xref: /openbmc/linux/net/tipc/socket.c (revision 5b448065)
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  * @conn_type: TIPC type used when connection was established
77  * @conn_instance: TIPC instance used when connection was established
78  * @published: non-zero if port has one or more associated names
79  * @max_pkt: maximum packet size "hint" used when building messages sent by port
80  * @maxnagle: maximum size of msg which can be subject to nagle
81  * @portid: unique port identity in TIPC socket hash table
82  * @phdr: preformatted message header used when sending messages
83  * @cong_links: list of congested links
84  * @publications: list of publications for port
85  * @blocking_link: address of the congested link we are currently sleeping on
86  * @pub_count: total # of publications port has made during its lifetime
87  * @conn_timeout: the time we can wait for an unresponded setup request
88  * @probe_unacked: probe has not received ack yet
89  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
90  * @cong_link_cnt: number of congested links
91  * @snt_unacked: # messages sent by socket, and not yet acked by peer
92  * @snd_win: send window size
93  * @peer_caps: peer capabilities mask
94  * @rcv_unacked: # messages read by user, but not yet acked back to peer
95  * @rcv_win: receive window size
96  * @peer: 'connected' peer for dgram/rdm
97  * @node: hash table node
98  * @mc_method: cookie for use between socket and broadcast layer
99  * @rcu: rcu struct for tipc_sock
100  * @group: TIPC communications group
101  * @oneway: message count in one direction (FIXME)
102  * @nagle_start: current nagle value
103  * @snd_backlog: send backlog count
104  * @msg_acc: messages accepted; used in managing backlog and nagle
105  * @pkt_cnt: TIPC socket packet count
106  * @expect_ack: whether this TIPC socket is expecting an ack
107  * @nodelay: setsockopt() TIPC_NODELAY setting
108  * @group_is_open: TIPC socket group is fully open (FIXME)
109  */
110 struct tipc_sock {
111 	struct sock sk;
112 	u32 conn_type;
113 	u32 conn_instance;
114 	u32 max_pkt;
115 	u32 maxnagle;
116 	u32 portid;
117 	struct tipc_msg phdr;
118 	struct list_head cong_links;
119 	struct list_head publications;
120 	u32 pub_count;
121 	atomic_t dupl_rcvcnt;
122 	u16 conn_timeout;
123 	bool probe_unacked;
124 	u16 cong_link_cnt;
125 	u16 snt_unacked;
126 	u16 snd_win;
127 	u16 peer_caps;
128 	u16 rcv_unacked;
129 	u16 rcv_win;
130 	struct sockaddr_tipc peer;
131 	struct rhash_head node;
132 	struct tipc_mc_method mc_method;
133 	struct rcu_head rcu;
134 	struct tipc_group *group;
135 	u32 oneway;
136 	u32 nagle_start;
137 	u16 snd_backlog;
138 	u16 msg_acc;
139 	u16 pkt_cnt;
140 	bool expect_ack;
141 	bool nodelay;
142 	bool group_is_open;
143 	bool published;
144 };
145 
146 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
147 static void tipc_data_ready(struct sock *sk);
148 static void tipc_write_space(struct sock *sk);
149 static void tipc_sock_destruct(struct sock *sk);
150 static int tipc_release(struct socket *sock);
151 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
152 		       bool kern);
153 static void tipc_sk_timeout(struct timer_list *t);
154 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
155 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
156 static int tipc_sk_leave(struct tipc_sock *tsk);
157 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
158 static int tipc_sk_insert(struct tipc_sock *tsk);
159 static void tipc_sk_remove(struct tipc_sock *tsk);
160 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
161 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
162 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
163 
164 static const struct proto_ops packet_ops;
165 static const struct proto_ops stream_ops;
166 static const struct proto_ops msg_ops;
167 static struct proto tipc_proto;
168 static const struct rhashtable_params tsk_rht_params;
169 
170 static u32 tsk_own_node(struct tipc_sock *tsk)
171 {
172 	return msg_prevnode(&tsk->phdr);
173 }
174 
175 static u32 tsk_peer_node(struct tipc_sock *tsk)
176 {
177 	return msg_destnode(&tsk->phdr);
178 }
179 
180 static u32 tsk_peer_port(struct tipc_sock *tsk)
181 {
182 	return msg_destport(&tsk->phdr);
183 }
184 
185 static  bool tsk_unreliable(struct tipc_sock *tsk)
186 {
187 	return msg_src_droppable(&tsk->phdr) != 0;
188 }
189 
190 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
191 {
192 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
193 }
194 
195 static bool tsk_unreturnable(struct tipc_sock *tsk)
196 {
197 	return msg_dest_droppable(&tsk->phdr) != 0;
198 }
199 
200 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
201 {
202 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
203 }
204 
205 static int tsk_importance(struct tipc_sock *tsk)
206 {
207 	return msg_importance(&tsk->phdr);
208 }
209 
210 static struct tipc_sock *tipc_sk(const struct sock *sk)
211 {
212 	return container_of(sk, struct tipc_sock, sk);
213 }
214 
215 int tsk_set_importance(struct sock *sk, int imp)
216 {
217 	if (imp > TIPC_CRITICAL_IMPORTANCE)
218 		return -EINVAL;
219 	msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
220 	return 0;
221 }
222 
223 static bool tsk_conn_cong(struct tipc_sock *tsk)
224 {
225 	return tsk->snt_unacked > tsk->snd_win;
226 }
227 
228 static u16 tsk_blocks(int len)
229 {
230 	return ((len / FLOWCTL_BLK_SZ) + 1);
231 }
232 
233 /* tsk_blocks(): translate a buffer size in bytes to number of
234  * advertisable blocks, taking into account the ratio truesize(len)/len
235  * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
236  */
237 static u16 tsk_adv_blocks(int len)
238 {
239 	return len / FLOWCTL_BLK_SZ / 4;
240 }
241 
242 /* tsk_inc(): increment counter for sent or received data
243  * - If block based flow control is not supported by peer we
244  *   fall back to message based ditto, incrementing the counter
245  */
246 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
247 {
248 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
249 		return ((msglen / FLOWCTL_BLK_SZ) + 1);
250 	return 1;
251 }
252 
253 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
254  */
255 static void tsk_set_nagle(struct tipc_sock *tsk)
256 {
257 	struct sock *sk = &tsk->sk;
258 
259 	tsk->maxnagle = 0;
260 	if (sk->sk_type != SOCK_STREAM)
261 		return;
262 	if (tsk->nodelay)
263 		return;
264 	if (!(tsk->peer_caps & TIPC_NAGLE))
265 		return;
266 	/* Limit node local buffer size to avoid receive queue overflow */
267 	if (tsk->max_pkt == MAX_MSG_SIZE)
268 		tsk->maxnagle = 1500;
269 	else
270 		tsk->maxnagle = tsk->max_pkt;
271 }
272 
273 /**
274  * tsk_advance_rx_queue - discard first buffer in socket receive queue
275  * @sk: network socket
276  *
277  * Caller must hold socket lock
278  */
279 static void tsk_advance_rx_queue(struct sock *sk)
280 {
281 	trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
282 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
283 }
284 
285 /* tipc_sk_respond() : send response message back to sender
286  */
287 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
288 {
289 	u32 selector;
290 	u32 dnode;
291 	u32 onode = tipc_own_addr(sock_net(sk));
292 
293 	if (!tipc_msg_reverse(onode, &skb, err))
294 		return;
295 
296 	trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
297 	dnode = msg_destnode(buf_msg(skb));
298 	selector = msg_origport(buf_msg(skb));
299 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
300 }
301 
302 /**
303  * tsk_rej_rx_queue - reject all buffers in socket receive queue
304  * @sk: network socket
305  * @error: response error code
306  *
307  * Caller must hold socket lock
308  */
309 static void tsk_rej_rx_queue(struct sock *sk, int error)
310 {
311 	struct sk_buff *skb;
312 
313 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
314 		tipc_sk_respond(sk, skb, error);
315 }
316 
317 static bool tipc_sk_connected(struct sock *sk)
318 {
319 	return sk->sk_state == TIPC_ESTABLISHED;
320 }
321 
322 /* tipc_sk_type_connectionless - check if the socket is datagram socket
323  * @sk: socket
324  *
325  * Returns true if connection less, false otherwise
326  */
327 static bool tipc_sk_type_connectionless(struct sock *sk)
328 {
329 	return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
330 }
331 
332 /* tsk_peer_msg - verify if message was sent by connected port's peer
333  *
334  * Handles cases where the node's network address has changed from
335  * the default of <0.0.0> to its configured setting.
336  */
337 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
338 {
339 	struct sock *sk = &tsk->sk;
340 	u32 self = tipc_own_addr(sock_net(sk));
341 	u32 peer_port = tsk_peer_port(tsk);
342 	u32 orig_node, peer_node;
343 
344 	if (unlikely(!tipc_sk_connected(sk)))
345 		return false;
346 
347 	if (unlikely(msg_origport(msg) != peer_port))
348 		return false;
349 
350 	orig_node = msg_orignode(msg);
351 	peer_node = tsk_peer_node(tsk);
352 
353 	if (likely(orig_node == peer_node))
354 		return true;
355 
356 	if (!orig_node && peer_node == self)
357 		return true;
358 
359 	if (!peer_node && orig_node == self)
360 		return true;
361 
362 	return false;
363 }
364 
365 /* tipc_set_sk_state - set the sk_state of the socket
366  * @sk: socket
367  *
368  * Caller must hold socket lock
369  *
370  * Returns 0 on success, errno otherwise
371  */
372 static int tipc_set_sk_state(struct sock *sk, int state)
373 {
374 	int oldsk_state = sk->sk_state;
375 	int res = -EINVAL;
376 
377 	switch (state) {
378 	case TIPC_OPEN:
379 		res = 0;
380 		break;
381 	case TIPC_LISTEN:
382 	case TIPC_CONNECTING:
383 		if (oldsk_state == TIPC_OPEN)
384 			res = 0;
385 		break;
386 	case TIPC_ESTABLISHED:
387 		if (oldsk_state == TIPC_CONNECTING ||
388 		    oldsk_state == TIPC_OPEN)
389 			res = 0;
390 		break;
391 	case TIPC_DISCONNECTING:
392 		if (oldsk_state == TIPC_CONNECTING ||
393 		    oldsk_state == TIPC_ESTABLISHED)
394 			res = 0;
395 		break;
396 	}
397 
398 	if (!res)
399 		sk->sk_state = state;
400 
401 	return res;
402 }
403 
404 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
405 {
406 	struct sock *sk = sock->sk;
407 	int err = sock_error(sk);
408 	int typ = sock->type;
409 
410 	if (err)
411 		return err;
412 	if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
413 		if (sk->sk_state == TIPC_DISCONNECTING)
414 			return -EPIPE;
415 		else if (!tipc_sk_connected(sk))
416 			return -ENOTCONN;
417 	}
418 	if (!*timeout)
419 		return -EAGAIN;
420 	if (signal_pending(current))
421 		return sock_intr_errno(*timeout);
422 
423 	return 0;
424 }
425 
426 #define tipc_wait_for_cond(sock_, timeo_, condition_)			       \
427 ({                                                                             \
428 	DEFINE_WAIT_FUNC(wait_, woken_wake_function);                          \
429 	struct sock *sk_;						       \
430 	int rc_;							       \
431 									       \
432 	while ((rc_ = !(condition_))) {					       \
433 		/* coupled with smp_wmb() in tipc_sk_proto_rcv() */            \
434 		smp_rmb();                                                     \
435 		sk_ = (sock_)->sk;					       \
436 		rc_ = tipc_sk_sock_err((sock_), timeo_);		       \
437 		if (rc_)						       \
438 			break;						       \
439 		add_wait_queue(sk_sleep(sk_), &wait_);                         \
440 		release_sock(sk_);					       \
441 		*(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
442 		sched_annotate_sleep();				               \
443 		lock_sock(sk_);						       \
444 		remove_wait_queue(sk_sleep(sk_), &wait_);		       \
445 	}								       \
446 	rc_;								       \
447 })
448 
449 /**
450  * tipc_sk_create - create a TIPC socket
451  * @net: network namespace (must be default network)
452  * @sock: pre-allocated socket structure
453  * @protocol: protocol indicator (must be 0)
454  * @kern: caused by kernel or by userspace?
455  *
456  * This routine creates additional data structures used by the TIPC socket,
457  * initializes them, and links them together.
458  *
459  * Return: 0 on success, errno otherwise
460  */
461 static int tipc_sk_create(struct net *net, struct socket *sock,
462 			  int protocol, int kern)
463 {
464 	const struct proto_ops *ops;
465 	struct sock *sk;
466 	struct tipc_sock *tsk;
467 	struct tipc_msg *msg;
468 
469 	/* Validate arguments */
470 	if (unlikely(protocol != 0))
471 		return -EPROTONOSUPPORT;
472 
473 	switch (sock->type) {
474 	case SOCK_STREAM:
475 		ops = &stream_ops;
476 		break;
477 	case SOCK_SEQPACKET:
478 		ops = &packet_ops;
479 		break;
480 	case SOCK_DGRAM:
481 	case SOCK_RDM:
482 		ops = &msg_ops;
483 		break;
484 	default:
485 		return -EPROTOTYPE;
486 	}
487 
488 	/* Allocate socket's protocol area */
489 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
490 	if (sk == NULL)
491 		return -ENOMEM;
492 
493 	tsk = tipc_sk(sk);
494 	tsk->max_pkt = MAX_PKT_DEFAULT;
495 	tsk->maxnagle = 0;
496 	tsk->nagle_start = NAGLE_START_INIT;
497 	INIT_LIST_HEAD(&tsk->publications);
498 	INIT_LIST_HEAD(&tsk->cong_links);
499 	msg = &tsk->phdr;
500 
501 	/* Finish initializing socket data structures */
502 	sock->ops = ops;
503 	sock_init_data(sock, sk);
504 	tipc_set_sk_state(sk, TIPC_OPEN);
505 	if (tipc_sk_insert(tsk)) {
506 		pr_warn("Socket create failed; port number exhausted\n");
507 		return -EINVAL;
508 	}
509 
510 	/* Ensure tsk is visible before we read own_addr. */
511 	smp_mb();
512 
513 	tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
514 		      TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
515 
516 	msg_set_origport(msg, tsk->portid);
517 	timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
518 	sk->sk_shutdown = 0;
519 	sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
520 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
521 	sk->sk_data_ready = tipc_data_ready;
522 	sk->sk_write_space = tipc_write_space;
523 	sk->sk_destruct = tipc_sock_destruct;
524 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
525 	tsk->group_is_open = true;
526 	atomic_set(&tsk->dupl_rcvcnt, 0);
527 
528 	/* Start out with safe limits until we receive an advertised window */
529 	tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
530 	tsk->rcv_win = tsk->snd_win;
531 
532 	if (tipc_sk_type_connectionless(sk)) {
533 		tsk_set_unreturnable(tsk, true);
534 		if (sock->type == SOCK_DGRAM)
535 			tsk_set_unreliable(tsk, true);
536 	}
537 	__skb_queue_head_init(&tsk->mc_method.deferredq);
538 	trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
539 	return 0;
540 }
541 
542 static void tipc_sk_callback(struct rcu_head *head)
543 {
544 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
545 
546 	sock_put(&tsk->sk);
547 }
548 
549 /* Caller should hold socket lock for the socket. */
550 static void __tipc_shutdown(struct socket *sock, int error)
551 {
552 	struct sock *sk = sock->sk;
553 	struct tipc_sock *tsk = tipc_sk(sk);
554 	struct net *net = sock_net(sk);
555 	long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
556 	u32 dnode = tsk_peer_node(tsk);
557 	struct sk_buff *skb;
558 
559 	/* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
560 	tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
561 					    !tsk_conn_cong(tsk)));
562 
563 	/* Push out delayed messages if in Nagle mode */
564 	tipc_sk_push_backlog(tsk, false);
565 	/* Remove pending SYN */
566 	__skb_queue_purge(&sk->sk_write_queue);
567 
568 	/* Remove partially received buffer if any */
569 	skb = skb_peek(&sk->sk_receive_queue);
570 	if (skb && TIPC_SKB_CB(skb)->bytes_read) {
571 		__skb_unlink(skb, &sk->sk_receive_queue);
572 		kfree_skb(skb);
573 	}
574 
575 	/* Reject all unreceived messages if connectionless */
576 	if (tipc_sk_type_connectionless(sk)) {
577 		tsk_rej_rx_queue(sk, error);
578 		return;
579 	}
580 
581 	switch (sk->sk_state) {
582 	case TIPC_CONNECTING:
583 	case TIPC_ESTABLISHED:
584 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
585 		tipc_node_remove_conn(net, dnode, tsk->portid);
586 		/* Send a FIN+/- to its peer */
587 		skb = __skb_dequeue(&sk->sk_receive_queue);
588 		if (skb) {
589 			__skb_queue_purge(&sk->sk_receive_queue);
590 			tipc_sk_respond(sk, skb, error);
591 			break;
592 		}
593 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
594 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
595 				      tsk_own_node(tsk), tsk_peer_port(tsk),
596 				      tsk->portid, error);
597 		if (skb)
598 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
599 		break;
600 	case TIPC_LISTEN:
601 		/* Reject all SYN messages */
602 		tsk_rej_rx_queue(sk, error);
603 		break;
604 	default:
605 		__skb_queue_purge(&sk->sk_receive_queue);
606 		break;
607 	}
608 }
609 
610 /**
611  * tipc_release - destroy a TIPC socket
612  * @sock: socket to destroy
613  *
614  * This routine cleans up any messages that are still queued on the socket.
615  * For DGRAM and RDM socket types, all queued messages are rejected.
616  * For SEQPACKET and STREAM socket types, the first message is rejected
617  * and any others are discarded.  (If the first message on a STREAM socket
618  * is partially-read, it is discarded and the next one is rejected instead.)
619  *
620  * NOTE: Rejected messages are not necessarily returned to the sender!  They
621  * are returned or discarded according to the "destination droppable" setting
622  * specified for the message by the sender.
623  *
624  * Return: 0 on success, errno otherwise
625  */
626 static int tipc_release(struct socket *sock)
627 {
628 	struct sock *sk = sock->sk;
629 	struct tipc_sock *tsk;
630 
631 	/*
632 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
633 	 * releases a pre-allocated child socket that was never used)
634 	 */
635 	if (sk == NULL)
636 		return 0;
637 
638 	tsk = tipc_sk(sk);
639 	lock_sock(sk);
640 
641 	trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
642 	__tipc_shutdown(sock, TIPC_ERR_NO_PORT);
643 	sk->sk_shutdown = SHUTDOWN_MASK;
644 	tipc_sk_leave(tsk);
645 	tipc_sk_withdraw(tsk, NULL);
646 	__skb_queue_purge(&tsk->mc_method.deferredq);
647 	sk_stop_timer(sk, &sk->sk_timer);
648 	tipc_sk_remove(tsk);
649 
650 	sock_orphan(sk);
651 	/* Reject any messages that accumulated in backlog queue */
652 	release_sock(sk);
653 	tipc_dest_list_purge(&tsk->cong_links);
654 	tsk->cong_link_cnt = 0;
655 	call_rcu(&tsk->rcu, tipc_sk_callback);
656 	sock->sk = NULL;
657 
658 	return 0;
659 }
660 
661 /**
662  * __tipc_bind - associate or disassocate TIPC name(s) with a socket
663  * @sock: socket structure
664  * @skaddr: socket address describing name(s) and desired operation
665  * @alen: size of socket address data structure
666  *
667  * Name and name sequence binding is indicated using a positive scope value;
668  * a negative scope value unbinds the specified name.  Specifying no name
669  * (i.e. a socket address length of 0) unbinds all names from the socket.
670  *
671  * Return: 0 on success, errno otherwise
672  *
673  * NOTE: This routine doesn't need to take the socket lock since it doesn't
674  *       access any non-constant socket information.
675  */
676 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
677 {
678 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
679 	struct tipc_sock *tsk = tipc_sk(sock->sk);
680 	bool unbind = false;
681 
682 	if (unlikely(!alen))
683 		return tipc_sk_withdraw(tsk, NULL);
684 
685 	if (ua->addrtype == TIPC_SERVICE_ADDR) {
686 		ua->addrtype = TIPC_SERVICE_RANGE;
687 		ua->sr.upper = ua->sr.lower;
688 	}
689 	if (ua->scope < 0) {
690 		unbind = true;
691 		ua->scope = -ua->scope;
692 	}
693 	/* Users may still use deprecated TIPC_ZONE_SCOPE */
694 	if (ua->scope != TIPC_NODE_SCOPE)
695 		ua->scope = TIPC_CLUSTER_SCOPE;
696 
697 	if (tsk->group)
698 		return -EACCES;
699 
700 	if (unbind)
701 		return tipc_sk_withdraw(tsk, ua);
702 	return tipc_sk_publish(tsk, ua);
703 }
704 
705 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
706 {
707 	int res;
708 
709 	lock_sock(sock->sk);
710 	res = __tipc_bind(sock, skaddr, alen);
711 	release_sock(sock->sk);
712 	return res;
713 }
714 
715 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
716 {
717 	struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
718 	u32 atype = ua->addrtype;
719 
720 	if (alen) {
721 		if (!tipc_uaddr_valid(ua, alen))
722 			return -EINVAL;
723 		if (atype == TIPC_SOCKET_ADDR)
724 			return -EAFNOSUPPORT;
725 		if (ua->sr.type < TIPC_RESERVED_TYPES) {
726 			pr_warn_once("Can't bind to reserved service type %u\n",
727 				     ua->sr.type);
728 			return -EACCES;
729 		}
730 	}
731 	return tipc_sk_bind(sock, skaddr, alen);
732 }
733 
734 /**
735  * tipc_getname - get port ID of socket or peer socket
736  * @sock: socket structure
737  * @uaddr: area for returned socket address
738  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
739  *
740  * Return: 0 on success, errno otherwise
741  *
742  * NOTE: This routine doesn't need to take the socket lock since it only
743  *       accesses socket information that is unchanging (or which changes in
744  *       a completely predictable manner).
745  */
746 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
747 			int peer)
748 {
749 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
750 	struct sock *sk = sock->sk;
751 	struct tipc_sock *tsk = tipc_sk(sk);
752 
753 	memset(addr, 0, sizeof(*addr));
754 	if (peer) {
755 		if ((!tipc_sk_connected(sk)) &&
756 		    ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
757 			return -ENOTCONN;
758 		addr->addr.id.ref = tsk_peer_port(tsk);
759 		addr->addr.id.node = tsk_peer_node(tsk);
760 	} else {
761 		addr->addr.id.ref = tsk->portid;
762 		addr->addr.id.node = tipc_own_addr(sock_net(sk));
763 	}
764 
765 	addr->addrtype = TIPC_SOCKET_ADDR;
766 	addr->family = AF_TIPC;
767 	addr->scope = 0;
768 	addr->addr.name.domain = 0;
769 
770 	return sizeof(*addr);
771 }
772 
773 /**
774  * tipc_poll - read and possibly block on pollmask
775  * @file: file structure associated with the socket
776  * @sock: socket for which to calculate the poll bits
777  * @wait: ???
778  *
779  * Return: pollmask value
780  *
781  * COMMENTARY:
782  * It appears that the usual socket locking mechanisms are not useful here
783  * since the pollmask info is potentially out-of-date the moment this routine
784  * exits.  TCP and other protocols seem to rely on higher level poll routines
785  * to handle any preventable race conditions, so TIPC will do the same ...
786  *
787  * IMPORTANT: The fact that a read or write operation is indicated does NOT
788  * imply that the operation will succeed, merely that it should be performed
789  * and will not block.
790  */
791 static __poll_t tipc_poll(struct file *file, struct socket *sock,
792 			      poll_table *wait)
793 {
794 	struct sock *sk = sock->sk;
795 	struct tipc_sock *tsk = tipc_sk(sk);
796 	__poll_t revents = 0;
797 
798 	sock_poll_wait(file, sock, wait);
799 	trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
800 
801 	if (sk->sk_shutdown & RCV_SHUTDOWN)
802 		revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
803 	if (sk->sk_shutdown == SHUTDOWN_MASK)
804 		revents |= EPOLLHUP;
805 
806 	switch (sk->sk_state) {
807 	case TIPC_ESTABLISHED:
808 		if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
809 			revents |= EPOLLOUT;
810 		fallthrough;
811 	case TIPC_LISTEN:
812 	case TIPC_CONNECTING:
813 		if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
814 			revents |= EPOLLIN | EPOLLRDNORM;
815 		break;
816 	case TIPC_OPEN:
817 		if (tsk->group_is_open && !tsk->cong_link_cnt)
818 			revents |= EPOLLOUT;
819 		if (!tipc_sk_type_connectionless(sk))
820 			break;
821 		if (skb_queue_empty_lockless(&sk->sk_receive_queue))
822 			break;
823 		revents |= EPOLLIN | EPOLLRDNORM;
824 		break;
825 	case TIPC_DISCONNECTING:
826 		revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
827 		break;
828 	}
829 	return revents;
830 }
831 
832 /**
833  * tipc_sendmcast - send multicast message
834  * @sock: socket structure
835  * @ua: destination address struct
836  * @msg: message to send
837  * @dlen: length of data to send
838  * @timeout: timeout to wait for wakeup
839  *
840  * Called from function tipc_sendmsg(), which has done all sanity checks
841  * Return: the number of bytes sent on success, or errno
842  */
843 static int tipc_sendmcast(struct  socket *sock, struct tipc_uaddr *ua,
844 			  struct msghdr *msg, size_t dlen, long timeout)
845 {
846 	struct sock *sk = sock->sk;
847 	struct tipc_sock *tsk = tipc_sk(sk);
848 	struct tipc_msg *hdr = &tsk->phdr;
849 	struct net *net = sock_net(sk);
850 	int mtu = tipc_bcast_get_mtu(net);
851 	struct sk_buff_head pkts;
852 	struct tipc_nlist dsts;
853 	int rc;
854 
855 	if (tsk->group)
856 		return -EACCES;
857 
858 	/* Block or return if any destination link is congested */
859 	rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
860 	if (unlikely(rc))
861 		return rc;
862 
863 	/* Lookup destination nodes */
864 	tipc_nlist_init(&dsts, tipc_own_addr(net));
865 	tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
866 	if (!dsts.local && !dsts.remote)
867 		return -EHOSTUNREACH;
868 
869 	/* Build message header */
870 	msg_set_type(hdr, TIPC_MCAST_MSG);
871 	msg_set_hdr_sz(hdr, MCAST_H_SIZE);
872 	msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
873 	msg_set_destport(hdr, 0);
874 	msg_set_destnode(hdr, 0);
875 	msg_set_nametype(hdr, ua->sr.type);
876 	msg_set_namelower(hdr, ua->sr.lower);
877 	msg_set_nameupper(hdr, ua->sr.upper);
878 
879 	/* Build message as chain of buffers */
880 	__skb_queue_head_init(&pkts);
881 	rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
882 
883 	/* Send message if build was successful */
884 	if (unlikely(rc == dlen)) {
885 		trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
886 					TIPC_DUMP_SK_SNDQ, " ");
887 		rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
888 				     &tsk->cong_link_cnt);
889 	}
890 
891 	tipc_nlist_purge(&dsts);
892 
893 	return rc ? rc : dlen;
894 }
895 
896 /**
897  * tipc_send_group_msg - send a message to a member in the group
898  * @net: network namespace
899  * @tsk: tipc socket
900  * @m: message to send
901  * @mb: group member
902  * @dnode: destination node
903  * @dport: destination port
904  * @dlen: total length of message data
905  */
906 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
907 			       struct msghdr *m, struct tipc_member *mb,
908 			       u32 dnode, u32 dport, int dlen)
909 {
910 	u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
911 	struct tipc_mc_method *method = &tsk->mc_method;
912 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
913 	struct tipc_msg *hdr = &tsk->phdr;
914 	struct sk_buff_head pkts;
915 	int mtu, rc;
916 
917 	/* Complete message header */
918 	msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
919 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
920 	msg_set_destport(hdr, dport);
921 	msg_set_destnode(hdr, dnode);
922 	msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
923 
924 	/* Build message as chain of buffers */
925 	__skb_queue_head_init(&pkts);
926 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
927 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
928 	if (unlikely(rc != dlen))
929 		return rc;
930 
931 	/* Send message */
932 	rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
933 	if (unlikely(rc == -ELINKCONG)) {
934 		tipc_dest_push(&tsk->cong_links, dnode, 0);
935 		tsk->cong_link_cnt++;
936 	}
937 
938 	/* Update send window */
939 	tipc_group_update_member(mb, blks);
940 
941 	/* A broadcast sent within next EXPIRE period must follow same path */
942 	method->rcast = true;
943 	method->mandatory = true;
944 	return dlen;
945 }
946 
947 /**
948  * tipc_send_group_unicast - send message to a member in the group
949  * @sock: socket structure
950  * @m: message to send
951  * @dlen: total length of message data
952  * @timeout: timeout to wait for wakeup
953  *
954  * Called from function tipc_sendmsg(), which has done all sanity checks
955  * Return: the number of bytes sent on success, or errno
956  */
957 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
958 				   int dlen, long timeout)
959 {
960 	struct sock *sk = sock->sk;
961 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
962 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
963 	struct tipc_sock *tsk = tipc_sk(sk);
964 	struct net *net = sock_net(sk);
965 	struct tipc_member *mb = NULL;
966 	u32 node, port;
967 	int rc;
968 
969 	node = ua->sk.node;
970 	port = ua->sk.ref;
971 	if (!port && !node)
972 		return -EHOSTUNREACH;
973 
974 	/* Block or return if destination link or member is congested */
975 	rc = tipc_wait_for_cond(sock, &timeout,
976 				!tipc_dest_find(&tsk->cong_links, node, 0) &&
977 				tsk->group &&
978 				!tipc_group_cong(tsk->group, node, port, blks,
979 						 &mb));
980 	if (unlikely(rc))
981 		return rc;
982 
983 	if (unlikely(!mb))
984 		return -EHOSTUNREACH;
985 
986 	rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
987 
988 	return rc ? rc : dlen;
989 }
990 
991 /**
992  * tipc_send_group_anycast - send message to any member with given identity
993  * @sock: socket structure
994  * @m: message to send
995  * @dlen: total length of message data
996  * @timeout: timeout to wait for wakeup
997  *
998  * Called from function tipc_sendmsg(), which has done all sanity checks
999  * Return: the number of bytes sent on success, or errno
1000  */
1001 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1002 				   int dlen, long timeout)
1003 {
1004 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1005 	struct sock *sk = sock->sk;
1006 	struct tipc_sock *tsk = tipc_sk(sk);
1007 	struct list_head *cong_links = &tsk->cong_links;
1008 	int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1009 	struct tipc_msg *hdr = &tsk->phdr;
1010 	struct tipc_member *first = NULL;
1011 	struct tipc_member *mbr = NULL;
1012 	struct net *net = sock_net(sk);
1013 	u32 node, port, exclude;
1014 	struct list_head dsts;
1015 	int lookups = 0;
1016 	int dstcnt, rc;
1017 	bool cong;
1018 
1019 	INIT_LIST_HEAD(&dsts);
1020 	ua->sa.type = msg_nametype(hdr);
1021 	ua->scope = msg_lookup_scope(hdr);
1022 
1023 	while (++lookups < 4) {
1024 		exclude = tipc_group_exclude(tsk->group);
1025 
1026 		first = NULL;
1027 
1028 		/* Look for a non-congested destination member, if any */
1029 		while (1) {
1030 			if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1031 						       exclude, false))
1032 				return -EHOSTUNREACH;
1033 			tipc_dest_pop(&dsts, &node, &port);
1034 			cong = tipc_group_cong(tsk->group, node, port, blks,
1035 					       &mbr);
1036 			if (!cong)
1037 				break;
1038 			if (mbr == first)
1039 				break;
1040 			if (!first)
1041 				first = mbr;
1042 		}
1043 
1044 		/* Start over if destination was not in member list */
1045 		if (unlikely(!mbr))
1046 			continue;
1047 
1048 		if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1049 			break;
1050 
1051 		/* Block or return if destination link or member is congested */
1052 		rc = tipc_wait_for_cond(sock, &timeout,
1053 					!tipc_dest_find(cong_links, node, 0) &&
1054 					tsk->group &&
1055 					!tipc_group_cong(tsk->group, node, port,
1056 							 blks, &mbr));
1057 		if (unlikely(rc))
1058 			return rc;
1059 
1060 		/* Send, unless destination disappeared while waiting */
1061 		if (likely(mbr))
1062 			break;
1063 	}
1064 
1065 	if (unlikely(lookups >= 4))
1066 		return -EHOSTUNREACH;
1067 
1068 	rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1069 
1070 	return rc ? rc : dlen;
1071 }
1072 
1073 /**
1074  * tipc_send_group_bcast - send message to all members in communication group
1075  * @sock: socket structure
1076  * @m: message to send
1077  * @dlen: total length of message data
1078  * @timeout: timeout to wait for wakeup
1079  *
1080  * Called from function tipc_sendmsg(), which has done all sanity checks
1081  * Return: the number of bytes sent on success, or errno
1082  */
1083 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1084 				 int dlen, long timeout)
1085 {
1086 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1087 	struct sock *sk = sock->sk;
1088 	struct net *net = sock_net(sk);
1089 	struct tipc_sock *tsk = tipc_sk(sk);
1090 	struct tipc_nlist *dsts;
1091 	struct tipc_mc_method *method = &tsk->mc_method;
1092 	bool ack = method->mandatory && method->rcast;
1093 	int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1094 	struct tipc_msg *hdr = &tsk->phdr;
1095 	int mtu = tipc_bcast_get_mtu(net);
1096 	struct sk_buff_head pkts;
1097 	int rc = -EHOSTUNREACH;
1098 
1099 	/* Block or return if any destination link or member is congested */
1100 	rc = tipc_wait_for_cond(sock, &timeout,
1101 				!tsk->cong_link_cnt && tsk->group &&
1102 				!tipc_group_bc_cong(tsk->group, blks));
1103 	if (unlikely(rc))
1104 		return rc;
1105 
1106 	dsts = tipc_group_dests(tsk->group);
1107 	if (!dsts->local && !dsts->remote)
1108 		return -EHOSTUNREACH;
1109 
1110 	/* Complete message header */
1111 	if (ua) {
1112 		msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1113 		msg_set_nameinst(hdr, ua->sa.instance);
1114 	} else {
1115 		msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1116 		msg_set_nameinst(hdr, 0);
1117 	}
1118 	msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1119 	msg_set_destport(hdr, 0);
1120 	msg_set_destnode(hdr, 0);
1121 	msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1122 
1123 	/* Avoid getting stuck with repeated forced replicasts */
1124 	msg_set_grp_bc_ack_req(hdr, ack);
1125 
1126 	/* Build message as chain of buffers */
1127 	__skb_queue_head_init(&pkts);
1128 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1129 	if (unlikely(rc != dlen))
1130 		return rc;
1131 
1132 	/* Send message */
1133 	rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1134 	if (unlikely(rc))
1135 		return rc;
1136 
1137 	/* Update broadcast sequence number and send windows */
1138 	tipc_group_update_bc_members(tsk->group, blks, ack);
1139 
1140 	/* Broadcast link is now free to choose method for next broadcast */
1141 	method->mandatory = false;
1142 	method->expires = jiffies;
1143 
1144 	return dlen;
1145 }
1146 
1147 /**
1148  * tipc_send_group_mcast - send message to all members with given identity
1149  * @sock: socket structure
1150  * @m: message to send
1151  * @dlen: total length of message data
1152  * @timeout: timeout to wait for wakeup
1153  *
1154  * Called from function tipc_sendmsg(), which has done all sanity checks
1155  * Return: the number of bytes sent on success, or errno
1156  */
1157 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1158 				 int dlen, long timeout)
1159 {
1160 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1161 	struct sock *sk = sock->sk;
1162 	struct tipc_sock *tsk = tipc_sk(sk);
1163 	struct tipc_group *grp = tsk->group;
1164 	struct tipc_msg *hdr = &tsk->phdr;
1165 	struct net *net = sock_net(sk);
1166 	struct list_head dsts;
1167 	u32 dstcnt, exclude;
1168 
1169 	INIT_LIST_HEAD(&dsts);
1170 	ua->sa.type = msg_nametype(hdr);
1171 	ua->scope = msg_lookup_scope(hdr);
1172 	exclude = tipc_group_exclude(grp);
1173 
1174 	if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1175 		return -EHOSTUNREACH;
1176 
1177 	if (dstcnt == 1) {
1178 		tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1179 		return tipc_send_group_unicast(sock, m, dlen, timeout);
1180 	}
1181 
1182 	tipc_dest_list_purge(&dsts);
1183 	return tipc_send_group_bcast(sock, m, dlen, timeout);
1184 }
1185 
1186 /**
1187  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1188  * @net: the associated network namespace
1189  * @arrvq: queue with arriving messages, to be cloned after destination lookup
1190  * @inputq: queue with cloned messages, delivered to socket after dest lookup
1191  *
1192  * Multi-threaded: parallel calls with reference to same queues may occur
1193  */
1194 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1195 		       struct sk_buff_head *inputq)
1196 {
1197 	u32 self = tipc_own_addr(net);
1198 	struct sk_buff *skb, *_skb;
1199 	u32 portid, onode;
1200 	struct sk_buff_head tmpq;
1201 	struct list_head dports;
1202 	struct tipc_msg *hdr;
1203 	struct tipc_uaddr ua;
1204 	int user, mtyp, hlen;
1205 	bool exact;
1206 
1207 	__skb_queue_head_init(&tmpq);
1208 	INIT_LIST_HEAD(&dports);
1209 	ua.addrtype = TIPC_SERVICE_RANGE;
1210 
1211 	skb = tipc_skb_peek(arrvq, &inputq->lock);
1212 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1213 		hdr = buf_msg(skb);
1214 		user = msg_user(hdr);
1215 		mtyp = msg_type(hdr);
1216 		hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1217 		onode = msg_orignode(hdr);
1218 		ua.sr.type = msg_nametype(hdr);
1219 
1220 		if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1221 			spin_lock_bh(&inputq->lock);
1222 			if (skb_peek(arrvq) == skb) {
1223 				__skb_dequeue(arrvq);
1224 				__skb_queue_tail(inputq, skb);
1225 			}
1226 			kfree_skb(skb);
1227 			spin_unlock_bh(&inputq->lock);
1228 			continue;
1229 		}
1230 
1231 		/* Group messages require exact scope match */
1232 		if (msg_in_group(hdr)) {
1233 			ua.sr.lower = 0;
1234 			ua.sr.upper = ~0;
1235 			ua.scope = msg_lookup_scope(hdr);
1236 			exact = true;
1237 		} else {
1238 			/* TIPC_NODE_SCOPE means "any scope" in this context */
1239 			if (onode == self)
1240 				ua.scope = TIPC_NODE_SCOPE;
1241 			else
1242 				ua.scope = TIPC_CLUSTER_SCOPE;
1243 			exact = false;
1244 			ua.sr.lower = msg_namelower(hdr);
1245 			ua.sr.upper = msg_nameupper(hdr);
1246 		}
1247 
1248 		/* Create destination port list: */
1249 		tipc_nametbl_lookup_mcast_sockets(net, &ua, exact, &dports);
1250 
1251 		/* Clone message per destination */
1252 		while (tipc_dest_pop(&dports, NULL, &portid)) {
1253 			_skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1254 			if (_skb) {
1255 				msg_set_destport(buf_msg(_skb), portid);
1256 				__skb_queue_tail(&tmpq, _skb);
1257 				continue;
1258 			}
1259 			pr_warn("Failed to clone mcast rcv buffer\n");
1260 		}
1261 		/* Append to inputq if not already done by other thread */
1262 		spin_lock_bh(&inputq->lock);
1263 		if (skb_peek(arrvq) == skb) {
1264 			skb_queue_splice_tail_init(&tmpq, inputq);
1265 			/* Decrease the skb's refcnt as increasing in the
1266 			 * function tipc_skb_peek
1267 			 */
1268 			kfree_skb(__skb_dequeue(arrvq));
1269 		}
1270 		spin_unlock_bh(&inputq->lock);
1271 		__skb_queue_purge(&tmpq);
1272 		kfree_skb(skb);
1273 	}
1274 	tipc_sk_rcv(net, inputq);
1275 }
1276 
1277 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1278  *                         when socket is in Nagle mode
1279  */
1280 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1281 {
1282 	struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1283 	struct sk_buff *skb = skb_peek_tail(txq);
1284 	struct net *net = sock_net(&tsk->sk);
1285 	u32 dnode = tsk_peer_node(tsk);
1286 	int rc;
1287 
1288 	if (nagle_ack) {
1289 		tsk->pkt_cnt += skb_queue_len(txq);
1290 		if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1291 			tsk->oneway = 0;
1292 			if (tsk->nagle_start < NAGLE_START_MAX)
1293 				tsk->nagle_start *= 2;
1294 			tsk->expect_ack = false;
1295 			pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1296 				 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1297 				 tsk->nagle_start);
1298 		} else {
1299 			tsk->nagle_start = NAGLE_START_INIT;
1300 			if (skb) {
1301 				msg_set_ack_required(buf_msg(skb));
1302 				tsk->expect_ack = true;
1303 			} else {
1304 				tsk->expect_ack = false;
1305 			}
1306 		}
1307 		tsk->msg_acc = 0;
1308 		tsk->pkt_cnt = 0;
1309 	}
1310 
1311 	if (!skb || tsk->cong_link_cnt)
1312 		return;
1313 
1314 	/* Do not send SYN again after congestion */
1315 	if (msg_is_syn(buf_msg(skb)))
1316 		return;
1317 
1318 	if (tsk->msg_acc)
1319 		tsk->pkt_cnt += skb_queue_len(txq);
1320 	tsk->snt_unacked += tsk->snd_backlog;
1321 	tsk->snd_backlog = 0;
1322 	rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1323 	if (rc == -ELINKCONG)
1324 		tsk->cong_link_cnt = 1;
1325 }
1326 
1327 /**
1328  * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1329  * @tsk: receiving socket
1330  * @skb: pointer to message buffer.
1331  * @inputq: buffer list containing the buffers
1332  * @xmitq: output message area
1333  */
1334 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1335 				   struct sk_buff_head *inputq,
1336 				   struct sk_buff_head *xmitq)
1337 {
1338 	struct tipc_msg *hdr = buf_msg(skb);
1339 	u32 onode = tsk_own_node(tsk);
1340 	struct sock *sk = &tsk->sk;
1341 	int mtyp = msg_type(hdr);
1342 	bool was_cong;
1343 
1344 	/* Ignore if connection cannot be validated: */
1345 	if (!tsk_peer_msg(tsk, hdr)) {
1346 		trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1347 		goto exit;
1348 	}
1349 
1350 	if (unlikely(msg_errcode(hdr))) {
1351 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1352 		tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1353 				      tsk_peer_port(tsk));
1354 		sk->sk_state_change(sk);
1355 
1356 		/* State change is ignored if socket already awake,
1357 		 * - convert msg to abort msg and add to inqueue
1358 		 */
1359 		msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1360 		msg_set_type(hdr, TIPC_CONN_MSG);
1361 		msg_set_size(hdr, BASIC_H_SIZE);
1362 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1363 		__skb_queue_tail(inputq, skb);
1364 		return;
1365 	}
1366 
1367 	tsk->probe_unacked = false;
1368 
1369 	if (mtyp == CONN_PROBE) {
1370 		msg_set_type(hdr, CONN_PROBE_REPLY);
1371 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1372 			__skb_queue_tail(xmitq, skb);
1373 		return;
1374 	} else if (mtyp == CONN_ACK) {
1375 		was_cong = tsk_conn_cong(tsk);
1376 		tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1377 		tsk->snt_unacked -= msg_conn_ack(hdr);
1378 		if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1379 			tsk->snd_win = msg_adv_win(hdr);
1380 		if (was_cong && !tsk_conn_cong(tsk))
1381 			sk->sk_write_space(sk);
1382 	} else if (mtyp != CONN_PROBE_REPLY) {
1383 		pr_warn("Received unknown CONN_PROTO msg\n");
1384 	}
1385 exit:
1386 	kfree_skb(skb);
1387 }
1388 
1389 /**
1390  * tipc_sendmsg - send message in connectionless manner
1391  * @sock: socket structure
1392  * @m: message to send
1393  * @dsz: amount of user data to be sent
1394  *
1395  * Message must have an destination specified explicitly.
1396  * Used for SOCK_RDM and SOCK_DGRAM messages,
1397  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1398  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1399  *
1400  * Return: the number of bytes sent on success, or errno otherwise
1401  */
1402 static int tipc_sendmsg(struct socket *sock,
1403 			struct msghdr *m, size_t dsz)
1404 {
1405 	struct sock *sk = sock->sk;
1406 	int ret;
1407 
1408 	lock_sock(sk);
1409 	ret = __tipc_sendmsg(sock, m, dsz);
1410 	release_sock(sk);
1411 
1412 	return ret;
1413 }
1414 
1415 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1416 {
1417 	struct sock *sk = sock->sk;
1418 	struct net *net = sock_net(sk);
1419 	struct tipc_sock *tsk = tipc_sk(sk);
1420 	struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1421 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1422 	struct list_head *clinks = &tsk->cong_links;
1423 	bool syn = !tipc_sk_type_connectionless(sk);
1424 	struct tipc_group *grp = tsk->group;
1425 	struct tipc_msg *hdr = &tsk->phdr;
1426 	struct tipc_socket_addr skaddr;
1427 	struct sk_buff_head pkts;
1428 	int atype, mtu, rc;
1429 
1430 	if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1431 		return -EMSGSIZE;
1432 
1433 	if (ua) {
1434 		if (!tipc_uaddr_valid(ua, m->msg_namelen))
1435 			return -EINVAL;
1436 		 atype = ua->addrtype;
1437 	}
1438 
1439 	/* If socket belongs to a communication group follow other paths */
1440 	if (grp) {
1441 		if (!ua)
1442 			return tipc_send_group_bcast(sock, m, dlen, timeout);
1443 		if (atype == TIPC_SERVICE_ADDR)
1444 			return tipc_send_group_anycast(sock, m, dlen, timeout);
1445 		if (atype == TIPC_SOCKET_ADDR)
1446 			return tipc_send_group_unicast(sock, m, dlen, timeout);
1447 		if (atype == TIPC_SERVICE_RANGE)
1448 			return tipc_send_group_mcast(sock, m, dlen, timeout);
1449 		return -EINVAL;
1450 	}
1451 
1452 	if (!ua) {
1453 		ua = (struct tipc_uaddr *)&tsk->peer;
1454 		if (!syn && ua->family != AF_TIPC)
1455 			return -EDESTADDRREQ;
1456 		atype = ua->addrtype;
1457 	}
1458 
1459 	if (unlikely(syn)) {
1460 		if (sk->sk_state == TIPC_LISTEN)
1461 			return -EPIPE;
1462 		if (sk->sk_state != TIPC_OPEN)
1463 			return -EISCONN;
1464 		if (tsk->published)
1465 			return -EOPNOTSUPP;
1466 		if (atype == TIPC_SERVICE_ADDR) {
1467 			tsk->conn_type = ua->sa.type;
1468 			tsk->conn_instance = ua->sa.instance;
1469 		}
1470 		msg_set_syn(hdr, 1);
1471 	}
1472 
1473 	/* Determine destination */
1474 	if (atype == TIPC_SERVICE_RANGE) {
1475 		return tipc_sendmcast(sock, ua, m, dlen, timeout);
1476 	} else if (atype == TIPC_SERVICE_ADDR) {
1477 		skaddr.node = ua->lookup_node;
1478 		ua->scope = tipc_node2scope(skaddr.node);
1479 		if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1480 			return -EHOSTUNREACH;
1481 	} else if (atype == TIPC_SOCKET_ADDR) {
1482 		skaddr = ua->sk;
1483 	} else {
1484 		return -EINVAL;
1485 	}
1486 
1487 	/* Block or return if destination link is congested */
1488 	rc = tipc_wait_for_cond(sock, &timeout,
1489 				!tipc_dest_find(clinks, skaddr.node, 0));
1490 	if (unlikely(rc))
1491 		return rc;
1492 
1493 	/* Finally build message header */
1494 	msg_set_destnode(hdr, skaddr.node);
1495 	msg_set_destport(hdr, skaddr.ref);
1496 	if (atype == TIPC_SERVICE_ADDR) {
1497 		msg_set_type(hdr, TIPC_NAMED_MSG);
1498 		msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1499 		msg_set_nametype(hdr, ua->sa.type);
1500 		msg_set_nameinst(hdr, ua->sa.instance);
1501 		msg_set_lookup_scope(hdr, ua->scope);
1502 	} else { /* TIPC_SOCKET_ADDR */
1503 		msg_set_type(hdr, TIPC_DIRECT_MSG);
1504 		msg_set_lookup_scope(hdr, 0);
1505 		msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1506 	}
1507 
1508 	/* Add message body */
1509 	__skb_queue_head_init(&pkts);
1510 	mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1511 	rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1512 	if (unlikely(rc != dlen))
1513 		return rc;
1514 	if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1515 		__skb_queue_purge(&pkts);
1516 		return -ENOMEM;
1517 	}
1518 
1519 	/* Send message */
1520 	trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1521 	rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1522 	if (unlikely(rc == -ELINKCONG)) {
1523 		tipc_dest_push(clinks, skaddr.node, 0);
1524 		tsk->cong_link_cnt++;
1525 		rc = 0;
1526 	}
1527 
1528 	if (unlikely(syn && !rc))
1529 		tipc_set_sk_state(sk, TIPC_CONNECTING);
1530 
1531 	return rc ? rc : dlen;
1532 }
1533 
1534 /**
1535  * tipc_sendstream - send stream-oriented data
1536  * @sock: socket structure
1537  * @m: data to send
1538  * @dsz: total length of data to be transmitted
1539  *
1540  * Used for SOCK_STREAM data.
1541  *
1542  * Return: the number of bytes sent on success (or partial success),
1543  * or errno if no data sent
1544  */
1545 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1546 {
1547 	struct sock *sk = sock->sk;
1548 	int ret;
1549 
1550 	lock_sock(sk);
1551 	ret = __tipc_sendstream(sock, m, dsz);
1552 	release_sock(sk);
1553 
1554 	return ret;
1555 }
1556 
1557 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1558 {
1559 	struct sock *sk = sock->sk;
1560 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1561 	long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1562 	struct sk_buff_head *txq = &sk->sk_write_queue;
1563 	struct tipc_sock *tsk = tipc_sk(sk);
1564 	struct tipc_msg *hdr = &tsk->phdr;
1565 	struct net *net = sock_net(sk);
1566 	struct sk_buff *skb;
1567 	u32 dnode = tsk_peer_node(tsk);
1568 	int maxnagle = tsk->maxnagle;
1569 	int maxpkt = tsk->max_pkt;
1570 	int send, sent = 0;
1571 	int blocks, rc = 0;
1572 
1573 	if (unlikely(dlen > INT_MAX))
1574 		return -EMSGSIZE;
1575 
1576 	/* Handle implicit connection setup */
1577 	if (unlikely(dest)) {
1578 		rc = __tipc_sendmsg(sock, m, dlen);
1579 		if (dlen && dlen == rc) {
1580 			tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1581 			tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1582 		}
1583 		return rc;
1584 	}
1585 
1586 	do {
1587 		rc = tipc_wait_for_cond(sock, &timeout,
1588 					(!tsk->cong_link_cnt &&
1589 					 !tsk_conn_cong(tsk) &&
1590 					 tipc_sk_connected(sk)));
1591 		if (unlikely(rc))
1592 			break;
1593 		send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1594 		blocks = tsk->snd_backlog;
1595 		if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1596 		    send <= maxnagle) {
1597 			rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1598 			if (unlikely(rc < 0))
1599 				break;
1600 			blocks += rc;
1601 			tsk->msg_acc++;
1602 			if (blocks <= 64 && tsk->expect_ack) {
1603 				tsk->snd_backlog = blocks;
1604 				sent += send;
1605 				break;
1606 			} else if (blocks > 64) {
1607 				tsk->pkt_cnt += skb_queue_len(txq);
1608 			} else {
1609 				skb = skb_peek_tail(txq);
1610 				if (skb) {
1611 					msg_set_ack_required(buf_msg(skb));
1612 					tsk->expect_ack = true;
1613 				} else {
1614 					tsk->expect_ack = false;
1615 				}
1616 				tsk->msg_acc = 0;
1617 				tsk->pkt_cnt = 0;
1618 			}
1619 		} else {
1620 			rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1621 			if (unlikely(rc != send))
1622 				break;
1623 			blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1624 		}
1625 		trace_tipc_sk_sendstream(sk, skb_peek(txq),
1626 					 TIPC_DUMP_SK_SNDQ, " ");
1627 		rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1628 		if (unlikely(rc == -ELINKCONG)) {
1629 			tsk->cong_link_cnt = 1;
1630 			rc = 0;
1631 		}
1632 		if (likely(!rc)) {
1633 			tsk->snt_unacked += blocks;
1634 			tsk->snd_backlog = 0;
1635 			sent += send;
1636 		}
1637 	} while (sent < dlen && !rc);
1638 
1639 	return sent ? sent : rc;
1640 }
1641 
1642 /**
1643  * tipc_send_packet - send a connection-oriented message
1644  * @sock: socket structure
1645  * @m: message to send
1646  * @dsz: length of data to be transmitted
1647  *
1648  * Used for SOCK_SEQPACKET messages.
1649  *
1650  * Return: the number of bytes sent on success, or errno otherwise
1651  */
1652 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1653 {
1654 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1655 		return -EMSGSIZE;
1656 
1657 	return tipc_sendstream(sock, m, dsz);
1658 }
1659 
1660 /* tipc_sk_finish_conn - complete the setup of a connection
1661  */
1662 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1663 				u32 peer_node)
1664 {
1665 	struct sock *sk = &tsk->sk;
1666 	struct net *net = sock_net(sk);
1667 	struct tipc_msg *msg = &tsk->phdr;
1668 
1669 	msg_set_syn(msg, 0);
1670 	msg_set_destnode(msg, peer_node);
1671 	msg_set_destport(msg, peer_port);
1672 	msg_set_type(msg, TIPC_CONN_MSG);
1673 	msg_set_lookup_scope(msg, 0);
1674 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1675 
1676 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1677 	tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1678 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1679 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1680 	tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1681 	tsk_set_nagle(tsk);
1682 	__skb_queue_purge(&sk->sk_write_queue);
1683 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1684 		return;
1685 
1686 	/* Fall back to message based flow control */
1687 	tsk->rcv_win = FLOWCTL_MSG_WIN;
1688 	tsk->snd_win = FLOWCTL_MSG_WIN;
1689 }
1690 
1691 /**
1692  * tipc_sk_set_orig_addr - capture sender's address for received message
1693  * @m: descriptor for message info
1694  * @skb: received message
1695  *
1696  * Note: Address is not captured if not requested by receiver.
1697  */
1698 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1699 {
1700 	DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1701 	struct tipc_msg *hdr = buf_msg(skb);
1702 
1703 	if (!srcaddr)
1704 		return;
1705 
1706 	srcaddr->sock.family = AF_TIPC;
1707 	srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1708 	srcaddr->sock.scope = 0;
1709 	srcaddr->sock.addr.id.ref = msg_origport(hdr);
1710 	srcaddr->sock.addr.id.node = msg_orignode(hdr);
1711 	srcaddr->sock.addr.name.domain = 0;
1712 	m->msg_namelen = sizeof(struct sockaddr_tipc);
1713 
1714 	if (!msg_in_group(hdr))
1715 		return;
1716 
1717 	/* Group message users may also want to know sending member's id */
1718 	srcaddr->member.family = AF_TIPC;
1719 	srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1720 	srcaddr->member.scope = 0;
1721 	srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1722 	srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1723 	srcaddr->member.addr.name.domain = 0;
1724 	m->msg_namelen = sizeof(*srcaddr);
1725 }
1726 
1727 /**
1728  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1729  * @m: descriptor for message info
1730  * @skb: received message buffer
1731  * @tsk: TIPC port associated with message
1732  *
1733  * Note: Ancillary data is not captured if not requested by receiver.
1734  *
1735  * Return: 0 if successful, otherwise errno
1736  */
1737 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1738 				 struct tipc_sock *tsk)
1739 {
1740 	struct tipc_msg *msg;
1741 	u32 anc_data[3];
1742 	u32 err;
1743 	u32 dest_type;
1744 	int has_name;
1745 	int res;
1746 
1747 	if (likely(m->msg_controllen == 0))
1748 		return 0;
1749 	msg = buf_msg(skb);
1750 
1751 	/* Optionally capture errored message object(s) */
1752 	err = msg ? msg_errcode(msg) : 0;
1753 	if (unlikely(err)) {
1754 		anc_data[0] = err;
1755 		anc_data[1] = msg_data_sz(msg);
1756 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1757 		if (res)
1758 			return res;
1759 		if (anc_data[1]) {
1760 			if (skb_linearize(skb))
1761 				return -ENOMEM;
1762 			msg = buf_msg(skb);
1763 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1764 				       msg_data(msg));
1765 			if (res)
1766 				return res;
1767 		}
1768 	}
1769 
1770 	/* Optionally capture message destination object */
1771 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1772 	switch (dest_type) {
1773 	case TIPC_NAMED_MSG:
1774 		has_name = 1;
1775 		anc_data[0] = msg_nametype(msg);
1776 		anc_data[1] = msg_namelower(msg);
1777 		anc_data[2] = msg_namelower(msg);
1778 		break;
1779 	case TIPC_MCAST_MSG:
1780 		has_name = 1;
1781 		anc_data[0] = msg_nametype(msg);
1782 		anc_data[1] = msg_namelower(msg);
1783 		anc_data[2] = msg_nameupper(msg);
1784 		break;
1785 	case TIPC_CONN_MSG:
1786 		has_name = (tsk->conn_type != 0);
1787 		anc_data[0] = tsk->conn_type;
1788 		anc_data[1] = tsk->conn_instance;
1789 		anc_data[2] = tsk->conn_instance;
1790 		break;
1791 	default:
1792 		has_name = 0;
1793 	}
1794 	if (has_name) {
1795 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1796 		if (res)
1797 			return res;
1798 	}
1799 
1800 	return 0;
1801 }
1802 
1803 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1804 {
1805 	struct sock *sk = &tsk->sk;
1806 	struct sk_buff *skb = NULL;
1807 	struct tipc_msg *msg;
1808 	u32 peer_port = tsk_peer_port(tsk);
1809 	u32 dnode = tsk_peer_node(tsk);
1810 
1811 	if (!tipc_sk_connected(sk))
1812 		return NULL;
1813 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1814 			      dnode, tsk_own_node(tsk), peer_port,
1815 			      tsk->portid, TIPC_OK);
1816 	if (!skb)
1817 		return NULL;
1818 	msg = buf_msg(skb);
1819 	msg_set_conn_ack(msg, tsk->rcv_unacked);
1820 	tsk->rcv_unacked = 0;
1821 
1822 	/* Adjust to and advertize the correct window limit */
1823 	if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1824 		tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1825 		msg_set_adv_win(msg, tsk->rcv_win);
1826 	}
1827 	return skb;
1828 }
1829 
1830 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1831 {
1832 	struct sk_buff *skb;
1833 
1834 	skb = tipc_sk_build_ack(tsk);
1835 	if (!skb)
1836 		return;
1837 
1838 	tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1839 			   msg_link_selector(buf_msg(skb)));
1840 }
1841 
1842 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1843 {
1844 	struct sock *sk = sock->sk;
1845 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1846 	long timeo = *timeop;
1847 	int err = sock_error(sk);
1848 
1849 	if (err)
1850 		return err;
1851 
1852 	for (;;) {
1853 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1854 			if (sk->sk_shutdown & RCV_SHUTDOWN) {
1855 				err = -ENOTCONN;
1856 				break;
1857 			}
1858 			add_wait_queue(sk_sleep(sk), &wait);
1859 			release_sock(sk);
1860 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1861 			sched_annotate_sleep();
1862 			lock_sock(sk);
1863 			remove_wait_queue(sk_sleep(sk), &wait);
1864 		}
1865 		err = 0;
1866 		if (!skb_queue_empty(&sk->sk_receive_queue))
1867 			break;
1868 		err = -EAGAIN;
1869 		if (!timeo)
1870 			break;
1871 		err = sock_intr_errno(timeo);
1872 		if (signal_pending(current))
1873 			break;
1874 
1875 		err = sock_error(sk);
1876 		if (err)
1877 			break;
1878 	}
1879 	*timeop = timeo;
1880 	return err;
1881 }
1882 
1883 /**
1884  * tipc_recvmsg - receive packet-oriented message
1885  * @sock: network socket
1886  * @m: descriptor for message info
1887  * @buflen: length of user buffer area
1888  * @flags: receive flags
1889  *
1890  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1891  * If the complete message doesn't fit in user area, truncate it.
1892  *
1893  * Return: size of returned message data, errno otherwise
1894  */
1895 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1896 			size_t buflen,	int flags)
1897 {
1898 	struct sock *sk = sock->sk;
1899 	bool connected = !tipc_sk_type_connectionless(sk);
1900 	struct tipc_sock *tsk = tipc_sk(sk);
1901 	int rc, err, hlen, dlen, copy;
1902 	struct sk_buff_head xmitq;
1903 	struct tipc_msg *hdr;
1904 	struct sk_buff *skb;
1905 	bool grp_evt;
1906 	long timeout;
1907 
1908 	/* Catch invalid receive requests */
1909 	if (unlikely(!buflen))
1910 		return -EINVAL;
1911 
1912 	lock_sock(sk);
1913 	if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1914 		rc = -ENOTCONN;
1915 		goto exit;
1916 	}
1917 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1918 
1919 	/* Step rcv queue to first msg with data or error; wait if necessary */
1920 	do {
1921 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
1922 		if (unlikely(rc))
1923 			goto exit;
1924 		skb = skb_peek(&sk->sk_receive_queue);
1925 		hdr = buf_msg(skb);
1926 		dlen = msg_data_sz(hdr);
1927 		hlen = msg_hdr_sz(hdr);
1928 		err = msg_errcode(hdr);
1929 		grp_evt = msg_is_grp_evt(hdr);
1930 		if (likely(dlen || err))
1931 			break;
1932 		tsk_advance_rx_queue(sk);
1933 	} while (1);
1934 
1935 	/* Collect msg meta data, including error code and rejected data */
1936 	tipc_sk_set_orig_addr(m, skb);
1937 	rc = tipc_sk_anc_data_recv(m, skb, tsk);
1938 	if (unlikely(rc))
1939 		goto exit;
1940 	hdr = buf_msg(skb);
1941 
1942 	/* Capture data if non-error msg, otherwise just set return value */
1943 	if (likely(!err)) {
1944 		copy = min_t(int, dlen, buflen);
1945 		if (unlikely(copy != dlen))
1946 			m->msg_flags |= MSG_TRUNC;
1947 		rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1948 	} else {
1949 		copy = 0;
1950 		rc = 0;
1951 		if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1952 			rc = -ECONNRESET;
1953 	}
1954 	if (unlikely(rc))
1955 		goto exit;
1956 
1957 	/* Mark message as group event if applicable */
1958 	if (unlikely(grp_evt)) {
1959 		if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1960 			m->msg_flags |= MSG_EOR;
1961 		m->msg_flags |= MSG_OOB;
1962 		copy = 0;
1963 	}
1964 
1965 	/* Caption of data or error code/rejected data was successful */
1966 	if (unlikely(flags & MSG_PEEK))
1967 		goto exit;
1968 
1969 	/* Send group flow control advertisement when applicable */
1970 	if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1971 		__skb_queue_head_init(&xmitq);
1972 		tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1973 					  msg_orignode(hdr), msg_origport(hdr),
1974 					  &xmitq);
1975 		tipc_node_distr_xmit(sock_net(sk), &xmitq);
1976 	}
1977 
1978 	tsk_advance_rx_queue(sk);
1979 
1980 	if (likely(!connected))
1981 		goto exit;
1982 
1983 	/* Send connection flow control advertisement when applicable */
1984 	tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1985 	if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1986 		tipc_sk_send_ack(tsk);
1987 exit:
1988 	release_sock(sk);
1989 	return rc ? rc : copy;
1990 }
1991 
1992 /**
1993  * tipc_recvstream - receive stream-oriented data
1994  * @sock: network socket
1995  * @m: descriptor for message info
1996  * @buflen: total size of user buffer area
1997  * @flags: receive flags
1998  *
1999  * Used for SOCK_STREAM messages only.  If not enough data is available
2000  * will optionally wait for more; never truncates data.
2001  *
2002  * Return: size of returned message data, errno otherwise
2003  */
2004 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2005 			   size_t buflen, int flags)
2006 {
2007 	struct sock *sk = sock->sk;
2008 	struct tipc_sock *tsk = tipc_sk(sk);
2009 	struct sk_buff *skb;
2010 	struct tipc_msg *hdr;
2011 	struct tipc_skb_cb *skb_cb;
2012 	bool peek = flags & MSG_PEEK;
2013 	int offset, required, copy, copied = 0;
2014 	int hlen, dlen, err, rc;
2015 	long timeout;
2016 
2017 	/* Catch invalid receive attempts */
2018 	if (unlikely(!buflen))
2019 		return -EINVAL;
2020 
2021 	lock_sock(sk);
2022 
2023 	if (unlikely(sk->sk_state == TIPC_OPEN)) {
2024 		rc = -ENOTCONN;
2025 		goto exit;
2026 	}
2027 	required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2028 	timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2029 
2030 	do {
2031 		/* Look at first msg in receive queue; wait if necessary */
2032 		rc = tipc_wait_for_rcvmsg(sock, &timeout);
2033 		if (unlikely(rc))
2034 			break;
2035 		skb = skb_peek(&sk->sk_receive_queue);
2036 		skb_cb = TIPC_SKB_CB(skb);
2037 		hdr = buf_msg(skb);
2038 		dlen = msg_data_sz(hdr);
2039 		hlen = msg_hdr_sz(hdr);
2040 		err = msg_errcode(hdr);
2041 
2042 		/* Discard any empty non-errored (SYN-) message */
2043 		if (unlikely(!dlen && !err)) {
2044 			tsk_advance_rx_queue(sk);
2045 			continue;
2046 		}
2047 
2048 		/* Collect msg meta data, incl. error code and rejected data */
2049 		if (!copied) {
2050 			tipc_sk_set_orig_addr(m, skb);
2051 			rc = tipc_sk_anc_data_recv(m, skb, tsk);
2052 			if (rc)
2053 				break;
2054 			hdr = buf_msg(skb);
2055 		}
2056 
2057 		/* Copy data if msg ok, otherwise return error/partial data */
2058 		if (likely(!err)) {
2059 			offset = skb_cb->bytes_read;
2060 			copy = min_t(int, dlen - offset, buflen - copied);
2061 			rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2062 			if (unlikely(rc))
2063 				break;
2064 			copied += copy;
2065 			offset += copy;
2066 			if (unlikely(offset < dlen)) {
2067 				if (!peek)
2068 					skb_cb->bytes_read = offset;
2069 				break;
2070 			}
2071 		} else {
2072 			rc = 0;
2073 			if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2074 				rc = -ECONNRESET;
2075 			if (copied || rc)
2076 				break;
2077 		}
2078 
2079 		if (unlikely(peek))
2080 			break;
2081 
2082 		tsk_advance_rx_queue(sk);
2083 
2084 		/* Send connection flow control advertisement when applicable */
2085 		tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2086 		if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2087 			tipc_sk_send_ack(tsk);
2088 
2089 		/* Exit if all requested data or FIN/error received */
2090 		if (copied == buflen || err)
2091 			break;
2092 
2093 	} while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2094 exit:
2095 	release_sock(sk);
2096 	return copied ? copied : rc;
2097 }
2098 
2099 /**
2100  * tipc_write_space - wake up thread if port congestion is released
2101  * @sk: socket
2102  */
2103 static void tipc_write_space(struct sock *sk)
2104 {
2105 	struct socket_wq *wq;
2106 
2107 	rcu_read_lock();
2108 	wq = rcu_dereference(sk->sk_wq);
2109 	if (skwq_has_sleeper(wq))
2110 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2111 						EPOLLWRNORM | EPOLLWRBAND);
2112 	rcu_read_unlock();
2113 }
2114 
2115 /**
2116  * tipc_data_ready - wake up threads to indicate messages have been received
2117  * @sk: socket
2118  */
2119 static void tipc_data_ready(struct sock *sk)
2120 {
2121 	struct socket_wq *wq;
2122 
2123 	rcu_read_lock();
2124 	wq = rcu_dereference(sk->sk_wq);
2125 	if (skwq_has_sleeper(wq))
2126 		wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2127 						EPOLLRDNORM | EPOLLRDBAND);
2128 	rcu_read_unlock();
2129 }
2130 
2131 static void tipc_sock_destruct(struct sock *sk)
2132 {
2133 	__skb_queue_purge(&sk->sk_receive_queue);
2134 }
2135 
2136 static void tipc_sk_proto_rcv(struct sock *sk,
2137 			      struct sk_buff_head *inputq,
2138 			      struct sk_buff_head *xmitq)
2139 {
2140 	struct sk_buff *skb = __skb_dequeue(inputq);
2141 	struct tipc_sock *tsk = tipc_sk(sk);
2142 	struct tipc_msg *hdr = buf_msg(skb);
2143 	struct tipc_group *grp = tsk->group;
2144 	bool wakeup = false;
2145 
2146 	switch (msg_user(hdr)) {
2147 	case CONN_MANAGER:
2148 		tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2149 		return;
2150 	case SOCK_WAKEUP:
2151 		tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2152 		/* coupled with smp_rmb() in tipc_wait_for_cond() */
2153 		smp_wmb();
2154 		tsk->cong_link_cnt--;
2155 		wakeup = true;
2156 		tipc_sk_push_backlog(tsk, false);
2157 		break;
2158 	case GROUP_PROTOCOL:
2159 		tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2160 		break;
2161 	case TOP_SRV:
2162 		tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2163 				      hdr, inputq, xmitq);
2164 		break;
2165 	default:
2166 		break;
2167 	}
2168 
2169 	if (wakeup)
2170 		sk->sk_write_space(sk);
2171 
2172 	kfree_skb(skb);
2173 }
2174 
2175 /**
2176  * tipc_sk_filter_connect - check incoming message for a connection-based socket
2177  * @tsk: TIPC socket
2178  * @skb: pointer to message buffer.
2179  * @xmitq: for Nagle ACK if any
2180  * Return: true if message should be added to receive queue, false otherwise
2181  */
2182 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2183 				   struct sk_buff_head *xmitq)
2184 {
2185 	struct sock *sk = &tsk->sk;
2186 	struct net *net = sock_net(sk);
2187 	struct tipc_msg *hdr = buf_msg(skb);
2188 	bool con_msg = msg_connected(hdr);
2189 	u32 pport = tsk_peer_port(tsk);
2190 	u32 pnode = tsk_peer_node(tsk);
2191 	u32 oport = msg_origport(hdr);
2192 	u32 onode = msg_orignode(hdr);
2193 	int err = msg_errcode(hdr);
2194 	unsigned long delay;
2195 
2196 	if (unlikely(msg_mcast(hdr)))
2197 		return false;
2198 	tsk->oneway = 0;
2199 
2200 	switch (sk->sk_state) {
2201 	case TIPC_CONNECTING:
2202 		/* Setup ACK */
2203 		if (likely(con_msg)) {
2204 			if (err)
2205 				break;
2206 			tipc_sk_finish_conn(tsk, oport, onode);
2207 			msg_set_importance(&tsk->phdr, msg_importance(hdr));
2208 			/* ACK+ message with data is added to receive queue */
2209 			if (msg_data_sz(hdr))
2210 				return true;
2211 			/* Empty ACK-, - wake up sleeping connect() and drop */
2212 			sk->sk_state_change(sk);
2213 			msg_set_dest_droppable(hdr, 1);
2214 			return false;
2215 		}
2216 		/* Ignore connectionless message if not from listening socket */
2217 		if (oport != pport || onode != pnode)
2218 			return false;
2219 
2220 		/* Rejected SYN */
2221 		if (err != TIPC_ERR_OVERLOAD)
2222 			break;
2223 
2224 		/* Prepare for new setup attempt if we have a SYN clone */
2225 		if (skb_queue_empty(&sk->sk_write_queue))
2226 			break;
2227 		get_random_bytes(&delay, 2);
2228 		delay %= (tsk->conn_timeout / 4);
2229 		delay = msecs_to_jiffies(delay + 100);
2230 		sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2231 		return false;
2232 	case TIPC_OPEN:
2233 	case TIPC_DISCONNECTING:
2234 		return false;
2235 	case TIPC_LISTEN:
2236 		/* Accept only SYN message */
2237 		if (!msg_is_syn(hdr) &&
2238 		    tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2239 			return false;
2240 		if (!con_msg && !err)
2241 			return true;
2242 		return false;
2243 	case TIPC_ESTABLISHED:
2244 		if (!skb_queue_empty(&sk->sk_write_queue))
2245 			tipc_sk_push_backlog(tsk, false);
2246 		/* Accept only connection-based messages sent by peer */
2247 		if (likely(con_msg && !err && pport == oport &&
2248 			   pnode == onode)) {
2249 			if (msg_ack_required(hdr)) {
2250 				struct sk_buff *skb;
2251 
2252 				skb = tipc_sk_build_ack(tsk);
2253 				if (skb) {
2254 					msg_set_nagle_ack(buf_msg(skb));
2255 					__skb_queue_tail(xmitq, skb);
2256 				}
2257 			}
2258 			return true;
2259 		}
2260 		if (!tsk_peer_msg(tsk, hdr))
2261 			return false;
2262 		if (!err)
2263 			return true;
2264 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2265 		tipc_node_remove_conn(net, pnode, tsk->portid);
2266 		sk->sk_state_change(sk);
2267 		return true;
2268 	default:
2269 		pr_err("Unknown sk_state %u\n", sk->sk_state);
2270 	}
2271 	/* Abort connection setup attempt */
2272 	tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2273 	sk->sk_err = ECONNREFUSED;
2274 	sk->sk_state_change(sk);
2275 	return true;
2276 }
2277 
2278 /**
2279  * rcvbuf_limit - get proper overload limit of socket receive queue
2280  * @sk: socket
2281  * @skb: message
2282  *
2283  * For connection oriented messages, irrespective of importance,
2284  * default queue limit is 2 MB.
2285  *
2286  * For connectionless messages, queue limits are based on message
2287  * importance as follows:
2288  *
2289  * TIPC_LOW_IMPORTANCE       (2 MB)
2290  * TIPC_MEDIUM_IMPORTANCE    (4 MB)
2291  * TIPC_HIGH_IMPORTANCE      (8 MB)
2292  * TIPC_CRITICAL_IMPORTANCE  (16 MB)
2293  *
2294  * Return: overload limit according to corresponding message importance
2295  */
2296 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2297 {
2298 	struct tipc_sock *tsk = tipc_sk(sk);
2299 	struct tipc_msg *hdr = buf_msg(skb);
2300 
2301 	if (unlikely(msg_in_group(hdr)))
2302 		return READ_ONCE(sk->sk_rcvbuf);
2303 
2304 	if (unlikely(!msg_connected(hdr)))
2305 		return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2306 
2307 	if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2308 		return READ_ONCE(sk->sk_rcvbuf);
2309 
2310 	return FLOWCTL_MSG_LIM;
2311 }
2312 
2313 /**
2314  * tipc_sk_filter_rcv - validate incoming message
2315  * @sk: socket
2316  * @skb: pointer to message.
2317  * @xmitq: output message area (FIXME)
2318  *
2319  * Enqueues message on receive queue if acceptable; optionally handles
2320  * disconnect indication for a connected socket.
2321  *
2322  * Called with socket lock already taken
2323  */
2324 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2325 			       struct sk_buff_head *xmitq)
2326 {
2327 	bool sk_conn = !tipc_sk_type_connectionless(sk);
2328 	struct tipc_sock *tsk = tipc_sk(sk);
2329 	struct tipc_group *grp = tsk->group;
2330 	struct tipc_msg *hdr = buf_msg(skb);
2331 	struct net *net = sock_net(sk);
2332 	struct sk_buff_head inputq;
2333 	int mtyp = msg_type(hdr);
2334 	int limit, err = TIPC_OK;
2335 
2336 	trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2337 	TIPC_SKB_CB(skb)->bytes_read = 0;
2338 	__skb_queue_head_init(&inputq);
2339 	__skb_queue_tail(&inputq, skb);
2340 
2341 	if (unlikely(!msg_isdata(hdr)))
2342 		tipc_sk_proto_rcv(sk, &inputq, xmitq);
2343 
2344 	if (unlikely(grp))
2345 		tipc_group_filter_msg(grp, &inputq, xmitq);
2346 
2347 	if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2348 		tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2349 
2350 	/* Validate and add to receive buffer if there is space */
2351 	while ((skb = __skb_dequeue(&inputq))) {
2352 		hdr = buf_msg(skb);
2353 		limit = rcvbuf_limit(sk, skb);
2354 		if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2355 		    (!sk_conn && msg_connected(hdr)) ||
2356 		    (!grp && msg_in_group(hdr)))
2357 			err = TIPC_ERR_NO_PORT;
2358 		else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2359 			trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2360 					   "err_overload2!");
2361 			atomic_inc(&sk->sk_drops);
2362 			err = TIPC_ERR_OVERLOAD;
2363 		}
2364 
2365 		if (unlikely(err)) {
2366 			if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2367 				trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2368 						      "@filter_rcv!");
2369 				__skb_queue_tail(xmitq, skb);
2370 			}
2371 			err = TIPC_OK;
2372 			continue;
2373 		}
2374 		__skb_queue_tail(&sk->sk_receive_queue, skb);
2375 		skb_set_owner_r(skb, sk);
2376 		trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2377 					 "rcvq >90% allocated!");
2378 		sk->sk_data_ready(sk);
2379 	}
2380 }
2381 
2382 /**
2383  * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2384  * @sk: socket
2385  * @skb: message
2386  *
2387  * Caller must hold socket lock
2388  */
2389 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2390 {
2391 	unsigned int before = sk_rmem_alloc_get(sk);
2392 	struct sk_buff_head xmitq;
2393 	unsigned int added;
2394 
2395 	__skb_queue_head_init(&xmitq);
2396 
2397 	tipc_sk_filter_rcv(sk, skb, &xmitq);
2398 	added = sk_rmem_alloc_get(sk) - before;
2399 	atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2400 
2401 	/* Send pending response/rejected messages, if any */
2402 	tipc_node_distr_xmit(sock_net(sk), &xmitq);
2403 	return 0;
2404 }
2405 
2406 /**
2407  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2408  *                   inputq and try adding them to socket or backlog queue
2409  * @inputq: list of incoming buffers with potentially different destinations
2410  * @sk: socket where the buffers should be enqueued
2411  * @dport: port number for the socket
2412  * @xmitq: output queue
2413  *
2414  * Caller must hold socket lock
2415  */
2416 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2417 			    u32 dport, struct sk_buff_head *xmitq)
2418 {
2419 	unsigned long time_limit = jiffies + 2;
2420 	struct sk_buff *skb;
2421 	unsigned int lim;
2422 	atomic_t *dcnt;
2423 	u32 onode;
2424 
2425 	while (skb_queue_len(inputq)) {
2426 		if (unlikely(time_after_eq(jiffies, time_limit)))
2427 			return;
2428 
2429 		skb = tipc_skb_dequeue(inputq, dport);
2430 		if (unlikely(!skb))
2431 			return;
2432 
2433 		/* Add message directly to receive queue if possible */
2434 		if (!sock_owned_by_user(sk)) {
2435 			tipc_sk_filter_rcv(sk, skb, xmitq);
2436 			continue;
2437 		}
2438 
2439 		/* Try backlog, compensating for double-counted bytes */
2440 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2441 		if (!sk->sk_backlog.len)
2442 			atomic_set(dcnt, 0);
2443 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2444 		if (likely(!sk_add_backlog(sk, skb, lim))) {
2445 			trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2446 						 "bklg & rcvq >90% allocated!");
2447 			continue;
2448 		}
2449 
2450 		trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2451 		/* Overload => reject message back to sender */
2452 		onode = tipc_own_addr(sock_net(sk));
2453 		atomic_inc(&sk->sk_drops);
2454 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2455 			trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2456 					      "@sk_enqueue!");
2457 			__skb_queue_tail(xmitq, skb);
2458 		}
2459 		break;
2460 	}
2461 }
2462 
2463 /**
2464  * tipc_sk_rcv - handle a chain of incoming buffers
2465  * @net: the associated network namespace
2466  * @inputq: buffer list containing the buffers
2467  * Consumes all buffers in list until inputq is empty
2468  * Note: may be called in multiple threads referring to the same queue
2469  */
2470 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2471 {
2472 	struct sk_buff_head xmitq;
2473 	u32 dnode, dport = 0;
2474 	int err;
2475 	struct tipc_sock *tsk;
2476 	struct sock *sk;
2477 	struct sk_buff *skb;
2478 
2479 	__skb_queue_head_init(&xmitq);
2480 	while (skb_queue_len(inputq)) {
2481 		dport = tipc_skb_peek_port(inputq, dport);
2482 		tsk = tipc_sk_lookup(net, dport);
2483 
2484 		if (likely(tsk)) {
2485 			sk = &tsk->sk;
2486 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2487 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2488 				spin_unlock_bh(&sk->sk_lock.slock);
2489 			}
2490 			/* Send pending response/rejected messages, if any */
2491 			tipc_node_distr_xmit(sock_net(sk), &xmitq);
2492 			sock_put(sk);
2493 			continue;
2494 		}
2495 		/* No destination socket => dequeue skb if still there */
2496 		skb = tipc_skb_dequeue(inputq, dport);
2497 		if (!skb)
2498 			return;
2499 
2500 		/* Try secondary lookup if unresolved named message */
2501 		err = TIPC_ERR_NO_PORT;
2502 		if (tipc_msg_lookup_dest(net, skb, &err))
2503 			goto xmit;
2504 
2505 		/* Prepare for message rejection */
2506 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2507 			continue;
2508 
2509 		trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2510 xmit:
2511 		dnode = msg_destnode(buf_msg(skb));
2512 		tipc_node_xmit_skb(net, skb, dnode, dport);
2513 	}
2514 }
2515 
2516 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2517 {
2518 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
2519 	struct sock *sk = sock->sk;
2520 	int done;
2521 
2522 	do {
2523 		int err = sock_error(sk);
2524 		if (err)
2525 			return err;
2526 		if (!*timeo_p)
2527 			return -ETIMEDOUT;
2528 		if (signal_pending(current))
2529 			return sock_intr_errno(*timeo_p);
2530 		if (sk->sk_state == TIPC_DISCONNECTING)
2531 			break;
2532 
2533 		add_wait_queue(sk_sleep(sk), &wait);
2534 		done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2535 				     &wait);
2536 		remove_wait_queue(sk_sleep(sk), &wait);
2537 	} while (!done);
2538 	return 0;
2539 }
2540 
2541 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2542 {
2543 	if (addr->family != AF_TIPC)
2544 		return false;
2545 	if (addr->addrtype == TIPC_SERVICE_RANGE)
2546 		return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2547 	return (addr->addrtype == TIPC_SERVICE_ADDR ||
2548 		addr->addrtype == TIPC_SOCKET_ADDR);
2549 }
2550 
2551 /**
2552  * tipc_connect - establish a connection to another TIPC port
2553  * @sock: socket structure
2554  * @dest: socket address for destination port
2555  * @destlen: size of socket address data structure
2556  * @flags: file-related flags associated with socket
2557  *
2558  * Return: 0 on success, errno otherwise
2559  */
2560 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2561 			int destlen, int flags)
2562 {
2563 	struct sock *sk = sock->sk;
2564 	struct tipc_sock *tsk = tipc_sk(sk);
2565 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2566 	struct msghdr m = {NULL,};
2567 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2568 	int previous;
2569 	int res = 0;
2570 
2571 	if (destlen != sizeof(struct sockaddr_tipc))
2572 		return -EINVAL;
2573 
2574 	lock_sock(sk);
2575 
2576 	if (tsk->group) {
2577 		res = -EINVAL;
2578 		goto exit;
2579 	}
2580 
2581 	if (dst->family == AF_UNSPEC) {
2582 		memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2583 		if (!tipc_sk_type_connectionless(sk))
2584 			res = -EINVAL;
2585 		goto exit;
2586 	}
2587 	if (!tipc_sockaddr_is_sane(dst)) {
2588 		res = -EINVAL;
2589 		goto exit;
2590 	}
2591 	/* DGRAM/RDM connect(), just save the destaddr */
2592 	if (tipc_sk_type_connectionless(sk)) {
2593 		memcpy(&tsk->peer, dest, destlen);
2594 		goto exit;
2595 	} else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2596 		res = -EINVAL;
2597 		goto exit;
2598 	}
2599 
2600 	previous = sk->sk_state;
2601 
2602 	switch (sk->sk_state) {
2603 	case TIPC_OPEN:
2604 		/* Send a 'SYN-' to destination */
2605 		m.msg_name = dest;
2606 		m.msg_namelen = destlen;
2607 
2608 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
2609 		 * indicate send_msg() is never blocked.
2610 		 */
2611 		if (!timeout)
2612 			m.msg_flags = MSG_DONTWAIT;
2613 
2614 		res = __tipc_sendmsg(sock, &m, 0);
2615 		if ((res < 0) && (res != -EWOULDBLOCK))
2616 			goto exit;
2617 
2618 		/* Just entered TIPC_CONNECTING state; the only
2619 		 * difference is that return value in non-blocking
2620 		 * case is EINPROGRESS, rather than EALREADY.
2621 		 */
2622 		res = -EINPROGRESS;
2623 		fallthrough;
2624 	case TIPC_CONNECTING:
2625 		if (!timeout) {
2626 			if (previous == TIPC_CONNECTING)
2627 				res = -EALREADY;
2628 			goto exit;
2629 		}
2630 		timeout = msecs_to_jiffies(timeout);
2631 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2632 		res = tipc_wait_for_connect(sock, &timeout);
2633 		break;
2634 	case TIPC_ESTABLISHED:
2635 		res = -EISCONN;
2636 		break;
2637 	default:
2638 		res = -EINVAL;
2639 	}
2640 
2641 exit:
2642 	release_sock(sk);
2643 	return res;
2644 }
2645 
2646 /**
2647  * tipc_listen - allow socket to listen for incoming connections
2648  * @sock: socket structure
2649  * @len: (unused)
2650  *
2651  * Return: 0 on success, errno otherwise
2652  */
2653 static int tipc_listen(struct socket *sock, int len)
2654 {
2655 	struct sock *sk = sock->sk;
2656 	int res;
2657 
2658 	lock_sock(sk);
2659 	res = tipc_set_sk_state(sk, TIPC_LISTEN);
2660 	release_sock(sk);
2661 
2662 	return res;
2663 }
2664 
2665 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2666 {
2667 	struct sock *sk = sock->sk;
2668 	DEFINE_WAIT(wait);
2669 	int err;
2670 
2671 	/* True wake-one mechanism for incoming connections: only
2672 	 * one process gets woken up, not the 'whole herd'.
2673 	 * Since we do not 'race & poll' for established sockets
2674 	 * anymore, the common case will execute the loop only once.
2675 	*/
2676 	for (;;) {
2677 		prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2678 					  TASK_INTERRUPTIBLE);
2679 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2680 			release_sock(sk);
2681 			timeo = schedule_timeout(timeo);
2682 			lock_sock(sk);
2683 		}
2684 		err = 0;
2685 		if (!skb_queue_empty(&sk->sk_receive_queue))
2686 			break;
2687 		err = -EAGAIN;
2688 		if (!timeo)
2689 			break;
2690 		err = sock_intr_errno(timeo);
2691 		if (signal_pending(current))
2692 			break;
2693 	}
2694 	finish_wait(sk_sleep(sk), &wait);
2695 	return err;
2696 }
2697 
2698 /**
2699  * tipc_accept - wait for connection request
2700  * @sock: listening socket
2701  * @new_sock: new socket that is to be connected
2702  * @flags: file-related flags associated with socket
2703  * @kern: caused by kernel or by userspace?
2704  *
2705  * Return: 0 on success, errno otherwise
2706  */
2707 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2708 		       bool kern)
2709 {
2710 	struct sock *new_sk, *sk = sock->sk;
2711 	struct sk_buff *buf;
2712 	struct tipc_sock *new_tsock;
2713 	struct tipc_msg *msg;
2714 	long timeo;
2715 	int res;
2716 
2717 	lock_sock(sk);
2718 
2719 	if (sk->sk_state != TIPC_LISTEN) {
2720 		res = -EINVAL;
2721 		goto exit;
2722 	}
2723 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2724 	res = tipc_wait_for_accept(sock, timeo);
2725 	if (res)
2726 		goto exit;
2727 
2728 	buf = skb_peek(&sk->sk_receive_queue);
2729 
2730 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2731 	if (res)
2732 		goto exit;
2733 	security_sk_clone(sock->sk, new_sock->sk);
2734 
2735 	new_sk = new_sock->sk;
2736 	new_tsock = tipc_sk(new_sk);
2737 	msg = buf_msg(buf);
2738 
2739 	/* we lock on new_sk; but lockdep sees the lock on sk */
2740 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2741 
2742 	/*
2743 	 * Reject any stray messages received by new socket
2744 	 * before the socket lock was taken (very, very unlikely)
2745 	 */
2746 	tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2747 
2748 	/* Connect new socket to it's peer */
2749 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2750 
2751 	tsk_set_importance(new_sk, msg_importance(msg));
2752 	if (msg_named(msg)) {
2753 		new_tsock->conn_type = msg_nametype(msg);
2754 		new_tsock->conn_instance = msg_nameinst(msg);
2755 	}
2756 
2757 	/*
2758 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2759 	 * Respond to 'SYN+' by queuing it on new socket.
2760 	 */
2761 	if (!msg_data_sz(msg)) {
2762 		struct msghdr m = {NULL,};
2763 
2764 		tsk_advance_rx_queue(sk);
2765 		__tipc_sendstream(new_sock, &m, 0);
2766 	} else {
2767 		__skb_dequeue(&sk->sk_receive_queue);
2768 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2769 		skb_set_owner_r(buf, new_sk);
2770 	}
2771 	release_sock(new_sk);
2772 exit:
2773 	release_sock(sk);
2774 	return res;
2775 }
2776 
2777 /**
2778  * tipc_shutdown - shutdown socket connection
2779  * @sock: socket structure
2780  * @how: direction to close (must be SHUT_RDWR)
2781  *
2782  * Terminates connection (if necessary), then purges socket's receive queue.
2783  *
2784  * Return: 0 on success, errno otherwise
2785  */
2786 static int tipc_shutdown(struct socket *sock, int how)
2787 {
2788 	struct sock *sk = sock->sk;
2789 	int res;
2790 
2791 	if (how != SHUT_RDWR)
2792 		return -EINVAL;
2793 
2794 	lock_sock(sk);
2795 
2796 	trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2797 	__tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2798 	sk->sk_shutdown = SHUTDOWN_MASK;
2799 
2800 	if (sk->sk_state == TIPC_DISCONNECTING) {
2801 		/* Discard any unreceived messages */
2802 		__skb_queue_purge(&sk->sk_receive_queue);
2803 
2804 		res = 0;
2805 	} else {
2806 		res = -ENOTCONN;
2807 	}
2808 	/* Wake up anyone sleeping in poll. */
2809 	sk->sk_state_change(sk);
2810 
2811 	release_sock(sk);
2812 	return res;
2813 }
2814 
2815 static void tipc_sk_check_probing_state(struct sock *sk,
2816 					struct sk_buff_head *list)
2817 {
2818 	struct tipc_sock *tsk = tipc_sk(sk);
2819 	u32 pnode = tsk_peer_node(tsk);
2820 	u32 pport = tsk_peer_port(tsk);
2821 	u32 self = tsk_own_node(tsk);
2822 	u32 oport = tsk->portid;
2823 	struct sk_buff *skb;
2824 
2825 	if (tsk->probe_unacked) {
2826 		tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2827 		sk->sk_err = ECONNABORTED;
2828 		tipc_node_remove_conn(sock_net(sk), pnode, pport);
2829 		sk->sk_state_change(sk);
2830 		return;
2831 	}
2832 	/* Prepare new probe */
2833 	skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2834 			      pnode, self, pport, oport, TIPC_OK);
2835 	if (skb)
2836 		__skb_queue_tail(list, skb);
2837 	tsk->probe_unacked = true;
2838 	sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2839 }
2840 
2841 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2842 {
2843 	struct tipc_sock *tsk = tipc_sk(sk);
2844 
2845 	/* Try again later if dest link is congested */
2846 	if (tsk->cong_link_cnt) {
2847 		sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2848 		return;
2849 	}
2850 	/* Prepare SYN for retransmit */
2851 	tipc_msg_skb_clone(&sk->sk_write_queue, list);
2852 }
2853 
2854 static void tipc_sk_timeout(struct timer_list *t)
2855 {
2856 	struct sock *sk = from_timer(sk, t, sk_timer);
2857 	struct tipc_sock *tsk = tipc_sk(sk);
2858 	u32 pnode = tsk_peer_node(tsk);
2859 	struct sk_buff_head list;
2860 	int rc = 0;
2861 
2862 	__skb_queue_head_init(&list);
2863 	bh_lock_sock(sk);
2864 
2865 	/* Try again later if socket is busy */
2866 	if (sock_owned_by_user(sk)) {
2867 		sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2868 		bh_unlock_sock(sk);
2869 		sock_put(sk);
2870 		return;
2871 	}
2872 
2873 	if (sk->sk_state == TIPC_ESTABLISHED)
2874 		tipc_sk_check_probing_state(sk, &list);
2875 	else if (sk->sk_state == TIPC_CONNECTING)
2876 		tipc_sk_retry_connect(sk, &list);
2877 
2878 	bh_unlock_sock(sk);
2879 
2880 	if (!skb_queue_empty(&list))
2881 		rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2882 
2883 	/* SYN messages may cause link congestion */
2884 	if (rc == -ELINKCONG) {
2885 		tipc_dest_push(&tsk->cong_links, pnode, 0);
2886 		tsk->cong_link_cnt = 1;
2887 	}
2888 	sock_put(sk);
2889 }
2890 
2891 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2892 {
2893 	struct sock *sk = &tsk->sk;
2894 	struct net *net = sock_net(sk);
2895 	struct tipc_socket_addr skaddr;
2896 	struct publication *p;
2897 	u32 key;
2898 
2899 	if (tipc_sk_connected(sk))
2900 		return -EINVAL;
2901 	key = tsk->portid + tsk->pub_count + 1;
2902 	if (key == tsk->portid)
2903 		return -EADDRINUSE;
2904 	skaddr.ref = tsk->portid;
2905 	skaddr.node = tipc_own_addr(net);
2906 	p = tipc_nametbl_publish(net, ua, &skaddr, key);
2907 	if (unlikely(!p))
2908 		return -EINVAL;
2909 
2910 	list_add(&p->binding_sock, &tsk->publications);
2911 	tsk->pub_count++;
2912 	tsk->published = true;
2913 	return 0;
2914 }
2915 
2916 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2917 {
2918 	struct net *net = sock_net(&tsk->sk);
2919 	struct publication *safe, *p;
2920 	struct tipc_uaddr _ua;
2921 	int rc = -EINVAL;
2922 
2923 	list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2924 		if (!ua) {
2925 			tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2926 				   p->sr.type, p->sr.lower, p->sr.upper);
2927 			tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2928 			continue;
2929 		}
2930 		/* Unbind specific publication */
2931 		if (p->scope != ua->scope)
2932 			continue;
2933 		if (p->sr.type != ua->sr.type)
2934 			continue;
2935 		if (p->sr.lower != ua->sr.lower)
2936 			continue;
2937 		if (p->sr.upper != ua->sr.upper)
2938 			break;
2939 		tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2940 		rc = 0;
2941 		break;
2942 	}
2943 	if (list_empty(&tsk->publications)) {
2944 		tsk->published = 0;
2945 		rc = 0;
2946 	}
2947 	return rc;
2948 }
2949 
2950 /* tipc_sk_reinit: set non-zero address in all existing sockets
2951  *                 when we go from standalone to network mode.
2952  */
2953 void tipc_sk_reinit(struct net *net)
2954 {
2955 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2956 	struct rhashtable_iter iter;
2957 	struct tipc_sock *tsk;
2958 	struct tipc_msg *msg;
2959 
2960 	rhashtable_walk_enter(&tn->sk_rht, &iter);
2961 
2962 	do {
2963 		rhashtable_walk_start(&iter);
2964 
2965 		while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2966 			sock_hold(&tsk->sk);
2967 			rhashtable_walk_stop(&iter);
2968 			lock_sock(&tsk->sk);
2969 			msg = &tsk->phdr;
2970 			msg_set_prevnode(msg, tipc_own_addr(net));
2971 			msg_set_orignode(msg, tipc_own_addr(net));
2972 			release_sock(&tsk->sk);
2973 			rhashtable_walk_start(&iter);
2974 			sock_put(&tsk->sk);
2975 		}
2976 
2977 		rhashtable_walk_stop(&iter);
2978 	} while (tsk == ERR_PTR(-EAGAIN));
2979 
2980 	rhashtable_walk_exit(&iter);
2981 }
2982 
2983 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2984 {
2985 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2986 	struct tipc_sock *tsk;
2987 
2988 	rcu_read_lock();
2989 	tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2990 	if (tsk)
2991 		sock_hold(&tsk->sk);
2992 	rcu_read_unlock();
2993 
2994 	return tsk;
2995 }
2996 
2997 static int tipc_sk_insert(struct tipc_sock *tsk)
2998 {
2999 	struct sock *sk = &tsk->sk;
3000 	struct net *net = sock_net(sk);
3001 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3002 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3003 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3004 
3005 	while (remaining--) {
3006 		portid++;
3007 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3008 			portid = TIPC_MIN_PORT;
3009 		tsk->portid = portid;
3010 		sock_hold(&tsk->sk);
3011 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3012 						   tsk_rht_params))
3013 			return 0;
3014 		sock_put(&tsk->sk);
3015 	}
3016 
3017 	return -1;
3018 }
3019 
3020 static void tipc_sk_remove(struct tipc_sock *tsk)
3021 {
3022 	struct sock *sk = &tsk->sk;
3023 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3024 
3025 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3026 		WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3027 		__sock_put(sk);
3028 	}
3029 }
3030 
3031 static const struct rhashtable_params tsk_rht_params = {
3032 	.nelem_hint = 192,
3033 	.head_offset = offsetof(struct tipc_sock, node),
3034 	.key_offset = offsetof(struct tipc_sock, portid),
3035 	.key_len = sizeof(u32), /* portid */
3036 	.max_size = 1048576,
3037 	.min_size = 256,
3038 	.automatic_shrinking = true,
3039 };
3040 
3041 int tipc_sk_rht_init(struct net *net)
3042 {
3043 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3044 
3045 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3046 }
3047 
3048 void tipc_sk_rht_destroy(struct net *net)
3049 {
3050 	struct tipc_net *tn = net_generic(net, tipc_net_id);
3051 
3052 	/* Wait for socket readers to complete */
3053 	synchronize_net();
3054 
3055 	rhashtable_destroy(&tn->sk_rht);
3056 }
3057 
3058 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3059 {
3060 	struct net *net = sock_net(&tsk->sk);
3061 	struct tipc_group *grp = tsk->group;
3062 	struct tipc_msg *hdr = &tsk->phdr;
3063 	struct tipc_uaddr ua;
3064 	int rc;
3065 
3066 	if (mreq->type < TIPC_RESERVED_TYPES)
3067 		return -EACCES;
3068 	if (mreq->scope > TIPC_NODE_SCOPE)
3069 		return -EINVAL;
3070 	if (mreq->scope != TIPC_NODE_SCOPE)
3071 		mreq->scope = TIPC_CLUSTER_SCOPE;
3072 	if (grp)
3073 		return -EACCES;
3074 	grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3075 	if (!grp)
3076 		return -ENOMEM;
3077 	tsk->group = grp;
3078 	msg_set_lookup_scope(hdr, mreq->scope);
3079 	msg_set_nametype(hdr, mreq->type);
3080 	msg_set_dest_droppable(hdr, true);
3081 	tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3082 		   mreq->type, mreq->instance, mreq->instance);
3083 	tipc_nametbl_build_group(net, grp, &ua);
3084 	rc = tipc_sk_publish(tsk, &ua);
3085 	if (rc) {
3086 		tipc_group_delete(net, grp);
3087 		tsk->group = NULL;
3088 		return rc;
3089 	}
3090 	/* Eliminate any risk that a broadcast overtakes sent JOINs */
3091 	tsk->mc_method.rcast = true;
3092 	tsk->mc_method.mandatory = true;
3093 	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3094 	return rc;
3095 }
3096 
3097 static int tipc_sk_leave(struct tipc_sock *tsk)
3098 {
3099 	struct net *net = sock_net(&tsk->sk);
3100 	struct tipc_group *grp = tsk->group;
3101 	struct tipc_uaddr ua;
3102 	int scope;
3103 
3104 	if (!grp)
3105 		return -EINVAL;
3106 	ua.addrtype = TIPC_SERVICE_RANGE;
3107 	tipc_group_self(grp, &ua.sr, &scope);
3108 	ua.scope = scope;
3109 	tipc_group_delete(net, grp);
3110 	tsk->group = NULL;
3111 	tipc_sk_withdraw(tsk, &ua);
3112 	return 0;
3113 }
3114 
3115 /**
3116  * tipc_setsockopt - set socket option
3117  * @sock: socket structure
3118  * @lvl: option level
3119  * @opt: option identifier
3120  * @ov: pointer to new option value
3121  * @ol: length of option value
3122  *
3123  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3124  * (to ease compatibility).
3125  *
3126  * Return: 0 on success, errno otherwise
3127  */
3128 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3129 			   sockptr_t ov, unsigned int ol)
3130 {
3131 	struct sock *sk = sock->sk;
3132 	struct tipc_sock *tsk = tipc_sk(sk);
3133 	struct tipc_group_req mreq;
3134 	u32 value = 0;
3135 	int res = 0;
3136 
3137 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3138 		return 0;
3139 	if (lvl != SOL_TIPC)
3140 		return -ENOPROTOOPT;
3141 
3142 	switch (opt) {
3143 	case TIPC_IMPORTANCE:
3144 	case TIPC_SRC_DROPPABLE:
3145 	case TIPC_DEST_DROPPABLE:
3146 	case TIPC_CONN_TIMEOUT:
3147 	case TIPC_NODELAY:
3148 		if (ol < sizeof(value))
3149 			return -EINVAL;
3150 		if (copy_from_sockptr(&value, ov, sizeof(u32)))
3151 			return -EFAULT;
3152 		break;
3153 	case TIPC_GROUP_JOIN:
3154 		if (ol < sizeof(mreq))
3155 			return -EINVAL;
3156 		if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3157 			return -EFAULT;
3158 		break;
3159 	default:
3160 		if (!sockptr_is_null(ov) || ol)
3161 			return -EINVAL;
3162 	}
3163 
3164 	lock_sock(sk);
3165 
3166 	switch (opt) {
3167 	case TIPC_IMPORTANCE:
3168 		res = tsk_set_importance(sk, value);
3169 		break;
3170 	case TIPC_SRC_DROPPABLE:
3171 		if (sock->type != SOCK_STREAM)
3172 			tsk_set_unreliable(tsk, value);
3173 		else
3174 			res = -ENOPROTOOPT;
3175 		break;
3176 	case TIPC_DEST_DROPPABLE:
3177 		tsk_set_unreturnable(tsk, value);
3178 		break;
3179 	case TIPC_CONN_TIMEOUT:
3180 		tipc_sk(sk)->conn_timeout = value;
3181 		break;
3182 	case TIPC_MCAST_BROADCAST:
3183 		tsk->mc_method.rcast = false;
3184 		tsk->mc_method.mandatory = true;
3185 		break;
3186 	case TIPC_MCAST_REPLICAST:
3187 		tsk->mc_method.rcast = true;
3188 		tsk->mc_method.mandatory = true;
3189 		break;
3190 	case TIPC_GROUP_JOIN:
3191 		res = tipc_sk_join(tsk, &mreq);
3192 		break;
3193 	case TIPC_GROUP_LEAVE:
3194 		res = tipc_sk_leave(tsk);
3195 		break;
3196 	case TIPC_NODELAY:
3197 		tsk->nodelay = !!value;
3198 		tsk_set_nagle(tsk);
3199 		break;
3200 	default:
3201 		res = -EINVAL;
3202 	}
3203 
3204 	release_sock(sk);
3205 
3206 	return res;
3207 }
3208 
3209 /**
3210  * tipc_getsockopt - get socket option
3211  * @sock: socket structure
3212  * @lvl: option level
3213  * @opt: option identifier
3214  * @ov: receptacle for option value
3215  * @ol: receptacle for length of option value
3216  *
3217  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3218  * (to ease compatibility).
3219  *
3220  * Return: 0 on success, errno otherwise
3221  */
3222 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3223 			   char __user *ov, int __user *ol)
3224 {
3225 	struct sock *sk = sock->sk;
3226 	struct tipc_sock *tsk = tipc_sk(sk);
3227 	struct tipc_service_range seq;
3228 	int len, scope;
3229 	u32 value;
3230 	int res;
3231 
3232 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3233 		return put_user(0, ol);
3234 	if (lvl != SOL_TIPC)
3235 		return -ENOPROTOOPT;
3236 	res = get_user(len, ol);
3237 	if (res)
3238 		return res;
3239 
3240 	lock_sock(sk);
3241 
3242 	switch (opt) {
3243 	case TIPC_IMPORTANCE:
3244 		value = tsk_importance(tsk);
3245 		break;
3246 	case TIPC_SRC_DROPPABLE:
3247 		value = tsk_unreliable(tsk);
3248 		break;
3249 	case TIPC_DEST_DROPPABLE:
3250 		value = tsk_unreturnable(tsk);
3251 		break;
3252 	case TIPC_CONN_TIMEOUT:
3253 		value = tsk->conn_timeout;
3254 		/* no need to set "res", since already 0 at this point */
3255 		break;
3256 	case TIPC_NODE_RECVQ_DEPTH:
3257 		value = 0; /* was tipc_queue_size, now obsolete */
3258 		break;
3259 	case TIPC_SOCK_RECVQ_DEPTH:
3260 		value = skb_queue_len(&sk->sk_receive_queue);
3261 		break;
3262 	case TIPC_SOCK_RECVQ_USED:
3263 		value = sk_rmem_alloc_get(sk);
3264 		break;
3265 	case TIPC_GROUP_JOIN:
3266 		seq.type = 0;
3267 		if (tsk->group)
3268 			tipc_group_self(tsk->group, &seq, &scope);
3269 		value = seq.type;
3270 		break;
3271 	default:
3272 		res = -EINVAL;
3273 	}
3274 
3275 	release_sock(sk);
3276 
3277 	if (res)
3278 		return res;	/* "get" failed */
3279 
3280 	if (len < sizeof(value))
3281 		return -EINVAL;
3282 
3283 	if (copy_to_user(ov, &value, sizeof(value)))
3284 		return -EFAULT;
3285 
3286 	return put_user(sizeof(value), ol);
3287 }
3288 
3289 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3290 {
3291 	struct net *net = sock_net(sock->sk);
3292 	struct tipc_sioc_nodeid_req nr = {0};
3293 	struct tipc_sioc_ln_req lnr;
3294 	void __user *argp = (void __user *)arg;
3295 
3296 	switch (cmd) {
3297 	case SIOCGETLINKNAME:
3298 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
3299 			return -EFAULT;
3300 		if (!tipc_node_get_linkname(net,
3301 					    lnr.bearer_id & 0xffff, lnr.peer,
3302 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
3303 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
3304 				return -EFAULT;
3305 			return 0;
3306 		}
3307 		return -EADDRNOTAVAIL;
3308 	case SIOCGETNODEID:
3309 		if (copy_from_user(&nr, argp, sizeof(nr)))
3310 			return -EFAULT;
3311 		if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3312 			return -EADDRNOTAVAIL;
3313 		if (copy_to_user(argp, &nr, sizeof(nr)))
3314 			return -EFAULT;
3315 		return 0;
3316 	default:
3317 		return -ENOIOCTLCMD;
3318 	}
3319 }
3320 
3321 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3322 {
3323 	struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3324 	struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3325 	u32 onode = tipc_own_addr(sock_net(sock1->sk));
3326 
3327 	tsk1->peer.family = AF_TIPC;
3328 	tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3329 	tsk1->peer.scope = TIPC_NODE_SCOPE;
3330 	tsk1->peer.addr.id.ref = tsk2->portid;
3331 	tsk1->peer.addr.id.node = onode;
3332 	tsk2->peer.family = AF_TIPC;
3333 	tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3334 	tsk2->peer.scope = TIPC_NODE_SCOPE;
3335 	tsk2->peer.addr.id.ref = tsk1->portid;
3336 	tsk2->peer.addr.id.node = onode;
3337 
3338 	tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3339 	tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3340 	return 0;
3341 }
3342 
3343 /* Protocol switches for the various types of TIPC sockets */
3344 
3345 static const struct proto_ops msg_ops = {
3346 	.owner		= THIS_MODULE,
3347 	.family		= AF_TIPC,
3348 	.release	= tipc_release,
3349 	.bind		= tipc_bind,
3350 	.connect	= tipc_connect,
3351 	.socketpair	= tipc_socketpair,
3352 	.accept		= sock_no_accept,
3353 	.getname	= tipc_getname,
3354 	.poll		= tipc_poll,
3355 	.ioctl		= tipc_ioctl,
3356 	.listen		= sock_no_listen,
3357 	.shutdown	= tipc_shutdown,
3358 	.setsockopt	= tipc_setsockopt,
3359 	.getsockopt	= tipc_getsockopt,
3360 	.sendmsg	= tipc_sendmsg,
3361 	.recvmsg	= tipc_recvmsg,
3362 	.mmap		= sock_no_mmap,
3363 	.sendpage	= sock_no_sendpage
3364 };
3365 
3366 static const struct proto_ops packet_ops = {
3367 	.owner		= THIS_MODULE,
3368 	.family		= AF_TIPC,
3369 	.release	= tipc_release,
3370 	.bind		= tipc_bind,
3371 	.connect	= tipc_connect,
3372 	.socketpair	= tipc_socketpair,
3373 	.accept		= tipc_accept,
3374 	.getname	= tipc_getname,
3375 	.poll		= tipc_poll,
3376 	.ioctl		= tipc_ioctl,
3377 	.listen		= tipc_listen,
3378 	.shutdown	= tipc_shutdown,
3379 	.setsockopt	= tipc_setsockopt,
3380 	.getsockopt	= tipc_getsockopt,
3381 	.sendmsg	= tipc_send_packet,
3382 	.recvmsg	= tipc_recvmsg,
3383 	.mmap		= sock_no_mmap,
3384 	.sendpage	= sock_no_sendpage
3385 };
3386 
3387 static const struct proto_ops stream_ops = {
3388 	.owner		= THIS_MODULE,
3389 	.family		= AF_TIPC,
3390 	.release	= tipc_release,
3391 	.bind		= tipc_bind,
3392 	.connect	= tipc_connect,
3393 	.socketpair	= tipc_socketpair,
3394 	.accept		= tipc_accept,
3395 	.getname	= tipc_getname,
3396 	.poll		= tipc_poll,
3397 	.ioctl		= tipc_ioctl,
3398 	.listen		= tipc_listen,
3399 	.shutdown	= tipc_shutdown,
3400 	.setsockopt	= tipc_setsockopt,
3401 	.getsockopt	= tipc_getsockopt,
3402 	.sendmsg	= tipc_sendstream,
3403 	.recvmsg	= tipc_recvstream,
3404 	.mmap		= sock_no_mmap,
3405 	.sendpage	= sock_no_sendpage
3406 };
3407 
3408 static const struct net_proto_family tipc_family_ops = {
3409 	.owner		= THIS_MODULE,
3410 	.family		= AF_TIPC,
3411 	.create		= tipc_sk_create
3412 };
3413 
3414 static struct proto tipc_proto = {
3415 	.name		= "TIPC",
3416 	.owner		= THIS_MODULE,
3417 	.obj_size	= sizeof(struct tipc_sock),
3418 	.sysctl_rmem	= sysctl_tipc_rmem
3419 };
3420 
3421 /**
3422  * tipc_socket_init - initialize TIPC socket interface
3423  *
3424  * Return: 0 on success, errno otherwise
3425  */
3426 int tipc_socket_init(void)
3427 {
3428 	int res;
3429 
3430 	res = proto_register(&tipc_proto, 1);
3431 	if (res) {
3432 		pr_err("Failed to register TIPC protocol type\n");
3433 		goto out;
3434 	}
3435 
3436 	res = sock_register(&tipc_family_ops);
3437 	if (res) {
3438 		pr_err("Failed to register TIPC socket type\n");
3439 		proto_unregister(&tipc_proto);
3440 		goto out;
3441 	}
3442  out:
3443 	return res;
3444 }
3445 
3446 /**
3447  * tipc_socket_stop - stop TIPC socket interface
3448  */
3449 void tipc_socket_stop(void)
3450 {
3451 	sock_unregister(tipc_family_ops.family);
3452 	proto_unregister(&tipc_proto);
3453 }
3454 
3455 /* Caller should hold socket lock for the passed tipc socket. */
3456 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3457 {
3458 	u32 peer_node;
3459 	u32 peer_port;
3460 	struct nlattr *nest;
3461 
3462 	peer_node = tsk_peer_node(tsk);
3463 	peer_port = tsk_peer_port(tsk);
3464 
3465 	nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3466 	if (!nest)
3467 		return -EMSGSIZE;
3468 
3469 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3470 		goto msg_full;
3471 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3472 		goto msg_full;
3473 
3474 	if (tsk->conn_type != 0) {
3475 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3476 			goto msg_full;
3477 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3478 			goto msg_full;
3479 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3480 			goto msg_full;
3481 	}
3482 	nla_nest_end(skb, nest);
3483 
3484 	return 0;
3485 
3486 msg_full:
3487 	nla_nest_cancel(skb, nest);
3488 
3489 	return -EMSGSIZE;
3490 }
3491 
3492 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3493 			  *tsk)
3494 {
3495 	struct net *net = sock_net(skb->sk);
3496 	struct sock *sk = &tsk->sk;
3497 
3498 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3499 	    nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3500 		return -EMSGSIZE;
3501 
3502 	if (tipc_sk_connected(sk)) {
3503 		if (__tipc_nl_add_sk_con(skb, tsk))
3504 			return -EMSGSIZE;
3505 	} else if (!list_empty(&tsk->publications)) {
3506 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3507 			return -EMSGSIZE;
3508 	}
3509 	return 0;
3510 }
3511 
3512 /* Caller should hold socket lock for the passed tipc socket. */
3513 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3514 			    struct tipc_sock *tsk)
3515 {
3516 	struct nlattr *attrs;
3517 	void *hdr;
3518 
3519 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3520 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3521 	if (!hdr)
3522 		goto msg_cancel;
3523 
3524 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3525 	if (!attrs)
3526 		goto genlmsg_cancel;
3527 
3528 	if (__tipc_nl_add_sk_info(skb, tsk))
3529 		goto attr_msg_cancel;
3530 
3531 	nla_nest_end(skb, attrs);
3532 	genlmsg_end(skb, hdr);
3533 
3534 	return 0;
3535 
3536 attr_msg_cancel:
3537 	nla_nest_cancel(skb, attrs);
3538 genlmsg_cancel:
3539 	genlmsg_cancel(skb, hdr);
3540 msg_cancel:
3541 	return -EMSGSIZE;
3542 }
3543 
3544 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3545 		    int (*skb_handler)(struct sk_buff *skb,
3546 				       struct netlink_callback *cb,
3547 				       struct tipc_sock *tsk))
3548 {
3549 	struct rhashtable_iter *iter = (void *)cb->args[4];
3550 	struct tipc_sock *tsk;
3551 	int err;
3552 
3553 	rhashtable_walk_start(iter);
3554 	while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3555 		if (IS_ERR(tsk)) {
3556 			err = PTR_ERR(tsk);
3557 			if (err == -EAGAIN) {
3558 				err = 0;
3559 				continue;
3560 			}
3561 			break;
3562 		}
3563 
3564 		sock_hold(&tsk->sk);
3565 		rhashtable_walk_stop(iter);
3566 		lock_sock(&tsk->sk);
3567 		err = skb_handler(skb, cb, tsk);
3568 		if (err) {
3569 			release_sock(&tsk->sk);
3570 			sock_put(&tsk->sk);
3571 			goto out;
3572 		}
3573 		release_sock(&tsk->sk);
3574 		rhashtable_walk_start(iter);
3575 		sock_put(&tsk->sk);
3576 	}
3577 	rhashtable_walk_stop(iter);
3578 out:
3579 	return skb->len;
3580 }
3581 EXPORT_SYMBOL(tipc_nl_sk_walk);
3582 
3583 int tipc_dump_start(struct netlink_callback *cb)
3584 {
3585 	return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3586 }
3587 EXPORT_SYMBOL(tipc_dump_start);
3588 
3589 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3590 {
3591 	/* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3592 	struct rhashtable_iter *iter = (void *)cb->args[4];
3593 	struct tipc_net *tn = tipc_net(net);
3594 
3595 	if (!iter) {
3596 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3597 		if (!iter)
3598 			return -ENOMEM;
3599 
3600 		cb->args[4] = (long)iter;
3601 	}
3602 
3603 	rhashtable_walk_enter(&tn->sk_rht, iter);
3604 	return 0;
3605 }
3606 
3607 int tipc_dump_done(struct netlink_callback *cb)
3608 {
3609 	struct rhashtable_iter *hti = (void *)cb->args[4];
3610 
3611 	rhashtable_walk_exit(hti);
3612 	kfree(hti);
3613 	return 0;
3614 }
3615 EXPORT_SYMBOL(tipc_dump_done);
3616 
3617 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3618 			   struct tipc_sock *tsk, u32 sk_filter_state,
3619 			   u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3620 {
3621 	struct sock *sk = &tsk->sk;
3622 	struct nlattr *attrs;
3623 	struct nlattr *stat;
3624 
3625 	/*filter response w.r.t sk_state*/
3626 	if (!(sk_filter_state & (1 << sk->sk_state)))
3627 		return 0;
3628 
3629 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3630 	if (!attrs)
3631 		goto msg_cancel;
3632 
3633 	if (__tipc_nl_add_sk_info(skb, tsk))
3634 		goto attr_msg_cancel;
3635 
3636 	if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3637 	    nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3638 	    nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3639 	    nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3640 			from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3641 					 sock_i_uid(sk))) ||
3642 	    nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3643 			      tipc_diag_gen_cookie(sk),
3644 			      TIPC_NLA_SOCK_PAD))
3645 		goto attr_msg_cancel;
3646 
3647 	stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3648 	if (!stat)
3649 		goto attr_msg_cancel;
3650 
3651 	if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3652 			skb_queue_len(&sk->sk_receive_queue)) ||
3653 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3654 			skb_queue_len(&sk->sk_write_queue)) ||
3655 	    nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3656 			atomic_read(&sk->sk_drops)))
3657 		goto stat_msg_cancel;
3658 
3659 	if (tsk->cong_link_cnt &&
3660 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3661 		goto stat_msg_cancel;
3662 
3663 	if (tsk_conn_cong(tsk) &&
3664 	    nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3665 		goto stat_msg_cancel;
3666 
3667 	nla_nest_end(skb, stat);
3668 
3669 	if (tsk->group)
3670 		if (tipc_group_fill_sock_diag(tsk->group, skb))
3671 			goto stat_msg_cancel;
3672 
3673 	nla_nest_end(skb, attrs);
3674 
3675 	return 0;
3676 
3677 stat_msg_cancel:
3678 	nla_nest_cancel(skb, stat);
3679 attr_msg_cancel:
3680 	nla_nest_cancel(skb, attrs);
3681 msg_cancel:
3682 	return -EMSGSIZE;
3683 }
3684 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3685 
3686 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3687 {
3688 	return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3689 }
3690 
3691 /* Caller should hold socket lock for the passed tipc socket. */
3692 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3693 				 struct netlink_callback *cb,
3694 				 struct publication *publ)
3695 {
3696 	void *hdr;
3697 	struct nlattr *attrs;
3698 
3699 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3700 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3701 	if (!hdr)
3702 		goto msg_cancel;
3703 
3704 	attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3705 	if (!attrs)
3706 		goto genlmsg_cancel;
3707 
3708 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3709 		goto attr_msg_cancel;
3710 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3711 		goto attr_msg_cancel;
3712 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3713 		goto attr_msg_cancel;
3714 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3715 		goto attr_msg_cancel;
3716 
3717 	nla_nest_end(skb, attrs);
3718 	genlmsg_end(skb, hdr);
3719 
3720 	return 0;
3721 
3722 attr_msg_cancel:
3723 	nla_nest_cancel(skb, attrs);
3724 genlmsg_cancel:
3725 	genlmsg_cancel(skb, hdr);
3726 msg_cancel:
3727 	return -EMSGSIZE;
3728 }
3729 
3730 /* Caller should hold socket lock for the passed tipc socket. */
3731 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3732 				  struct netlink_callback *cb,
3733 				  struct tipc_sock *tsk, u32 *last_publ)
3734 {
3735 	int err;
3736 	struct publication *p;
3737 
3738 	if (*last_publ) {
3739 		list_for_each_entry(p, &tsk->publications, binding_sock) {
3740 			if (p->key == *last_publ)
3741 				break;
3742 		}
3743 		if (p->key != *last_publ) {
3744 			/* We never set seq or call nl_dump_check_consistent()
3745 			 * this means that setting prev_seq here will cause the
3746 			 * consistence check to fail in the netlink callback
3747 			 * handler. Resulting in the last NLMSG_DONE message
3748 			 * having the NLM_F_DUMP_INTR flag set.
3749 			 */
3750 			cb->prev_seq = 1;
3751 			*last_publ = 0;
3752 			return -EPIPE;
3753 		}
3754 	} else {
3755 		p = list_first_entry(&tsk->publications, struct publication,
3756 				     binding_sock);
3757 	}
3758 
3759 	list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3760 		err = __tipc_nl_add_sk_publ(skb, cb, p);
3761 		if (err) {
3762 			*last_publ = p->key;
3763 			return err;
3764 		}
3765 	}
3766 	*last_publ = 0;
3767 
3768 	return 0;
3769 }
3770 
3771 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3772 {
3773 	int err;
3774 	u32 tsk_portid = cb->args[0];
3775 	u32 last_publ = cb->args[1];
3776 	u32 done = cb->args[2];
3777 	struct net *net = sock_net(skb->sk);
3778 	struct tipc_sock *tsk;
3779 
3780 	if (!tsk_portid) {
3781 		struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3782 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3783 
3784 		if (!attrs[TIPC_NLA_SOCK])
3785 			return -EINVAL;
3786 
3787 		err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3788 						  attrs[TIPC_NLA_SOCK],
3789 						  tipc_nl_sock_policy, NULL);
3790 		if (err)
3791 			return err;
3792 
3793 		if (!sock[TIPC_NLA_SOCK_REF])
3794 			return -EINVAL;
3795 
3796 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3797 	}
3798 
3799 	if (done)
3800 		return 0;
3801 
3802 	tsk = tipc_sk_lookup(net, tsk_portid);
3803 	if (!tsk)
3804 		return -EINVAL;
3805 
3806 	lock_sock(&tsk->sk);
3807 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3808 	if (!err)
3809 		done = 1;
3810 	release_sock(&tsk->sk);
3811 	sock_put(&tsk->sk);
3812 
3813 	cb->args[0] = tsk_portid;
3814 	cb->args[1] = last_publ;
3815 	cb->args[2] = done;
3816 
3817 	return skb->len;
3818 }
3819 
3820 /**
3821  * tipc_sk_filtering - check if a socket should be traced
3822  * @sk: the socket to be examined
3823  *
3824  * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3825  * (portid, sock type, name type, name lower, name upper)
3826  *
3827  * Return: true if the socket meets the socket tuple data
3828  * (value 0 = 'any') or when there is no tuple set (all = 0),
3829  * otherwise false
3830  */
3831 bool tipc_sk_filtering(struct sock *sk)
3832 {
3833 	struct tipc_sock *tsk;
3834 	struct publication *p;
3835 	u32 _port, _sktype, _type, _lower, _upper;
3836 	u32 type = 0, lower = 0, upper = 0;
3837 
3838 	if (!sk)
3839 		return true;
3840 
3841 	tsk = tipc_sk(sk);
3842 
3843 	_port = sysctl_tipc_sk_filter[0];
3844 	_sktype = sysctl_tipc_sk_filter[1];
3845 	_type = sysctl_tipc_sk_filter[2];
3846 	_lower = sysctl_tipc_sk_filter[3];
3847 	_upper = sysctl_tipc_sk_filter[4];
3848 
3849 	if (!_port && !_sktype && !_type && !_lower && !_upper)
3850 		return true;
3851 
3852 	if (_port)
3853 		return (_port == tsk->portid);
3854 
3855 	if (_sktype && _sktype != sk->sk_type)
3856 		return false;
3857 
3858 	if (tsk->published) {
3859 		p = list_first_entry_or_null(&tsk->publications,
3860 					     struct publication, binding_sock);
3861 		if (p) {
3862 			type = p->sr.type;
3863 			lower = p->sr.lower;
3864 			upper = p->sr.upper;
3865 		}
3866 	}
3867 
3868 	if (!tipc_sk_type_connectionless(sk)) {
3869 		type = tsk->conn_type;
3870 		lower = tsk->conn_instance;
3871 		upper = tsk->conn_instance;
3872 	}
3873 
3874 	if ((_type && _type != type) || (_lower && _lower != lower) ||
3875 	    (_upper && _upper != upper))
3876 		return false;
3877 
3878 	return true;
3879 }
3880 
3881 u32 tipc_sock_get_portid(struct sock *sk)
3882 {
3883 	return (sk) ? (tipc_sk(sk))->portid : 0;
3884 }
3885 
3886 /**
3887  * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3888  *			both the rcv and backlog queues are considered
3889  * @sk: tipc sk to be checked
3890  * @skb: tipc msg to be checked
3891  *
3892  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3893  */
3894 
3895 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3896 {
3897 	atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3898 	unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3899 	unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3900 
3901 	return (qsize > lim * 90 / 100);
3902 }
3903 
3904 /**
3905  * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3906  *			only the rcv queue is considered
3907  * @sk: tipc sk to be checked
3908  * @skb: tipc msg to be checked
3909  *
3910  * Return: true if the socket rx queue allocation is > 90%, otherwise false
3911  */
3912 
3913 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3914 {
3915 	unsigned int lim = rcvbuf_limit(sk, skb);
3916 	unsigned int qsize = sk_rmem_alloc_get(sk);
3917 
3918 	return (qsize > lim * 90 / 100);
3919 }
3920 
3921 /**
3922  * tipc_sk_dump - dump TIPC socket
3923  * @sk: tipc sk to be dumped
3924  * @dqueues: bitmask to decide if any socket queue to be dumped?
3925  *           - TIPC_DUMP_NONE: don't dump socket queues
3926  *           - TIPC_DUMP_SK_SNDQ: dump socket send queue
3927  *           - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3928  *           - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3929  *           - TIPC_DUMP_ALL: dump all the socket queues above
3930  * @buf: returned buffer of dump data in format
3931  */
3932 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3933 {
3934 	int i = 0;
3935 	size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3936 	struct tipc_sock *tsk;
3937 	struct publication *p;
3938 	bool tsk_connected;
3939 
3940 	if (!sk) {
3941 		i += scnprintf(buf, sz, "sk data: (null)\n");
3942 		return i;
3943 	}
3944 
3945 	tsk = tipc_sk(sk);
3946 	tsk_connected = !tipc_sk_type_connectionless(sk);
3947 
3948 	i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3949 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3950 	i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3951 	i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3952 	i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3953 	if (tsk_connected) {
3954 		i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3955 		i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3956 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3957 		i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3958 	}
3959 	i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3960 	if (tsk->published) {
3961 		p = list_first_entry_or_null(&tsk->publications,
3962 					     struct publication, binding_sock);
3963 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3964 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3965 		i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3966 	}
3967 	i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3968 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3969 	i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3970 	i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3971 	i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3972 	i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3973 	i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3974 	i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3975 	i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3976 	i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3977 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3978 	i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3979 	i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3980 	i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3981 
3982 	if (dqueues & TIPC_DUMP_SK_SNDQ) {
3983 		i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3984 		i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3985 	}
3986 
3987 	if (dqueues & TIPC_DUMP_SK_RCVQ) {
3988 		i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3989 		i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3990 	}
3991 
3992 	if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3993 		i += scnprintf(buf + i, sz - i, "sk_backlog:\n  head ");
3994 		i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3995 		if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3996 			i += scnprintf(buf + i, sz - i, "  tail ");
3997 			i += tipc_skb_dump(sk->sk_backlog.tail, false,
3998 					   buf + i);
3999 		}
4000 	}
4001 
4002 	return i;
4003 }
4004