tsk_set_nagle(struct tipc_sock *tsk)
254 {
255 	struct sock *sk = &tsk->sk;
256 
257 	tsk->maxnagle = 0;
258 	if (sk->sk_type != SOCK_STREAM)
259 		return;
260 	if (tsk->nodelay)
261 		return;
262 	if (!(tsk->peer_caps & TIPC_NAGLE))
263 		return;
264 	/* Limit node local buffer size to avoid receive queue overflow */
265 	if (tsk->max_pkt == MAX_MSG_SIZE)
266 		tsk->maxnagle = 1500;
267 	else
268 		tsk->maxnagle = tsk->max_pkt;
269 }
270 
271 /**
272  * tsk_advance_rx_queue - discard first buffer in socket receive queue
273  * @sk: network socket
274  *
275  * Caller must hold socket lock
276  */
277 static void tsk_advance_rx_queue(struct sock *sk)
278 {
279 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
280 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
281 }
282 
283 /* tipc_sk_respond() : send response message back to sender
284  */
285 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
286 {
287 	u32 selector;
288 	u32 dnode;
289 	u32 onode = tipc_own_addr(sock_net(sk));
290 
291 	if (!tipc_msg_reverse(onode, &skb, err))
292 		return;
293 
294 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
295 	dnode = msg_destnode(buf_msg(skb));
296 	selector = msg_origport(buf_msg(skb));
297 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
298 }
299 
300 /**
301  * tsk_rej_rx_queue - reject all buffers in socket receive queue
302  * @sk: network socket
303  * @error: response error code
304  *
305  * Caller must hold socket lock
306  */
307 static void tsk_rej_rx_queue(struct sock *sk, int error)
308 {
309 	struct sk_buff *skb;
310 
311 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
312 		tipc_sk_respond(sk, skb, error);
313 }
314 
315 static bool tipc_sk_connected(struct sock *sk)
316 {
317 	return sk->sk_state == TIPC_ESTABLISHED;
318 }
319 
320 /* tipc_sk_type_connectionless - check if the socket is datagram socket
321  * @sk: socket
322  *
323  * Returns true if connection less, false otherwise
324  */
325 static bool tipc_sk_type_connectionless(struct sock *sk)
326 {
327 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
328 }
329 
330 /* tsk_peer_msg - verify if message was sent by connected port's peer
331  *
332  * Handles cases where the node's network address has changed from
333  * the default of <0.0.0> to its configured setting.
334  */
335 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
336 {
337 	struct sock *sk = &tsk->sk;
338 	u32 self = tipc_own_addr(sock_net(sk));
339 	u32 peer_port = tsk_peer_port(tsk);
340 	u32 orig_node, peer_node;
341 
342 	if (unlikely(!tipc_sk_connected(sk)))
343 		return false;
344 
345 	if (unlikely(msg_origport(msg) != peer_port))
346 		return false;
347 
348 	orig_node = msg_orignode(msg);
349 	peer_node = tsk_peer_node(tsk);
350 
351 	if (likely(orig_node == peer_node))
352 		return true;
353 
354 	if (!orig_node && peer_node == self)
355 		return true;
356 
357 	if (!peer_node && orig_node == self)
358 		return true;
359 
360 	return false;
361 }
362 
363 /* tipc_set_sk_state - set the sk_state of the socket
364  * @sk: socket
365  *
366  * Caller must hold socket lock
367  *
368  * Returns 0 on success, errno otherwise
369  */
370 static int tipc_set_sk_state(struct sock *sk, int state)
371 {
372 	int oldsk_state = sk->sk_state;
373 	int res = -EINVAL;
374 
375 	switch (state) {
376 	case TIPC_OPEN:
377 		res = 0;
378 		break;
379 	case TIPC_LISTEN:
380 	case TIPC_CONNECTING:
381 		if (oldsk_state == TIPC_OPEN)
382 			res = 0;
383 		break;
384 	case TIPC_ESTABLISHED:
385 		if (oldsk_state == TIPC_CONNECTING ||
386 		    oldsk_state == TIPC_OPEN)
387 			res = 0;
388 		break;
389 	case TIPC_DISCONNECTING:
390 		if (oldsk_state == TIPC_CONNECTING ||
391 		    oldsk_state == TIPC_ESTABLISHED)
392 			res = 0;
393 		break;
394 	}
395 
396 	if (!res)
397 		sk->sk_state = state;
398 
399 	return res;
400 }
401 
402 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
403 {
404 	struct sock *sk = sock->sk;
405 	int err = sock_error(sk);
406 	int typ = sock->type;
407 
408 	if (err)
409 		return err;
410 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
411 		if (sk->sk_state == TIPC_DISCONNECTING)
412 			return -EPIPE;
413 		else if (!tipc_sk_connected(sk))
414 			return -ENOTCONN;
415 	}
416 	if (!*timeout)
417 		return -EAGAIN;
418 	if (signal_pending(current))
419 		return sock_intr_errno(*timeout);
420 
421 	return 0;
422 }
423 
424 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
425 ({                                                                             \
426 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
427 	struct sock *sk_;						       \
428 	int rc_;							       \
429 									       \
430 	while ((rc_ = !(condition_))) {					       \
431 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
432 		smp_rmb();                                                     \
433 		sk_ = (sock_)->sk;					       \
434 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
435 		if (rc_)						       \
436 			break;						       \
437 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
438 		release_sock(sk_);					       \
439 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
440 		sched_annotate_sleep();				               \
441 		lock_sock(sk_);						       \
442 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
443 	}								       \
444 	rc_;								       \
445 })
446 
447 /**
448  * tipc_sk_create - create a TIPC socket
449  * @net: network namespace (must be default network)
450  * @sock: pre-allocated socket structure
451  * @protocol: protocol indicator (must be 0)
452  * @kern: caused by kernel or by userspace?
453  *
454  * This routine creates additional data structures used by the TIPC socket,
455  * initializes them, and links them together.
456  *
457  * Return: 0 on success, errno otherwise
458  */
459 static int tipc_sk_create(struct net *net, struct socket *sock,
460 			  int protocol, int kern)
461 {
462 	const struct proto_ops *ops;
463 	struct sock *sk;
464 	struct tipc_sock *tsk;
465 	struct tipc_msg *msg;
466 
467 	/* Validate arguments */
468 	if (unlikely(protocol != 0))
469 		return -EPROTONOSUPPORT;
470 
471 	switch (sock->type) {
472 	case SOCK_STREAM:
473 		ops = &stream_ops;
474 		break;
475 	case SOCK_SEQPACKET:
476 		ops = &packet_ops;
477 		break;
478 	case SOCK_DGRAM:
479 	case SOCK_RDM:
480 		ops = &msg_ops;
481 		break;
482 	default:
483 		return -EPROTOTYPE;
484 	}
485 
486 	/* Allocate socket's protocol area */
487 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
488 	if (sk == NULL)
489 		return -ENOMEM;
490 
491 	tsk = tipc_sk(sk);
492 	tsk->max_pkt = MAX_PKT_DEFAULT;
493 	tsk->maxnagle = 0;
494 	tsk->nagle_start = NAGLE_START_INIT;
495 	INIT_LIST_HEAD(&tsk->publications);
496 	INIT_LIST_HEAD(&tsk->cong_links);
497 	msg = &tsk->phdr;
498 
499 	/* Finish initializing socket data structures */
500 	sock->ops = ops;
501 	sock_init_data(sock, sk);
502 	tipc_set_sk_state(sk, TIPC_OPEN);
503 	if (tipc_sk_insert(tsk)) {
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 = 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 	/* Determine destination */
1464 	if (atype == TIPC_SERVICE_RANGE) {
1465 		return tipc_sendmcast(sock, ua, m, dlen, timeout);
1466 	} else if (atype == TIPC_SERVICE_ADDR) {
1467 		skaddr.node = ua->lookup_node;
1468 		ua->scope = tipc_node2scope(skaddr.node);
1469 		if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1470 			return -EHOSTUNREACH;
1471 	} else if (atype == TIPC_SOCKET_ADDR) {
1472 		skaddr = ua->sk;
1473 	} else {
1474 		return -EINVAL;
1475 	}
1476 
1477 	/* Block or return if destination link is congested */
1478 	rc = tipc_wait_for_cond(sock, &timeout,
1479 				!tipc_dest_find(clinks, skaddr.node, 0));
1480 	if (unlikely(rc))
1481 		return rc;
1482 
1483 	/* Finally build message header */
1484 	msg_set_destnode(hdr, skaddr.node);
1485 	msg_set_destport(hdr, skaddr.ref);
1486 	if (atype == TIPC_SERVICE_ADDR) {
1487 		msg_set_type(hdr, TIPC_NAMED_MSG);
1488 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1489 		msg_set_nametype(hdr, ua->sa.type);
1490 		msg_set_nameinst(hdr, ua->sa.instance);
1491 		msg_set_lookup_scope(hdr, ua->scope);
1492 	} else { /* TIPC_SOCKET_ADDR */
1493 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1494 		msg_set_lookup_scope(hdr, 0);
1495 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1496 	}
1497 
1498 	/* Add message body */
1499 	__skb_queue_head_init(&pkts);
1500 	mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1501 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1502 	if (unlikely(rc != dlen))
1503 		return rc;
1504 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1505 		__skb_queue_purge(&pkts);
1506 		return -ENOMEM;
1507 	}
1508 
1509 	/* Send message */
1510 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1511 	rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1512 	if (unlikely(rc == -ELINKCONG)) {
1513 		tipc_dest_push(clinks, skaddr.node, 0);
1514 		tsk->cong_link_cnt++;
1515 		rc = 0;
1516 	}
1517 
1518 	if (unlikely(syn && !rc))
1519 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1520 
1521 	return rc ? rc : dlen;
1522 }
1523 
1524 /**
1525  * tipc_sendstream - send stream-oriented data
1526  * @sock: socket structure
1527  * @m: data to send
1528  * @dsz: total length of data to be transmitted
1529  *
1530  * Used for SOCK_STREAM data.
1531  *
1532  * Return: the number of bytes sent on success (or partial success),
1533  * or errno if no data sent
1534  */
1535 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1536 {
1537 	struct sock *sk = sock->sk;
1538 	int ret;
1539 
1540 	lock_sock(sk);
1541 	ret = __tipc_sendstream(sock, m, dsz);
1542 	release_sock(sk);
1543 
1544 	return ret;
1545 }
1546 
1547 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1548 {
1549 	struct sock *sk = sock->sk;
1550 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1551 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1552 	struct sk_buff_head *txq = &sk->sk_write_queue;
1553 	struct tipc_sock *tsk = tipc_sk(sk);
1554 	struct tipc_msg *hdr = &tsk->phdr;
1555 	struct net *net = sock_net(sk);
1556 	struct sk_buff *skb;
1557 	u32 dnode = tsk_peer_node(tsk);
1558 	int maxnagle = tsk->maxnagle;
1559 	int maxpkt = tsk->max_pkt;
1560 	int send, sent = 0;
1561 	int blocks, rc = 0;
1562 
1563 	if (unlikely(dlen > INT_MAX))
1564 		return -EMSGSIZE;
1565 
1566 	/* Handle implicit connection setup */
1567 	if (unlikely(dest)) {
1568 		rc = __tipc_sendmsg(sock, m, dlen);
1569 		if (dlen && dlen == rc) {
1570 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1571 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1572 		}
1573 		return rc;
1574 	}
1575 
1576 	do {
1577 		rc = tipc_wait_for_cond(sock, &timeout,
1578 					(!tsk->cong_link_cnt &&
1579 					 !tsk_conn_cong(tsk) &&
1580 					 tipc_sk_connected(sk)));
1581 		if (unlikely(rc))
1582 			break;
1583 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1584 		blocks = tsk->snd_backlog;
1585 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1586 		    send <= maxnagle) {
1587 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1588 			if (unlikely(rc < 0))
1589 				break;
1590 			blocks += rc;
1591 			tsk->msg_acc++;
1592 			if (blocks <= 64 && tsk->expect_ack) {
1593 				tsk->snd_backlog = blocks;
1594 				sent += send;
1595 				break;
1596 			} else if (blocks > 64) {
1597 				tsk->pkt_cnt += skb_queue_len(txq);
1598 			} else {
1599 				skb = skb_peek_tail(txq);
1600 				if (skb) {
1601 					msg_set_ack_required(buf_msg(skb));
1602 					tsk->expect_ack = true;
1603 				} else {
1604 					tsk->expect_ack = false;
1605 				}
1606 				tsk->msg_acc = 0;
1607 				tsk->pkt_cnt = 0;
1608 			}
1609 		} else {
1610 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1611 			if (unlikely(rc != send))
1612 				break;
1613 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1614 		}
1615 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1616 					 TIPC_DUMP_SK_SNDQ, " ");
1617 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1618 		if (unlikely(rc == -ELINKCONG)) {
1619 			tsk->cong_link_cnt = 1;
1620 			rc = 0;
1621 		}
1622 		if (likely(!rc)) {
1623 			tsk->snt_unacked += blocks;
1624 			tsk->snd_backlog = 0;
1625 			sent += send;
1626 		}
1627 	} while (sent < dlen && !rc);
1628 
1629 	return sent ? sent : rc;
1630 }
1631 
1632 /**
1633  * tipc_send_packet - send a connection-oriented message
1634  * @sock: socket structure
1635  * @m: message to send
1636  * @dsz: length of data to be transmitted
1637  *
1638  * Used for SOCK_SEQPACKET messages.
1639  *
1640  * Return: the number of bytes sent on success, or errno otherwise
1641  */
1642 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1643 {
1644 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1645 		return -EMSGSIZE;
1646 
1647 	return tipc_sendstream(sock, m, dsz);
1648 }
1649 
1650 /* tipc_sk_finish_conn - complete the setup of a connection
1651  */
1652 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1653 				u32 peer_node)
1654 {
1655 	struct sock *sk = &tsk->sk;
1656 	struct net *net = sock_net(sk);
1657 	struct tipc_msg *msg = &tsk->phdr;
1658 
1659 	msg_set_syn(msg, 0);
1660 	msg_set_destnode(msg, peer_node);
1661 	msg_set_destport(msg, peer_port);
1662 	msg_set_type(msg, TIPC_CONN_MSG);
1663 	msg_set_lookup_scope(msg, 0);
1664 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1665 
1666 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1667 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1668 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1669 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1670 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1671 	tsk_set_nagle(tsk);
1672 	__skb_queue_purge(&sk->sk_write_queue);
1673 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1674 		return;
1675 
1676 	/* Fall back to message based flow control */
1677 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1678 	tsk->snd_win = FLOWCTL_MSG_WIN;
1679 }
1680 
1681 /**
1682  * tipc_sk_set_orig_addr - capture sender's address for received message
1683  * @m: descriptor for message info
1684  * @skb: received message
1685  *
1686  * Note: Address is not captured if not requested by receiver.
1687  */
1688 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1689 {
1690 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1691 	struct tipc_msg *hdr = buf_msg(skb);
1692 
1693 	if (!srcaddr)
1694 		return;
1695 
1696 	srcaddr->sock.family = AF_TIPC;
1697 	srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1698 	srcaddr->sock.scope = 0;
1699 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1700 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1701 	srcaddr->sock.addr.name.domain = 0;
1702 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1703 
1704 	if (!msg_in_group(hdr))
1705 		return;
1706 
1707 	/* Group message users may also want to know sending member's id */
1708 	srcaddr->member.family = AF_TIPC;
1709 	srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1710 	srcaddr->member.scope = 0;
1711 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1712 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1713 	srcaddr->member.addr.name.domain = 0;
1714 	m->msg_namelen = sizeof(*srcaddr);
1715 }
1716 
1717 /**
1718  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1719  * @m: descriptor for message info
1720  * @skb: received message buffer
1721  * @tsk: TIPC port associated with message
1722  *
1723  * Note: Ancillary data is not captured if not requested by receiver.
1724  *
1725  * Return: 0 if successful, otherwise errno
1726  */
1727 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1728 				 struct tipc_sock *tsk)
1729 {
1730 	struct tipc_msg *hdr;
1731 	u32 data[3] = {0,};
1732 	bool has_addr;
1733 	int dlen, rc;
1734 
1735 	if (likely(m->msg_controllen == 0))
1736 		return 0;
1737 
1738 	hdr = buf_msg(skb);
1739 	dlen = msg_data_sz(hdr);
1740 
1741 	/* Capture errored message object, if any */
1742 	if (msg_errcode(hdr)) {
1743 		if (skb_linearize(skb))
1744 			return -ENOMEM;
1745 		hdr = buf_msg(skb);
1746 		data[0] = msg_errcode(hdr);
1747 		data[1] = dlen;
1748 		rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1749 		if (rc || !dlen)
1750 			return rc;
1751 		rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1752 		if (rc)
1753 			return rc;
1754 	}
1755 
1756 	/* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1757 	switch (msg_type(hdr)) {
1758 	case TIPC_NAMED_MSG:
1759 		has_addr = true;
1760 		data[0] = msg_nametype(hdr);
1761 		data[1] = msg_namelower(hdr);
1762 		data[2] = data[1];
1763 		break;
1764 	case TIPC_MCAST_MSG:
1765 		has_addr = true;
1766 		data[0] = msg_nametype(hdr);
1767 		data[1] = msg_namelower(hdr);
1768 		data[2] = msg_nameupper(hdr);
1769 		break;
1770 	case TIPC_CONN_MSG:
1771 		has_addr = !!tsk->conn_addrtype;
1772 		data[0] = msg_nametype(&tsk->phdr);
1773 		data[1] = msg_nameinst(&tsk->phdr);
1774 		data[2] = data[1];
1775 		break;
1776 	default:
1777 		has_addr = false;
1778 	}
1779 	if (!has_addr)
1780 		return 0;
1781 	return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1782 }
1783 
1784 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1785 {
1786 	struct sock *sk = &tsk->sk;
1787 	struct sk_buff *skb = NULL;
1788 	struct tipc_msg *msg;
1789 	u32 peer_port = tsk_peer_port(tsk);
1790 	u32 dnode = tsk_peer_node(tsk);
1791 
1792 	if (!tipc_sk_connected(sk))
1793 		return NULL;
1794 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1795 			      dnode, tsk_own_node(tsk), peer_port,
1796 			      tsk->portid, TIPC_OK);
1797 	if (!skb)
1798 		return NULL;
1799 	msg = buf_msg(skb);
1800 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1801 	tsk->rcv_unacked = 0;
1802 
1803 	/* Adjust to and advertize the correct window limit */
1804 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1805 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1806 		msg_set_adv_win(msg, tsk->rcv_win);
1807 	}
1808 	return skb;
1809 }
1810 
1811 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1812 {
1813 	struct sk_buff *skb;
1814 
1815 	skb = tipc_sk_build_ack(tsk);
1816 	if (!skb)
1817 		return;
1818 
1819 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1820 			   msg_link_selector(buf_msg(skb)));
1821 }
1822 
1823 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1824 {
1825 	struct sock *sk = sock->sk;
1826 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1827 	long timeo = *timeop;
1828 	int err = sock_error(sk);
1829 
1830 	if (err)
1831 		return err;
1832 
1833 	for (;;) {
1834 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1835 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1836 				err = -ENOTCONN;
1837 				break;
1838 			}
1839 			add_wait_queue(sk_sleep(sk), &wait);
1840 			release_sock(sk);
1841 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1842 			sched_annotate_sleep();
1843 			lock_sock(sk);
1844 			remove_wait_queue(sk_sleep(sk), &wait);
1845 		}
1846 		err = 0;
1847 		if (!skb_queue_empty(&sk->sk_receive_queue))
1848 			break;
1849 		err = -EAGAIN;
1850 		if (!timeo)
1851 			break;
1852 		err = sock_intr_errno(timeo);
1853 		if (signal_pending(current))
1854 			break;
1855 
1856 		err = sock_error(sk);
1857 		if (err)
1858 			break;
1859 	}
1860 	*timeop = timeo;
1861 	return err;
1862 }
1863 
1864 /**
1865  * tipc_recvmsg - receive packet-oriented message
1866  * @sock: network socket
1867  * @m: descriptor for message info
1868  * @buflen: length of user buffer area
1869  * @flags: receive flags
1870  *
1871  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1872  * If the complete message doesn't fit in user area, truncate it.
1873  *
1874  * Return: size of returned message data, errno otherwise
1875  */
1876 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1877 			size_t buflen,	int flags)
1878 {
1879 	struct sock *sk = sock->sk;
1880 	bool connected = !tipc_sk_type_connectionless(sk);
1881 	struct tipc_sock *tsk = tipc_sk(sk);
1882 	int rc, err, hlen, dlen, copy;
1883 	struct tipc_skb_cb *skb_cb;
1884 	struct sk_buff_head xmitq;
1885 	struct tipc_msg *hdr;
1886 	struct sk_buff *skb;
1887 	bool grp_evt;
1888 	long timeout;
1889 
1890 	/* Catch invalid receive requests */
1891 	if (unlikely(!buflen))
1892 		return -EINVAL;
1893 
1894 	lock_sock(sk);
1895 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1896 		rc = -ENOTCONN;
1897 		goto exit;
1898 	}
1899 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1900 
1901 	/* Step rcv queue to first msg with data or error; wait if necessary */
1902 	do {
1903 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1904 		if (unlikely(rc))
1905 			goto exit;
1906 		skb = skb_peek(&sk->sk_receive_queue);
1907 		skb_cb = TIPC_SKB_CB(skb);
1908 		hdr = buf_msg(skb);
1909 		dlen = msg_data_sz(hdr);
1910 		hlen = msg_hdr_sz(hdr);
1911 		err = msg_errcode(hdr);
1912 		grp_evt = msg_is_grp_evt(hdr);
1913 		if (likely(dlen || err))
1914 			break;
1915 		tsk_advance_rx_queue(sk);
1916 	} while (1);
1917 
1918 	/* Collect msg meta data, including error code and rejected data */
1919 	tipc_sk_set_orig_addr(m, skb);
1920 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1921 	if (unlikely(rc))
1922 		goto exit;
1923 	hdr = buf_msg(skb);
1924 
1925 	/* Capture data if non-error msg, otherwise just set return value */
1926 	if (likely(!err)) {
1927 		int offset = skb_cb->bytes_read;
1928 
1929 		copy = min_t(int, dlen - offset, buflen);
1930 		rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1931 		if (unlikely(rc))
1932 			goto exit;
1933 		if (unlikely(offset + copy < dlen)) {
1934 			if (flags & MSG_EOR) {
1935 				if (!(flags & MSG_PEEK))
1936 					skb_cb->bytes_read = offset + copy;
1937 			} else {
1938 				m->msg_flags |= MSG_TRUNC;
1939 				skb_cb->bytes_read = 0;
1940 			}
1941 		} else {
1942 			if (flags & MSG_EOR)
1943 				m->msg_flags |= MSG_EOR;
1944 			skb_cb->bytes_read = 0;
1945 		}
1946 	} else {
1947 		copy = 0;
1948 		rc = 0;
1949 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1950 			rc = -ECONNRESET;
1951 			goto exit;
1952 		}
1953 	}
1954 
1955 	/* Mark message as group event if applicable */
1956 	if (unlikely(grp_evt)) {
1957 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1958 			m->msg_flags |= MSG_EOR;
1959 		m->msg_flags |= MSG_OOB;
1960 		copy = 0;
1961 	}
1962 
1963 	/* Caption of data or error code/rejected data was successful */
1964 	if (unlikely(flags & MSG_PEEK))
1965 		goto exit;
1966 
1967 	/* Send group flow control advertisement when applicable */
1968 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1969 		__skb_queue_head_init(&xmitq);
1970 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1971 					  msg_orignode(hdr), msg_origport(hdr),
1972 					  &xmitq);
1973 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1974 	}
1975 
1976 	if (!skb_cb->bytes_read)
1977 		tsk_advance_rx_queue(sk);
1978 
1979 	if (likely(!connected) || skb_cb->bytes_read)
1980 		goto exit;
1981 
1982 	/* Send connection flow control advertisement when applicable */
1983 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1984 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1985 		tipc_sk_send_ack(tsk);
1986 exit:
1987 	release_sock(sk);
1988 	return rc ? rc : copy;
1989 }
1990 
1991 /**
1992  * tipc_recvstream - receive stream-oriented data
1993  * @sock: network socket
1994  * @m: descriptor for message info
1995  * @buflen: total size of user buffer area
1996  * @flags: receive flags
1997  *
1998  * Used for SOCK_STREAM messages only.  If not enough data is available
1999  * will optionally wait for more; never truncates data.
2000  *
2001  * Return: size of returned message data, errno otherwise
2002  */
2003 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2004 			   size_t buflen, int flags)
2005 {
2006 	struct sock *sk = sock->sk;
2007 	struct tipc_sock *tsk = tipc_sk(sk);
2008 	struct sk_buff *skb;
2009 	struct tipc_msg *hdr;
2010 	struct tipc_skb_cb *skb_cb;
2011 	bool peek = flags & MSG_PEEK;
2012 	int offset, required, copy, copied = 0;
2013 	int hlen, dlen, err, rc;
2014 	long timeout;
2015 
2016 	/* Catch invalid receive attempts */
2017 	if (unlikely(!buflen))
2018 		return -EINVAL;
2019 
2020 	lock_sock(sk);
2021 
2022 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2023 		rc = -ENOTCONN;
2024 		goto exit;
2025 	}
2026 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2027 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2028 
2029 	do {
2030 		/* Look at first msg in receive queue; wait if necessary */
2031 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2032 		if (unlikely(rc))
2033 			break;
2034 		skb = skb_peek(&sk->sk_receive_queue);
2035 		skb_cb = TIPC_SKB_CB(skb);
2036 		hdr = buf_msg(skb);
2037 		dlen = msg_data_sz(hdr);
2038 		hlen = msg_hdr_sz(hdr);
2039 		err = msg_errcode(hdr);
2040 
2041 		/* Discard any empty non-errored (SYN-) message */
2042 		if (unlikely(!dlen && !err)) {
2043 			tsk_advance_rx_queue(sk);
2044 			continue;
2045 		}
2046 
2047 		/* Collect msg meta data, incl. error code and rejected data */
2048 		if (!copied) {
2049 			tipc_sk_set_orig_addr(m, skb);
2050 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2051 			if (rc)
2052 				break;
2053 			hdr = buf_msg(skb);
2054 		}
2055 
2056 		/* Copy data if msg ok, otherwise return error/partial data */
2057 		if (likely(!err)) {
2058 			offset = skb_cb->bytes_read;
2059 			copy = min_t(int, dlen - offset, buflen - copied);
2060 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2061 			if (unlikely(rc))
2062 				break;
2063 			copied += copy;
2064 			offset += copy;
2065 			if (unlikely(offset < dlen)) {
2066 				if (!peek)
2067 					skb_cb->bytes_read = offset;
2068 				break;
2069 			}
2070 		} else {
2071 			rc = 0;
2072 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2073 				rc = -ECONNRESET;
2074 			if (copied || rc)
2075 				break;
2076 		}
2077 
2078 		if (unlikely(peek))
2079 			break;
2080 
2081 		tsk_advance_rx_queue(sk);
2082 
2083 		/* Send connection flow control advertisement when applicable */
2084 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2085 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2086 			tipc_sk_send_ack(tsk);
2087 
2088 		/* Exit if all requested data or FIN/error received */
2089 		if (copied == buflen || err)
2090 			break;
2091 
2092 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2093 exit:
2094 	release_sock(sk);
2095 	return copied ? copied : rc;
2096 }
2097 
2098 /**
2099  * tipc_write_space - wake up thread if port congestion is released
2100  * @sk: socket
2101  */
2102 static void tipc_write_space(struct sock *sk)
2103 {
2104 	struct socket_wq *wq;
2105 
2106 	rcu_read_lock();
2107 	wq = rcu_dereference(sk->sk_wq);
2108 	if (skwq_has_sleeper(wq))
2109 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2110 						EPOLLWRNORM | EPOLLWRBAND);
2111 	rcu_read_unlock();
2112 }
2113 
2114 /**
2115  * tipc_data_ready - wake up threads to indicate messages have been received
2116  * @sk: socket
2117  */
2118 static void tipc_data_ready(struct sock *sk)
2119 {
2120 	struct socket_wq *wq;
2121 
2122 	rcu_read_lock();
2123 	wq = rcu_dereference(sk->sk_wq);
2124 	if (skwq_has_sleeper(wq))
2125 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2126 						EPOLLRDNORM | EPOLLRDBAND);
2127 	rcu_read_unlock();
2128 }
2129 
2130 static void tipc_sock_destruct(struct sock *sk)
2131 {
2132 	__skb_queue_purge(&sk->sk_receive_queue);
2133 }
2134 
2135 static void tipc_sk_proto_rcv(struct sock *sk,
2136 			      struct sk_buff_head *inputq,
2137 			      struct sk_buff_head *xmitq)
2138 {
2139 	struct sk_buff *skb = __skb_dequeue(inputq);
2140 	struct tipc_sock *tsk = tipc_sk(sk);
2141 	struct tipc_msg *hdr = buf_msg(skb);
2142 	struct tipc_group *grp = tsk->group;
2143 	bool wakeup = false;
2144 
2145 	switch (msg_user(hdr)) {
2146 	case CONN_MANAGER:
2147 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2148 		return;
2149 	case SOCK_WAKEUP:
2150 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2151 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2152 		smp_wmb();
2153 		tsk->cong_link_cnt--;
2154 		wakeup = true;
2155 		tipc_sk_push_backlog(tsk, false);
2156 		break;
2157 	case GROUP_PROTOCOL:
2158 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2159 		break;
2160 	case TOP_SRV:
2161 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2162 				      hdr, inputq, xmitq);
2163 		break;
2164 	default:
2165 		break;
2166 	}
2167 
2168 	if (wakeup)
2169 		sk->sk_write_space(sk);
2170 
2171 	kfree_skb(skb);
2172 }
2173 
2174 /**
2175  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2176  * @tsk: TIPC socket
2177  * @skb: pointer to message buffer.
2178  * @xmitq: for Nagle ACK if any
2179  * Return: true if message should be added to receive queue, false otherwise
2180  */
2181 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2182 				   struct sk_buff_head *xmitq)
2183 {
2184 	struct sock *sk = &tsk->sk;
2185 	struct net *net = sock_net(sk);
2186 	struct tipc_msg *hdr = buf_msg(skb);
2187 	bool con_msg = msg_connected(hdr);
2188 	u32 pport = tsk_peer_port(tsk);
2189 	u32 pnode = tsk_peer_node(tsk);
2190 	u32 oport = msg_origport(hdr);
2191 	u32 onode = msg_orignode(hdr);
2192 	int err = msg_errcode(hdr);
2193 	unsigned long delay;
2194 
2195 	if (unlikely(msg_mcast(hdr)))
2196 		return false;
2197 	tsk->oneway = 0;
2198 
2199 	switch (sk->sk_state) {
2200 	case TIPC_CONNECTING:
2201 		/* Setup ACK */
2202 		if (likely(con_msg)) {
2203 			if (err)
2204 				break;
2205 			tipc_sk_finish_conn(tsk, oport, onode);
2206 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2207 			/* ACK+ message with data is added to receive queue */
2208 			if (msg_data_sz(hdr))
2209 				return true;
2210 			/* Empty ACK-, - wake up sleeping connect() and drop */
2211 			sk->sk_state_change(sk);
2212 			msg_set_dest_droppable(hdr, 1);
2213 			return false;
2214 		}
2215 		/* Ignore connectionless message if not from listening socket */
2216 		if (oport != pport || onode != pnode)
2217 			return false;
2218 
2219 		/* Rejected SYN */
2220 		if (err != TIPC_ERR_OVERLOAD)
2221 			break;
2222 
2223 		/* Prepare for new setup attempt if we have a SYN clone */
2224 		if (skb_queue_empty(&sk->sk_write_queue))
2225 			break;
2226 		get_random_bytes(&delay, 2);
2227 		delay %= (tsk->conn_timeout / 4);
2228 		delay = msecs_to_jiffies(delay + 100);
2229 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2230 		return false;
2231 	case TIPC_OPEN:
2232 	case TIPC_DISCONNECTING:
2233 		return false;
2234 	case TIPC_LISTEN:
2235 		/* Accept only SYN message */
2236 		if (!msg_is_syn(hdr) &&
2237 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2238 			return false;
2239 		if (!con_msg && !err)
2240 			return true;
2241 		return false;
2242 	case TIPC_ESTABLISHED:
2243 		if (!skb_queue_empty(&sk->sk_write_queue))
2244 			tipc_sk_push_backlog(tsk, false);
2245 		/* Accept only connection-based messages sent by peer */
2246 		if (likely(con_msg && !err && pport == oport &&
2247 			   pnode == onode)) {
2248 			if (msg_ack_required(hdr)) {
2249 				struct sk_buff *skb;
2250 
2251 				skb = tipc_sk_build_ack(tsk);
2252 				if (skb) {
2253 					msg_set_nagle_ack(buf_msg(skb));
2254 					__skb_queue_tail(xmitq, skb);
2255 				}
2256 			}
2257 			return true;
2258 		}
2259 		if (!tsk_peer_msg(tsk, hdr))
2260 			return false;
2261 		if (!err)
2262 			return true;
2263 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2264 		tipc_node_remove_conn(net, pnode, tsk->portid);
2265 		sk->sk_state_change(sk);
2266 		return true;
2267 	default:
2268 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2269 	}
2270 	/* Abort connection setup attempt */
2271 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2272 	sk->sk_err = ECONNREFUSED;
2273 	sk->sk_state_change(sk);
2274 	return true;
2275 }
2276 
2277 /**
2278  * rcvbuf_limit - get proper overload limit of socket receive queue
2279  * @sk: socket
2280  * @skb: message
2281  *
2282  * For connection oriented messages, irrespective of importance,
2283  * default queue limit is 2 MB.
2284  *
2285  * For connectionless messages, queue limits are based on message
2286  * importance as follows:
2287  *
2288  * TIPC_LOW_IMPORTANCE       (2 MB)
2289  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2290  * TIPC_HIGH_IMPORTANCE      (8 MB)
2291  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2292  *
2293  * Return: overload limit according to corresponding message importance
2294  */
2295 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2296 {
2297 	struct tipc_sock *tsk = tipc_sk(sk);
2298 	struct tipc_msg *hdr = buf_msg(skb);
2299 
2300 	if (unlikely(msg_in_group(hdr)))
2301 		return READ_ONCE(sk->sk_rcvbuf);
2302 
2303 	if (unlikely(!msg_connected(hdr)))
2304 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2305 
2306 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2307 		return READ_ONCE(sk->sk_rcvbuf);
2308 
2309 	return FLOWCTL_MSG_LIM;
2310 }
2311 
2312 /**
2313  * tipc_sk_filter_rcv - validate incoming message
2314  * @sk: socket
2315  * @skb: pointer to message.
2316  * @xmitq: output message area (FIXME)
2317  *
2318  * Enqueues message on receive queue if acceptable; optionally handles
2319  * disconnect indication for a connected socket.
2320  *
2321  * Called with socket lock already taken
2322  */
2323 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2324 			       struct sk_buff_head *xmitq)
2325 {
2326 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2327 	struct tipc_sock *tsk = tipc_sk(sk);
2328 	struct tipc_group *grp = tsk->group;
2329 	struct tipc_msg *hdr = buf_msg(skb);
2330 	struct net *net = sock_net(sk);
2331 	struct sk_buff_head inputq;
2332 	int mtyp = msg_type(hdr);
2333 	int limit, err = TIPC_OK;
2334 
2335 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2336 	TIPC_SKB_CB(skb)->bytes_read = 0;
2337 	__skb_queue_head_init(&inputq);
2338 	__skb_queue_tail(&inputq, skb);
2339 
2340 	if (unlikely(!msg_isdata(hdr)))
2341 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2342 
2343 	if (unlikely(grp))
2344 		tipc_group_filter_msg(grp, &inputq, xmitq);
2345 
2346 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2347 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2348 
2349 	/* Validate and add to receive buffer if there is space */
2350 	while ((skb = __skb_dequeue(&inputq))) {
2351 		hdr = buf_msg(skb);
2352 		limit = rcvbuf_limit(sk, skb);
2353 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2354 		    (!sk_conn && msg_connected(hdr)) ||
2355 		    (!grp && msg_in_group(hdr)))
2356 			err = TIPC_ERR_NO_PORT;
2357 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2358 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2359 					   "err_overload2!");
2360 			atomic_inc(&sk->sk_drops);
2361 			err = TIPC_ERR_OVERLOAD;
2362 		}
2363 
2364 		if (unlikely(err)) {
2365 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2366 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2367 						      "@filter_rcv!");
2368 				__skb_queue_tail(xmitq, skb);
2369 			}
2370 			err = TIPC_OK;
2371 			continue;
2372 		}
2373 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2374 		skb_set_owner_r(skb, sk);
2375 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2376 					 "rcvq >90% allocated!");
2377 		sk->sk_data_ready(sk);
2378 	}
2379 }
2380 
2381 /**
2382  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2383  * @sk: socket
2384  * @skb: message
2385  *
2386  * Caller must hold socket lock
2387  */
2388 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2389 {
2390 	unsigned int before = sk_rmem_alloc_get(sk);
2391 	struct sk_buff_head xmitq;
2392 	unsigned int added;
2393 
2394 	__skb_queue_head_init(&xmitq);
2395 
2396 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2397 	added = sk_rmem_alloc_get(sk) - before;
2398 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2399 
2400 	/* Send pending response/rejected messages, if any */
2401 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2402 	return 0;
2403 }
2404 
2405 /**
2406  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2407  *                   inputq and try adding them to socket or backlog queue
2408  * @inputq: list of incoming buffers with potentially different destinations
2409  * @sk: socket where the buffers should be enqueued
2410  * @dport: port number for the socket
2411  * @xmitq: output queue
2412  *
2413  * Caller must hold socket lock
2414  */
2415 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2416 			    u32 dport, struct sk_buff_head *xmitq)
2417 {
2418 	unsigned long time_limit = jiffies + 2;
2419 	struct sk_buff *skb;
2420 	unsigned int lim;
2421 	atomic_t *dcnt;
2422 	u32 onode;
2423 
2424 	while (skb_queue_len(inputq)) {
2425 		if (unlikely(time_after_eq(jiffies, time_limit)))
2426 			return;
2427 
2428 		skb = tipc_skb_dequeue(inputq, dport);
2429 		if (unlikely(!skb))
2430 			return;
2431 
2432 		/* Add message directly to receive queue if possible */
2433 		if (!sock_owned_by_user(sk)) {
2434 			tipc_sk_filter_rcv(sk, skb, xmitq);
2435 			continue;
2436 		}
2437 
2438 		/* Try backlog, compensating for double-counted bytes */
2439 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2440 		if (!sk->sk_backlog.len)
2441 			atomic_set(dcnt, 0);
2442 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2443 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2444 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2445 						 "bklg & rcvq >90% allocated!");
2446 			continue;
2447 		}
2448 
2449 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2450 		/* Overload => reject message back to sender */
2451 		onode = tipc_own_addr(sock_net(sk));
2452 		atomic_inc(&sk->sk_drops);
2453 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2454 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2455 					      "@sk_enqueue!");
2456 			__skb_queue_tail(xmitq, skb);
2457 		}
2458 		break;
2459 	}
2460 }
2461 
2462 /**
2463  * tipc_sk_rcv - handle a chain of incoming buffers
2464  * @net: the associated network namespace
2465  * @inputq: buffer list containing the buffers
2466  * Consumes all buffers in list until inputq is empty
2467  * Note: may be called in multiple threads referring to the same queue
2468  */
2469 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2470 {
2471 	struct sk_buff_head xmitq;
2472 	u32 dnode, dport = 0;
2473 	int err;
2474 	struct tipc_sock *tsk;
2475 	struct sock *sk;
2476 	struct sk_buff *skb;
2477 
2478 	__skb_queue_head_init(&xmitq);
2479 	while (skb_queue_len(inputq)) {
2480 		dport = tipc_skb_peek_port(inputq, dport);
2481 		tsk = tipc_sk_lookup(net, dport);
2482 
2483 		if (likely(tsk)) {
2484 			sk = &tsk->sk;
2485 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2486 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2487 				spin_unlock_bh(&sk->sk_lock.slock);
2488 			}
2489 			/* Send pending response/rejected messages, if any */
2490 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2491 			sock_put(sk);
2492 			continue;
2493 		}
2494 		/* No destination socket => dequeue skb if still there */
2495 		skb = tipc_skb_dequeue(inputq, dport);
2496 		if (!skb)
2497 			return;
2498 
2499 		/* Try secondary lookup if unresolved named message */
2500 		err = TIPC_ERR_NO_PORT;
2501 		if (tipc_msg_lookup_dest(net, skb, &err))
2502 			goto xmit;
2503 
2504 		/* Prepare for message rejection */
2505 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2506 			continue;
2507 
2508 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2509 xmit:
2510 		dnode = msg_destnode(buf_msg(skb));
2511 		tipc_node_xmit_skb(net, skb, dnode, dport);
2512 	}
2513 }
2514 
2515 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2516 {
2517 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2518 	struct sock *sk = sock->sk;
2519 	int done;
2520 
2521 	do {
2522 		int err = sock_error(sk);
2523 		if (err)
2524 			return err;
2525 		if (!*timeo_p)
2526 			return -ETIMEDOUT;
2527 		if (signal_pending(current))
2528 			return sock_intr_errno(*timeo_p);
2529 		if (sk->sk_state == TIPC_DISCONNECTING)
2530 			break;
2531 
2532 		add_wait_queue(sk_sleep(sk), &wait);
2533 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2534 				     &wait);
2535 		remove_wait_queue(sk_sleep(sk), &wait);
2536 	} while (!done);
2537 	return 0;
2538 }
2539 
2540 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2541 {
2542 	if (addr->family != AF_TIPC)
2543 		return false;
2544 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2545 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2546 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2547 		addr->addrtype == TIPC_SOCKET_ADDR);
2548 }
2549 
2550 /**
2551  * tipc_connect - establish a connection to another TIPC port
2552  * @sock: socket structure
2553  * @dest: socket address for destination port
2554  * @destlen: size of socket address data structure
2555  * @flags: file-related flags associated with socket
2556  *
2557  * Return: 0 on success, errno otherwise
2558  */
2559 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2560 			int destlen, int flags)
2561 {
2562 	struct sock *sk = sock->sk;
2563 	struct tipc_sock *tsk = tipc_sk(sk);
2564 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2565 	struct msghdr m = {NULL,};
2566 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2567 	int previous;
2568 	int res = 0;
2569 
2570 	if (destlen != sizeof(struct sockaddr_tipc))
2571 		return -EINVAL;
2572 
2573 	lock_sock(sk);
2574 
2575 	if (tsk->group) {
2576 		res = -EINVAL;
2577 		goto exit;
2578 	}
2579 
2580 	if (dst->family == AF_UNSPEC) {
2581 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2582 		if (!tipc_sk_type_connectionless(sk))
2583 			res = -EINVAL;
2584 		goto exit;
2585 	}
2586 	if (!tipc_sockaddr_is_sane(dst)) {
2587 		res = -EINVAL;
2588 		goto exit;
2589 	}
2590 	/* DGRAM/RDM connect(), just save the destaddr */
2591 	if (tipc_sk_type_connectionless(sk)) {
2592 		memcpy(&tsk->peer, dest, destlen);
2593 		goto exit;
2594 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2595 		res = -EINVAL;
2596 		goto exit;
2597 	}
2598 
2599 	previous = sk->sk_state;
2600 
2601 	switch (sk->sk_state) {
2602 	case TIPC_OPEN:
2603 		/* Send a 'SYN-' to destination */
2604 		m.msg_name = dest;
2605 		m.msg_namelen = destlen;
2606 
2607 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2608 		 * indicate send_msg() is never blocked.
2609 		 */
2610 		if (!timeout)
2611 			m.msg_flags = MSG_DONTWAIT;
2612 
2613 		res = __tipc_sendmsg(sock, &m, 0);
2614 		if ((res < 0) && (res != -EWOULDBLOCK))
2615 			goto exit;
2616 
2617 		/* Just entered TIPC_CONNECTING state; the only
2618 		 * difference is that return value in non-blocking
2619 		 * case is EINPROGRESS, rather than EALREADY.
2620 		 */
2621 		res = -EINPROGRESS;
2622 		fallthrough;
2623 	case TIPC_CONNECTING:
2624 		if (!timeout) {
2625 			if (previous == TIPC_CONNECTING)
2626 				res = -EALREADY;
2627 			goto exit;
2628 		}
2629 		timeout = msecs_to_jiffies(timeout);
2630 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2631 		res = tipc_wait_for_connect(sock, &timeout);
2632 		break;
2633 	case TIPC_ESTABLISHED:
2634 		res = -EISCONN;
2635 		break;
2636 	default:
2637 		res = -EINVAL;
2638 	}
2639 
2640 exit:
2641 	release_sock(sk);
2642 	return res;
2643 }
2644 
2645 /**
2646  * tipc_listen - allow socket to listen for incoming connections
2647  * @sock: socket structure
2648  * @len: (unused)
2649  *
2650  * Return: 0 on success, errno otherwise
2651  */
2652 static int tipc_listen(struct socket *sock, int len)
2653 {
2654 	struct sock *sk = sock->sk;
2655 	int res;
2656 
2657 	lock_sock(sk);
2658 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2659 	release_sock(sk);
2660 
2661 	return res;
2662 }
2663 
2664 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2665 {
2666 	struct sock *sk = sock->sk;
2667 	DEFINE_WAIT(wait);
2668 	int err;
2669 
2670 	/* True wake-one mechanism for incoming connections: only
2671 	 * one process gets woken up, not the 'whole herd'.
2672 	 * Since we do not 'race & poll' for established sockets
2673 	 * anymore, the common case will execute the loop only once.
2674 	*/
2675 	for (;;) {
2676 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2677 					  TASK_INTERRUPTIBLE);
2678 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2679 			release_sock(sk);
2680 			timeo = schedule_timeout(timeo);
2681 			lock_sock(sk);
2682 		}
2683 		err = 0;
2684 		if (!skb_queue_empty(&sk->sk_receive_queue))
2685 			break;
2686 		err = -EAGAIN;
2687 		if (!timeo)
2688 			break;
2689 		err = sock_intr_errno(timeo);
2690 		if (signal_pending(current))
2691 			break;
2692 	}
2693 	finish_wait(sk_sleep(sk), &wait);
2694 	return err;
2695 }
2696 
2697 /**
2698  * tipc_accept - wait for connection request
2699  * @sock: listening socket
2700  * @new_sock: new socket that is to be connected
2701  * @flags: file-related flags associated with socket
2702  * @kern: caused by kernel or by userspace?
2703  *
2704  * Return: 0 on success, errno otherwise
2705  */
2706 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2707 		       bool kern)
2708 {
2709 	struct sock *new_sk, *sk = sock->sk;
2710 	struct sk_buff *buf;
2711 	struct tipc_sock *new_tsock;
2712 	struct tipc_msg *msg;
2713 	long timeo;
2714 	int res;
2715 
2716 	lock_sock(sk);
2717 
2718 	if (sk->sk_state != TIPC_LISTEN) {
2719 		res = -EINVAL;
2720 		goto exit;
2721 	}
2722 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2723 	res = tipc_wait_for_accept(sock, timeo);
2724 	if (res)
2725 		goto exit;
2726 
2727 	buf = skb_peek(&sk->sk_receive_queue);
2728 
2729 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2730 	if (res)
2731 		goto exit;
2732 	security_sk_clone(sock->sk, new_sock->sk);
2733 
2734 	new_sk = new_sock->sk;
2735 	new_tsock = tipc_sk(new_sk);
2736 	msg = buf_msg(buf);
2737 
2738 	/* we lock on new_sk; but lockdep sees the lock on sk */
2739 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2740 
2741 	/*
2742 	 * Reject any stray messages received by new socket
2743 	 * before the socket lock was taken (very, very unlikely)
2744 	 */
2745 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2746 
2747 	/* Connect new socket to it's peer */
2748 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2749 
2750 	tsk_set_importance(new_sk, msg_importance(msg));
2751 	if (msg_named(msg)) {
2752 		new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2753 		msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2754 		msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2755 	}
2756 
2757 	/*
2758 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2759 	 * Respond to 'SYN+' by queuing it on new socket.
2760 	 */
2761 	if (!msg_data_sz(msg)) {
2762 		struct msghdr m = {NULL,};
2763 
2764 		tsk_advance_rx_queue(sk);
2765 		__tipc_sendstream(new_sock, &m, 0);
2766 	} else {
2767 		__skb_dequeue(&sk->sk_receive_queue);
2768 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2769 		skb_set_owner_r(buf, new_sk);
2770 	}
2771 	release_sock(new_sk);
2772 exit:
2773 	release_sock(sk);
2774 	return res;
2775 }
2776 
2777 /**
2778  * tipc_shutdown - shutdown socket connection
2779  * @sock: socket structure
2780  * @how: direction to close (must be SHUT_RDWR)
2781  *
2782  * Terminates connection (if necessary), then purges socket's receive queue.
2783  *
2784  * Return: 0 on success, errno otherwise
2785  */
2786 static int tipc_shutdown(struct socket *sock, int how)
2787 {
2788 	struct sock *sk = sock->sk;
2789 	int res;
2790 
2791 	if (how != SHUT_RDWR)
2792 		return -EINVAL;
2793 
2794 	lock_sock(sk);
2795 
2796 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2797 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2798 	sk->sk_shutdown = SHUTDOWN_MASK;
2799 
2800 	if (sk->sk_state == TIPC_DISCONNECTING) {
2801 		/* Discard any unreceived messages */
2802 		__skb_queue_purge(&sk->sk_receive_queue);
2803 
2804 		res = 0;
2805 	} else {
2806 		res = -ENOTCONN;
2807 	}
2808 	/* Wake up anyone sleeping in poll. */
2809 	sk->sk_state_change(sk);
2810 
2811 	release_sock(sk);
2812 	return res;
2813 }
2814 
2815 static void tipc_sk_check_probing_state(struct sock *sk,
2816 					struct sk_buff_head *list)
2817 {
2818 	struct tipc_sock *tsk = tipc_sk(sk);
2819 	u32 pnode = tsk_peer_node(tsk);
2820 	u32 pport = tsk_peer_port(tsk);
2821 	u32 self = tsk_own_node(tsk);
2822 	u32 oport = tsk->portid;
2823 	struct sk_buff *skb;
2824 
2825 	if (tsk->probe_unacked) {
2826 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2827 		sk->sk_err = ECONNABORTED;
2828 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2829 		sk->sk_state_change(sk);
2830 		return;
2831 	}
2832 	/* Prepare new probe */
2833 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2834 			      pnode, self, pport, oport, TIPC_OK);
2835 	if (skb)
2836 		__skb_queue_tail(list, skb);
2837 	tsk->probe_unacked = true;
2838 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2839 }
2840 
2841 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2842 {
2843 	struct tipc_sock *tsk = tipc_sk(sk);
2844 
2845 	/* Try again later if dest link is congested */
2846 	if (tsk->cong_link_cnt) {
2847 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2848 		return;
2849 	}
2850 	/* Prepare SYN for retransmit */
2851 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2852 }
2853 
2854 static void tipc_sk_timeout(struct timer_list *t)
2855 {
2856 	struct sock *sk = from_timer(sk, t, sk_timer);
2857 	struct tipc_sock *tsk = tipc_sk(sk);
2858 	u32 pnode = tsk_peer_node(tsk);
2859 	struct sk_buff_head list;
2860 	int rc = 0;
2861 
2862 	__skb_queue_head_init(&list);
2863 	bh_lock_sock(sk);
2864 
2865 	/* Try again later if socket is busy */
2866 	if (sock_owned_by_user(sk)) {
2867 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2868 		bh_unlock_sock(sk);
2869 		sock_put(sk);
2870 		return;
2871 	}
2872 
2873 	if (sk->sk_state == TIPC_ESTABLISHED)
2874 		tipc_sk_check_probing_state(sk, &list);
2875 	else if (sk->sk_state == TIPC_CONNECTING)
2876 		tipc_sk_retry_connect(sk, &list);
2877 
2878 	bh_unlock_sock(sk);
2879 
2880 	if (!skb_queue_empty(&list))
2881 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2882 
2883 	/* SYN messages may cause link congestion */
2884 	if (rc == -ELINKCONG) {
2885 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2886 		tsk->cong_link_cnt = 1;
2887 	}
2888 	sock_put(sk);
2889 }
2890 
2891 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2892 {
2893 	struct sock *sk = &tsk->sk;
2894 	struct net *net = sock_net(sk);
2895 	struct tipc_socket_addr skaddr;
2896 	struct publication *p;
2897 	u32 key;
2898 
2899 	if (tipc_sk_connected(sk))
2900 		return -EINVAL;
2901 	key = tsk->portid + tsk->pub_count + 1;
2902 	if (key == tsk->portid)
2903 		return -EADDRINUSE;
2904 	skaddr.ref = tsk->portid;
2905 	skaddr.node = tipc_own_addr(net);
2906 	p = tipc_nametbl_publish(net, ua, &skaddr, key);
2907 	if (unlikely(!p))
2908 		return -EINVAL;
2909 
2910 	list_add(&p->binding_sock, &tsk->publications);
2911 	tsk->pub_count++;
2912 	tsk->published = true;
2913 	return 0;
2914 }
2915 
2916 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2917 {
2918 	struct net *net = sock_net(&tsk->sk);
2919 	struct publication *safe, *p;
2920 	struct tipc_uaddr _ua;
2921 	int rc = -EINVAL;
2922 
2923 	list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2924 		if (!ua) {
2925 			tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2926 				   p->sr.type, p->sr.lower, p->sr.upper);
2927 			tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2928 			continue;
2929 		}
2930 		/* Unbind specific publication */
2931 		if (p->scope != ua->scope)
2932 			continue;
2933 		if (p->sr.type != ua->sr.type)
2934 			continue;
2935 		if (p->sr.lower != ua->sr.lower)
2936 			continue;
2937 		if (p->sr.upper != ua->sr.upper)
2938 			break;
2939 		tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2940 		rc = 0;
2941 		break;
2942 	}
2943 	if (list_empty(&tsk->publications)) {
2944 		tsk->published = 0;
2945 		rc = 0;
2946 	}
2947 	return rc;
2948 }
2949 
2950 /* tipc_sk_reinit: set non-zero address in all existing sockets
2951  *                 when we go from standalone to network mode.
2952  */
2953 void tipc_sk_reinit(struct net *net)
2954 {
2955 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2956 	struct rhashtable_iter iter;
2957 	struct tipc_sock *tsk;
2958 	struct tipc_msg *msg;
2959 
2960 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2961 
2962 	do {
2963 		rhashtable_walk_start(&iter);
2964 
2965 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2966 			sock_hold(&tsk->sk);
2967 			rhashtable_walk_stop(&iter);
2968 			lock_sock(&tsk->sk);
2969 			msg = &tsk->phdr;
2970 			msg_set_prevnode(msg, tipc_own_addr(net));
2971 			msg_set_orignode(msg, tipc_own_addr(net));
2972 			release_sock(&tsk->sk);
2973 			rhashtable_walk_start(&iter);
2974 			sock_put(&tsk->sk);
2975 		}
2976 
2977 		rhashtable_walk_stop(&iter);
2978 	} while (tsk == ERR_PTR(-EAGAIN));
2979 
2980 	rhashtable_walk_exit(&iter);
2981 }
2982 
2983 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2984 {
2985 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2986 	struct tipc_sock *tsk;
2987 
2988 	rcu_read_lock();
2989 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2990 	if (tsk)
2991 		sock_hold(&tsk->sk);
2992 	rcu_read_unlock();
2993 
2994 	return tsk;
2995 }
2996 
2997 static int tipc_sk_insert(struct tipc_sock *tsk)
2998 {
2999 	struct sock *sk = &tsk->sk;
3000 	struct net *net = sock_net(sk);
3001 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3002 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3003 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3004 
3005 	while (remaining--) {
3006 		portid++;
3007 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3008 			portid = TIPC_MIN_PORT;
3009 		tsk->portid = portid;
3010 		sock_hold(&tsk->sk);
3011 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3012 						   tsk_rht_params))
3013 			return 0;
3014 		sock_put(&tsk->sk);
3015 	}
3016 
3017 	return -1;
3018 }
3019 
3020 static void tipc_sk_remove(struct tipc_sock *tsk)
3021 {
3022 	struct sock *sk = &tsk->sk;
3023 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3024 
3025 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3026 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3027 		__sock_put(sk);
3028 	}
3029 }
3030 
3031 static const struct rhashtable_params tsk_rht_params = {
3032 	.nelem_hint = 192,
3033 	.head_offset = offsetof(struct tipc_sock, node),
3034 	.key_offset = offsetof(struct tipc_sock, portid),
3035 	.key_len = sizeof(u32), /* portid */
3036 	.max_size = 1048576,
3037 	.min_size = 256,
3038 	.automatic_shrinking = true,
3039 };
3040 
3041 int tipc_sk_rht_init(struct net *net)
3042 {
3043 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3044 
3045 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3046 }
3047 
3048 void tipc_sk_rht_destroy(struct net *net)
3049 {
3050 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3051 
3052 	/* Wait for socket readers to complete */
3053 	synchronize_net();
3054 
3055 	rhashtable_destroy(&tn->sk_rht);
3056 }
3057 
3058 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3059 {
3060 	struct net *net = sock_net(&tsk->sk);
3061 	struct tipc_group *grp = tsk->group;
3062 	struct tipc_msg *hdr = &tsk->phdr;
3063 	struct tipc_uaddr ua;
3064 	int rc;
3065 
3066 	if (mreq->type < TIPC_RESERVED_TYPES)
3067 		return -EACCES;
3068 	if (mreq->scope > TIPC_NODE_SCOPE)
3069 		return -EINVAL;
3070 	if (mreq->scope != TIPC_NODE_SCOPE)
3071 		mreq->scope = TIPC_CLUSTER_SCOPE;
3072 	if (grp)
3073 		return -EACCES;
3074 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3075 	if (!grp)
3076 		return -ENOMEM;
3077 	tsk->group = grp;
3078 	msg_set_lookup_scope(hdr, mreq->scope);
3079 	msg_set_nametype(hdr, mreq->type);
3080 	msg_set_dest_droppable(hdr, true);
3081 	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3082 		   mreq->type, mreq->instance, mreq->instance);
3083 	tipc_nametbl_build_group(net, grp, &ua);
3084 	rc = tipc_sk_publish(tsk, &ua);
3085 	if (rc) {
3086 		tipc_group_delete(net, grp);
3087 		tsk->group = NULL;
3088 		return rc;
3089 	}
3090 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3091 	tsk->mc_method.rcast = true;
3092 	tsk->mc_method.mandatory = true;
3093 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3094 	return rc;
3095 }
3096 
3097 static int tipc_sk_leave(struct tipc_sock *tsk)
3098 {
3099 	struct net *net = sock_net(&tsk->sk);
3100 	struct tipc_group *grp = tsk->group;
3101 	struct tipc_uaddr ua;
3102 	int scope;
3103 
3104 	if (!grp)
3105 		return -EINVAL;
3106 	ua.addrtype = TIPC_SERVICE_RANGE;
3107 	tipc_group_self(grp, &ua.sr, &scope);
3108 	ua.scope = scope;
3109 	tipc_group_delete(net, grp);
3110 	tsk->group = NULL;
3111 	tipc_sk_withdraw(tsk, &ua);
3112 	return 0;
3113 }
3114 
3115 /**
3116  * tipc_setsockopt - set socket option
3117  * @sock: socket structure
3118  * @lvl: option level
3119  * @opt: option identifier
3120  * @ov: pointer to new option value
3121  * @ol: length of option value
3122  *
3123  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3124  * (to ease compatibility).
3125  *
3126  * Return: 0 on success, errno otherwise
3127  */
3128 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3129 			   sockptr_t ov, unsigned int ol)
3130 {
3131 	struct sock *sk = sock->sk;
3132 	struct tipc_sock *tsk = tipc_sk(sk);
3133 	struct tipc_group_req mreq;
3134 	u32 value = 0;
3135 	int res = 0;
3136 
3137 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3138 		return 0;
3139 	if (lvl != SOL_TIPC)
3140 		return -ENOPROTOOPT;
3141 
3142 	switch (opt) {
3143 	case TIPC_IMPORTANCE:
3144 	case TIPC_SRC_DROPPABLE:
3145 	case TIPC_DEST_DROPPABLE:
3146 	case TIPC_CONN_TIMEOUT:
3147 	case TIPC_NODELAY:
3148 		if (ol < sizeof(value))
3149 			return -EINVAL;
3150 		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3151 			return -EFAULT;
3152 		break;
3153 	case TIPC_GROUP_JOIN:
3154 		if (ol < sizeof(mreq))
3155 			return -EINVAL;
3156 		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3157 			return -EFAULT;
3158 		break;
3159 	default:
3160 		if (!sockptr_is_null(ov) || ol)
3161 			return -EINVAL;
3162 	}
3163 
3164 	lock_sock(sk);
3165 
3166 	switch (opt) {
3167 	case TIPC_IMPORTANCE:
3168 		res = tsk_set_importance(sk, value);
3169 		break;
3170 	case TIPC_SRC_DROPPABLE:
3171 		if (sock->type != SOCK_STREAM)
3172 			tsk_set_unreliable(tsk, value);
3173 		else
3174 			res = -ENOPROTOOPT;
3175 		break;
3176 	case TIPC_DEST_DROPPABLE:
3177 		tsk_set_unreturnable(tsk, value);
3178 		break;
3179 	case TIPC_CONN_TIMEOUT:
3180 		tipc_sk(sk)->conn_timeout = value;
3181 		break;
3182 	case TIPC_MCAST_BROADCAST:
3183 		tsk->mc_method.rcast = false;
3184 		tsk->mc_method.mandatory = true;
3185 		break;
3186 	case TIPC_MCAST_REPLICAST:
3187 		tsk->mc_method.rcast = true;
3188 		tsk->mc_method.mandatory = true;
3189 		break;
3190 	case TIPC_GROUP_JOIN:
3191 		res = tipc_sk_join(tsk, &mreq);
3192 		break;
3193 	case TIPC_GROUP_LEAVE:
3194 		res = tipc_sk_leave(tsk);
3195 		break;
3196 	case TIPC_NODELAY:
3197 		tsk->nodelay = !!value;
3198 		tsk_set_nagle(tsk);
3199 		break;
3200 	default:
3201 		res = -EINVAL;
3202 	}
3203 
3204 	release_sock(sk);
3205 
3206 	return res;
3207 }
3208 
3209 /**
3210  * tipc_getsockopt - get socket option
3211  * @sock: socket structure
3212  * @lvl: option level
3213  * @opt: option identifier
3214  * @ov: receptacle for option value
3215  * @ol: receptacle for length of option value
3216  *
3217  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3218  * (to ease compatibility).
3219  *
3220  * Return: 0 on success, errno otherwise
3221  */
3222 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3223 			   char __user *ov, int __user *ol)
3224 {
3225 	struct sock *sk = sock->sk;
3226 	struct tipc_sock *tsk = tipc_sk(sk);
3227 	struct tipc_service_range seq;
3228 	int len, scope;
3229 	u32 value;
3230 	int res;
3231 
3232 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3233 		return put_user(0, ol);
3234 	if (lvl != SOL_TIPC)
3235 		return -ENOPROTOOPT;
3236 	res = get_user(len, ol);
3237 	if (res)
3238 		return res;
3239 
3240 	lock_sock(sk);
3241 
3242 	switch (opt) {
3243 	case TIPC_IMPORTANCE:
3244 		value = tsk_importance(tsk);
3245 		break;
3246 	case TIPC_SRC_DROPPABLE:
3247 		value = tsk_unreliable(tsk);
3248 		break;
3249 	case TIPC_DEST_DROPPABLE:
3250 		value = tsk_unreturnable(tsk);
3251 		break;
3252 	case TIPC_CONN_TIMEOUT:
3253 		value = tsk->conn_timeout;
3254 		/* no need to set "res", since already 0 at this point */
3255 		break;
3256 	case TIPC_NODE_RECVQ_DEPTH:
3257 		value = 0; /* was tipc_queue_size, now obsolete */
3258 		break;
3259 	case TIPC_SOCK_RECVQ_DEPTH:
3260 		value = skb_queue_len(&sk->sk_receive_queue);
3261 		break;
3262 	case TIPC_SOCK_RECVQ_USED:
3263 		value = sk_rmem_alloc_get(sk);
3264 		break;
3265 	case TIPC_GROUP_JOIN:
3266 		seq.type = 0;
3267 		if (tsk->group)
3268 			tipc_group_self(tsk->group, &seq, &scope);
3269 		value = seq.type;
3270 		break;
3271 	default:
3272 		res = -EINVAL;
3273 	}
3274 
3275 	release_sock(sk);
3276 
3277 	if (res)
3278 		return res;	/* "get" failed */
3279 
3280 	if (len < sizeof(value))
3281 		return -EINVAL;
3282 
3283 	if (copy_to_user(ov, &value, sizeof(value)))
3284 		return -EFAULT;
3285 
3286 	return put_user(sizeof(value), ol);
3287 }
3288 
3289 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3290 {
3291 	struct net *net = sock_net(sock->sk);
3292 	struct tipc_sioc_nodeid_req nr = {0};
3293 	struct tipc_sioc_ln_req lnr;
3294 	void __user *argp = (void __user *)arg;
3295 
3296 	switch (cmd) {
3297 	case SIOCGETLINKNAME:
3298 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3299 			return -EFAULT;
3300 		if (!tipc_node_get_linkname(net,
3301 					    lnr.bearer_id & 0xffff, lnr.peer,
3302 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3303 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3304 				return -EFAULT;
3305 			return 0;
3306 		}
3307 		return -EADDRNOTAVAIL;
3308 	case SIOCGETNODEID:
3309 		if (copy_from_user(&nr, argp, sizeof(nr)))
3310 			return -EFAULT;
3311 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3312 			return -EADDRNOTAVAIL;
3313 		if (copy_to_user(argp, &nr, sizeof(nr)))
3314 			return -EFAULT;
3315 		return 0;
3316 	default:
3317 		return -ENOIOCTLCMD;
3318 	}
3319 }
3320 
3321 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3322 {
3323 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3324 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3325 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3326 
3327 	tsk1->peer.family = AF_TIPC;
3328 	tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3329 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3330 	tsk1->peer.addr.id.ref = tsk2->portid;
3331 	tsk1->peer.addr.id.node = onode;
3332 	tsk2->peer.family = AF_TIPC;
3333 	tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3334 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3335 	tsk2->peer.addr.id.ref = tsk1->portid;
3336 	tsk2->peer.addr.id.node = onode;
3337 
3338 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3339 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3340 	return 0;
3341 }
3342 
3343 /* Protocol switches for the various types of TIPC sockets */
3344 
3345 static const struct proto_ops msg_ops = {
3346 	.owner		= THIS_MODULE,
3347 	.family		= AF_TIPC,
3348 	.release	= tipc_release,
3349 	.bind		= tipc_bind,
3350 	.connect	= tipc_connect,
3351 	.socketpair	= tipc_socketpair,
3352 	.accept		= sock_no_accept,
3353 	.getname	= tipc_getname,
3354 	.poll		= tipc_poll,
3355 	.ioctl		= tipc_ioctl,
3356 	.listen		= sock_no_listen,
3357 	.shutdown	= tipc_shutdown,
3358 	.setsockopt	= tipc_setsockopt,
3359 	.getsockopt	= tipc_getsockopt,
3360 	.sendmsg	= tipc_sendmsg,
3361 	.recvmsg	= tipc_recvmsg,
3362 	.mmap		= sock_no_mmap,
3363 	.sendpage	= sock_no_sendpage
3364 };
3365 
3366 static const struct proto_ops packet_ops = {
3367 	.owner		= THIS_MODULE,
3368 	.family		= AF_TIPC,
3369 	.release	= tipc_release,
3370 	.bind		= tipc_bind,
3371 	.connect	= tipc_connect,
3372 	.socketpair	= tipc_socketpair,
3373 	.accept		= tipc_accept,
3374 	.getname	= tipc_getname,
3375 	.poll		= tipc_poll,
3376 	.ioctl		= tipc_ioctl,
3377 	.listen		= tipc_listen,
3378 	.shutdown	= tipc_shutdown,
3379 	.setsockopt	= tipc_setsockopt,
3380 	.getsockopt	= tipc_getsockopt,
3381 	.sendmsg	= tipc_send_packet,
3382 	.recvmsg	= tipc_recvmsg,
3383 	.mmap		= sock_no_mmap,
3384 	.sendpage	= sock_no_sendpage
3385 };
3386 
3387 static const struct proto_ops stream_ops = {
3388 	.owner		= THIS_MODULE,
3389 	.family		= AF_TIPC,
3390 	.release	= tipc_release,
3391 	.bind		= tipc_bind,
3392 	.connect	= tipc_connect,
3393 	.socketpair	= tipc_socketpair,
3394 	.accept		= tipc_accept,
3395 	.getname	= tipc_getname,
3396 	.poll		= tipc_poll,
3397 	.ioctl		= tipc_ioctl,
3398 	.listen		= tipc_listen,
3399 	.shutdown	= tipc_shutdown,
3400 	.setsockopt	= tipc_setsockopt,
3401 	.getsockopt	= tipc_getsockopt,
3402 	.sendmsg	= tipc_sendstream,
3403 	.recvmsg	= tipc_recvstream,
3404 	.mmap		= sock_no_mmap,
3405 	.sendpage	= sock_no_sendpage
3406 };
3407 
3408 static const struct net_proto_family tipc_family_ops = {
3409 	.owner		= THIS_MODULE,
3410 	.family		= AF_TIPC,
3411 	.create		= tipc_sk_create
3412 };
3413 
3414 static struct proto tipc_proto = {
3415 	.name		= "TIPC",
3416 	.owner		= THIS_MODULE,
3417 	.obj_size	= sizeof(struct tipc_sock),
3418 	.sysctl_rmem	= sysctl_tipc_rmem
3419 };
3420 
3421 /**
3422  * tipc_socket_init - initialize TIPC socket interface
3423  *
3424  * Return: 0 on success, errno otherwise
3425  */
3426 int tipc_socket_init(void)
3427 {
3428 	int res;
3429 
3430 	res = proto_register(&tipc_proto, 1);
3431 	if (res) {
3432 		pr_err("Failed to register TIPC protocol type\n");
3433 		goto out;
3434 	}
3435 
3436 	res = sock_register(&tipc_family_ops);
3437 	if (res) {
3438 		pr_err("Failed to register TIPC socket type\n");
3439 		proto_unregister(&tipc_proto);
3440 		goto out;
3441 	}
3442  out:
3443 	return res;
3444 }
3445 
3446 /**
3447  * tipc_socket_stop - stop TIPC socket interface
3448  */
3449 void tipc_socket_stop(void)
3450 {
3451 	sock_unregister(tipc_family_ops.family);
3452 	proto_unregister(&tipc_proto);
3453 }
3454 
3455 /* Caller should hold socket lock for the passed tipc socket. */
3456 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3457 {
3458 	u32 peer_node, peer_port;
3459 	u32 conn_type, conn_instance;
3460 	struct nlattr *nest;
3461 
3462 	peer_node = tsk_peer_node(tsk);
3463 	peer_port = tsk_peer_port(tsk);
3464 	conn_type = msg_nametype(&tsk->phdr);
3465 	conn_instance = msg_nameinst(&tsk->phdr);
3466 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3467 	if (!nest)
3468 		return -EMSGSIZE;
3469 
3470 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3471 		goto msg_full;
3472 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3473 		goto msg_full;
3474 
3475 	if (tsk->conn_addrtype != 0) {
3476 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3477 			goto msg_full;
3478 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3479 			goto msg_full;
3480 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3481 			goto msg_full;
3482 	}
3483 	nla_nest_end(skb, nest);
3484 
3485 	return 0;
3486 
3487 msg_full:
3488 	nla_nest_cancel(skb, nest);
3489 
3490 	return -EMSGSIZE;
3491 }
3492 
3493 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3494 			  *tsk)
3495 {
3496 	struct net *net = sock_net(skb->sk);
3497 	struct sock *sk = &tsk->sk;
3498 
3499 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3500 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3501 		return -EMSGSIZE;
3502 
3503 	if (tipc_sk_connected(sk)) {
3504 		if (__tipc_nl_add_sk_con(skb, tsk))
3505 			return -EMSGSIZE;
3506 	} else if (!list_empty(&tsk->publications)) {
3507 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3508 			return -EMSGSIZE;
3509 	}
3510 	return 0;
3511 }
3512 
3513 /* Caller should hold socket lock for the passed tipc socket. */
3514 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3515 			    struct tipc_sock *tsk)
3516 {
3517 	struct nlattr *attrs;
3518 	void *hdr;
3519 
3520 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3521 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3522 	if (!hdr)
3523 		goto msg_cancel;
3524 
3525 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3526 	if (!attrs)
3527 		goto genlmsg_cancel;
3528 
3529 	if (__tipc_nl_add_sk_info(skb, tsk))
3530 		goto attr_msg_cancel;
3531 
3532 	nla_nest_end(skb, attrs);
3533 	genlmsg_end(skb, hdr);
3534 
3535 	return 0;
3536 
3537 attr_msg_cancel:
3538 	nla_nest_cancel(skb, attrs);
3539 genlmsg_cancel:
3540 	genlmsg_cancel(skb, hdr);
3541 msg_cancel:
3542 	return -EMSGSIZE;
3543 }
3544 
3545 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3546 		    int (*skb_handler)(struct sk_buff *skb,
3547 				       struct netlink_callback *cb,
3548 				       struct tipc_sock *tsk))
3549 {
3550 	struct rhashtable_iter *iter = (void *)cb->args[4];
3551 	struct tipc_sock *tsk;
3552 	int err;
3553 
3554 	rhashtable_walk_start(iter);
3555 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3556 		if (IS_ERR(tsk)) {
3557 			err = PTR_ERR(tsk);
3558 			if (err == -EAGAIN) {
3559 				err = 0;
3560 				continue;
3561 			}
3562 			break;
3563 		}
3564 
3565 		sock_hold(&tsk->sk);
3566 		rhashtable_walk_stop(iter);
3567 		lock_sock(&tsk->sk);
3568 		err = skb_handler(skb, cb, tsk);
3569 		if (err) {
3570 			release_sock(&tsk->sk);
3571 			sock_put(&tsk->sk);
3572 			goto out;
3573 		}
3574 		release_sock(&tsk->sk);
3575 		rhashtable_walk_start(iter);
3576 		sock_put(&tsk->sk);
3577 	}
3578 	rhashtable_walk_stop(iter);
3579 out:
3580 	return skb->len;
3581 }
3582 EXPORT_SYMBOL(tipc_nl_sk_walk);
3583 
3584 int tipc_dump_start(struct netlink_callback *cb)
3585 {
3586 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3587 }
3588 EXPORT_SYMBOL(tipc_dump_start);
3589 
3590 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3591 {
3592 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3593 	struct rhashtable_iter *iter = (void *)cb->args[4];
3594 	struct tipc_net *tn = tipc_net(net);
3595 
3596 	if (!iter) {
3597 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3598 		if (!iter)
3599 			return -ENOMEM;
3600 
3601 		cb->args[4] = (long)iter;
3602 	}
3603 
3604 	rhashtable_walk_enter(&tn->sk_rht, iter);
3605 	return 0;
3606 }
3607 
3608 int tipc_dump_done(struct netlink_callback *cb)
3609 {
3610 	struct rhashtable_iter *hti = (void *)cb->args[4];
3611 
3612 	rhashtable_walk_exit(hti);
3613 	kfree(hti);
3614 	return 0;
3615 }
3616 EXPORT_SYMBOL(tipc_dump_done);
3617 
3618 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3619 			   struct tipc_sock *tsk, u32 sk_filter_state,
3620 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3621 {
3622 	struct sock *sk = &tsk->sk;
3623 	struct nlattr *attrs;
3624 	struct nlattr *stat;
3625 
3626 	/*filter response w.r.t sk_state*/
3627 	if (!(sk_filter_state & (1 << sk->sk_state)))
3628 		return 0;
3629 
3630 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3631 	if (!attrs)
3632 		goto msg_cancel;
3633 
3634 	if (__tipc_nl_add_sk_info(skb, tsk))
3635 		goto attr_msg_cancel;
3636 
3637 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3638 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3639 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3640 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3641 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3642 					 sock_i_uid(sk))) ||
3643 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3644 			      tipc_diag_gen_cookie(sk),
3645 			      TIPC_NLA_SOCK_PAD))
3646 		goto attr_msg_cancel;
3647 
3648 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3649 	if (!stat)
3650 		goto attr_msg_cancel;
3651 
3652 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3653 			skb_queue_len(&sk->sk_receive_queue)) ||
3654 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3655 			skb_queue_len(&sk->sk_write_queue)) ||
3656 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3657 			atomic_read(&sk->sk_drops)))
3658 		goto stat_msg_cancel;
3659 
3660 	if (tsk->cong_link_cnt &&
3661 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3662 		goto stat_msg_cancel;
3663 
3664 	if (tsk_conn_cong(tsk) &&
3665 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3666 		goto stat_msg_cancel;
3667 
3668 	nla_nest_end(skb, stat);
3669 
3670 	if (tsk->group)
3671 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3672 			goto stat_msg_cancel;
3673 
3674 	nla_nest_end(skb, attrs);
3675 
3676 	return 0;
3677 
3678 stat_msg_cancel:
3679 	nla_nest_cancel(skb, stat);
3680 attr_msg_cancel:
3681 	nla_nest_cancel(skb, attrs);
3682 msg_cancel:
3683 	return -EMSGSIZE;
3684 }
3685 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3686 
3687 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3688 {
3689 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3690 }
3691 
3692 /* Caller should hold socket lock for the passed tipc socket. */
3693 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3694 				 struct netlink_callback *cb,
3695 				 struct publication *publ)
3696 {
3697 	void *hdr;
3698 	struct nlattr *attrs;
3699 
3700 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3701 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3702 	if (!hdr)
3703 		goto msg_cancel;
3704 
3705 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3706 	if (!attrs)
3707 		goto genlmsg_cancel;
3708 
3709 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3710 		goto attr_msg_cancel;
3711 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3712 		goto attr_msg_cancel;
3713 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3714 		goto attr_msg_cancel;
3715 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3716 		goto attr_msg_cancel;
3717 
3718 	nla_nest_end(skb, attrs);
3719 	genlmsg_end(skb, hdr);
3720 
3721 	return 0;
3722 
3723 attr_msg_cancel:
3724 	nla_nest_cancel(skb, attrs);
3725 genlmsg_cancel:
3726 	genlmsg_cancel(skb, hdr);
3727 msg_cancel:
3728 	return -EMSGSIZE;
3729 }
3730 
3731 /* Caller should hold socket lock for the passed tipc socket. */
3732 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3733 				  struct netlink_callback *cb,
3734 				  struct tipc_sock *tsk, u32 *last_publ)
3735 {
3736 	int err;
3737 	struct publication *p;
3738 
3739 	if (*last_publ) {
3740 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3741 			if (p->key == *last_publ)
3742 				break;
3743 		}
3744 		if (p->key != *last_publ) {
3745 			/* We never set seq or call nl_dump_check_consistent()
3746 			 * this means that setting prev_seq here will cause the
3747 			 * consistence check to fail in the netlink callback
3748 			 * handler. Resulting in the last NLMSG_DONE message
3749 			 * having the NLM_F_DUMP_INTR flag set.
3750 			 */
3751 			cb->prev_seq = 1;
3752 			*last_publ = 0;
3753 			return -EPIPE;
3754 		}
3755 	} else {
3756 		p = list_first_entry(&tsk->publications, struct publication,
3757 				     binding_sock);
3758 	}
3759 
3760 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3761 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3762 		if (err) {
3763 			*last_publ = p->key;
3764 			return err;
3765 		}
3766 	}
3767 	*last_publ = 0;
3768 
3769 	return 0;
3770 }
3771 
3772 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3773 {
3774 	int err;
3775 	u32 tsk_portid = cb->args[0];
3776 	u32 last_publ = cb->args[1];
3777 	u32 done = cb->args[2];
3778 	struct net *net = sock_net(skb->sk);
3779 	struct tipc_sock *tsk;
3780 
3781 	if (!tsk_portid) {
3782 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3783 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3784 
3785 		if (!attrs[TIPC_NLA_SOCK])
3786 			return -EINVAL;
3787 
3788 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3789 						  attrs[TIPC_NLA_SOCK],
3790 						  tipc_nl_sock_policy, NULL);
3791 		if (err)
3792 			return err;
3793 
3794 		if (!sock[TIPC_NLA_SOCK_REF])
3795 			return -EINVAL;
3796 
3797 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3798 	}
3799 
3800 	if (done)
3801 		return 0;
3802 
3803 	tsk = tipc_sk_lookup(net, tsk_portid);
3804 	if (!tsk)
3805 		return -EINVAL;
3806 
3807 	lock_sock(&tsk->sk);
3808 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3809 	if (!err)
3810 		done = 1;
3811 	release_sock(&tsk->sk);
3812 	sock_put(&tsk->sk);
3813 
3814 	cb->args[0] = tsk_portid;
3815 	cb->args[1] = last_publ;
3816 	cb->args[2] = done;
3817 
3818 	return skb->len;
3819 }
3820 
3821 /**
3822  * tipc_sk_filtering - check if a socket should be traced
3823  * @sk: the socket to be examined
3824  *
3825  * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3826  * (portid, sock type, name type, name lower, name upper)
3827  *
3828  * Return: true if the socket meets the socket tuple data
3829  * (value 0 = 'any') or when there is no tuple set (all = 0),
3830  * otherwise false
3831  */
3832 bool tipc_sk_filtering(struct sock *sk)
3833 {
3834 	struct tipc_sock *tsk;
3835 	struct publication *p;
3836 	u32 _port, _sktype, _type, _lower, _upper;
3837 	u32 type = 0, lower = 0, upper = 0;
3838 
3839 	if (!sk)
3840 		return true;
3841 
3842 	tsk = tipc_sk(sk);
3843 
3844 	_port = sysctl_tipc_sk_filter[0];
3845 	_sktype = sysctl_tipc_sk_filter[1];
3846 	_type = sysctl_tipc_sk_filter[2];
3847 	_lower = sysctl_tipc_sk_filter[3];
3848 	_upper = sysctl_tipc_sk_filter[4];
3849 
3850 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3851 		return true;
3852 
3853 	if (_port)
3854 		return (_port == tsk->portid);
3855 
3856 	if (_sktype && _sktype != sk->sk_type)
3857 		return false;
3858 
3859 	if (tsk->published) {
3860 		p = list_first_entry_or_null(&tsk->publications,
3861 					     struct publication, binding_sock);
3862 		if (p) {
3863 			type = p->sr.type;
3864 			lower = p->sr.lower;
3865 			upper = p->sr.upper;
3866 		}
3867 	}
3868 
3869 	if (!tipc_sk_type_connectionless(sk)) {
3870 		type = msg_nametype(&tsk->phdr);
3871 		lower = msg_nameinst(&tsk->phdr);
3872 		upper = lower;
3873 	}
3874 
3875 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3876 	    (_upper && _upper != upper))
3877 		return false;
3878 
3879 	return true;
3880 }
3881 
3882 u32 tipc_sock_get_portid(struct sock *sk)
3883 {
3884 	return (sk) ? (tipc_sk(sk))->portid : 0;
3885 }
3886 
3887 /**
3888  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3889  *			both the rcv and backlog queues are considered
3890  * @sk: tipc sk to be checked
3891  * @skb: tipc msg to be checked
3892  *
3893  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3894  */
3895 
3896 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3897 {
3898 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3899 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3900 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3901 
3902 	return (qsize > lim * 90 / 100);
3903 }
3904 
3905 /**
3906  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3907  *			only the rcv queue is considered
3908  * @sk: tipc sk to be checked
3909  * @skb: tipc msg to be checked
3910  *
3911  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3912  */
3913 
3914 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3915 {
3916 	unsigned int lim = rcvbuf_limit(sk, skb);
3917 	unsigned int qsize = sk_rmem_alloc_get(sk);
3918 
3919 	return (qsize > lim * 90 / 100);
3920 }
3921 
3922 /**
3923  * tipc_sk_dump - dump TIPC socket
3924  * @sk: tipc sk to be dumped
3925  * @dqueues: bitmask to decide if any socket queue to be dumped?
3926  *           - TIPC_DUMP_NONE: don't dump socket queues
3927  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3928  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3929  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3930  *           - TIPC_DUMP_ALL: dump all the socket queues above
3931  * @buf: returned buffer of dump data in format
3932  */
3933 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3934 {
3935 	int i = 0;
3936 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3937 	u32 conn_type, conn_instance;
3938 	struct tipc_sock *tsk;
3939 	struct publication *p;
3940 	bool tsk_connected;
3941 
3942 	if (!sk) {
3943 		i += scnprintf(buf, sz, "sk data: (null)\n");
3944 		return i;
3945 	}
3946 
3947 	tsk = tipc_sk(sk);
3948 	tsk_connected = !tipc_sk_type_connectionless(sk);
3949 
3950 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3951 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3952 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3953 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3954 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3955 	if (tsk_connected) {
3956 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3957 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3958 		conn_type = msg_nametype(&tsk->phdr);
3959 		conn_instance = msg_nameinst(&tsk->phdr);
3960 		i += scnprintf(buf + i, sz - i, " %u", conn_type);
3961 		i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3962 	}
3963 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3964 	if (tsk->published) {
3965 		p = list_first_entry_or_null(&tsk->publications,
3966 					     struct publication, binding_sock);
3967 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3968 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3969 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3970 	}
3971 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3972 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3973 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3974 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3975 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3976 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3977 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3978 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3979 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3980 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3981 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3982 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3983 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3984 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3985 
3986 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3987 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3988 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3989 	}
3990 
3991 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3992 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3993 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3994 	}
3995 
3996 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3997 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3998 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3999 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4000 			i += scnprintf(buf + i, sz - i, "  tail ");
4001 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
4002 					   buf + i);
4003 		}
4004 	}
4005 
4006 	return i;
4007 }
4008