xref: /openbmc/linux/net/tipc/socket.c (revision d87c25e8)
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 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
162 
163 static const struct proto_ops packet_ops;
164 static const struct proto_ops stream_ops;
165 static const struct proto_ops msg_ops;
166 static struct proto tipc_proto;
167 static const struct rhashtable_params tsk_rht_params;
168 
169 static u32 tsk_own_node(struct tipc_sock *tsk)
170 {
171 	return msg_prevnode(&tsk->phdr);
172 }
173 
174 static u32 tsk_peer_node(struct tipc_sock *tsk)
175 {
176 	return msg_destnode(&tsk->phdr);
177 }
178 
179 static u32 tsk_peer_port(struct tipc_sock *tsk)
180 {
181 	return msg_destport(&tsk->phdr);
182 }
183 
184 static  bool tsk_unreliable(struct tipc_sock *tsk)
185 {
186 	return msg_src_droppable(&tsk->phdr) != 0;
187 }
188 
189 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
190 {
191 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
192 }
193 
194 static bool tsk_unreturnable(struct tipc_sock *tsk)
195 {
196 	return msg_dest_droppable(&tsk->phdr) != 0;
197 }
198 
199 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
200 {
201 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
202 }
203 
204 static int tsk_importance(struct tipc_sock *tsk)
205 {
206 	return msg_importance(&tsk->phdr);
207 }
208 
209 static struct tipc_sock *tipc_sk(const struct sock *sk)
210 {
211 	return container_of(sk, struct tipc_sock, sk);
212 }
213 
214 int tsk_set_importance(struct sock *sk, int imp)
215 {
216 	if (imp > TIPC_CRITICAL_IMPORTANCE)
217 		return -EINVAL;
218 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
219 	return 0;
220 }
221 
222 static bool tsk_conn_cong(struct tipc_sock *tsk)
223 {
224 	return tsk->snt_unacked > tsk->snd_win;
225 }
226 
227 static u16 tsk_blocks(int len)
228 {
229 	return ((len / FLOWCTL_BLK_SZ) + 1);
230 }
231 
232 /* tsk_blocks(): translate a buffer size in bytes to number of
233  * advertisable blocks, taking into account the ratio truesize(len)/len
234  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
235  */
236 static u16 tsk_adv_blocks(int len)
237 {
238 	return len / FLOWCTL_BLK_SZ / 4;
239 }
240 
241 /* tsk_inc(): increment counter for sent or received data
242  * - If block based flow control is not supported by peer we
243  *   fall back to message based ditto, incrementing the counter
244  */
245 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
246 {
247 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
248 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
249 	return 1;
250 }
251 
252 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
253  */
254 static void tsk_set_nagle(struct tipc_sock *tsk)
255 {
256 	struct sock *sk = &tsk->sk;
257 
258 	tsk->maxnagle = 0;
259 	if (sk->sk_type != SOCK_STREAM)
260 		return;
261 	if (tsk->nodelay)
262 		return;
263 	if (!(tsk->peer_caps & TIPC_NAGLE))
264 		return;
265 	/* Limit node local buffer size to avoid receive queue overflow */
266 	if (tsk->max_pkt == MAX_MSG_SIZE)
267 		tsk->maxnagle = 1500;
268 	else
269 		tsk->maxnagle = tsk->max_pkt;
270 }
271 
272 /**
273  * tsk_advance_rx_queue - discard first buffer in socket receive queue
274  * @sk: network socket
275  *
276  * Caller must hold socket lock
277  */
278 static void tsk_advance_rx_queue(struct sock *sk)
279 {
280 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
281 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
282 }
283 
284 /* tipc_sk_respond() : send response message back to sender
285  */
286 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
287 {
288 	u32 selector;
289 	u32 dnode;
290 	u32 onode = tipc_own_addr(sock_net(sk));
291 
292 	if (!tipc_msg_reverse(onode, &skb, err))
293 		return;
294 
295 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
296 	dnode = msg_destnode(buf_msg(skb));
297 	selector = msg_origport(buf_msg(skb));
298 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
299 }
300 
301 /**
302  * tsk_rej_rx_queue - reject all buffers in socket receive queue
303  * @sk: network socket
304  * @error: response error code
305  *
306  * Caller must hold socket lock
307  */
308 static void tsk_rej_rx_queue(struct sock *sk, int error)
309 {
310 	struct sk_buff *skb;
311 
312 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
313 		tipc_sk_respond(sk, skb, error);
314 }
315 
316 static bool tipc_sk_connected(struct sock *sk)
317 {
318 	return sk->sk_state == TIPC_ESTABLISHED;
319 }
320 
321 /* tipc_sk_type_connectionless - check if the socket is datagram socket
322  * @sk: socket
323  *
324  * Returns true if connection less, false otherwise
325  */
326 static bool tipc_sk_type_connectionless(struct sock *sk)
327 {
328 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
329 }
330 
331 /* tsk_peer_msg - verify if message was sent by connected port's peer
332  *
333  * Handles cases where the node's network address has changed from
334  * the default of <0.0.0> to its configured setting.
335  */
336 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
337 {
338 	struct sock *sk = &tsk->sk;
339 	u32 self = tipc_own_addr(sock_net(sk));
340 	u32 peer_port = tsk_peer_port(tsk);
341 	u32 orig_node, peer_node;
342 
343 	if (unlikely(!tipc_sk_connected(sk)))
344 		return false;
345 
346 	if (unlikely(msg_origport(msg) != peer_port))
347 		return false;
348 
349 	orig_node = msg_orignode(msg);
350 	peer_node = tsk_peer_node(tsk);
351 
352 	if (likely(orig_node == peer_node))
353 		return true;
354 
355 	if (!orig_node && peer_node == self)
356 		return true;
357 
358 	if (!peer_node && orig_node == self)
359 		return true;
360 
361 	return false;
362 }
363 
364 /* tipc_set_sk_state - set the sk_state of the socket
365  * @sk: socket
366  *
367  * Caller must hold socket lock
368  *
369  * Returns 0 on success, errno otherwise
370  */
371 static int tipc_set_sk_state(struct sock *sk, int state)
372 {
373 	int oldsk_state = sk->sk_state;
374 	int res = -EINVAL;
375 
376 	switch (state) {
377 	case TIPC_OPEN:
378 		res = 0;
379 		break;
380 	case TIPC_LISTEN:
381 	case TIPC_CONNECTING:
382 		if (oldsk_state == TIPC_OPEN)
383 			res = 0;
384 		break;
385 	case TIPC_ESTABLISHED:
386 		if (oldsk_state == TIPC_CONNECTING ||
387 		    oldsk_state == TIPC_OPEN)
388 			res = 0;
389 		break;
390 	case TIPC_DISCONNECTING:
391 		if (oldsk_state == TIPC_CONNECTING ||
392 		    oldsk_state == TIPC_ESTABLISHED)
393 			res = 0;
394 		break;
395 	}
396 
397 	if (!res)
398 		sk->sk_state = state;
399 
400 	return res;
401 }
402 
403 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
404 {
405 	struct sock *sk = sock->sk;
406 	int err = sock_error(sk);
407 	int typ = sock->type;
408 
409 	if (err)
410 		return err;
411 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
412 		if (sk->sk_state == TIPC_DISCONNECTING)
413 			return -EPIPE;
414 		else if (!tipc_sk_connected(sk))
415 			return -ENOTCONN;
416 	}
417 	if (!*timeout)
418 		return -EAGAIN;
419 	if (signal_pending(current))
420 		return sock_intr_errno(*timeout);
421 
422 	return 0;
423 }
424 
425 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
426 ({                                                                             \
427 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
428 	struct sock *sk_;						       \
429 	int rc_;							       \
430 									       \
431 	while ((rc_ = !(condition_))) {					       \
432 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
433 		smp_rmb();                                                     \
434 		sk_ = (sock_)->sk;					       \
435 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
436 		if (rc_)						       \
437 			break;						       \
438 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
439 		release_sock(sk_);					       \
440 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
441 		sched_annotate_sleep();				               \
442 		lock_sock(sk_);						       \
443 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
444 	}								       \
445 	rc_;								       \
446 })
447 
448 /**
449  * tipc_sk_create - create a TIPC socket
450  * @net: network namespace (must be default network)
451  * @sock: pre-allocated socket structure
452  * @protocol: protocol indicator (must be 0)
453  * @kern: caused by kernel or by userspace?
454  *
455  * This routine creates additional data structures used by the TIPC socket,
456  * initializes them, and links them together.
457  *
458  * Return: 0 on success, errno otherwise
459  */
460 static int tipc_sk_create(struct net *net, struct socket *sock,
461 			  int protocol, int kern)
462 {
463 	const struct proto_ops *ops;
464 	struct sock *sk;
465 	struct tipc_sock *tsk;
466 	struct tipc_msg *msg;
467 
468 	/* Validate arguments */
469 	if (unlikely(protocol != 0))
470 		return -EPROTONOSUPPORT;
471 
472 	switch (sock->type) {
473 	case SOCK_STREAM:
474 		ops = &stream_ops;
475 		break;
476 	case SOCK_SEQPACKET:
477 		ops = &packet_ops;
478 		break;
479 	case SOCK_DGRAM:
480 	case SOCK_RDM:
481 		ops = &msg_ops;
482 		break;
483 	default:
484 		return -EPROTOTYPE;
485 	}
486 
487 	/* Allocate socket's protocol area */
488 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
489 	if (sk == NULL)
490 		return -ENOMEM;
491 
492 	tsk = tipc_sk(sk);
493 	tsk->max_pkt = MAX_PKT_DEFAULT;
494 	tsk->maxnagle = 0;
495 	tsk->nagle_start = NAGLE_START_INIT;
496 	INIT_LIST_HEAD(&tsk->publications);
497 	INIT_LIST_HEAD(&tsk->cong_links);
498 	msg = &tsk->phdr;
499 
500 	/* Finish initializing socket data structures */
501 	sock->ops = ops;
502 	sock_init_data(sock, sk);
503 	tipc_set_sk_state(sk, TIPC_OPEN);
504 	if (tipc_sk_insert(tsk)) {
505 		pr_warn("Socket create failed; port number exhausted\n");
506 		return -EINVAL;
507 	}
508 
509 	/* Ensure tsk is visible before we read own_addr. */
510 	smp_mb();
511 
512 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
513 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
514 
515 	msg_set_origport(msg, tsk->portid);
516 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
517 	sk->sk_shutdown = 0;
518 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
519 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
520 	sk->sk_data_ready = tipc_data_ready;
521 	sk->sk_write_space = tipc_write_space;
522 	sk->sk_destruct = tipc_sock_destruct;
523 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
524 	tsk->group_is_open = true;
525 	atomic_set(&tsk->dupl_rcvcnt, 0);
526 
527 	/* Start out with safe limits until we receive an advertised window */
528 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
529 	tsk->rcv_win = tsk->snd_win;
530 
531 	if (tipc_sk_type_connectionless(sk)) {
532 		tsk_set_unreturnable(tsk, true);
533 		if (sock->type == SOCK_DGRAM)
534 			tsk_set_unreliable(tsk, true);
535 	}
536 	__skb_queue_head_init(&tsk->mc_method.deferredq);
537 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
538 	return 0;
539 }
540 
541 static void tipc_sk_callback(struct rcu_head *head)
542 {
543 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
544 
545 	sock_put(&tsk->sk);
546 }
547 
548 /* Caller should hold socket lock for the socket. */
549 static void __tipc_shutdown(struct socket *sock, int error)
550 {
551 	struct sock *sk = sock->sk;
552 	struct tipc_sock *tsk = tipc_sk(sk);
553 	struct net *net = sock_net(sk);
554 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
555 	u32 dnode = tsk_peer_node(tsk);
556 	struct sk_buff *skb;
557 
558 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
559 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
560 					    !tsk_conn_cong(tsk)));
561 
562 	/* Push out delayed messages if in Nagle mode */
563 	tipc_sk_push_backlog(tsk, false);
564 	/* Remove pending SYN */
565 	__skb_queue_purge(&sk->sk_write_queue);
566 
567 	/* Remove partially received buffer if any */
568 	skb = skb_peek(&sk->sk_receive_queue);
569 	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
570 		__skb_unlink(skb, &sk->sk_receive_queue);
571 		kfree_skb(skb);
572 	}
573 
574 	/* Reject all unreceived messages if connectionless */
575 	if (tipc_sk_type_connectionless(sk)) {
576 		tsk_rej_rx_queue(sk, error);
577 		return;
578 	}
579 
580 	switch (sk->sk_state) {
581 	case TIPC_CONNECTING:
582 	case TIPC_ESTABLISHED:
583 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
584 		tipc_node_remove_conn(net, dnode, tsk->portid);
585 		/* Send a FIN+/- to its peer */
586 		skb = __skb_dequeue(&sk->sk_receive_queue);
587 		if (skb) {
588 			__skb_queue_purge(&sk->sk_receive_queue);
589 			tipc_sk_respond(sk, skb, error);
590 			break;
591 		}
592 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
593 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
594 				      tsk_own_node(tsk), tsk_peer_port(tsk),
595 				      tsk->portid, error);
596 		if (skb)
597 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
598 		break;
599 	case TIPC_LISTEN:
600 		/* Reject all SYN messages */
601 		tsk_rej_rx_queue(sk, error);
602 		break;
603 	default:
604 		__skb_queue_purge(&sk->sk_receive_queue);
605 		break;
606 	}
607 }
608 
609 /**
610  * tipc_release - destroy a TIPC socket
611  * @sock: socket to destroy
612  *
613  * This routine cleans up any messages that are still queued on the socket.
614  * For DGRAM and RDM socket types, all queued messages are rejected.
615  * For SEQPACKET and STREAM socket types, the first message is rejected
616  * and any others are discarded.  (If the first message on a STREAM socket
617  * is partially-read, it is discarded and the next one is rejected instead.)
618  *
619  * NOTE: Rejected messages are not necessarily returned to the sender!  They
620  * are returned or discarded according to the "destination droppable" setting
621  * specified for the message by the sender.
622  *
623  * Return: 0 on success, errno otherwise
624  */
625 static int tipc_release(struct socket *sock)
626 {
627 	struct sock *sk = sock->sk;
628 	struct tipc_sock *tsk;
629 
630 	/*
631 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
632 	 * releases a pre-allocated child socket that was never used)
633 	 */
634 	if (sk == NULL)
635 		return 0;
636 
637 	tsk = tipc_sk(sk);
638 	lock_sock(sk);
639 
640 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
641 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
642 	sk->sk_shutdown = SHUTDOWN_MASK;
643 	tipc_sk_leave(tsk);
644 	tipc_sk_withdraw(tsk, NULL);
645 	__skb_queue_purge(&tsk->mc_method.deferredq);
646 	sk_stop_timer(sk, &sk->sk_timer);
647 	tipc_sk_remove(tsk);
648 
649 	sock_orphan(sk);
650 	/* Reject any messages that accumulated in backlog queue */
651 	release_sock(sk);
652 	tipc_dest_list_purge(&tsk->cong_links);
653 	tsk->cong_link_cnt = 0;
654 	call_rcu(&tsk->rcu, tipc_sk_callback);
655 	sock->sk = NULL;
656 
657 	return 0;
658 }
659 
660 /**
661  * __tipc_bind - associate or disassocate TIPC name(s) with a socket
662  * @sock: socket structure
663  * @skaddr: socket address describing name(s) and desired operation
664  * @alen: size of socket address data structure
665  *
666  * Name and name sequence binding are indicated using a positive scope value;
667  * a negative scope value unbinds the specified name.  Specifying no name
668  * (i.e. a socket address length of 0) unbinds all names from the socket.
669  *
670  * Return: 0 on success, errno otherwise
671  *
672  * NOTE: This routine doesn't need to take the socket lock since it doesn't
673  *       access any non-constant socket information.
674  */
675 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
676 {
677 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
678 	struct tipc_sock *tsk = tipc_sk(sock->sk);
679 	bool unbind = false;
680 
681 	if (unlikely(!alen))
682 		return tipc_sk_withdraw(tsk, NULL);
683 
684 	if (ua->addrtype == TIPC_SERVICE_ADDR) {
685 		ua->addrtype = TIPC_SERVICE_RANGE;
686 		ua->sr.upper = ua->sr.lower;
687 	}
688 	if (ua->scope < 0) {
689 		unbind = true;
690 		ua->scope = -ua->scope;
691 	}
692 	/* Users may still use deprecated TIPC_ZONE_SCOPE */
693 	if (ua->scope != TIPC_NODE_SCOPE)
694 		ua->scope = TIPC_CLUSTER_SCOPE;
695 
696 	if (tsk->group)
697 		return -EACCES;
698 
699 	if (unbind)
700 		return tipc_sk_withdraw(tsk, ua);
701 	return tipc_sk_publish(tsk, ua);
702 }
703 
704 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
705 {
706 	int res;
707 
708 	lock_sock(sock->sk);
709 	res = __tipc_bind(sock, skaddr, alen);
710 	release_sock(sock->sk);
711 	return res;
712 }
713 
714 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
715 {
716 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
717 	u32 atype = ua->addrtype;
718 
719 	if (alen) {
720 		if (!tipc_uaddr_valid(ua, alen))
721 			return -EINVAL;
722 		if (atype == TIPC_SOCKET_ADDR)
723 			return -EAFNOSUPPORT;
724 		if (ua->sr.type < TIPC_RESERVED_TYPES) {
725 			pr_warn_once("Can't bind to reserved service type %u\n",
726 				     ua->sr.type);
727 			return -EACCES;
728 		}
729 	}
730 	return tipc_sk_bind(sock, skaddr, alen);
731 }
732 
733 /**
734  * tipc_getname - get port ID of socket or peer socket
735  * @sock: socket structure
736  * @uaddr: area for returned socket address
737  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
738  *
739  * Return: 0 on success, errno otherwise
740  *
741  * NOTE: This routine doesn't need to take the socket lock since it only
742  *       accesses socket information that is unchanging (or which changes in
743  *       a completely predictable manner).
744  */
745 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
746 			int peer)
747 {
748 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
749 	struct sock *sk = sock->sk;
750 	struct tipc_sock *tsk = tipc_sk(sk);
751 
752 	memset(addr, 0, sizeof(*addr));
753 	if (peer) {
754 		if ((!tipc_sk_connected(sk)) &&
755 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
756 			return -ENOTCONN;
757 		addr->addr.id.ref = tsk_peer_port(tsk);
758 		addr->addr.id.node = tsk_peer_node(tsk);
759 	} else {
760 		addr->addr.id.ref = tsk->portid;
761 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
762 	}
763 
764 	addr->addrtype = TIPC_SOCKET_ADDR;
765 	addr->family = AF_TIPC;
766 	addr->scope = 0;
767 	addr->addr.name.domain = 0;
768 
769 	return sizeof(*addr);
770 }
771 
772 /**
773  * tipc_poll - read and possibly block on pollmask
774  * @file: file structure associated with the socket
775  * @sock: socket for which to calculate the poll bits
776  * @wait: ???
777  *
778  * Return: pollmask value
779  *
780  * COMMENTARY:
781  * It appears that the usual socket locking mechanisms are not useful here
782  * since the pollmask info is potentially out-of-date the moment this routine
783  * exits.  TCP and other protocols seem to rely on higher level poll routines
784  * to handle any preventable race conditions, so TIPC will do the same ...
785  *
786  * IMPORTANT: The fact that a read or write operation is indicated does NOT
787  * imply that the operation will succeed, merely that it should be performed
788  * and will not block.
789  */
790 static __poll_t tipc_poll(struct file *file, struct socket *sock,
791 			      poll_table *wait)
792 {
793 	struct sock *sk = sock->sk;
794 	struct tipc_sock *tsk = tipc_sk(sk);
795 	__poll_t revents = 0;
796 
797 	sock_poll_wait(file, sock, wait);
798 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
799 
800 	if (sk->sk_shutdown & RCV_SHUTDOWN)
801 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
802 	if (sk->sk_shutdown == SHUTDOWN_MASK)
803 		revents |= EPOLLHUP;
804 
805 	switch (sk->sk_state) {
806 	case TIPC_ESTABLISHED:
807 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
808 			revents |= EPOLLOUT;
809 		fallthrough;
810 	case TIPC_LISTEN:
811 	case TIPC_CONNECTING:
812 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
813 			revents |= EPOLLIN | EPOLLRDNORM;
814 		break;
815 	case TIPC_OPEN:
816 		if (tsk->group_is_open && !tsk->cong_link_cnt)
817 			revents |= EPOLLOUT;
818 		if (!tipc_sk_type_connectionless(sk))
819 			break;
820 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
821 			break;
822 		revents |= EPOLLIN | EPOLLRDNORM;
823 		break;
824 	case TIPC_DISCONNECTING:
825 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
826 		break;
827 	}
828 	return revents;
829 }
830 
831 /**
832  * tipc_sendmcast - send multicast message
833  * @sock: socket structure
834  * @ua: destination address struct
835  * @msg: message to send
836  * @dlen: length of data to send
837  * @timeout: timeout to wait for wakeup
838  *
839  * Called from function tipc_sendmsg(), which has done all sanity checks
840  * Return: the number of bytes sent on success, or errno
841  */
842 static int tipc_sendmcast(struct  socket *sock, struct tipc_uaddr *ua,
843 			  struct msghdr *msg, size_t dlen, long timeout)
844 {
845 	struct sock *sk = sock->sk;
846 	struct tipc_sock *tsk = tipc_sk(sk);
847 	struct tipc_msg *hdr = &tsk->phdr;
848 	struct net *net = sock_net(sk);
849 	int mtu = tipc_bcast_get_mtu(net);
850 	struct sk_buff_head pkts;
851 	struct tipc_nlist dsts;
852 	int rc;
853 
854 	if (tsk->group)
855 		return -EACCES;
856 
857 	/* Block or return if any destination link is congested */
858 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
859 	if (unlikely(rc))
860 		return rc;
861 
862 	/* Lookup destination nodes */
863 	tipc_nlist_init(&dsts, tipc_own_addr(net));
864 	tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
865 	if (!dsts.local && !dsts.remote)
866 		return -EHOSTUNREACH;
867 
868 	/* Build message header */
869 	msg_set_type(hdr, TIPC_MCAST_MSG);
870 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
871 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
872 	msg_set_destport(hdr, 0);
873 	msg_set_destnode(hdr, 0);
874 	msg_set_nametype(hdr, ua->sr.type);
875 	msg_set_namelower(hdr, ua->sr.lower);
876 	msg_set_nameupper(hdr, ua->sr.upper);
877 
878 	/* Build message as chain of buffers */
879 	__skb_queue_head_init(&pkts);
880 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
881 
882 	/* Send message if build was successful */
883 	if (unlikely(rc == dlen)) {
884 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
885 					TIPC_DUMP_SK_SNDQ, " ");
886 		rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
887 				     &tsk->cong_link_cnt);
888 	}
889 
890 	tipc_nlist_purge(&dsts);
891 
892 	return rc ? rc : dlen;
893 }
894 
895 /**
896  * tipc_send_group_msg - send a message to a member in the group
897  * @net: network namespace
898  * @tsk: tipc socket
899  * @m: message to send
900  * @mb: group member
901  * @dnode: destination node
902  * @dport: destination port
903  * @dlen: total length of message data
904  */
905 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
906 			       struct msghdr *m, struct tipc_member *mb,
907 			       u32 dnode, u32 dport, int dlen)
908 {
909 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
910 	struct tipc_mc_method *method = &tsk->mc_method;
911 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
912 	struct tipc_msg *hdr = &tsk->phdr;
913 	struct sk_buff_head pkts;
914 	int mtu, rc;
915 
916 	/* Complete message header */
917 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
918 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
919 	msg_set_destport(hdr, dport);
920 	msg_set_destnode(hdr, dnode);
921 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
922 
923 	/* Build message as chain of buffers */
924 	__skb_queue_head_init(&pkts);
925 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
926 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
927 	if (unlikely(rc != dlen))
928 		return rc;
929 
930 	/* Send message */
931 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
932 	if (unlikely(rc == -ELINKCONG)) {
933 		tipc_dest_push(&tsk->cong_links, dnode, 0);
934 		tsk->cong_link_cnt++;
935 	}
936 
937 	/* Update send window */
938 	tipc_group_update_member(mb, blks);
939 
940 	/* A broadcast sent within next EXPIRE period must follow same path */
941 	method->rcast = true;
942 	method->mandatory = true;
943 	return dlen;
944 }
945 
946 /**
947  * tipc_send_group_unicast - send message to a member in the group
948  * @sock: socket structure
949  * @m: message to send
950  * @dlen: total length of message data
951  * @timeout: timeout to wait for wakeup
952  *
953  * Called from function tipc_sendmsg(), which has done all sanity checks
954  * Return: the number of bytes sent on success, or errno
955  */
956 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
957 				   int dlen, long timeout)
958 {
959 	struct sock *sk = sock->sk;
960 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
961 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
962 	struct tipc_sock *tsk = tipc_sk(sk);
963 	struct net *net = sock_net(sk);
964 	struct tipc_member *mb = NULL;
965 	u32 node, port;
966 	int rc;
967 
968 	node = ua->sk.node;
969 	port = ua->sk.ref;
970 	if (!port && !node)
971 		return -EHOSTUNREACH;
972 
973 	/* Block or return if destination link or member is congested */
974 	rc = tipc_wait_for_cond(sock, &timeout,
975 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
976 				tsk->group &&
977 				!tipc_group_cong(tsk->group, node, port, blks,
978 						 &mb));
979 	if (unlikely(rc))
980 		return rc;
981 
982 	if (unlikely(!mb))
983 		return -EHOSTUNREACH;
984 
985 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
986 
987 	return rc ? rc : dlen;
988 }
989 
990 /**
991  * tipc_send_group_anycast - send message to any member with given identity
992  * @sock: socket structure
993  * @m: message to send
994  * @dlen: total length of message data
995  * @timeout: timeout to wait for wakeup
996  *
997  * Called from function tipc_sendmsg(), which has done all sanity checks
998  * Return: the number of bytes sent on success, or errno
999  */
1000 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1001 				   int dlen, long timeout)
1002 {
1003 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1004 	struct sock *sk = sock->sk;
1005 	struct tipc_sock *tsk = tipc_sk(sk);
1006 	struct list_head *cong_links = &tsk->cong_links;
1007 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1008 	struct tipc_msg *hdr = &tsk->phdr;
1009 	struct tipc_member *first = NULL;
1010 	struct tipc_member *mbr = NULL;
1011 	struct net *net = sock_net(sk);
1012 	u32 node, port, exclude;
1013 	struct list_head dsts;
1014 	int lookups = 0;
1015 	int dstcnt, rc;
1016 	bool cong;
1017 
1018 	INIT_LIST_HEAD(&dsts);
1019 	ua->sa.type = msg_nametype(hdr);
1020 	ua->scope = msg_lookup_scope(hdr);
1021 
1022 	while (++lookups < 4) {
1023 		exclude = tipc_group_exclude(tsk->group);
1024 
1025 		first = NULL;
1026 
1027 		/* Look for a non-congested destination member, if any */
1028 		while (1) {
1029 			if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1030 						       exclude, false))
1031 				return -EHOSTUNREACH;
1032 			tipc_dest_pop(&dsts, &node, &port);
1033 			cong = tipc_group_cong(tsk->group, node, port, blks,
1034 					       &mbr);
1035 			if (!cong)
1036 				break;
1037 			if (mbr == first)
1038 				break;
1039 			if (!first)
1040 				first = mbr;
1041 		}
1042 
1043 		/* Start over if destination was not in member list */
1044 		if (unlikely(!mbr))
1045 			continue;
1046 
1047 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1048 			break;
1049 
1050 		/* Block or return if destination link or member is congested */
1051 		rc = tipc_wait_for_cond(sock, &timeout,
1052 					!tipc_dest_find(cong_links, node, 0) &&
1053 					tsk->group &&
1054 					!tipc_group_cong(tsk->group, node, port,
1055 							 blks, &mbr));
1056 		if (unlikely(rc))
1057 			return rc;
1058 
1059 		/* Send, unless destination disappeared while waiting */
1060 		if (likely(mbr))
1061 			break;
1062 	}
1063 
1064 	if (unlikely(lookups >= 4))
1065 		return -EHOSTUNREACH;
1066 
1067 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1068 
1069 	return rc ? rc : dlen;
1070 }
1071 
1072 /**
1073  * tipc_send_group_bcast - send message to all members in communication group
1074  * @sock: socket structure
1075  * @m: message to send
1076  * @dlen: total length of message data
1077  * @timeout: timeout to wait for wakeup
1078  *
1079  * Called from function tipc_sendmsg(), which has done all sanity checks
1080  * Return: the number of bytes sent on success, or errno
1081  */
1082 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1083 				 int dlen, long timeout)
1084 {
1085 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1086 	struct sock *sk = sock->sk;
1087 	struct net *net = sock_net(sk);
1088 	struct tipc_sock *tsk = tipc_sk(sk);
1089 	struct tipc_nlist *dsts;
1090 	struct tipc_mc_method *method = &tsk->mc_method;
1091 	bool ack = method->mandatory && method->rcast;
1092 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1093 	struct tipc_msg *hdr = &tsk->phdr;
1094 	int mtu = tipc_bcast_get_mtu(net);
1095 	struct sk_buff_head pkts;
1096 	int rc = -EHOSTUNREACH;
1097 
1098 	/* Block or return if any destination link or member is congested */
1099 	rc = tipc_wait_for_cond(sock, &timeout,
1100 				!tsk->cong_link_cnt && tsk->group &&
1101 				!tipc_group_bc_cong(tsk->group, blks));
1102 	if (unlikely(rc))
1103 		return rc;
1104 
1105 	dsts = tipc_group_dests(tsk->group);
1106 	if (!dsts->local && !dsts->remote)
1107 		return -EHOSTUNREACH;
1108 
1109 	/* Complete message header */
1110 	if (ua) {
1111 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1112 		msg_set_nameinst(hdr, ua->sa.instance);
1113 	} else {
1114 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1115 		msg_set_nameinst(hdr, 0);
1116 	}
1117 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1118 	msg_set_destport(hdr, 0);
1119 	msg_set_destnode(hdr, 0);
1120 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1121 
1122 	/* Avoid getting stuck with repeated forced replicasts */
1123 	msg_set_grp_bc_ack_req(hdr, ack);
1124 
1125 	/* Build message as chain of buffers */
1126 	__skb_queue_head_init(&pkts);
1127 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1128 	if (unlikely(rc != dlen))
1129 		return rc;
1130 
1131 	/* Send message */
1132 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1133 	if (unlikely(rc))
1134 		return rc;
1135 
1136 	/* Update broadcast sequence number and send windows */
1137 	tipc_group_update_bc_members(tsk->group, blks, ack);
1138 
1139 	/* Broadcast link is now free to choose method for next broadcast */
1140 	method->mandatory = false;
1141 	method->expires = jiffies;
1142 
1143 	return dlen;
1144 }
1145 
1146 /**
1147  * tipc_send_group_mcast - send message to all members with given identity
1148  * @sock: socket structure
1149  * @m: message to send
1150  * @dlen: total length of message data
1151  * @timeout: timeout to wait for wakeup
1152  *
1153  * Called from function tipc_sendmsg(), which has done all sanity checks
1154  * Return: the number of bytes sent on success, or errno
1155  */
1156 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1157 				 int dlen, long timeout)
1158 {
1159 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1160 	struct sock *sk = sock->sk;
1161 	struct tipc_sock *tsk = tipc_sk(sk);
1162 	struct tipc_group *grp = tsk->group;
1163 	struct tipc_msg *hdr = &tsk->phdr;
1164 	struct net *net = sock_net(sk);
1165 	struct list_head dsts;
1166 	u32 dstcnt, exclude;
1167 
1168 	INIT_LIST_HEAD(&dsts);
1169 	ua->sa.type = msg_nametype(hdr);
1170 	ua->scope = msg_lookup_scope(hdr);
1171 	exclude = tipc_group_exclude(grp);
1172 
1173 	if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1174 		return -EHOSTUNREACH;
1175 
1176 	if (dstcnt == 1) {
1177 		tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1178 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1179 	}
1180 
1181 	tipc_dest_list_purge(&dsts);
1182 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1183 }
1184 
1185 /**
1186  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1187  * @net: the associated network namespace
1188  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1189  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1190  *
1191  * Multi-threaded: parallel calls with reference to same queues may occur
1192  */
1193 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1194 		       struct sk_buff_head *inputq)
1195 {
1196 	u32 self = tipc_own_addr(net);
1197 	struct sk_buff *skb, *_skb;
1198 	u32 portid, onode;
1199 	struct sk_buff_head tmpq;
1200 	struct list_head dports;
1201 	struct tipc_msg *hdr;
1202 	struct tipc_uaddr ua;
1203 	int user, mtyp, hlen;
1204 
1205 	__skb_queue_head_init(&tmpq);
1206 	INIT_LIST_HEAD(&dports);
1207 	ua.addrtype = TIPC_SERVICE_RANGE;
1208 
1209 	/* tipc_skb_peek() increments the head skb's reference counter */
1210 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1211 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1212 		hdr = buf_msg(skb);
1213 		user = msg_user(hdr);
1214 		mtyp = msg_type(hdr);
1215 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1216 		onode = msg_orignode(hdr);
1217 		ua.sr.type = msg_nametype(hdr);
1218 		ua.sr.lower = msg_namelower(hdr);
1219 		ua.sr.upper = msg_nameupper(hdr);
1220 		if (onode == self)
1221 			ua.scope = TIPC_ANY_SCOPE;
1222 		else
1223 			ua.scope = TIPC_CLUSTER_SCOPE;
1224 
1225 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1226 			spin_lock_bh(&inputq->lock);
1227 			if (skb_peek(arrvq) == skb) {
1228 				__skb_dequeue(arrvq);
1229 				__skb_queue_tail(inputq, skb);
1230 			}
1231 			kfree_skb(skb);
1232 			spin_unlock_bh(&inputq->lock);
1233 			continue;
1234 		}
1235 
1236 		/* Group messages require exact scope match */
1237 		if (msg_in_group(hdr)) {
1238 			ua.sr.lower = 0;
1239 			ua.sr.upper = ~0;
1240 			ua.scope = msg_lookup_scope(hdr);
1241 		}
1242 
1243 		/* Create destination port list: */
1244 		tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1245 
1246 		/* Clone message per destination */
1247 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1248 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1249 			if (_skb) {
1250 				msg_set_destport(buf_msg(_skb), portid);
1251 				__skb_queue_tail(&tmpq, _skb);
1252 				continue;
1253 			}
1254 			pr_warn("Failed to clone mcast rcv buffer\n");
1255 		}
1256 		/* Append clones to inputq only if skb is still head of arrvq */
1257 		spin_lock_bh(&inputq->lock);
1258 		if (skb_peek(arrvq) == skb) {
1259 			skb_queue_splice_tail_init(&tmpq, inputq);
1260 			/* Decrement the skb's refcnt */
1261 			kfree_skb(__skb_dequeue(arrvq));
1262 		}
1263 		spin_unlock_bh(&inputq->lock);
1264 		__skb_queue_purge(&tmpq);
1265 		kfree_skb(skb);
1266 	}
1267 	tipc_sk_rcv(net, inputq);
1268 }
1269 
1270 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1271  *                         when socket is in Nagle mode
1272  */
1273 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1274 {
1275 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1276 	struct sk_buff *skb = skb_peek_tail(txq);
1277 	struct net *net = sock_net(&tsk->sk);
1278 	u32 dnode = tsk_peer_node(tsk);
1279 	int rc;
1280 
1281 	if (nagle_ack) {
1282 		tsk->pkt_cnt += skb_queue_len(txq);
1283 		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1284 			tsk->oneway = 0;
1285 			if (tsk->nagle_start < NAGLE_START_MAX)
1286 				tsk->nagle_start *= 2;
1287 			tsk->expect_ack = false;
1288 			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1289 				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1290 				 tsk->nagle_start);
1291 		} else {
1292 			tsk->nagle_start = NAGLE_START_INIT;
1293 			if (skb) {
1294 				msg_set_ack_required(buf_msg(skb));
1295 				tsk->expect_ack = true;
1296 			} else {
1297 				tsk->expect_ack = false;
1298 			}
1299 		}
1300 		tsk->msg_acc = 0;
1301 		tsk->pkt_cnt = 0;
1302 	}
1303 
1304 	if (!skb || tsk->cong_link_cnt)
1305 		return;
1306 
1307 	/* Do not send SYN again after congestion */
1308 	if (msg_is_syn(buf_msg(skb)))
1309 		return;
1310 
1311 	if (tsk->msg_acc)
1312 		tsk->pkt_cnt += skb_queue_len(txq);
1313 	tsk->snt_unacked += tsk->snd_backlog;
1314 	tsk->snd_backlog = 0;
1315 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1316 	if (rc == -ELINKCONG)
1317 		tsk->cong_link_cnt = 1;
1318 }
1319 
1320 /**
1321  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1322  * @tsk: receiving socket
1323  * @skb: pointer to message buffer.
1324  * @inputq: buffer list containing the buffers
1325  * @xmitq: output message area
1326  */
1327 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1328 				   struct sk_buff_head *inputq,
1329 				   struct sk_buff_head *xmitq)
1330 {
1331 	struct tipc_msg *hdr = buf_msg(skb);
1332 	u32 onode = tsk_own_node(tsk);
1333 	struct sock *sk = &tsk->sk;
1334 	int mtyp = msg_type(hdr);
1335 	bool was_cong;
1336 
1337 	/* Ignore if connection cannot be validated: */
1338 	if (!tsk_peer_msg(tsk, hdr)) {
1339 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1340 		goto exit;
1341 	}
1342 
1343 	if (unlikely(msg_errcode(hdr))) {
1344 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1345 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1346 				      tsk_peer_port(tsk));
1347 		sk->sk_state_change(sk);
1348 
1349 		/* State change is ignored if socket already awake,
1350 		 * - convert msg to abort msg and add to inqueue
1351 		 */
1352 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1353 		msg_set_type(hdr, TIPC_CONN_MSG);
1354 		msg_set_size(hdr, BASIC_H_SIZE);
1355 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1356 		__skb_queue_tail(inputq, skb);
1357 		return;
1358 	}
1359 
1360 	tsk->probe_unacked = false;
1361 
1362 	if (mtyp == CONN_PROBE) {
1363 		msg_set_type(hdr, CONN_PROBE_REPLY);
1364 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1365 			__skb_queue_tail(xmitq, skb);
1366 		return;
1367 	} else if (mtyp == CONN_ACK) {
1368 		was_cong = tsk_conn_cong(tsk);
1369 		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1370 		tsk->snt_unacked -= msg_conn_ack(hdr);
1371 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1372 			tsk->snd_win = msg_adv_win(hdr);
1373 		if (was_cong && !tsk_conn_cong(tsk))
1374 			sk->sk_write_space(sk);
1375 	} else if (mtyp != CONN_PROBE_REPLY) {
1376 		pr_warn("Received unknown CONN_PROTO msg\n");
1377 	}
1378 exit:
1379 	kfree_skb(skb);
1380 }
1381 
1382 /**
1383  * tipc_sendmsg - send message in connectionless manner
1384  * @sock: socket structure
1385  * @m: message to send
1386  * @dsz: amount of user data to be sent
1387  *
1388  * Message must have an destination specified explicitly.
1389  * Used for SOCK_RDM and SOCK_DGRAM messages,
1390  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1391  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1392  *
1393  * Return: the number of bytes sent on success, or errno otherwise
1394  */
1395 static int tipc_sendmsg(struct socket *sock,
1396 			struct msghdr *m, size_t dsz)
1397 {
1398 	struct sock *sk = sock->sk;
1399 	int ret;
1400 
1401 	lock_sock(sk);
1402 	ret = __tipc_sendmsg(sock, m, dsz);
1403 	release_sock(sk);
1404 
1405 	return ret;
1406 }
1407 
1408 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1409 {
1410 	struct sock *sk = sock->sk;
1411 	struct net *net = sock_net(sk);
1412 	struct tipc_sock *tsk = tipc_sk(sk);
1413 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1414 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1415 	struct list_head *clinks = &tsk->cong_links;
1416 	bool syn = !tipc_sk_type_connectionless(sk);
1417 	struct tipc_group *grp = tsk->group;
1418 	struct tipc_msg *hdr = &tsk->phdr;
1419 	struct tipc_socket_addr skaddr;
1420 	struct sk_buff_head pkts;
1421 	int atype, mtu, rc;
1422 
1423 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1424 		return -EMSGSIZE;
1425 
1426 	if (ua) {
1427 		if (!tipc_uaddr_valid(ua, m->msg_namelen))
1428 			return -EINVAL;
1429 		atype = ua->addrtype;
1430 	}
1431 
1432 	/* If socket belongs to a communication group follow other paths */
1433 	if (grp) {
1434 		if (!ua)
1435 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1436 		if (atype == TIPC_SERVICE_ADDR)
1437 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1438 		if (atype == TIPC_SOCKET_ADDR)
1439 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1440 		if (atype == TIPC_SERVICE_RANGE)
1441 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1442 		return -EINVAL;
1443 	}
1444 
1445 	if (!ua) {
1446 		ua = (struct tipc_uaddr *)&tsk->peer;
1447 		if (!syn && ua->family != AF_TIPC)
1448 			return -EDESTADDRREQ;
1449 		atype = ua->addrtype;
1450 	}
1451 
1452 	if (unlikely(syn)) {
1453 		if (sk->sk_state == TIPC_LISTEN)
1454 			return -EPIPE;
1455 		if (sk->sk_state != TIPC_OPEN)
1456 			return -EISCONN;
1457 		if (tsk->published)
1458 			return -EOPNOTSUPP;
1459 		if (atype == TIPC_SERVICE_ADDR)
1460 			tsk->conn_addrtype = atype;
1461 		msg_set_syn(hdr, 1);
1462 	}
1463 
1464 	memset(&skaddr, 0, sizeof(skaddr));
1465 
1466 	/* Determine destination */
1467 	if (atype == TIPC_SERVICE_RANGE) {
1468 		return tipc_sendmcast(sock, ua, m, dlen, timeout);
1469 	} else if (atype == TIPC_SERVICE_ADDR) {
1470 		skaddr.node = ua->lookup_node;
1471 		ua->scope = tipc_node2scope(skaddr.node);
1472 		if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1473 			return -EHOSTUNREACH;
1474 	} else if (atype == TIPC_SOCKET_ADDR) {
1475 		skaddr = ua->sk;
1476 	} else {
1477 		return -EINVAL;
1478 	}
1479 
1480 	/* Block or return if destination link is congested */
1481 	rc = tipc_wait_for_cond(sock, &timeout,
1482 				!tipc_dest_find(clinks, skaddr.node, 0));
1483 	if (unlikely(rc))
1484 		return rc;
1485 
1486 	/* Finally build message header */
1487 	msg_set_destnode(hdr, skaddr.node);
1488 	msg_set_destport(hdr, skaddr.ref);
1489 	if (atype == TIPC_SERVICE_ADDR) {
1490 		msg_set_type(hdr, TIPC_NAMED_MSG);
1491 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1492 		msg_set_nametype(hdr, ua->sa.type);
1493 		msg_set_nameinst(hdr, ua->sa.instance);
1494 		msg_set_lookup_scope(hdr, ua->scope);
1495 	} else { /* TIPC_SOCKET_ADDR */
1496 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1497 		msg_set_lookup_scope(hdr, 0);
1498 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1499 	}
1500 
1501 	/* Add message body */
1502 	__skb_queue_head_init(&pkts);
1503 	mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1504 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1505 	if (unlikely(rc != dlen))
1506 		return rc;
1507 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1508 		__skb_queue_purge(&pkts);
1509 		return -ENOMEM;
1510 	}
1511 
1512 	/* Send message */
1513 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1514 	rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1515 	if (unlikely(rc == -ELINKCONG)) {
1516 		tipc_dest_push(clinks, skaddr.node, 0);
1517 		tsk->cong_link_cnt++;
1518 		rc = 0;
1519 	}
1520 
1521 	if (unlikely(syn && !rc)) {
1522 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1523 		if (dlen && timeout) {
1524 			timeout = msecs_to_jiffies(timeout);
1525 			tipc_wait_for_connect(sock, &timeout);
1526 		}
1527 	}
1528 
1529 	return rc ? rc : dlen;
1530 }
1531 
1532 /**
1533  * tipc_sendstream - send stream-oriented data
1534  * @sock: socket structure
1535  * @m: data to send
1536  * @dsz: total length of data to be transmitted
1537  *
1538  * Used for SOCK_STREAM data.
1539  *
1540  * Return: the number of bytes sent on success (or partial success),
1541  * or errno if no data sent
1542  */
1543 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1544 {
1545 	struct sock *sk = sock->sk;
1546 	int ret;
1547 
1548 	lock_sock(sk);
1549 	ret = __tipc_sendstream(sock, m, dsz);
1550 	release_sock(sk);
1551 
1552 	return ret;
1553 }
1554 
1555 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1556 {
1557 	struct sock *sk = sock->sk;
1558 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1559 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1560 	struct sk_buff_head *txq = &sk->sk_write_queue;
1561 	struct tipc_sock *tsk = tipc_sk(sk);
1562 	struct tipc_msg *hdr = &tsk->phdr;
1563 	struct net *net = sock_net(sk);
1564 	struct sk_buff *skb;
1565 	u32 dnode = tsk_peer_node(tsk);
1566 	int maxnagle = tsk->maxnagle;
1567 	int maxpkt = tsk->max_pkt;
1568 	int send, sent = 0;
1569 	int blocks, rc = 0;
1570 
1571 	if (unlikely(dlen > INT_MAX))
1572 		return -EMSGSIZE;
1573 
1574 	/* Handle implicit connection setup */
1575 	if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1576 		rc = __tipc_sendmsg(sock, m, dlen);
1577 		if (dlen && dlen == rc) {
1578 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1579 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1580 		}
1581 		return rc;
1582 	}
1583 
1584 	do {
1585 		rc = tipc_wait_for_cond(sock, &timeout,
1586 					(!tsk->cong_link_cnt &&
1587 					 !tsk_conn_cong(tsk) &&
1588 					 tipc_sk_connected(sk)));
1589 		if (unlikely(rc))
1590 			break;
1591 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1592 		blocks = tsk->snd_backlog;
1593 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1594 		    send <= maxnagle) {
1595 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1596 			if (unlikely(rc < 0))
1597 				break;
1598 			blocks += rc;
1599 			tsk->msg_acc++;
1600 			if (blocks <= 64 && tsk->expect_ack) {
1601 				tsk->snd_backlog = blocks;
1602 				sent += send;
1603 				break;
1604 			} else if (blocks > 64) {
1605 				tsk->pkt_cnt += skb_queue_len(txq);
1606 			} else {
1607 				skb = skb_peek_tail(txq);
1608 				if (skb) {
1609 					msg_set_ack_required(buf_msg(skb));
1610 					tsk->expect_ack = true;
1611 				} else {
1612 					tsk->expect_ack = false;
1613 				}
1614 				tsk->msg_acc = 0;
1615 				tsk->pkt_cnt = 0;
1616 			}
1617 		} else {
1618 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1619 			if (unlikely(rc != send))
1620 				break;
1621 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1622 		}
1623 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1624 					 TIPC_DUMP_SK_SNDQ, " ");
1625 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1626 		if (unlikely(rc == -ELINKCONG)) {
1627 			tsk->cong_link_cnt = 1;
1628 			rc = 0;
1629 		}
1630 		if (likely(!rc)) {
1631 			tsk->snt_unacked += blocks;
1632 			tsk->snd_backlog = 0;
1633 			sent += send;
1634 		}
1635 	} while (sent < dlen && !rc);
1636 
1637 	return sent ? sent : rc;
1638 }
1639 
1640 /**
1641  * tipc_send_packet - send a connection-oriented message
1642  * @sock: socket structure
1643  * @m: message to send
1644  * @dsz: length of data to be transmitted
1645  *
1646  * Used for SOCK_SEQPACKET messages.
1647  *
1648  * Return: the number of bytes sent on success, or errno otherwise
1649  */
1650 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1651 {
1652 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1653 		return -EMSGSIZE;
1654 
1655 	return tipc_sendstream(sock, m, dsz);
1656 }
1657 
1658 /* tipc_sk_finish_conn - complete the setup of a connection
1659  */
1660 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1661 				u32 peer_node)
1662 {
1663 	struct sock *sk = &tsk->sk;
1664 	struct net *net = sock_net(sk);
1665 	struct tipc_msg *msg = &tsk->phdr;
1666 
1667 	msg_set_syn(msg, 0);
1668 	msg_set_destnode(msg, peer_node);
1669 	msg_set_destport(msg, peer_port);
1670 	msg_set_type(msg, TIPC_CONN_MSG);
1671 	msg_set_lookup_scope(msg, 0);
1672 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1673 
1674 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1675 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1676 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1677 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1678 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1679 	tsk_set_nagle(tsk);
1680 	__skb_queue_purge(&sk->sk_write_queue);
1681 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1682 		return;
1683 
1684 	/* Fall back to message based flow control */
1685 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1686 	tsk->snd_win = FLOWCTL_MSG_WIN;
1687 }
1688 
1689 /**
1690  * tipc_sk_set_orig_addr - capture sender's address for received message
1691  * @m: descriptor for message info
1692  * @skb: received message
1693  *
1694  * Note: Address is not captured if not requested by receiver.
1695  */
1696 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1697 {
1698 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1699 	struct tipc_msg *hdr = buf_msg(skb);
1700 
1701 	if (!srcaddr)
1702 		return;
1703 
1704 	srcaddr->sock.family = AF_TIPC;
1705 	srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1706 	srcaddr->sock.scope = 0;
1707 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1708 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1709 	srcaddr->sock.addr.name.domain = 0;
1710 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1711 
1712 	if (!msg_in_group(hdr))
1713 		return;
1714 
1715 	/* Group message users may also want to know sending member's id */
1716 	srcaddr->member.family = AF_TIPC;
1717 	srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1718 	srcaddr->member.scope = 0;
1719 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1720 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1721 	srcaddr->member.addr.name.domain = 0;
1722 	m->msg_namelen = sizeof(*srcaddr);
1723 }
1724 
1725 /**
1726  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1727  * @m: descriptor for message info
1728  * @skb: received message buffer
1729  * @tsk: TIPC port associated with message
1730  *
1731  * Note: Ancillary data is not captured if not requested by receiver.
1732  *
1733  * Return: 0 if successful, otherwise errno
1734  */
1735 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1736 				 struct tipc_sock *tsk)
1737 {
1738 	struct tipc_msg *hdr;
1739 	u32 data[3] = {0,};
1740 	bool has_addr;
1741 	int dlen, rc;
1742 
1743 	if (likely(m->msg_controllen == 0))
1744 		return 0;
1745 
1746 	hdr = buf_msg(skb);
1747 	dlen = msg_data_sz(hdr);
1748 
1749 	/* Capture errored message object, if any */
1750 	if (msg_errcode(hdr)) {
1751 		if (skb_linearize(skb))
1752 			return -ENOMEM;
1753 		hdr = buf_msg(skb);
1754 		data[0] = msg_errcode(hdr);
1755 		data[1] = dlen;
1756 		rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1757 		if (rc || !dlen)
1758 			return rc;
1759 		rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1760 		if (rc)
1761 			return rc;
1762 	}
1763 
1764 	/* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1765 	switch (msg_type(hdr)) {
1766 	case TIPC_NAMED_MSG:
1767 		has_addr = true;
1768 		data[0] = msg_nametype(hdr);
1769 		data[1] = msg_namelower(hdr);
1770 		data[2] = data[1];
1771 		break;
1772 	case TIPC_MCAST_MSG:
1773 		has_addr = true;
1774 		data[0] = msg_nametype(hdr);
1775 		data[1] = msg_namelower(hdr);
1776 		data[2] = msg_nameupper(hdr);
1777 		break;
1778 	case TIPC_CONN_MSG:
1779 		has_addr = !!tsk->conn_addrtype;
1780 		data[0] = msg_nametype(&tsk->phdr);
1781 		data[1] = msg_nameinst(&tsk->phdr);
1782 		data[2] = data[1];
1783 		break;
1784 	default:
1785 		has_addr = false;
1786 	}
1787 	if (!has_addr)
1788 		return 0;
1789 	return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1790 }
1791 
1792 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1793 {
1794 	struct sock *sk = &tsk->sk;
1795 	struct sk_buff *skb = NULL;
1796 	struct tipc_msg *msg;
1797 	u32 peer_port = tsk_peer_port(tsk);
1798 	u32 dnode = tsk_peer_node(tsk);
1799 
1800 	if (!tipc_sk_connected(sk))
1801 		return NULL;
1802 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1803 			      dnode, tsk_own_node(tsk), peer_port,
1804 			      tsk->portid, TIPC_OK);
1805 	if (!skb)
1806 		return NULL;
1807 	msg = buf_msg(skb);
1808 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1809 	tsk->rcv_unacked = 0;
1810 
1811 	/* Adjust to and advertize the correct window limit */
1812 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1813 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1814 		msg_set_adv_win(msg, tsk->rcv_win);
1815 	}
1816 	return skb;
1817 }
1818 
1819 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1820 {
1821 	struct sk_buff *skb;
1822 
1823 	skb = tipc_sk_build_ack(tsk);
1824 	if (!skb)
1825 		return;
1826 
1827 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1828 			   msg_link_selector(buf_msg(skb)));
1829 }
1830 
1831 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1832 {
1833 	struct sock *sk = sock->sk;
1834 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1835 	long timeo = *timeop;
1836 	int err = sock_error(sk);
1837 
1838 	if (err)
1839 		return err;
1840 
1841 	for (;;) {
1842 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1843 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1844 				err = -ENOTCONN;
1845 				break;
1846 			}
1847 			add_wait_queue(sk_sleep(sk), &wait);
1848 			release_sock(sk);
1849 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1850 			sched_annotate_sleep();
1851 			lock_sock(sk);
1852 			remove_wait_queue(sk_sleep(sk), &wait);
1853 		}
1854 		err = 0;
1855 		if (!skb_queue_empty(&sk->sk_receive_queue))
1856 			break;
1857 		err = -EAGAIN;
1858 		if (!timeo)
1859 			break;
1860 		err = sock_intr_errno(timeo);
1861 		if (signal_pending(current))
1862 			break;
1863 
1864 		err = sock_error(sk);
1865 		if (err)
1866 			break;
1867 	}
1868 	*timeop = timeo;
1869 	return err;
1870 }
1871 
1872 /**
1873  * tipc_recvmsg - receive packet-oriented message
1874  * @sock: network socket
1875  * @m: descriptor for message info
1876  * @buflen: length of user buffer area
1877  * @flags: receive flags
1878  *
1879  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1880  * If the complete message doesn't fit in user area, truncate it.
1881  *
1882  * Return: size of returned message data, errno otherwise
1883  */
1884 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1885 			size_t buflen,	int flags)
1886 {
1887 	struct sock *sk = sock->sk;
1888 	bool connected = !tipc_sk_type_connectionless(sk);
1889 	struct tipc_sock *tsk = tipc_sk(sk);
1890 	int rc, err, hlen, dlen, copy;
1891 	struct tipc_skb_cb *skb_cb;
1892 	struct sk_buff_head xmitq;
1893 	struct tipc_msg *hdr;
1894 	struct sk_buff *skb;
1895 	bool grp_evt;
1896 	long timeout;
1897 
1898 	/* Catch invalid receive requests */
1899 	if (unlikely(!buflen))
1900 		return -EINVAL;
1901 
1902 	lock_sock(sk);
1903 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1904 		rc = -ENOTCONN;
1905 		goto exit;
1906 	}
1907 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1908 
1909 	/* Step rcv queue to first msg with data or error; wait if necessary */
1910 	do {
1911 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1912 		if (unlikely(rc))
1913 			goto exit;
1914 		skb = skb_peek(&sk->sk_receive_queue);
1915 		skb_cb = TIPC_SKB_CB(skb);
1916 		hdr = buf_msg(skb);
1917 		dlen = msg_data_sz(hdr);
1918 		hlen = msg_hdr_sz(hdr);
1919 		err = msg_errcode(hdr);
1920 		grp_evt = msg_is_grp_evt(hdr);
1921 		if (likely(dlen || err))
1922 			break;
1923 		tsk_advance_rx_queue(sk);
1924 	} while (1);
1925 
1926 	/* Collect msg meta data, including error code and rejected data */
1927 	tipc_sk_set_orig_addr(m, skb);
1928 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1929 	if (unlikely(rc))
1930 		goto exit;
1931 	hdr = buf_msg(skb);
1932 
1933 	/* Capture data if non-error msg, otherwise just set return value */
1934 	if (likely(!err)) {
1935 		int offset = skb_cb->bytes_read;
1936 
1937 		copy = min_t(int, dlen - offset, buflen);
1938 		rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1939 		if (unlikely(rc))
1940 			goto exit;
1941 		if (unlikely(offset + copy < dlen)) {
1942 			if (flags & MSG_EOR) {
1943 				if (!(flags & MSG_PEEK))
1944 					skb_cb->bytes_read = offset + copy;
1945 			} else {
1946 				m->msg_flags |= MSG_TRUNC;
1947 				skb_cb->bytes_read = 0;
1948 			}
1949 		} else {
1950 			if (flags & MSG_EOR)
1951 				m->msg_flags |= MSG_EOR;
1952 			skb_cb->bytes_read = 0;
1953 		}
1954 	} else {
1955 		copy = 0;
1956 		rc = 0;
1957 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1958 			rc = -ECONNRESET;
1959 			goto exit;
1960 		}
1961 	}
1962 
1963 	/* Mark message as group event if applicable */
1964 	if (unlikely(grp_evt)) {
1965 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1966 			m->msg_flags |= MSG_EOR;
1967 		m->msg_flags |= MSG_OOB;
1968 		copy = 0;
1969 	}
1970 
1971 	/* Caption of data or error code/rejected data was successful */
1972 	if (unlikely(flags & MSG_PEEK))
1973 		goto exit;
1974 
1975 	/* Send group flow control advertisement when applicable */
1976 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1977 		__skb_queue_head_init(&xmitq);
1978 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1979 					  msg_orignode(hdr), msg_origport(hdr),
1980 					  &xmitq);
1981 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1982 	}
1983 
1984 	if (skb_cb->bytes_read)
1985 		goto exit;
1986 
1987 	tsk_advance_rx_queue(sk);
1988 
1989 	if (likely(!connected))
1990 		goto exit;
1991 
1992 	/* Send connection flow control advertisement when applicable */
1993 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1994 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1995 		tipc_sk_send_ack(tsk);
1996 exit:
1997 	release_sock(sk);
1998 	return rc ? rc : copy;
1999 }
2000 
2001 /**
2002  * tipc_recvstream - receive stream-oriented data
2003  * @sock: network socket
2004  * @m: descriptor for message info
2005  * @buflen: total size of user buffer area
2006  * @flags: receive flags
2007  *
2008  * Used for SOCK_STREAM messages only.  If not enough data is available
2009  * will optionally wait for more; never truncates data.
2010  *
2011  * Return: size of returned message data, errno otherwise
2012  */
2013 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2014 			   size_t buflen, int flags)
2015 {
2016 	struct sock *sk = sock->sk;
2017 	struct tipc_sock *tsk = tipc_sk(sk);
2018 	struct sk_buff *skb;
2019 	struct tipc_msg *hdr;
2020 	struct tipc_skb_cb *skb_cb;
2021 	bool peek = flags & MSG_PEEK;
2022 	int offset, required, copy, copied = 0;
2023 	int hlen, dlen, err, rc;
2024 	long timeout;
2025 
2026 	/* Catch invalid receive attempts */
2027 	if (unlikely(!buflen))
2028 		return -EINVAL;
2029 
2030 	lock_sock(sk);
2031 
2032 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2033 		rc = -ENOTCONN;
2034 		goto exit;
2035 	}
2036 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2037 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2038 
2039 	do {
2040 		/* Look at first msg in receive queue; wait if necessary */
2041 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2042 		if (unlikely(rc))
2043 			break;
2044 		skb = skb_peek(&sk->sk_receive_queue);
2045 		skb_cb = TIPC_SKB_CB(skb);
2046 		hdr = buf_msg(skb);
2047 		dlen = msg_data_sz(hdr);
2048 		hlen = msg_hdr_sz(hdr);
2049 		err = msg_errcode(hdr);
2050 
2051 		/* Discard any empty non-errored (SYN-) message */
2052 		if (unlikely(!dlen && !err)) {
2053 			tsk_advance_rx_queue(sk);
2054 			continue;
2055 		}
2056 
2057 		/* Collect msg meta data, incl. error code and rejected data */
2058 		if (!copied) {
2059 			tipc_sk_set_orig_addr(m, skb);
2060 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2061 			if (rc)
2062 				break;
2063 			hdr = buf_msg(skb);
2064 		}
2065 
2066 		/* Copy data if msg ok, otherwise return error/partial data */
2067 		if (likely(!err)) {
2068 			offset = skb_cb->bytes_read;
2069 			copy = min_t(int, dlen - offset, buflen - copied);
2070 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2071 			if (unlikely(rc))
2072 				break;
2073 			copied += copy;
2074 			offset += copy;
2075 			if (unlikely(offset < dlen)) {
2076 				if (!peek)
2077 					skb_cb->bytes_read = offset;
2078 				break;
2079 			}
2080 		} else {
2081 			rc = 0;
2082 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2083 				rc = -ECONNRESET;
2084 			if (copied || rc)
2085 				break;
2086 		}
2087 
2088 		if (unlikely(peek))
2089 			break;
2090 
2091 		tsk_advance_rx_queue(sk);
2092 
2093 		/* Send connection flow control advertisement when applicable */
2094 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2095 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2096 			tipc_sk_send_ack(tsk);
2097 
2098 		/* Exit if all requested data or FIN/error received */
2099 		if (copied == buflen || err)
2100 			break;
2101 
2102 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2103 exit:
2104 	release_sock(sk);
2105 	return copied ? copied : rc;
2106 }
2107 
2108 /**
2109  * tipc_write_space - wake up thread if port congestion is released
2110  * @sk: socket
2111  */
2112 static void tipc_write_space(struct sock *sk)
2113 {
2114 	struct socket_wq *wq;
2115 
2116 	rcu_read_lock();
2117 	wq = rcu_dereference(sk->sk_wq);
2118 	if (skwq_has_sleeper(wq))
2119 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2120 						EPOLLWRNORM | EPOLLWRBAND);
2121 	rcu_read_unlock();
2122 }
2123 
2124 /**
2125  * tipc_data_ready - wake up threads to indicate messages have been received
2126  * @sk: socket
2127  */
2128 static void tipc_data_ready(struct sock *sk)
2129 {
2130 	struct socket_wq *wq;
2131 
2132 	rcu_read_lock();
2133 	wq = rcu_dereference(sk->sk_wq);
2134 	if (skwq_has_sleeper(wq))
2135 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2136 						EPOLLRDNORM | EPOLLRDBAND);
2137 	rcu_read_unlock();
2138 }
2139 
2140 static void tipc_sock_destruct(struct sock *sk)
2141 {
2142 	__skb_queue_purge(&sk->sk_receive_queue);
2143 }
2144 
2145 static void tipc_sk_proto_rcv(struct sock *sk,
2146 			      struct sk_buff_head *inputq,
2147 			      struct sk_buff_head *xmitq)
2148 {
2149 	struct sk_buff *skb = __skb_dequeue(inputq);
2150 	struct tipc_sock *tsk = tipc_sk(sk);
2151 	struct tipc_msg *hdr = buf_msg(skb);
2152 	struct tipc_group *grp = tsk->group;
2153 	bool wakeup = false;
2154 
2155 	switch (msg_user(hdr)) {
2156 	case CONN_MANAGER:
2157 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2158 		return;
2159 	case SOCK_WAKEUP:
2160 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2161 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2162 		smp_wmb();
2163 		tsk->cong_link_cnt--;
2164 		wakeup = true;
2165 		tipc_sk_push_backlog(tsk, false);
2166 		break;
2167 	case GROUP_PROTOCOL:
2168 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2169 		break;
2170 	case TOP_SRV:
2171 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2172 				      hdr, inputq, xmitq);
2173 		break;
2174 	default:
2175 		break;
2176 	}
2177 
2178 	if (wakeup)
2179 		sk->sk_write_space(sk);
2180 
2181 	kfree_skb(skb);
2182 }
2183 
2184 /**
2185  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2186  * @tsk: TIPC socket
2187  * @skb: pointer to message buffer.
2188  * @xmitq: for Nagle ACK if any
2189  * Return: true if message should be added to receive queue, false otherwise
2190  */
2191 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2192 				   struct sk_buff_head *xmitq)
2193 {
2194 	struct sock *sk = &tsk->sk;
2195 	struct net *net = sock_net(sk);
2196 	struct tipc_msg *hdr = buf_msg(skb);
2197 	bool con_msg = msg_connected(hdr);
2198 	u32 pport = tsk_peer_port(tsk);
2199 	u32 pnode = tsk_peer_node(tsk);
2200 	u32 oport = msg_origport(hdr);
2201 	u32 onode = msg_orignode(hdr);
2202 	int err = msg_errcode(hdr);
2203 	unsigned long delay;
2204 
2205 	if (unlikely(msg_mcast(hdr)))
2206 		return false;
2207 	tsk->oneway = 0;
2208 
2209 	switch (sk->sk_state) {
2210 	case TIPC_CONNECTING:
2211 		/* Setup ACK */
2212 		if (likely(con_msg)) {
2213 			if (err)
2214 				break;
2215 			tipc_sk_finish_conn(tsk, oport, onode);
2216 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2217 			/* ACK+ message with data is added to receive queue */
2218 			if (msg_data_sz(hdr))
2219 				return true;
2220 			/* Empty ACK-, - wake up sleeping connect() and drop */
2221 			sk->sk_state_change(sk);
2222 			msg_set_dest_droppable(hdr, 1);
2223 			return false;
2224 		}
2225 		/* Ignore connectionless message if not from listening socket */
2226 		if (oport != pport || onode != pnode)
2227 			return false;
2228 
2229 		/* Rejected SYN */
2230 		if (err != TIPC_ERR_OVERLOAD)
2231 			break;
2232 
2233 		/* Prepare for new setup attempt if we have a SYN clone */
2234 		if (skb_queue_empty(&sk->sk_write_queue))
2235 			break;
2236 		get_random_bytes(&delay, 2);
2237 		delay %= (tsk->conn_timeout / 4);
2238 		delay = msecs_to_jiffies(delay + 100);
2239 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2240 		return false;
2241 	case TIPC_OPEN:
2242 	case TIPC_DISCONNECTING:
2243 		return false;
2244 	case TIPC_LISTEN:
2245 		/* Accept only SYN message */
2246 		if (!msg_is_syn(hdr) &&
2247 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2248 			return false;
2249 		if (!con_msg && !err)
2250 			return true;
2251 		return false;
2252 	case TIPC_ESTABLISHED:
2253 		if (!skb_queue_empty(&sk->sk_write_queue))
2254 			tipc_sk_push_backlog(tsk, false);
2255 		/* Accept only connection-based messages sent by peer */
2256 		if (likely(con_msg && !err && pport == oport &&
2257 			   pnode == onode)) {
2258 			if (msg_ack_required(hdr)) {
2259 				struct sk_buff *skb;
2260 
2261 				skb = tipc_sk_build_ack(tsk);
2262 				if (skb) {
2263 					msg_set_nagle_ack(buf_msg(skb));
2264 					__skb_queue_tail(xmitq, skb);
2265 				}
2266 			}
2267 			return true;
2268 		}
2269 		if (!tsk_peer_msg(tsk, hdr))
2270 			return false;
2271 		if (!err)
2272 			return true;
2273 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2274 		tipc_node_remove_conn(net, pnode, tsk->portid);
2275 		sk->sk_state_change(sk);
2276 		return true;
2277 	default:
2278 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2279 	}
2280 	/* Abort connection setup attempt */
2281 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2282 	sk->sk_err = ECONNREFUSED;
2283 	sk->sk_state_change(sk);
2284 	return true;
2285 }
2286 
2287 /**
2288  * rcvbuf_limit - get proper overload limit of socket receive queue
2289  * @sk: socket
2290  * @skb: message
2291  *
2292  * For connection oriented messages, irrespective of importance,
2293  * default queue limit is 2 MB.
2294  *
2295  * For connectionless messages, queue limits are based on message
2296  * importance as follows:
2297  *
2298  * TIPC_LOW_IMPORTANCE       (2 MB)
2299  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2300  * TIPC_HIGH_IMPORTANCE      (8 MB)
2301  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2302  *
2303  * Return: overload limit according to corresponding message importance
2304  */
2305 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2306 {
2307 	struct tipc_sock *tsk = tipc_sk(sk);
2308 	struct tipc_msg *hdr = buf_msg(skb);
2309 
2310 	if (unlikely(msg_in_group(hdr)))
2311 		return READ_ONCE(sk->sk_rcvbuf);
2312 
2313 	if (unlikely(!msg_connected(hdr)))
2314 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2315 
2316 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2317 		return READ_ONCE(sk->sk_rcvbuf);
2318 
2319 	return FLOWCTL_MSG_LIM;
2320 }
2321 
2322 /**
2323  * tipc_sk_filter_rcv - validate incoming message
2324  * @sk: socket
2325  * @skb: pointer to message.
2326  * @xmitq: output message area (FIXME)
2327  *
2328  * Enqueues message on receive queue if acceptable; optionally handles
2329  * disconnect indication for a connected socket.
2330  *
2331  * Called with socket lock already taken
2332  */
2333 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2334 			       struct sk_buff_head *xmitq)
2335 {
2336 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2337 	struct tipc_sock *tsk = tipc_sk(sk);
2338 	struct tipc_group *grp = tsk->group;
2339 	struct tipc_msg *hdr = buf_msg(skb);
2340 	struct net *net = sock_net(sk);
2341 	struct sk_buff_head inputq;
2342 	int mtyp = msg_type(hdr);
2343 	int limit, err = TIPC_OK;
2344 
2345 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2346 	TIPC_SKB_CB(skb)->bytes_read = 0;
2347 	__skb_queue_head_init(&inputq);
2348 	__skb_queue_tail(&inputq, skb);
2349 
2350 	if (unlikely(!msg_isdata(hdr)))
2351 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2352 
2353 	if (unlikely(grp))
2354 		tipc_group_filter_msg(grp, &inputq, xmitq);
2355 
2356 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2357 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2358 
2359 	/* Validate and add to receive buffer if there is space */
2360 	while ((skb = __skb_dequeue(&inputq))) {
2361 		hdr = buf_msg(skb);
2362 		limit = rcvbuf_limit(sk, skb);
2363 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2364 		    (!sk_conn && msg_connected(hdr)) ||
2365 		    (!grp && msg_in_group(hdr)))
2366 			err = TIPC_ERR_NO_PORT;
2367 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2368 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2369 					   "err_overload2!");
2370 			atomic_inc(&sk->sk_drops);
2371 			err = TIPC_ERR_OVERLOAD;
2372 		}
2373 
2374 		if (unlikely(err)) {
2375 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2376 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2377 						      "@filter_rcv!");
2378 				__skb_queue_tail(xmitq, skb);
2379 			}
2380 			err = TIPC_OK;
2381 			continue;
2382 		}
2383 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2384 		skb_set_owner_r(skb, sk);
2385 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2386 					 "rcvq >90% allocated!");
2387 		sk->sk_data_ready(sk);
2388 	}
2389 }
2390 
2391 /**
2392  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2393  * @sk: socket
2394  * @skb: message
2395  *
2396  * Caller must hold socket lock
2397  */
2398 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2399 {
2400 	unsigned int before = sk_rmem_alloc_get(sk);
2401 	struct sk_buff_head xmitq;
2402 	unsigned int added;
2403 
2404 	__skb_queue_head_init(&xmitq);
2405 
2406 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2407 	added = sk_rmem_alloc_get(sk) - before;
2408 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2409 
2410 	/* Send pending response/rejected messages, if any */
2411 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2412 	return 0;
2413 }
2414 
2415 /**
2416  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2417  *                   inputq and try adding them to socket or backlog queue
2418  * @inputq: list of incoming buffers with potentially different destinations
2419  * @sk: socket where the buffers should be enqueued
2420  * @dport: port number for the socket
2421  * @xmitq: output queue
2422  *
2423  * Caller must hold socket lock
2424  */
2425 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2426 			    u32 dport, struct sk_buff_head *xmitq)
2427 {
2428 	unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2429 	struct sk_buff *skb;
2430 	unsigned int lim;
2431 	atomic_t *dcnt;
2432 	u32 onode;
2433 
2434 	while (skb_queue_len(inputq)) {
2435 		if (unlikely(time_after_eq(jiffies, time_limit)))
2436 			return;
2437 
2438 		skb = tipc_skb_dequeue(inputq, dport);
2439 		if (unlikely(!skb))
2440 			return;
2441 
2442 		/* Add message directly to receive queue if possible */
2443 		if (!sock_owned_by_user(sk)) {
2444 			tipc_sk_filter_rcv(sk, skb, xmitq);
2445 			continue;
2446 		}
2447 
2448 		/* Try backlog, compensating for double-counted bytes */
2449 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2450 		if (!sk->sk_backlog.len)
2451 			atomic_set(dcnt, 0);
2452 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2453 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2454 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2455 						 "bklg & rcvq >90% allocated!");
2456 			continue;
2457 		}
2458 
2459 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2460 		/* Overload => reject message back to sender */
2461 		onode = tipc_own_addr(sock_net(sk));
2462 		atomic_inc(&sk->sk_drops);
2463 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2464 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2465 					      "@sk_enqueue!");
2466 			__skb_queue_tail(xmitq, skb);
2467 		}
2468 		break;
2469 	}
2470 }
2471 
2472 /**
2473  * tipc_sk_rcv - handle a chain of incoming buffers
2474  * @net: the associated network namespace
2475  * @inputq: buffer list containing the buffers
2476  * Consumes all buffers in list until inputq is empty
2477  * Note: may be called in multiple threads referring to the same queue
2478  */
2479 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2480 {
2481 	struct sk_buff_head xmitq;
2482 	u32 dnode, dport = 0;
2483 	int err;
2484 	struct tipc_sock *tsk;
2485 	struct sock *sk;
2486 	struct sk_buff *skb;
2487 
2488 	__skb_queue_head_init(&xmitq);
2489 	while (skb_queue_len(inputq)) {
2490 		dport = tipc_skb_peek_port(inputq, dport);
2491 		tsk = tipc_sk_lookup(net, dport);
2492 
2493 		if (likely(tsk)) {
2494 			sk = &tsk->sk;
2495 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2496 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2497 				spin_unlock_bh(&sk->sk_lock.slock);
2498 			}
2499 			/* Send pending response/rejected messages, if any */
2500 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2501 			sock_put(sk);
2502 			continue;
2503 		}
2504 		/* No destination socket => dequeue skb if still there */
2505 		skb = tipc_skb_dequeue(inputq, dport);
2506 		if (!skb)
2507 			return;
2508 
2509 		/* Try secondary lookup if unresolved named message */
2510 		err = TIPC_ERR_NO_PORT;
2511 		if (tipc_msg_lookup_dest(net, skb, &err))
2512 			goto xmit;
2513 
2514 		/* Prepare for message rejection */
2515 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2516 			continue;
2517 
2518 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2519 xmit:
2520 		dnode = msg_destnode(buf_msg(skb));
2521 		tipc_node_xmit_skb(net, skb, dnode, dport);
2522 	}
2523 }
2524 
2525 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2526 {
2527 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2528 	struct sock *sk = sock->sk;
2529 	int done;
2530 
2531 	do {
2532 		int err = sock_error(sk);
2533 		if (err)
2534 			return err;
2535 		if (!*timeo_p)
2536 			return -ETIMEDOUT;
2537 		if (signal_pending(current))
2538 			return sock_intr_errno(*timeo_p);
2539 		if (sk->sk_state == TIPC_DISCONNECTING)
2540 			break;
2541 
2542 		add_wait_queue(sk_sleep(sk), &wait);
2543 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2544 				     &wait);
2545 		remove_wait_queue(sk_sleep(sk), &wait);
2546 	} while (!done);
2547 	return 0;
2548 }
2549 
2550 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2551 {
2552 	if (addr->family != AF_TIPC)
2553 		return false;
2554 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2555 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2556 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2557 		addr->addrtype == TIPC_SOCKET_ADDR);
2558 }
2559 
2560 /**
2561  * tipc_connect - establish a connection to another TIPC port
2562  * @sock: socket structure
2563  * @dest: socket address for destination port
2564  * @destlen: size of socket address data structure
2565  * @flags: file-related flags associated with socket
2566  *
2567  * Return: 0 on success, errno otherwise
2568  */
2569 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2570 			int destlen, int flags)
2571 {
2572 	struct sock *sk = sock->sk;
2573 	struct tipc_sock *tsk = tipc_sk(sk);
2574 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2575 	struct msghdr m = {NULL,};
2576 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2577 	int previous;
2578 	int res = 0;
2579 
2580 	if (destlen != sizeof(struct sockaddr_tipc))
2581 		return -EINVAL;
2582 
2583 	lock_sock(sk);
2584 
2585 	if (tsk->group) {
2586 		res = -EINVAL;
2587 		goto exit;
2588 	}
2589 
2590 	if (dst->family == AF_UNSPEC) {
2591 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2592 		if (!tipc_sk_type_connectionless(sk))
2593 			res = -EINVAL;
2594 		goto exit;
2595 	}
2596 	if (!tipc_sockaddr_is_sane(dst)) {
2597 		res = -EINVAL;
2598 		goto exit;
2599 	}
2600 	/* DGRAM/RDM connect(), just save the destaddr */
2601 	if (tipc_sk_type_connectionless(sk)) {
2602 		memcpy(&tsk->peer, dest, destlen);
2603 		goto exit;
2604 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2605 		res = -EINVAL;
2606 		goto exit;
2607 	}
2608 
2609 	previous = sk->sk_state;
2610 
2611 	switch (sk->sk_state) {
2612 	case TIPC_OPEN:
2613 		/* Send a 'SYN-' to destination */
2614 		m.msg_name = dest;
2615 		m.msg_namelen = destlen;
2616 
2617 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2618 		 * indicate send_msg() is never blocked.
2619 		 */
2620 		if (!timeout)
2621 			m.msg_flags = MSG_DONTWAIT;
2622 
2623 		res = __tipc_sendmsg(sock, &m, 0);
2624 		if ((res < 0) && (res != -EWOULDBLOCK))
2625 			goto exit;
2626 
2627 		/* Just entered TIPC_CONNECTING state; the only
2628 		 * difference is that return value in non-blocking
2629 		 * case is EINPROGRESS, rather than EALREADY.
2630 		 */
2631 		res = -EINPROGRESS;
2632 		fallthrough;
2633 	case TIPC_CONNECTING:
2634 		if (!timeout) {
2635 			if (previous == TIPC_CONNECTING)
2636 				res = -EALREADY;
2637 			goto exit;
2638 		}
2639 		timeout = msecs_to_jiffies(timeout);
2640 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2641 		res = tipc_wait_for_connect(sock, &timeout);
2642 		break;
2643 	case TIPC_ESTABLISHED:
2644 		res = -EISCONN;
2645 		break;
2646 	default:
2647 		res = -EINVAL;
2648 	}
2649 
2650 exit:
2651 	release_sock(sk);
2652 	return res;
2653 }
2654 
2655 /**
2656  * tipc_listen - allow socket to listen for incoming connections
2657  * @sock: socket structure
2658  * @len: (unused)
2659  *
2660  * Return: 0 on success, errno otherwise
2661  */
2662 static int tipc_listen(struct socket *sock, int len)
2663 {
2664 	struct sock *sk = sock->sk;
2665 	int res;
2666 
2667 	lock_sock(sk);
2668 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2669 	release_sock(sk);
2670 
2671 	return res;
2672 }
2673 
2674 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2675 {
2676 	struct sock *sk = sock->sk;
2677 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2678 	int err;
2679 
2680 	/* True wake-one mechanism for incoming connections: only
2681 	 * one process gets woken up, not the 'whole herd'.
2682 	 * Since we do not 'race & poll' for established sockets
2683 	 * anymore, the common case will execute the loop only once.
2684 	*/
2685 	for (;;) {
2686 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2687 			add_wait_queue(sk_sleep(sk), &wait);
2688 			release_sock(sk);
2689 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2690 			lock_sock(sk);
2691 			remove_wait_queue(sk_sleep(sk), &wait);
2692 		}
2693 		err = 0;
2694 		if (!skb_queue_empty(&sk->sk_receive_queue))
2695 			break;
2696 		err = -EAGAIN;
2697 		if (!timeo)
2698 			break;
2699 		err = sock_intr_errno(timeo);
2700 		if (signal_pending(current))
2701 			break;
2702 	}
2703 	return err;
2704 }
2705 
2706 /**
2707  * tipc_accept - wait for connection request
2708  * @sock: listening socket
2709  * @new_sock: new socket that is to be connected
2710  * @flags: file-related flags associated with socket
2711  * @kern: caused by kernel or by userspace?
2712  *
2713  * Return: 0 on success, errno otherwise
2714  */
2715 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2716 		       bool kern)
2717 {
2718 	struct sock *new_sk, *sk = sock->sk;
2719 	struct tipc_sock *new_tsock;
2720 	struct msghdr m = {NULL,};
2721 	struct tipc_msg *msg;
2722 	struct sk_buff *buf;
2723 	long timeo;
2724 	int res;
2725 
2726 	lock_sock(sk);
2727 
2728 	if (sk->sk_state != TIPC_LISTEN) {
2729 		res = -EINVAL;
2730 		goto exit;
2731 	}
2732 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2733 	res = tipc_wait_for_accept(sock, timeo);
2734 	if (res)
2735 		goto exit;
2736 
2737 	buf = skb_peek(&sk->sk_receive_queue);
2738 
2739 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2740 	if (res)
2741 		goto exit;
2742 	security_sk_clone(sock->sk, new_sock->sk);
2743 
2744 	new_sk = new_sock->sk;
2745 	new_tsock = tipc_sk(new_sk);
2746 	msg = buf_msg(buf);
2747 
2748 	/* we lock on new_sk; but lockdep sees the lock on sk */
2749 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2750 
2751 	/*
2752 	 * Reject any stray messages received by new socket
2753 	 * before the socket lock was taken (very, very unlikely)
2754 	 */
2755 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2756 
2757 	/* Connect new socket to it's peer */
2758 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2759 
2760 	tsk_set_importance(new_sk, msg_importance(msg));
2761 	if (msg_named(msg)) {
2762 		new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2763 		msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2764 		msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2765 	}
2766 
2767 	/*
2768 	 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2769 	 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2770 	 */
2771 	if (!msg_data_sz(msg)) {
2772 		tsk_advance_rx_queue(sk);
2773 	} else {
2774 		__skb_dequeue(&sk->sk_receive_queue);
2775 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2776 		skb_set_owner_r(buf, new_sk);
2777 	}
2778 	__tipc_sendstream(new_sock, &m, 0);
2779 	release_sock(new_sk);
2780 exit:
2781 	release_sock(sk);
2782 	return res;
2783 }
2784 
2785 /**
2786  * tipc_shutdown - shutdown socket connection
2787  * @sock: socket structure
2788  * @how: direction to close (must be SHUT_RDWR)
2789  *
2790  * Terminates connection (if necessary), then purges socket's receive queue.
2791  *
2792  * Return: 0 on success, errno otherwise
2793  */
2794 static int tipc_shutdown(struct socket *sock, int how)
2795 {
2796 	struct sock *sk = sock->sk;
2797 	int res;
2798 
2799 	if (how != SHUT_RDWR)
2800 		return -EINVAL;
2801 
2802 	lock_sock(sk);
2803 
2804 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2805 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2806 	sk->sk_shutdown = SHUTDOWN_MASK;
2807 
2808 	if (sk->sk_state == TIPC_DISCONNECTING) {
2809 		/* Discard any unreceived messages */
2810 		__skb_queue_purge(&sk->sk_receive_queue);
2811 
2812 		res = 0;
2813 	} else {
2814 		res = -ENOTCONN;
2815 	}
2816 	/* Wake up anyone sleeping in poll. */
2817 	sk->sk_state_change(sk);
2818 
2819 	release_sock(sk);
2820 	return res;
2821 }
2822 
2823 static void tipc_sk_check_probing_state(struct sock *sk,
2824 					struct sk_buff_head *list)
2825 {
2826 	struct tipc_sock *tsk = tipc_sk(sk);
2827 	u32 pnode = tsk_peer_node(tsk);
2828 	u32 pport = tsk_peer_port(tsk);
2829 	u32 self = tsk_own_node(tsk);
2830 	u32 oport = tsk->portid;
2831 	struct sk_buff *skb;
2832 
2833 	if (tsk->probe_unacked) {
2834 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2835 		sk->sk_err = ECONNABORTED;
2836 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2837 		sk->sk_state_change(sk);
2838 		return;
2839 	}
2840 	/* Prepare new probe */
2841 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2842 			      pnode, self, pport, oport, TIPC_OK);
2843 	if (skb)
2844 		__skb_queue_tail(list, skb);
2845 	tsk->probe_unacked = true;
2846 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2847 }
2848 
2849 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2850 {
2851 	struct tipc_sock *tsk = tipc_sk(sk);
2852 
2853 	/* Try again later if dest link is congested */
2854 	if (tsk->cong_link_cnt) {
2855 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2856 		return;
2857 	}
2858 	/* Prepare SYN for retransmit */
2859 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2860 }
2861 
2862 static void tipc_sk_timeout(struct timer_list *t)
2863 {
2864 	struct sock *sk = from_timer(sk, t, sk_timer);
2865 	struct tipc_sock *tsk = tipc_sk(sk);
2866 	u32 pnode = tsk_peer_node(tsk);
2867 	struct sk_buff_head list;
2868 	int rc = 0;
2869 
2870 	__skb_queue_head_init(&list);
2871 	bh_lock_sock(sk);
2872 
2873 	/* Try again later if socket is busy */
2874 	if (sock_owned_by_user(sk)) {
2875 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2876 		bh_unlock_sock(sk);
2877 		sock_put(sk);
2878 		return;
2879 	}
2880 
2881 	if (sk->sk_state == TIPC_ESTABLISHED)
2882 		tipc_sk_check_probing_state(sk, &list);
2883 	else if (sk->sk_state == TIPC_CONNECTING)
2884 		tipc_sk_retry_connect(sk, &list);
2885 
2886 	bh_unlock_sock(sk);
2887 
2888 	if (!skb_queue_empty(&list))
2889 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2890 
2891 	/* SYN messages may cause link congestion */
2892 	if (rc == -ELINKCONG) {
2893 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2894 		tsk->cong_link_cnt = 1;
2895 	}
2896 	sock_put(sk);
2897 }
2898 
2899 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2900 {
2901 	struct sock *sk = &tsk->sk;
2902 	struct net *net = sock_net(sk);
2903 	struct tipc_socket_addr skaddr;
2904 	struct publication *p;
2905 	u32 key;
2906 
2907 	if (tipc_sk_connected(sk))
2908 		return -EINVAL;
2909 	key = tsk->portid + tsk->pub_count + 1;
2910 	if (key == tsk->portid)
2911 		return -EADDRINUSE;
2912 	skaddr.ref = tsk->portid;
2913 	skaddr.node = tipc_own_addr(net);
2914 	p = tipc_nametbl_publish(net, ua, &skaddr, key);
2915 	if (unlikely(!p))
2916 		return -EINVAL;
2917 
2918 	list_add(&p->binding_sock, &tsk->publications);
2919 	tsk->pub_count++;
2920 	tsk->published = true;
2921 	return 0;
2922 }
2923 
2924 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2925 {
2926 	struct net *net = sock_net(&tsk->sk);
2927 	struct publication *safe, *p;
2928 	struct tipc_uaddr _ua;
2929 	int rc = -EINVAL;
2930 
2931 	list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2932 		if (!ua) {
2933 			tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2934 				   p->sr.type, p->sr.lower, p->sr.upper);
2935 			tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2936 			continue;
2937 		}
2938 		/* Unbind specific publication */
2939 		if (p->scope != ua->scope)
2940 			continue;
2941 		if (p->sr.type != ua->sr.type)
2942 			continue;
2943 		if (p->sr.lower != ua->sr.lower)
2944 			continue;
2945 		if (p->sr.upper != ua->sr.upper)
2946 			break;
2947 		tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2948 		rc = 0;
2949 		break;
2950 	}
2951 	if (list_empty(&tsk->publications)) {
2952 		tsk->published = 0;
2953 		rc = 0;
2954 	}
2955 	return rc;
2956 }
2957 
2958 /* tipc_sk_reinit: set non-zero address in all existing sockets
2959  *                 when we go from standalone to network mode.
2960  */
2961 void tipc_sk_reinit(struct net *net)
2962 {
2963 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2964 	struct rhashtable_iter iter;
2965 	struct tipc_sock *tsk;
2966 	struct tipc_msg *msg;
2967 
2968 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2969 
2970 	do {
2971 		rhashtable_walk_start(&iter);
2972 
2973 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2974 			sock_hold(&tsk->sk);
2975 			rhashtable_walk_stop(&iter);
2976 			lock_sock(&tsk->sk);
2977 			msg = &tsk->phdr;
2978 			msg_set_prevnode(msg, tipc_own_addr(net));
2979 			msg_set_orignode(msg, tipc_own_addr(net));
2980 			release_sock(&tsk->sk);
2981 			rhashtable_walk_start(&iter);
2982 			sock_put(&tsk->sk);
2983 		}
2984 
2985 		rhashtable_walk_stop(&iter);
2986 	} while (tsk == ERR_PTR(-EAGAIN));
2987 
2988 	rhashtable_walk_exit(&iter);
2989 }
2990 
2991 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2992 {
2993 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2994 	struct tipc_sock *tsk;
2995 
2996 	rcu_read_lock();
2997 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2998 	if (tsk)
2999 		sock_hold(&tsk->sk);
3000 	rcu_read_unlock();
3001 
3002 	return tsk;
3003 }
3004 
3005 static int tipc_sk_insert(struct tipc_sock *tsk)
3006 {
3007 	struct sock *sk = &tsk->sk;
3008 	struct net *net = sock_net(sk);
3009 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3010 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3011 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3012 
3013 	while (remaining--) {
3014 		portid++;
3015 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3016 			portid = TIPC_MIN_PORT;
3017 		tsk->portid = portid;
3018 		sock_hold(&tsk->sk);
3019 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3020 						   tsk_rht_params))
3021 			return 0;
3022 		sock_put(&tsk->sk);
3023 	}
3024 
3025 	return -1;
3026 }
3027 
3028 static void tipc_sk_remove(struct tipc_sock *tsk)
3029 {
3030 	struct sock *sk = &tsk->sk;
3031 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3032 
3033 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3034 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3035 		__sock_put(sk);
3036 	}
3037 }
3038 
3039 static const struct rhashtable_params tsk_rht_params = {
3040 	.nelem_hint = 192,
3041 	.head_offset = offsetof(struct tipc_sock, node),
3042 	.key_offset = offsetof(struct tipc_sock, portid),
3043 	.key_len = sizeof(u32), /* portid */
3044 	.max_size = 1048576,
3045 	.min_size = 256,
3046 	.automatic_shrinking = true,
3047 };
3048 
3049 int tipc_sk_rht_init(struct net *net)
3050 {
3051 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3052 
3053 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3054 }
3055 
3056 void tipc_sk_rht_destroy(struct net *net)
3057 {
3058 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3059 
3060 	/* Wait for socket readers to complete */
3061 	synchronize_net();
3062 
3063 	rhashtable_destroy(&tn->sk_rht);
3064 }
3065 
3066 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3067 {
3068 	struct net *net = sock_net(&tsk->sk);
3069 	struct tipc_group *grp = tsk->group;
3070 	struct tipc_msg *hdr = &tsk->phdr;
3071 	struct tipc_uaddr ua;
3072 	int rc;
3073 
3074 	if (mreq->type < TIPC_RESERVED_TYPES)
3075 		return -EACCES;
3076 	if (mreq->scope > TIPC_NODE_SCOPE)
3077 		return -EINVAL;
3078 	if (mreq->scope != TIPC_NODE_SCOPE)
3079 		mreq->scope = TIPC_CLUSTER_SCOPE;
3080 	if (grp)
3081 		return -EACCES;
3082 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3083 	if (!grp)
3084 		return -ENOMEM;
3085 	tsk->group = grp;
3086 	msg_set_lookup_scope(hdr, mreq->scope);
3087 	msg_set_nametype(hdr, mreq->type);
3088 	msg_set_dest_droppable(hdr, true);
3089 	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3090 		   mreq->type, mreq->instance, mreq->instance);
3091 	tipc_nametbl_build_group(net, grp, &ua);
3092 	rc = tipc_sk_publish(tsk, &ua);
3093 	if (rc) {
3094 		tipc_group_delete(net, grp);
3095 		tsk->group = NULL;
3096 		return rc;
3097 	}
3098 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3099 	tsk->mc_method.rcast = true;
3100 	tsk->mc_method.mandatory = true;
3101 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3102 	return rc;
3103 }
3104 
3105 static int tipc_sk_leave(struct tipc_sock *tsk)
3106 {
3107 	struct net *net = sock_net(&tsk->sk);
3108 	struct tipc_group *grp = tsk->group;
3109 	struct tipc_uaddr ua;
3110 	int scope;
3111 
3112 	if (!grp)
3113 		return -EINVAL;
3114 	ua.addrtype = TIPC_SERVICE_RANGE;
3115 	tipc_group_self(grp, &ua.sr, &scope);
3116 	ua.scope = scope;
3117 	tipc_group_delete(net, grp);
3118 	tsk->group = NULL;
3119 	tipc_sk_withdraw(tsk, &ua);
3120 	return 0;
3121 }
3122 
3123 /**
3124  * tipc_setsockopt - set socket option
3125  * @sock: socket structure
3126  * @lvl: option level
3127  * @opt: option identifier
3128  * @ov: pointer to new option value
3129  * @ol: length of option value
3130  *
3131  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3132  * (to ease compatibility).
3133  *
3134  * Return: 0 on success, errno otherwise
3135  */
3136 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3137 			   sockptr_t ov, unsigned int ol)
3138 {
3139 	struct sock *sk = sock->sk;
3140 	struct tipc_sock *tsk = tipc_sk(sk);
3141 	struct tipc_group_req mreq;
3142 	u32 value = 0;
3143 	int res = 0;
3144 
3145 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3146 		return 0;
3147 	if (lvl != SOL_TIPC)
3148 		return -ENOPROTOOPT;
3149 
3150 	switch (opt) {
3151 	case TIPC_IMPORTANCE:
3152 	case TIPC_SRC_DROPPABLE:
3153 	case TIPC_DEST_DROPPABLE:
3154 	case TIPC_CONN_TIMEOUT:
3155 	case TIPC_NODELAY:
3156 		if (ol < sizeof(value))
3157 			return -EINVAL;
3158 		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3159 			return -EFAULT;
3160 		break;
3161 	case TIPC_GROUP_JOIN:
3162 		if (ol < sizeof(mreq))
3163 			return -EINVAL;
3164 		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3165 			return -EFAULT;
3166 		break;
3167 	default:
3168 		if (!sockptr_is_null(ov) || ol)
3169 			return -EINVAL;
3170 	}
3171 
3172 	lock_sock(sk);
3173 
3174 	switch (opt) {
3175 	case TIPC_IMPORTANCE:
3176 		res = tsk_set_importance(sk, value);
3177 		break;
3178 	case TIPC_SRC_DROPPABLE:
3179 		if (sock->type != SOCK_STREAM)
3180 			tsk_set_unreliable(tsk, value);
3181 		else
3182 			res = -ENOPROTOOPT;
3183 		break;
3184 	case TIPC_DEST_DROPPABLE:
3185 		tsk_set_unreturnable(tsk, value);
3186 		break;
3187 	case TIPC_CONN_TIMEOUT:
3188 		tipc_sk(sk)->conn_timeout = value;
3189 		break;
3190 	case TIPC_MCAST_BROADCAST:
3191 		tsk->mc_method.rcast = false;
3192 		tsk->mc_method.mandatory = true;
3193 		break;
3194 	case TIPC_MCAST_REPLICAST:
3195 		tsk->mc_method.rcast = true;
3196 		tsk->mc_method.mandatory = true;
3197 		break;
3198 	case TIPC_GROUP_JOIN:
3199 		res = tipc_sk_join(tsk, &mreq);
3200 		break;
3201 	case TIPC_GROUP_LEAVE:
3202 		res = tipc_sk_leave(tsk);
3203 		break;
3204 	case TIPC_NODELAY:
3205 		tsk->nodelay = !!value;
3206 		tsk_set_nagle(tsk);
3207 		break;
3208 	default:
3209 		res = -EINVAL;
3210 	}
3211 
3212 	release_sock(sk);
3213 
3214 	return res;
3215 }
3216 
3217 /**
3218  * tipc_getsockopt - get socket option
3219  * @sock: socket structure
3220  * @lvl: option level
3221  * @opt: option identifier
3222  * @ov: receptacle for option value
3223  * @ol: receptacle for length of option value
3224  *
3225  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3226  * (to ease compatibility).
3227  *
3228  * Return: 0 on success, errno otherwise
3229  */
3230 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3231 			   char __user *ov, int __user *ol)
3232 {
3233 	struct sock *sk = sock->sk;
3234 	struct tipc_sock *tsk = tipc_sk(sk);
3235 	struct tipc_service_range seq;
3236 	int len, scope;
3237 	u32 value;
3238 	int res;
3239 
3240 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3241 		return put_user(0, ol);
3242 	if (lvl != SOL_TIPC)
3243 		return -ENOPROTOOPT;
3244 	res = get_user(len, ol);
3245 	if (res)
3246 		return res;
3247 
3248 	lock_sock(sk);
3249 
3250 	switch (opt) {
3251 	case TIPC_IMPORTANCE:
3252 		value = tsk_importance(tsk);
3253 		break;
3254 	case TIPC_SRC_DROPPABLE:
3255 		value = tsk_unreliable(tsk);
3256 		break;
3257 	case TIPC_DEST_DROPPABLE:
3258 		value = tsk_unreturnable(tsk);
3259 		break;
3260 	case TIPC_CONN_TIMEOUT:
3261 		value = tsk->conn_timeout;
3262 		/* no need to set "res", since already 0 at this point */
3263 		break;
3264 	case TIPC_NODE_RECVQ_DEPTH:
3265 		value = 0; /* was tipc_queue_size, now obsolete */
3266 		break;
3267 	case TIPC_SOCK_RECVQ_DEPTH:
3268 		value = skb_queue_len(&sk->sk_receive_queue);
3269 		break;
3270 	case TIPC_SOCK_RECVQ_USED:
3271 		value = sk_rmem_alloc_get(sk);
3272 		break;
3273 	case TIPC_GROUP_JOIN:
3274 		seq.type = 0;
3275 		if (tsk->group)
3276 			tipc_group_self(tsk->group, &seq, &scope);
3277 		value = seq.type;
3278 		break;
3279 	default:
3280 		res = -EINVAL;
3281 	}
3282 
3283 	release_sock(sk);
3284 
3285 	if (res)
3286 		return res;	/* "get" failed */
3287 
3288 	if (len < sizeof(value))
3289 		return -EINVAL;
3290 
3291 	if (copy_to_user(ov, &value, sizeof(value)))
3292 		return -EFAULT;
3293 
3294 	return put_user(sizeof(value), ol);
3295 }
3296 
3297 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3298 {
3299 	struct net *net = sock_net(sock->sk);
3300 	struct tipc_sioc_nodeid_req nr = {0};
3301 	struct tipc_sioc_ln_req lnr;
3302 	void __user *argp = (void __user *)arg;
3303 
3304 	switch (cmd) {
3305 	case SIOCGETLINKNAME:
3306 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3307 			return -EFAULT;
3308 		if (!tipc_node_get_linkname(net,
3309 					    lnr.bearer_id & 0xffff, lnr.peer,
3310 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3311 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3312 				return -EFAULT;
3313 			return 0;
3314 		}
3315 		return -EADDRNOTAVAIL;
3316 	case SIOCGETNODEID:
3317 		if (copy_from_user(&nr, argp, sizeof(nr)))
3318 			return -EFAULT;
3319 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3320 			return -EADDRNOTAVAIL;
3321 		if (copy_to_user(argp, &nr, sizeof(nr)))
3322 			return -EFAULT;
3323 		return 0;
3324 	default:
3325 		return -ENOIOCTLCMD;
3326 	}
3327 }
3328 
3329 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3330 {
3331 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3332 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3333 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3334 
3335 	tsk1->peer.family = AF_TIPC;
3336 	tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3337 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3338 	tsk1->peer.addr.id.ref = tsk2->portid;
3339 	tsk1->peer.addr.id.node = onode;
3340 	tsk2->peer.family = AF_TIPC;
3341 	tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3342 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3343 	tsk2->peer.addr.id.ref = tsk1->portid;
3344 	tsk2->peer.addr.id.node = onode;
3345 
3346 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3347 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3348 	return 0;
3349 }
3350 
3351 /* Protocol switches for the various types of TIPC sockets */
3352 
3353 static const struct proto_ops msg_ops = {
3354 	.owner		= THIS_MODULE,
3355 	.family		= AF_TIPC,
3356 	.release	= tipc_release,
3357 	.bind		= tipc_bind,
3358 	.connect	= tipc_connect,
3359 	.socketpair	= tipc_socketpair,
3360 	.accept		= sock_no_accept,
3361 	.getname	= tipc_getname,
3362 	.poll		= tipc_poll,
3363 	.ioctl		= tipc_ioctl,
3364 	.listen		= sock_no_listen,
3365 	.shutdown	= tipc_shutdown,
3366 	.setsockopt	= tipc_setsockopt,
3367 	.getsockopt	= tipc_getsockopt,
3368 	.sendmsg	= tipc_sendmsg,
3369 	.recvmsg	= tipc_recvmsg,
3370 	.mmap		= sock_no_mmap,
3371 	.sendpage	= sock_no_sendpage
3372 };
3373 
3374 static const struct proto_ops packet_ops = {
3375 	.owner		= THIS_MODULE,
3376 	.family		= AF_TIPC,
3377 	.release	= tipc_release,
3378 	.bind		= tipc_bind,
3379 	.connect	= tipc_connect,
3380 	.socketpair	= tipc_socketpair,
3381 	.accept		= tipc_accept,
3382 	.getname	= tipc_getname,
3383 	.poll		= tipc_poll,
3384 	.ioctl		= tipc_ioctl,
3385 	.listen		= tipc_listen,
3386 	.shutdown	= tipc_shutdown,
3387 	.setsockopt	= tipc_setsockopt,
3388 	.getsockopt	= tipc_getsockopt,
3389 	.sendmsg	= tipc_send_packet,
3390 	.recvmsg	= tipc_recvmsg,
3391 	.mmap		= sock_no_mmap,
3392 	.sendpage	= sock_no_sendpage
3393 };
3394 
3395 static const struct proto_ops stream_ops = {
3396 	.owner		= THIS_MODULE,
3397 	.family		= AF_TIPC,
3398 	.release	= tipc_release,
3399 	.bind		= tipc_bind,
3400 	.connect	= tipc_connect,
3401 	.socketpair	= tipc_socketpair,
3402 	.accept		= tipc_accept,
3403 	.getname	= tipc_getname,
3404 	.poll		= tipc_poll,
3405 	.ioctl		= tipc_ioctl,
3406 	.listen		= tipc_listen,
3407 	.shutdown	= tipc_shutdown,
3408 	.setsockopt	= tipc_setsockopt,
3409 	.getsockopt	= tipc_getsockopt,
3410 	.sendmsg	= tipc_sendstream,
3411 	.recvmsg	= tipc_recvstream,
3412 	.mmap		= sock_no_mmap,
3413 	.sendpage	= sock_no_sendpage
3414 };
3415 
3416 static const struct net_proto_family tipc_family_ops = {
3417 	.owner		= THIS_MODULE,
3418 	.family		= AF_TIPC,
3419 	.create		= tipc_sk_create
3420 };
3421 
3422 static struct proto tipc_proto = {
3423 	.name		= "TIPC",
3424 	.owner		= THIS_MODULE,
3425 	.obj_size	= sizeof(struct tipc_sock),
3426 	.sysctl_rmem	= sysctl_tipc_rmem
3427 };
3428 
3429 /**
3430  * tipc_socket_init - initialize TIPC socket interface
3431  *
3432  * Return: 0 on success, errno otherwise
3433  */
3434 int tipc_socket_init(void)
3435 {
3436 	int res;
3437 
3438 	res = proto_register(&tipc_proto, 1);
3439 	if (res) {
3440 		pr_err("Failed to register TIPC protocol type\n");
3441 		goto out;
3442 	}
3443 
3444 	res = sock_register(&tipc_family_ops);
3445 	if (res) {
3446 		pr_err("Failed to register TIPC socket type\n");
3447 		proto_unregister(&tipc_proto);
3448 		goto out;
3449 	}
3450  out:
3451 	return res;
3452 }
3453 
3454 /**
3455  * tipc_socket_stop - stop TIPC socket interface
3456  */
3457 void tipc_socket_stop(void)
3458 {
3459 	sock_unregister(tipc_family_ops.family);
3460 	proto_unregister(&tipc_proto);
3461 }
3462 
3463 /* Caller should hold socket lock for the passed tipc socket. */
3464 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3465 {
3466 	u32 peer_node, peer_port;
3467 	u32 conn_type, conn_instance;
3468 	struct nlattr *nest;
3469 
3470 	peer_node = tsk_peer_node(tsk);
3471 	peer_port = tsk_peer_port(tsk);
3472 	conn_type = msg_nametype(&tsk->phdr);
3473 	conn_instance = msg_nameinst(&tsk->phdr);
3474 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3475 	if (!nest)
3476 		return -EMSGSIZE;
3477 
3478 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3479 		goto msg_full;
3480 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3481 		goto msg_full;
3482 
3483 	if (tsk->conn_addrtype != 0) {
3484 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3485 			goto msg_full;
3486 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3487 			goto msg_full;
3488 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3489 			goto msg_full;
3490 	}
3491 	nla_nest_end(skb, nest);
3492 
3493 	return 0;
3494 
3495 msg_full:
3496 	nla_nest_cancel(skb, nest);
3497 
3498 	return -EMSGSIZE;
3499 }
3500 
3501 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3502 			  *tsk)
3503 {
3504 	struct net *net = sock_net(skb->sk);
3505 	struct sock *sk = &tsk->sk;
3506 
3507 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3508 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3509 		return -EMSGSIZE;
3510 
3511 	if (tipc_sk_connected(sk)) {
3512 		if (__tipc_nl_add_sk_con(skb, tsk))
3513 			return -EMSGSIZE;
3514 	} else if (!list_empty(&tsk->publications)) {
3515 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3516 			return -EMSGSIZE;
3517 	}
3518 	return 0;
3519 }
3520 
3521 /* Caller should hold socket lock for the passed tipc socket. */
3522 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3523 			    struct tipc_sock *tsk)
3524 {
3525 	struct nlattr *attrs;
3526 	void *hdr;
3527 
3528 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3529 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3530 	if (!hdr)
3531 		goto msg_cancel;
3532 
3533 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3534 	if (!attrs)
3535 		goto genlmsg_cancel;
3536 
3537 	if (__tipc_nl_add_sk_info(skb, tsk))
3538 		goto attr_msg_cancel;
3539 
3540 	nla_nest_end(skb, attrs);
3541 	genlmsg_end(skb, hdr);
3542 
3543 	return 0;
3544 
3545 attr_msg_cancel:
3546 	nla_nest_cancel(skb, attrs);
3547 genlmsg_cancel:
3548 	genlmsg_cancel(skb, hdr);
3549 msg_cancel:
3550 	return -EMSGSIZE;
3551 }
3552 
3553 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3554 		    int (*skb_handler)(struct sk_buff *skb,
3555 				       struct netlink_callback *cb,
3556 				       struct tipc_sock *tsk))
3557 {
3558 	struct rhashtable_iter *iter = (void *)cb->args[4];
3559 	struct tipc_sock *tsk;
3560 	int err;
3561 
3562 	rhashtable_walk_start(iter);
3563 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3564 		if (IS_ERR(tsk)) {
3565 			err = PTR_ERR(tsk);
3566 			if (err == -EAGAIN) {
3567 				err = 0;
3568 				continue;
3569 			}
3570 			break;
3571 		}
3572 
3573 		sock_hold(&tsk->sk);
3574 		rhashtable_walk_stop(iter);
3575 		lock_sock(&tsk->sk);
3576 		err = skb_handler(skb, cb, tsk);
3577 		if (err) {
3578 			release_sock(&tsk->sk);
3579 			sock_put(&tsk->sk);
3580 			goto out;
3581 		}
3582 		release_sock(&tsk->sk);
3583 		rhashtable_walk_start(iter);
3584 		sock_put(&tsk->sk);
3585 	}
3586 	rhashtable_walk_stop(iter);
3587 out:
3588 	return skb->len;
3589 }
3590 EXPORT_SYMBOL(tipc_nl_sk_walk);
3591 
3592 int tipc_dump_start(struct netlink_callback *cb)
3593 {
3594 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3595 }
3596 EXPORT_SYMBOL(tipc_dump_start);
3597 
3598 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3599 {
3600 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3601 	struct rhashtable_iter *iter = (void *)cb->args[4];
3602 	struct tipc_net *tn = tipc_net(net);
3603 
3604 	if (!iter) {
3605 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3606 		if (!iter)
3607 			return -ENOMEM;
3608 
3609 		cb->args[4] = (long)iter;
3610 	}
3611 
3612 	rhashtable_walk_enter(&tn->sk_rht, iter);
3613 	return 0;
3614 }
3615 
3616 int tipc_dump_done(struct netlink_callback *cb)
3617 {
3618 	struct rhashtable_iter *hti = (void *)cb->args[4];
3619 
3620 	rhashtable_walk_exit(hti);
3621 	kfree(hti);
3622 	return 0;
3623 }
3624 EXPORT_SYMBOL(tipc_dump_done);
3625 
3626 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3627 			   struct tipc_sock *tsk, u32 sk_filter_state,
3628 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3629 {
3630 	struct sock *sk = &tsk->sk;
3631 	struct nlattr *attrs;
3632 	struct nlattr *stat;
3633 
3634 	/*filter response w.r.t sk_state*/
3635 	if (!(sk_filter_state & (1 << sk->sk_state)))
3636 		return 0;
3637 
3638 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3639 	if (!attrs)
3640 		goto msg_cancel;
3641 
3642 	if (__tipc_nl_add_sk_info(skb, tsk))
3643 		goto attr_msg_cancel;
3644 
3645 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3646 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3647 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3648 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3649 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3650 					 sock_i_uid(sk))) ||
3651 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3652 			      tipc_diag_gen_cookie(sk),
3653 			      TIPC_NLA_SOCK_PAD))
3654 		goto attr_msg_cancel;
3655 
3656 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3657 	if (!stat)
3658 		goto attr_msg_cancel;
3659 
3660 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3661 			skb_queue_len(&sk->sk_receive_queue)) ||
3662 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3663 			skb_queue_len(&sk->sk_write_queue)) ||
3664 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3665 			atomic_read(&sk->sk_drops)))
3666 		goto stat_msg_cancel;
3667 
3668 	if (tsk->cong_link_cnt &&
3669 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3670 		goto stat_msg_cancel;
3671 
3672 	if (tsk_conn_cong(tsk) &&
3673 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3674 		goto stat_msg_cancel;
3675 
3676 	nla_nest_end(skb, stat);
3677 
3678 	if (tsk->group)
3679 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3680 			goto stat_msg_cancel;
3681 
3682 	nla_nest_end(skb, attrs);
3683 
3684 	return 0;
3685 
3686 stat_msg_cancel:
3687 	nla_nest_cancel(skb, stat);
3688 attr_msg_cancel:
3689 	nla_nest_cancel(skb, attrs);
3690 msg_cancel:
3691 	return -EMSGSIZE;
3692 }
3693 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3694 
3695 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3696 {
3697 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3698 }
3699 
3700 /* Caller should hold socket lock for the passed tipc socket. */
3701 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3702 				 struct netlink_callback *cb,
3703 				 struct publication *publ)
3704 {
3705 	void *hdr;
3706 	struct nlattr *attrs;
3707 
3708 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3709 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3710 	if (!hdr)
3711 		goto msg_cancel;
3712 
3713 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3714 	if (!attrs)
3715 		goto genlmsg_cancel;
3716 
3717 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3718 		goto attr_msg_cancel;
3719 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3720 		goto attr_msg_cancel;
3721 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3722 		goto attr_msg_cancel;
3723 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3724 		goto attr_msg_cancel;
3725 
3726 	nla_nest_end(skb, attrs);
3727 	genlmsg_end(skb, hdr);
3728 
3729 	return 0;
3730 
3731 attr_msg_cancel:
3732 	nla_nest_cancel(skb, attrs);
3733 genlmsg_cancel:
3734 	genlmsg_cancel(skb, hdr);
3735 msg_cancel:
3736 	return -EMSGSIZE;
3737 }
3738 
3739 /* Caller should hold socket lock for the passed tipc socket. */
3740 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3741 				  struct netlink_callback *cb,
3742 				  struct tipc_sock *tsk, u32 *last_publ)
3743 {
3744 	int err;
3745 	struct publication *p;
3746 
3747 	if (*last_publ) {
3748 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3749 			if (p->key == *last_publ)
3750 				break;
3751 		}
3752 		if (p->key != *last_publ) {
3753 			/* We never set seq or call nl_dump_check_consistent()
3754 			 * this means that setting prev_seq here will cause the
3755 			 * consistence check to fail in the netlink callback
3756 			 * handler. Resulting in the last NLMSG_DONE message
3757 			 * having the NLM_F_DUMP_INTR flag set.
3758 			 */
3759 			cb->prev_seq = 1;
3760 			*last_publ = 0;
3761 			return -EPIPE;
3762 		}
3763 	} else {
3764 		p = list_first_entry(&tsk->publications, struct publication,
3765 				     binding_sock);
3766 	}
3767 
3768 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3769 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3770 		if (err) {
3771 			*last_publ = p->key;
3772 			return err;
3773 		}
3774 	}
3775 	*last_publ = 0;
3776 
3777 	return 0;
3778 }
3779 
3780 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3781 {
3782 	int err;
3783 	u32 tsk_portid = cb->args[0];
3784 	u32 last_publ = cb->args[1];
3785 	u32 done = cb->args[2];
3786 	struct net *net = sock_net(skb->sk);
3787 	struct tipc_sock *tsk;
3788 
3789 	if (!tsk_portid) {
3790 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3791 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3792 
3793 		if (!attrs[TIPC_NLA_SOCK])
3794 			return -EINVAL;
3795 
3796 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3797 						  attrs[TIPC_NLA_SOCK],
3798 						  tipc_nl_sock_policy, NULL);
3799 		if (err)
3800 			return err;
3801 
3802 		if (!sock[TIPC_NLA_SOCK_REF])
3803 			return -EINVAL;
3804 
3805 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3806 	}
3807 
3808 	if (done)
3809 		return 0;
3810 
3811 	tsk = tipc_sk_lookup(net, tsk_portid);
3812 	if (!tsk)
3813 		return -EINVAL;
3814 
3815 	lock_sock(&tsk->sk);
3816 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3817 	if (!err)
3818 		done = 1;
3819 	release_sock(&tsk->sk);
3820 	sock_put(&tsk->sk);
3821 
3822 	cb->args[0] = tsk_portid;
3823 	cb->args[1] = last_publ;
3824 	cb->args[2] = done;
3825 
3826 	return skb->len;
3827 }
3828 
3829 /**
3830  * tipc_sk_filtering - check if a socket should be traced
3831  * @sk: the socket to be examined
3832  *
3833  * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3834  * (portid, sock type, name type, name lower, name upper)
3835  *
3836  * Return: true if the socket meets the socket tuple data
3837  * (value 0 = 'any') or when there is no tuple set (all = 0),
3838  * otherwise false
3839  */
3840 bool tipc_sk_filtering(struct sock *sk)
3841 {
3842 	struct tipc_sock *tsk;
3843 	struct publication *p;
3844 	u32 _port, _sktype, _type, _lower, _upper;
3845 	u32 type = 0, lower = 0, upper = 0;
3846 
3847 	if (!sk)
3848 		return true;
3849 
3850 	tsk = tipc_sk(sk);
3851 
3852 	_port = sysctl_tipc_sk_filter[0];
3853 	_sktype = sysctl_tipc_sk_filter[1];
3854 	_type = sysctl_tipc_sk_filter[2];
3855 	_lower = sysctl_tipc_sk_filter[3];
3856 	_upper = sysctl_tipc_sk_filter[4];
3857 
3858 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3859 		return true;
3860 
3861 	if (_port)
3862 		return (_port == tsk->portid);
3863 
3864 	if (_sktype && _sktype != sk->sk_type)
3865 		return false;
3866 
3867 	if (tsk->published) {
3868 		p = list_first_entry_or_null(&tsk->publications,
3869 					     struct publication, binding_sock);
3870 		if (p) {
3871 			type = p->sr.type;
3872 			lower = p->sr.lower;
3873 			upper = p->sr.upper;
3874 		}
3875 	}
3876 
3877 	if (!tipc_sk_type_connectionless(sk)) {
3878 		type = msg_nametype(&tsk->phdr);
3879 		lower = msg_nameinst(&tsk->phdr);
3880 		upper = lower;
3881 	}
3882 
3883 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3884 	    (_upper && _upper != upper))
3885 		return false;
3886 
3887 	return true;
3888 }
3889 
3890 u32 tipc_sock_get_portid(struct sock *sk)
3891 {
3892 	return (sk) ? (tipc_sk(sk))->portid : 0;
3893 }
3894 
3895 /**
3896  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3897  *			both the rcv and backlog queues are considered
3898  * @sk: tipc sk to be checked
3899  * @skb: tipc msg to be checked
3900  *
3901  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3902  */
3903 
3904 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3905 {
3906 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3907 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3908 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3909 
3910 	return (qsize > lim * 90 / 100);
3911 }
3912 
3913 /**
3914  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3915  *			only the rcv queue is considered
3916  * @sk: tipc sk to be checked
3917  * @skb: tipc msg to be checked
3918  *
3919  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3920  */
3921 
3922 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3923 {
3924 	unsigned int lim = rcvbuf_limit(sk, skb);
3925 	unsigned int qsize = sk_rmem_alloc_get(sk);
3926 
3927 	return (qsize > lim * 90 / 100);
3928 }
3929 
3930 /**
3931  * tipc_sk_dump - dump TIPC socket
3932  * @sk: tipc sk to be dumped
3933  * @dqueues: bitmask to decide if any socket queue to be dumped?
3934  *           - TIPC_DUMP_NONE: don't dump socket queues
3935  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3936  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3937  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3938  *           - TIPC_DUMP_ALL: dump all the socket queues above
3939  * @buf: returned buffer of dump data in format
3940  */
3941 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3942 {
3943 	int i = 0;
3944 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3945 	u32 conn_type, conn_instance;
3946 	struct tipc_sock *tsk;
3947 	struct publication *p;
3948 	bool tsk_connected;
3949 
3950 	if (!sk) {
3951 		i += scnprintf(buf, sz, "sk data: (null)\n");
3952 		return i;
3953 	}
3954 
3955 	tsk = tipc_sk(sk);
3956 	tsk_connected = !tipc_sk_type_connectionless(sk);
3957 
3958 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3959 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3960 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3961 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3962 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3963 	if (tsk_connected) {
3964 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3965 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3966 		conn_type = msg_nametype(&tsk->phdr);
3967 		conn_instance = msg_nameinst(&tsk->phdr);
3968 		i += scnprintf(buf + i, sz - i, " %u", conn_type);
3969 		i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3970 	}
3971 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3972 	if (tsk->published) {
3973 		p = list_first_entry_or_null(&tsk->publications,
3974 					     struct publication, binding_sock);
3975 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3976 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3977 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3978 	}
3979 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3980 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3981 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3982 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3983 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3984 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3985 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3986 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3987 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3988 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3989 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3990 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3991 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3992 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3993 
3994 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3995 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3996 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3997 	}
3998 
3999 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
4000 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
4001 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
4002 	}
4003 
4004 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4005 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
4006 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4007 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4008 			i += scnprintf(buf + i, sz - i, "  tail ");
4009 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
4010 					   buf + i);
4011 		}
4012 	}
4013 
4014 	return i;
4015 }
4016