• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * net/tipc/socket.c: TIPC socket API
3  *
4  * Copyright (c) 2001-2007, 2012-2015, Ericsson AB
5  * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include <linux/rhashtable.h>
38 #include "core.h"
39 #include "name_table.h"
40 #include "node.h"
41 #include "link.h"
42 #include "name_distr.h"
43 #include "socket.h"
44 #include "bcast.h"
45 
46 #define SS_LISTENING		-1	/* socket is listening */
47 #define SS_READY		-2	/* socket is connectionless */
48 
49 #define CONN_TIMEOUT_DEFAULT	8000	/* default connect timeout = 8s */
50 #define CONN_PROBING_INTERVAL	msecs_to_jiffies(3600000)  /* [ms] => 1 h */
51 #define TIPC_FWD_MSG		1
52 #define TIPC_CONN_OK		0
53 #define TIPC_CONN_PROBING	1
54 #define TIPC_MAX_PORT		0xffffffff
55 #define TIPC_MIN_PORT		1
56 
57 /**
58  * struct tipc_sock - TIPC socket structure
59  * @sk: socket - interacts with 'port' and with user via the socket API
60  * @connected: non-zero if port is currently connected to a peer port
61  * @conn_type: TIPC type used when connection was established
62  * @conn_instance: TIPC instance used when connection was established
63  * @published: non-zero if port has one or more associated names
64  * @max_pkt: maximum packet size "hint" used when building messages sent by port
65  * @portid: unique port identity in TIPC socket hash table
66  * @phdr: preformatted message header used when sending messages
67  * @port_list: adjacent ports in TIPC's global list of ports
68  * @publications: list of publications for port
69  * @pub_count: total # of publications port has made during its lifetime
70  * @probing_state:
71  * @probing_intv:
72  * @conn_timeout: the time we can wait for an unresponded setup request
73  * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
74  * @link_cong: non-zero if owner must sleep because of link congestion
75  * @sent_unacked: # messages sent by socket, and not yet acked by peer
76  * @rcv_unacked: # messages read by user, but not yet acked back to peer
77  * @remote: 'connected' peer for dgram/rdm
78  * @node: hash table node
79  * @rcu: rcu struct for tipc_sock
80  */
81 struct tipc_sock {
82 	struct sock sk;
83 	int connected;
84 	u32 conn_type;
85 	u32 conn_instance;
86 	int published;
87 	u32 max_pkt;
88 	u32 portid;
89 	struct tipc_msg phdr;
90 	struct list_head sock_list;
91 	struct list_head publications;
92 	u32 pub_count;
93 	u32 probing_state;
94 	unsigned long probing_intv;
95 	uint conn_timeout;
96 	atomic_t dupl_rcvcnt;
97 	bool link_cong;
98 	uint sent_unacked;
99 	uint rcv_unacked;
100 	struct sockaddr_tipc remote;
101 	struct rhash_head node;
102 	struct rcu_head rcu;
103 };
104 
105 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
106 static void tipc_data_ready(struct sock *sk);
107 static void tipc_write_space(struct sock *sk);
108 static void tipc_sock_destruct(struct sock *sk);
109 static int tipc_release(struct socket *sock);
110 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
111 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
112 static void tipc_sk_timeout(unsigned long data);
113 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
114 			   struct tipc_name_seq const *seq);
115 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
116 			    struct tipc_name_seq const *seq);
117 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
118 static int tipc_sk_insert(struct tipc_sock *tsk);
119 static void tipc_sk_remove(struct tipc_sock *tsk);
120 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
121 			      size_t dsz);
122 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
123 
124 static const struct proto_ops packet_ops;
125 static const struct proto_ops stream_ops;
126 static const struct proto_ops msg_ops;
127 static struct proto tipc_proto;
128 
129 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
130 	[TIPC_NLA_SOCK_UNSPEC]		= { .type = NLA_UNSPEC },
131 	[TIPC_NLA_SOCK_ADDR]		= { .type = NLA_U32 },
132 	[TIPC_NLA_SOCK_REF]		= { .type = NLA_U32 },
133 	[TIPC_NLA_SOCK_CON]		= { .type = NLA_NESTED },
134 	[TIPC_NLA_SOCK_HAS_PUBL]	= { .type = NLA_FLAG }
135 };
136 
137 static const struct rhashtable_params tsk_rht_params;
138 
139 /*
140  * Revised TIPC socket locking policy:
141  *
142  * Most socket operations take the standard socket lock when they start
143  * and hold it until they finish (or until they need to sleep).  Acquiring
144  * this lock grants the owner exclusive access to the fields of the socket
145  * data structures, with the exception of the backlog queue.  A few socket
146  * operations can be done without taking the socket lock because they only
147  * read socket information that never changes during the life of the socket.
148  *
149  * Socket operations may acquire the lock for the associated TIPC port if they
150  * need to perform an operation on the port.  If any routine needs to acquire
151  * both the socket lock and the port lock it must take the socket lock first
152  * to avoid the risk of deadlock.
153  *
154  * The dispatcher handling incoming messages cannot grab the socket lock in
155  * the standard fashion, since invoked it runs at the BH level and cannot block.
156  * Instead, it checks to see if the socket lock is currently owned by someone,
157  * and either handles the message itself or adds it to the socket's backlog
158  * queue; in the latter case the queued message is processed once the process
159  * owning the socket lock releases it.
160  *
161  * NOTE: Releasing the socket lock while an operation is sleeping overcomes
162  * the problem of a blocked socket operation preventing any other operations
163  * from occurring.  However, applications must be careful if they have
164  * multiple threads trying to send (or receive) on the same socket, as these
165  * operations might interfere with each other.  For example, doing a connect
166  * and a receive at the same time might allow the receive to consume the
167  * ACK message meant for the connect.  While additional work could be done
168  * to try and overcome this, it doesn't seem to be worthwhile at the present.
169  *
170  * NOTE: Releasing the socket lock while an operation is sleeping also ensures
171  * that another operation that must be performed in a non-blocking manner is
172  * not delayed for very long because the lock has already been taken.
173  *
174  * NOTE: This code assumes that certain fields of a port/socket pair are
175  * constant over its lifetime; such fields can be examined without taking
176  * the socket lock and/or port lock, and do not need to be re-read even
177  * after resuming processing after waiting.  These fields include:
178  *   - socket type
179  *   - pointer to socket sk structure (aka tipc_sock structure)
180  *   - pointer to port structure
181  *   - port reference
182  */
183 
tsk_own_node(struct tipc_sock * tsk)184 static u32 tsk_own_node(struct tipc_sock *tsk)
185 {
186 	return msg_prevnode(&tsk->phdr);
187 }
188 
tsk_peer_node(struct tipc_sock * tsk)189 static u32 tsk_peer_node(struct tipc_sock *tsk)
190 {
191 	return msg_destnode(&tsk->phdr);
192 }
193 
tsk_peer_port(struct tipc_sock * tsk)194 static u32 tsk_peer_port(struct tipc_sock *tsk)
195 {
196 	return msg_destport(&tsk->phdr);
197 }
198 
tsk_unreliable(struct tipc_sock * tsk)199 static  bool tsk_unreliable(struct tipc_sock *tsk)
200 {
201 	return msg_src_droppable(&tsk->phdr) != 0;
202 }
203 
tsk_set_unreliable(struct tipc_sock * tsk,bool unreliable)204 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
205 {
206 	msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
207 }
208 
tsk_unreturnable(struct tipc_sock * tsk)209 static bool tsk_unreturnable(struct tipc_sock *tsk)
210 {
211 	return msg_dest_droppable(&tsk->phdr) != 0;
212 }
213 
tsk_set_unreturnable(struct tipc_sock * tsk,bool unreturnable)214 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
215 {
216 	msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
217 }
218 
tsk_importance(struct tipc_sock * tsk)219 static int tsk_importance(struct tipc_sock *tsk)
220 {
221 	return msg_importance(&tsk->phdr);
222 }
223 
tsk_set_importance(struct tipc_sock * tsk,int imp)224 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
225 {
226 	if (imp > TIPC_CRITICAL_IMPORTANCE)
227 		return -EINVAL;
228 	msg_set_importance(&tsk->phdr, (u32)imp);
229 	return 0;
230 }
231 
tipc_sk(const struct sock * sk)232 static struct tipc_sock *tipc_sk(const struct sock *sk)
233 {
234 	return container_of(sk, struct tipc_sock, sk);
235 }
236 
tsk_conn_cong(struct tipc_sock * tsk)237 static int tsk_conn_cong(struct tipc_sock *tsk)
238 {
239 	return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
240 }
241 
242 /**
243  * tsk_advance_rx_queue - discard first buffer in socket receive queue
244  *
245  * Caller must hold socket lock
246  */
tsk_advance_rx_queue(struct sock * sk)247 static void tsk_advance_rx_queue(struct sock *sk)
248 {
249 	kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
250 }
251 
252 /* tipc_sk_respond() : send response message back to sender
253  */
tipc_sk_respond(struct sock * sk,struct sk_buff * skb,int err)254 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
255 {
256 	u32 selector;
257 	u32 dnode;
258 	u32 onode = tipc_own_addr(sock_net(sk));
259 
260 	if (!tipc_msg_reverse(onode, &skb, err))
261 		return;
262 
263 	dnode = msg_destnode(buf_msg(skb));
264 	selector = msg_origport(buf_msg(skb));
265 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
266 }
267 
268 /**
269  * tsk_rej_rx_queue - reject all buffers in socket receive queue
270  *
271  * Caller must hold socket lock
272  */
tsk_rej_rx_queue(struct sock * sk)273 static void tsk_rej_rx_queue(struct sock *sk)
274 {
275 	struct sk_buff *skb;
276 
277 	while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
278 		tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
279 }
280 
281 /* tsk_peer_msg - verify if message was sent by connected port's peer
282  *
283  * Handles cases where the node's network address has changed from
284  * the default of <0.0.0> to its configured setting.
285  */
tsk_peer_msg(struct tipc_sock * tsk,struct tipc_msg * msg)286 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
287 {
288 	struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
289 	u32 peer_port = tsk_peer_port(tsk);
290 	u32 orig_node;
291 	u32 peer_node;
292 
293 	if (unlikely(!tsk->connected))
294 		return false;
295 
296 	if (unlikely(msg_origport(msg) != peer_port))
297 		return false;
298 
299 	orig_node = msg_orignode(msg);
300 	peer_node = tsk_peer_node(tsk);
301 
302 	if (likely(orig_node == peer_node))
303 		return true;
304 
305 	if (!orig_node && (peer_node == tn->own_addr))
306 		return true;
307 
308 	if (!peer_node && (orig_node == tn->own_addr))
309 		return true;
310 
311 	return false;
312 }
313 
314 /**
315  * tipc_sk_create - create a TIPC socket
316  * @net: network namespace (must be default network)
317  * @sock: pre-allocated socket structure
318  * @protocol: protocol indicator (must be 0)
319  * @kern: caused by kernel or by userspace?
320  *
321  * This routine creates additional data structures used by the TIPC socket,
322  * initializes them, and links them together.
323  *
324  * Returns 0 on success, errno otherwise
325  */
tipc_sk_create(struct net * net,struct socket * sock,int protocol,int kern)326 static int tipc_sk_create(struct net *net, struct socket *sock,
327 			  int protocol, int kern)
328 {
329 	struct tipc_net *tn;
330 	const struct proto_ops *ops;
331 	socket_state state;
332 	struct sock *sk;
333 	struct tipc_sock *tsk;
334 	struct tipc_msg *msg;
335 
336 	/* Validate arguments */
337 	if (unlikely(protocol != 0))
338 		return -EPROTONOSUPPORT;
339 
340 	switch (sock->type) {
341 	case SOCK_STREAM:
342 		ops = &stream_ops;
343 		state = SS_UNCONNECTED;
344 		break;
345 	case SOCK_SEQPACKET:
346 		ops = &packet_ops;
347 		state = SS_UNCONNECTED;
348 		break;
349 	case SOCK_DGRAM:
350 	case SOCK_RDM:
351 		ops = &msg_ops;
352 		state = SS_READY;
353 		break;
354 	default:
355 		return -EPROTOTYPE;
356 	}
357 
358 	/* Allocate socket's protocol area */
359 	sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
360 	if (sk == NULL)
361 		return -ENOMEM;
362 
363 	tsk = tipc_sk(sk);
364 	tsk->max_pkt = MAX_PKT_DEFAULT;
365 	INIT_LIST_HEAD(&tsk->publications);
366 	msg = &tsk->phdr;
367 	tn = net_generic(sock_net(sk), tipc_net_id);
368 	tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
369 		      NAMED_H_SIZE, 0);
370 
371 	/* Finish initializing socket data structures */
372 	sock->ops = ops;
373 	sock->state = state;
374 	sock_init_data(sock, sk);
375 	if (tipc_sk_insert(tsk)) {
376 		pr_warn("Socket create failed; port numbrer exhausted\n");
377 		return -EINVAL;
378 	}
379 	msg_set_origport(msg, tsk->portid);
380 	setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
381 	sk->sk_backlog_rcv = tipc_backlog_rcv;
382 	sk->sk_rcvbuf = sysctl_tipc_rmem[1];
383 	sk->sk_data_ready = tipc_data_ready;
384 	sk->sk_write_space = tipc_write_space;
385 	sk->sk_destruct = tipc_sock_destruct;
386 	tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
387 	tsk->sent_unacked = 0;
388 	atomic_set(&tsk->dupl_rcvcnt, 0);
389 
390 	if (sock->state == SS_READY) {
391 		tsk_set_unreturnable(tsk, true);
392 		if (sock->type == SOCK_DGRAM)
393 			tsk_set_unreliable(tsk, true);
394 	}
395 	return 0;
396 }
397 
tipc_sk_callback(struct rcu_head * head)398 static void tipc_sk_callback(struct rcu_head *head)
399 {
400 	struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
401 
402 	sock_put(&tsk->sk);
403 }
404 
405 /**
406  * tipc_release - destroy a TIPC socket
407  * @sock: socket to destroy
408  *
409  * This routine cleans up any messages that are still queued on the socket.
410  * For DGRAM and RDM socket types, all queued messages are rejected.
411  * For SEQPACKET and STREAM socket types, the first message is rejected
412  * and any others are discarded.  (If the first message on a STREAM socket
413  * is partially-read, it is discarded and the next one is rejected instead.)
414  *
415  * NOTE: Rejected messages are not necessarily returned to the sender!  They
416  * are returned or discarded according to the "destination droppable" setting
417  * specified for the message by the sender.
418  *
419  * Returns 0 on success, errno otherwise
420  */
tipc_release(struct socket * sock)421 static int tipc_release(struct socket *sock)
422 {
423 	struct sock *sk = sock->sk;
424 	struct net *net;
425 	struct tipc_sock *tsk;
426 	struct sk_buff *skb;
427 	u32 dnode;
428 
429 	/*
430 	 * Exit if socket isn't fully initialized (occurs when a failed accept()
431 	 * releases a pre-allocated child socket that was never used)
432 	 */
433 	if (sk == NULL)
434 		return 0;
435 
436 	net = sock_net(sk);
437 	tsk = tipc_sk(sk);
438 	lock_sock(sk);
439 
440 	/*
441 	 * Reject all unreceived messages, except on an active connection
442 	 * (which disconnects locally & sends a 'FIN+' to peer)
443 	 */
444 	dnode = tsk_peer_node(tsk);
445 	while (sock->state != SS_DISCONNECTING) {
446 		skb = __skb_dequeue(&sk->sk_receive_queue);
447 		if (skb == NULL)
448 			break;
449 		if (TIPC_SKB_CB(skb)->handle != NULL)
450 			kfree_skb(skb);
451 		else {
452 			if ((sock->state == SS_CONNECTING) ||
453 			    (sock->state == SS_CONNECTED)) {
454 				sock->state = SS_DISCONNECTING;
455 				tsk->connected = 0;
456 				tipc_node_remove_conn(net, dnode, tsk->portid);
457 			}
458 			tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
459 		}
460 	}
461 
462 	tipc_sk_withdraw(tsk, 0, NULL);
463 	sk_stop_timer(sk, &sk->sk_timer);
464 	tipc_sk_remove(tsk);
465 	if (tsk->connected) {
466 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
467 				      TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
468 				      tsk_own_node(tsk), tsk_peer_port(tsk),
469 				      tsk->portid, TIPC_ERR_NO_PORT);
470 		if (skb)
471 			tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
472 		tipc_node_remove_conn(net, dnode, tsk->portid);
473 	}
474 
475 	/* Reject any messages that accumulated in backlog queue */
476 	sock->state = SS_DISCONNECTING;
477 	release_sock(sk);
478 
479 	call_rcu(&tsk->rcu, tipc_sk_callback);
480 	sock->sk = NULL;
481 
482 	return 0;
483 }
484 
485 /**
486  * tipc_bind - associate or disassocate TIPC name(s) with a socket
487  * @sock: socket structure
488  * @uaddr: socket address describing name(s) and desired operation
489  * @uaddr_len: size of socket address data structure
490  *
491  * Name and name sequence binding is indicated using a positive scope value;
492  * a negative scope value unbinds the specified name.  Specifying no name
493  * (i.e. a socket address length of 0) unbinds all names from the socket.
494  *
495  * Returns 0 on success, errno otherwise
496  *
497  * NOTE: This routine doesn't need to take the socket lock since it doesn't
498  *       access any non-constant socket information.
499  */
tipc_bind(struct socket * sock,struct sockaddr * uaddr,int uaddr_len)500 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
501 		     int uaddr_len)
502 {
503 	struct sock *sk = sock->sk;
504 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
505 	struct tipc_sock *tsk = tipc_sk(sk);
506 	int res = -EINVAL;
507 
508 	lock_sock(sk);
509 	if (unlikely(!uaddr_len)) {
510 		res = tipc_sk_withdraw(tsk, 0, NULL);
511 		goto exit;
512 	}
513 
514 	if (uaddr_len < sizeof(struct sockaddr_tipc)) {
515 		res = -EINVAL;
516 		goto exit;
517 	}
518 	if (addr->family != AF_TIPC) {
519 		res = -EAFNOSUPPORT;
520 		goto exit;
521 	}
522 
523 	if (addr->addrtype == TIPC_ADDR_NAME)
524 		addr->addr.nameseq.upper = addr->addr.nameseq.lower;
525 	else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
526 		res = -EAFNOSUPPORT;
527 		goto exit;
528 	}
529 
530 	if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
531 	    (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
532 	    (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
533 		res = -EACCES;
534 		goto exit;
535 	}
536 
537 	res = (addr->scope > 0) ?
538 		tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
539 		tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
540 exit:
541 	release_sock(sk);
542 	return res;
543 }
544 
545 /**
546  * tipc_getname - get port ID of socket or peer socket
547  * @sock: socket structure
548  * @uaddr: area for returned socket address
549  * @uaddr_len: area for returned length of socket address
550  * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
551  *
552  * Returns 0 on success, errno otherwise
553  *
554  * NOTE: This routine doesn't need to take the socket lock since it only
555  *       accesses socket information that is unchanging (or which changes in
556  *       a completely predictable manner).
557  */
tipc_getname(struct socket * sock,struct sockaddr * uaddr,int * uaddr_len,int peer)558 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
559 			int *uaddr_len, int peer)
560 {
561 	struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
562 	struct tipc_sock *tsk = tipc_sk(sock->sk);
563 	struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
564 
565 	memset(addr, 0, sizeof(*addr));
566 	if (peer) {
567 		if ((sock->state != SS_CONNECTED) &&
568 			((peer != 2) || (sock->state != SS_DISCONNECTING)))
569 			return -ENOTCONN;
570 		addr->addr.id.ref = tsk_peer_port(tsk);
571 		addr->addr.id.node = tsk_peer_node(tsk);
572 	} else {
573 		addr->addr.id.ref = tsk->portid;
574 		addr->addr.id.node = tn->own_addr;
575 	}
576 
577 	*uaddr_len = sizeof(*addr);
578 	addr->addrtype = TIPC_ADDR_ID;
579 	addr->family = AF_TIPC;
580 	addr->scope = 0;
581 	addr->addr.name.domain = 0;
582 
583 	return 0;
584 }
585 
586 /**
587  * tipc_poll - read and possibly block on pollmask
588  * @file: file structure associated with the socket
589  * @sock: socket for which to calculate the poll bits
590  * @wait: ???
591  *
592  * Returns pollmask value
593  *
594  * COMMENTARY:
595  * It appears that the usual socket locking mechanisms are not useful here
596  * since the pollmask info is potentially out-of-date the moment this routine
597  * exits.  TCP and other protocols seem to rely on higher level poll routines
598  * to handle any preventable race conditions, so TIPC will do the same ...
599  *
600  * TIPC sets the returned events as follows:
601  *
602  * socket state		flags set
603  * ------------		---------
604  * unconnected		no read flags
605  *			POLLOUT if port is not congested
606  *
607  * connecting		POLLIN/POLLRDNORM if ACK/NACK in rx queue
608  *			no write flags
609  *
610  * connected		POLLIN/POLLRDNORM if data in rx queue
611  *			POLLOUT if port is not congested
612  *
613  * disconnecting	POLLIN/POLLRDNORM/POLLHUP
614  *			no write flags
615  *
616  * listening		POLLIN if SYN in rx queue
617  *			no write flags
618  *
619  * ready		POLLIN/POLLRDNORM if data in rx queue
620  * [connectionless]	POLLOUT (since port cannot be congested)
621  *
622  * IMPORTANT: The fact that a read or write operation is indicated does NOT
623  * imply that the operation will succeed, merely that it should be performed
624  * and will not block.
625  */
tipc_poll(struct file * file,struct socket * sock,poll_table * wait)626 static unsigned int tipc_poll(struct file *file, struct socket *sock,
627 			      poll_table *wait)
628 {
629 	struct sock *sk = sock->sk;
630 	struct tipc_sock *tsk = tipc_sk(sk);
631 	u32 mask = 0;
632 
633 	sock_poll_wait(file, sk_sleep(sk), wait);
634 
635 	switch ((int)sock->state) {
636 	case SS_UNCONNECTED:
637 		if (!tsk->link_cong)
638 			mask |= POLLOUT;
639 		break;
640 	case SS_READY:
641 	case SS_CONNECTED:
642 		if (!tsk->link_cong && !tsk_conn_cong(tsk))
643 			mask |= POLLOUT;
644 		/* fall thru' */
645 	case SS_CONNECTING:
646 	case SS_LISTENING:
647 		if (!skb_queue_empty(&sk->sk_receive_queue))
648 			mask |= (POLLIN | POLLRDNORM);
649 		break;
650 	case SS_DISCONNECTING:
651 		mask = (POLLIN | POLLRDNORM | POLLHUP);
652 		break;
653 	}
654 
655 	return mask;
656 }
657 
658 /**
659  * tipc_sendmcast - send multicast message
660  * @sock: socket structure
661  * @seq: destination address
662  * @msg: message to send
663  * @dsz: total length of message data
664  * @timeo: timeout to wait for wakeup
665  *
666  * Called from function tipc_sendmsg(), which has done all sanity checks
667  * Returns the number of bytes sent on success, or errno
668  */
tipc_sendmcast(struct socket * sock,struct tipc_name_seq * seq,struct msghdr * msg,size_t dsz,long timeo)669 static int tipc_sendmcast(struct  socket *sock, struct tipc_name_seq *seq,
670 			  struct msghdr *msg, size_t dsz, long timeo)
671 {
672 	struct sock *sk = sock->sk;
673 	struct tipc_sock *tsk = tipc_sk(sk);
674 	struct net *net = sock_net(sk);
675 	struct tipc_msg *mhdr = &tsk->phdr;
676 	struct sk_buff_head pktchain;
677 	struct iov_iter save = msg->msg_iter;
678 	uint mtu;
679 	int rc;
680 
681 	msg_set_type(mhdr, TIPC_MCAST_MSG);
682 	msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
683 	msg_set_destport(mhdr, 0);
684 	msg_set_destnode(mhdr, 0);
685 	msg_set_nametype(mhdr, seq->type);
686 	msg_set_namelower(mhdr, seq->lower);
687 	msg_set_nameupper(mhdr, seq->upper);
688 	msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
689 
690 	skb_queue_head_init(&pktchain);
691 
692 new_mtu:
693 	mtu = tipc_bcast_get_mtu(net);
694 	rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
695 	if (unlikely(rc < 0))
696 		return rc;
697 
698 	do {
699 		rc = tipc_bcast_xmit(net, &pktchain);
700 		if (likely(!rc))
701 			return dsz;
702 
703 		if (rc == -ELINKCONG) {
704 			tsk->link_cong = 1;
705 			rc = tipc_wait_for_sndmsg(sock, &timeo);
706 			if (!rc)
707 				continue;
708 		}
709 		__skb_queue_purge(&pktchain);
710 		if (rc == -EMSGSIZE) {
711 			msg->msg_iter = save;
712 			goto new_mtu;
713 		}
714 		break;
715 	} while (1);
716 	return rc;
717 }
718 
719 /**
720  * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
721  * @arrvq: queue with arriving messages, to be cloned after destination lookup
722  * @inputq: queue with cloned messages, delivered to socket after dest lookup
723  *
724  * Multi-threaded: parallel calls with reference to same queues may occur
725  */
tipc_sk_mcast_rcv(struct net * net,struct sk_buff_head * arrvq,struct sk_buff_head * inputq)726 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
727 		       struct sk_buff_head *inputq)
728 {
729 	struct tipc_msg *msg;
730 	struct tipc_plist dports;
731 	u32 portid;
732 	u32 scope = TIPC_CLUSTER_SCOPE;
733 	struct sk_buff_head tmpq;
734 	uint hsz;
735 	struct sk_buff *skb, *_skb;
736 
737 	__skb_queue_head_init(&tmpq);
738 	tipc_plist_init(&dports);
739 
740 	skb = tipc_skb_peek(arrvq, &inputq->lock);
741 	for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
742 		msg = buf_msg(skb);
743 		hsz = skb_headroom(skb) + msg_hdr_sz(msg);
744 
745 		if (in_own_node(net, msg_orignode(msg)))
746 			scope = TIPC_NODE_SCOPE;
747 
748 		/* Create destination port list and message clones: */
749 		tipc_nametbl_mc_translate(net,
750 					  msg_nametype(msg), msg_namelower(msg),
751 					  msg_nameupper(msg), scope, &dports);
752 		portid = tipc_plist_pop(&dports);
753 		for (; portid; portid = tipc_plist_pop(&dports)) {
754 			_skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
755 			if (_skb) {
756 				msg_set_destport(buf_msg(_skb), portid);
757 				__skb_queue_tail(&tmpq, _skb);
758 				continue;
759 			}
760 			pr_warn("Failed to clone mcast rcv buffer\n");
761 		}
762 		/* Append to inputq if not already done by other thread */
763 		spin_lock_bh(&inputq->lock);
764 		if (skb_peek(arrvq) == skb) {
765 			skb_queue_splice_tail_init(&tmpq, inputq);
766 			/* Decrease the skb's refcnt as increasing in the
767 			 * function tipc_skb_peek
768 			 */
769 			kfree_skb(__skb_dequeue(arrvq));
770 		}
771 		spin_unlock_bh(&inputq->lock);
772 		__skb_queue_purge(&tmpq);
773 		kfree_skb(skb);
774 	}
775 	tipc_sk_rcv(net, inputq);
776 }
777 
778 /**
779  * tipc_sk_proto_rcv - receive a connection mng protocol message
780  * @tsk: receiving socket
781  * @skb: pointer to message buffer.
782  */
tipc_sk_proto_rcv(struct tipc_sock * tsk,struct sk_buff * skb,struct sk_buff_head * xmitq)783 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
784 			      struct sk_buff_head *xmitq)
785 {
786 	struct sock *sk = &tsk->sk;
787 	u32 onode = tsk_own_node(tsk);
788 	struct tipc_msg *hdr = buf_msg(skb);
789 	int mtyp = msg_type(hdr);
790 	int conn_cong;
791 
792 	/* Ignore if connection cannot be validated: */
793 	if (!tsk_peer_msg(tsk, hdr))
794 		goto exit;
795 
796 	tsk->probing_state = TIPC_CONN_OK;
797 
798 	if (mtyp == CONN_PROBE) {
799 		msg_set_type(hdr, CONN_PROBE_REPLY);
800 		if (tipc_msg_reverse(onode, &skb, TIPC_OK))
801 			__skb_queue_tail(xmitq, skb);
802 		return;
803 	} else if (mtyp == CONN_ACK) {
804 		conn_cong = tsk_conn_cong(tsk);
805 		tsk->sent_unacked -= msg_msgcnt(hdr);
806 		if (conn_cong)
807 			sk->sk_write_space(sk);
808 	} else if (mtyp != CONN_PROBE_REPLY) {
809 		pr_warn("Received unknown CONN_PROTO msg\n");
810 	}
811 exit:
812 	kfree_skb(skb);
813 }
814 
tipc_wait_for_sndmsg(struct socket * sock,long * timeo_p)815 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
816 {
817 	struct sock *sk = sock->sk;
818 	struct tipc_sock *tsk = tipc_sk(sk);
819 	DEFINE_WAIT(wait);
820 	int done;
821 
822 	do {
823 		int err = sock_error(sk);
824 		if (err)
825 			return err;
826 		if (sock->state == SS_DISCONNECTING)
827 			return -EPIPE;
828 		if (!*timeo_p)
829 			return -EAGAIN;
830 		if (signal_pending(current))
831 			return sock_intr_errno(*timeo_p);
832 
833 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
834 		done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
835 		finish_wait(sk_sleep(sk), &wait);
836 	} while (!done);
837 	return 0;
838 }
839 
840 /**
841  * tipc_sendmsg - send message in connectionless manner
842  * @sock: socket structure
843  * @m: message to send
844  * @dsz: amount of user data to be sent
845  *
846  * Message must have an destination specified explicitly.
847  * Used for SOCK_RDM and SOCK_DGRAM messages,
848  * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
849  * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
850  *
851  * Returns the number of bytes sent on success, or errno otherwise
852  */
tipc_sendmsg(struct socket * sock,struct msghdr * m,size_t dsz)853 static int tipc_sendmsg(struct socket *sock,
854 			struct msghdr *m, size_t dsz)
855 {
856 	struct sock *sk = sock->sk;
857 	int ret;
858 
859 	lock_sock(sk);
860 	ret = __tipc_sendmsg(sock, m, dsz);
861 	release_sock(sk);
862 
863 	return ret;
864 }
865 
__tipc_sendmsg(struct socket * sock,struct msghdr * m,size_t dsz)866 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
867 {
868 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
869 	struct sock *sk = sock->sk;
870 	struct tipc_sock *tsk = tipc_sk(sk);
871 	struct net *net = sock_net(sk);
872 	struct tipc_msg *mhdr = &tsk->phdr;
873 	u32 dnode, dport;
874 	struct sk_buff_head pktchain;
875 	struct sk_buff *skb;
876 	struct tipc_name_seq *seq;
877 	struct iov_iter save;
878 	u32 mtu;
879 	long timeo;
880 	int rc;
881 
882 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
883 		return -EMSGSIZE;
884 	if (unlikely(!dest)) {
885 		if (tsk->connected && sock->state == SS_READY)
886 			dest = &tsk->remote;
887 		else
888 			return -EDESTADDRREQ;
889 	} else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
890 		   dest->family != AF_TIPC) {
891 		return -EINVAL;
892 	}
893 	if (unlikely(sock->state != SS_READY)) {
894 		if (sock->state == SS_LISTENING)
895 			return -EPIPE;
896 		if (sock->state != SS_UNCONNECTED)
897 			return -EISCONN;
898 		if (tsk->published)
899 			return -EOPNOTSUPP;
900 		if (dest->addrtype == TIPC_ADDR_NAME) {
901 			tsk->conn_type = dest->addr.name.name.type;
902 			tsk->conn_instance = dest->addr.name.name.instance;
903 		}
904 	}
905 	seq = &dest->addr.nameseq;
906 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
907 
908 	if (dest->addrtype == TIPC_ADDR_MCAST) {
909 		return tipc_sendmcast(sock, seq, m, dsz, timeo);
910 	} else if (dest->addrtype == TIPC_ADDR_NAME) {
911 		u32 type = dest->addr.name.name.type;
912 		u32 inst = dest->addr.name.name.instance;
913 		u32 domain = dest->addr.name.domain;
914 
915 		dnode = domain;
916 		msg_set_type(mhdr, TIPC_NAMED_MSG);
917 		msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
918 		msg_set_nametype(mhdr, type);
919 		msg_set_nameinst(mhdr, inst);
920 		msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
921 		dport = tipc_nametbl_translate(net, type, inst, &dnode);
922 		msg_set_destnode(mhdr, dnode);
923 		msg_set_destport(mhdr, dport);
924 		if (unlikely(!dport && !dnode))
925 			return -EHOSTUNREACH;
926 	} else if (dest->addrtype == TIPC_ADDR_ID) {
927 		dnode = dest->addr.id.node;
928 		msg_set_type(mhdr, TIPC_DIRECT_MSG);
929 		msg_set_lookup_scope(mhdr, 0);
930 		msg_set_destnode(mhdr, dnode);
931 		msg_set_destport(mhdr, dest->addr.id.ref);
932 		msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
933 	}
934 
935 	skb_queue_head_init(&pktchain);
936 	save = m->msg_iter;
937 new_mtu:
938 	mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
939 	rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
940 	if (rc < 0)
941 		return rc;
942 
943 	do {
944 		skb = skb_peek(&pktchain);
945 		TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
946 		rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
947 		if (likely(!rc)) {
948 			if (sock->state != SS_READY)
949 				sock->state = SS_CONNECTING;
950 			return dsz;
951 		}
952 		if (rc == -ELINKCONG) {
953 			tsk->link_cong = 1;
954 			rc = tipc_wait_for_sndmsg(sock, &timeo);
955 			if (!rc)
956 				continue;
957 		}
958 		__skb_queue_purge(&pktchain);
959 		if (rc == -EMSGSIZE) {
960 			m->msg_iter = save;
961 			goto new_mtu;
962 		}
963 		break;
964 	} while (1);
965 
966 	return rc;
967 }
968 
tipc_wait_for_sndpkt(struct socket * sock,long * timeo_p)969 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
970 {
971 	struct sock *sk = sock->sk;
972 	struct tipc_sock *tsk = tipc_sk(sk);
973 	DEFINE_WAIT(wait);
974 	int done;
975 
976 	do {
977 		int err = sock_error(sk);
978 		if (err)
979 			return err;
980 		if (sock->state == SS_DISCONNECTING)
981 			return -EPIPE;
982 		else if (sock->state != SS_CONNECTED)
983 			return -ENOTCONN;
984 		if (!*timeo_p)
985 			return -EAGAIN;
986 		if (signal_pending(current))
987 			return sock_intr_errno(*timeo_p);
988 
989 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
990 		done = sk_wait_event(sk, timeo_p,
991 				     (!tsk->link_cong &&
992 				      !tsk_conn_cong(tsk)) ||
993 				     !tsk->connected);
994 		finish_wait(sk_sleep(sk), &wait);
995 	} while (!done);
996 	return 0;
997 }
998 
999 /**
1000  * tipc_send_stream - send stream-oriented data
1001  * @sock: socket structure
1002  * @m: data to send
1003  * @dsz: total length of data to be transmitted
1004  *
1005  * Used for SOCK_STREAM data.
1006  *
1007  * Returns the number of bytes sent on success (or partial success),
1008  * or errno if no data sent
1009  */
tipc_send_stream(struct socket * sock,struct msghdr * m,size_t dsz)1010 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1011 {
1012 	struct sock *sk = sock->sk;
1013 	int ret;
1014 
1015 	lock_sock(sk);
1016 	ret = __tipc_send_stream(sock, m, dsz);
1017 	release_sock(sk);
1018 
1019 	return ret;
1020 }
1021 
__tipc_send_stream(struct socket * sock,struct msghdr * m,size_t dsz)1022 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1023 {
1024 	struct sock *sk = sock->sk;
1025 	struct net *net = sock_net(sk);
1026 	struct tipc_sock *tsk = tipc_sk(sk);
1027 	struct tipc_msg *mhdr = &tsk->phdr;
1028 	struct sk_buff_head pktchain;
1029 	DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1030 	u32 portid = tsk->portid;
1031 	int rc = -EINVAL;
1032 	long timeo;
1033 	u32 dnode;
1034 	uint mtu, send, sent = 0;
1035 	struct iov_iter save;
1036 
1037 	/* Handle implied connection establishment */
1038 	if (unlikely(dest)) {
1039 		rc = __tipc_sendmsg(sock, m, dsz);
1040 		if (dsz && (dsz == rc))
1041 			tsk->sent_unacked = 1;
1042 		return rc;
1043 	}
1044 	if (dsz > (uint)INT_MAX)
1045 		return -EMSGSIZE;
1046 
1047 	if (unlikely(sock->state != SS_CONNECTED)) {
1048 		if (sock->state == SS_DISCONNECTING)
1049 			return -EPIPE;
1050 		else
1051 			return -ENOTCONN;
1052 	}
1053 
1054 	timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1055 	dnode = tsk_peer_node(tsk);
1056 	skb_queue_head_init(&pktchain);
1057 
1058 next:
1059 	save = m->msg_iter;
1060 	mtu = tsk->max_pkt;
1061 	send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1062 	rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1063 	if (unlikely(rc < 0))
1064 		return rc;
1065 
1066 	do {
1067 		if (likely(!tsk_conn_cong(tsk))) {
1068 			rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1069 			if (likely(!rc)) {
1070 				tsk->sent_unacked++;
1071 				sent += send;
1072 				if (sent == dsz)
1073 					return dsz;
1074 				goto next;
1075 			}
1076 			if (rc == -EMSGSIZE) {
1077 				__skb_queue_purge(&pktchain);
1078 				tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1079 								 portid);
1080 				m->msg_iter = save;
1081 				goto next;
1082 			}
1083 			if (rc != -ELINKCONG)
1084 				break;
1085 
1086 			tsk->link_cong = 1;
1087 		}
1088 		rc = tipc_wait_for_sndpkt(sock, &timeo);
1089 	} while (!rc);
1090 
1091 	__skb_queue_purge(&pktchain);
1092 	return sent ? sent : rc;
1093 }
1094 
1095 /**
1096  * tipc_send_packet - send a connection-oriented message
1097  * @sock: socket structure
1098  * @m: message to send
1099  * @dsz: length of data to be transmitted
1100  *
1101  * Used for SOCK_SEQPACKET messages.
1102  *
1103  * Returns the number of bytes sent on success, or errno otherwise
1104  */
tipc_send_packet(struct socket * sock,struct msghdr * m,size_t dsz)1105 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1106 {
1107 	if (dsz > TIPC_MAX_USER_MSG_SIZE)
1108 		return -EMSGSIZE;
1109 
1110 	return tipc_send_stream(sock, m, dsz);
1111 }
1112 
1113 /* tipc_sk_finish_conn - complete the setup of a connection
1114  */
tipc_sk_finish_conn(struct tipc_sock * tsk,u32 peer_port,u32 peer_node)1115 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1116 				u32 peer_node)
1117 {
1118 	struct sock *sk = &tsk->sk;
1119 	struct net *net = sock_net(sk);
1120 	struct tipc_msg *msg = &tsk->phdr;
1121 
1122 	msg_set_destnode(msg, peer_node);
1123 	msg_set_destport(msg, peer_port);
1124 	msg_set_type(msg, TIPC_CONN_MSG);
1125 	msg_set_lookup_scope(msg, 0);
1126 	msg_set_hdr_sz(msg, SHORT_H_SIZE);
1127 
1128 	tsk->probing_intv = CONN_PROBING_INTERVAL;
1129 	tsk->probing_state = TIPC_CONN_OK;
1130 	tsk->connected = 1;
1131 	sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1132 	tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1133 	tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1134 }
1135 
1136 /**
1137  * set_orig_addr - capture sender's address for received message
1138  * @m: descriptor for message info
1139  * @msg: received message header
1140  *
1141  * Note: Address is not captured if not requested by receiver.
1142  */
set_orig_addr(struct msghdr * m,struct tipc_msg * msg)1143 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1144 {
1145 	DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1146 
1147 	if (addr) {
1148 		addr->family = AF_TIPC;
1149 		addr->addrtype = TIPC_ADDR_ID;
1150 		memset(&addr->addr, 0, sizeof(addr->addr));
1151 		addr->addr.id.ref = msg_origport(msg);
1152 		addr->addr.id.node = msg_orignode(msg);
1153 		addr->addr.name.domain = 0;	/* could leave uninitialized */
1154 		addr->scope = 0;		/* could leave uninitialized */
1155 		m->msg_namelen = sizeof(struct sockaddr_tipc);
1156 	}
1157 }
1158 
1159 /**
1160  * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1161  * @m: descriptor for message info
1162  * @msg: received message header
1163  * @tsk: TIPC port associated with message
1164  *
1165  * Note: Ancillary data is not captured if not requested by receiver.
1166  *
1167  * Returns 0 if successful, otherwise errno
1168  */
tipc_sk_anc_data_recv(struct msghdr * m,struct tipc_msg * msg,struct tipc_sock * tsk)1169 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1170 				 struct tipc_sock *tsk)
1171 {
1172 	u32 anc_data[3];
1173 	u32 err;
1174 	u32 dest_type;
1175 	int has_name;
1176 	int res;
1177 
1178 	if (likely(m->msg_controllen == 0))
1179 		return 0;
1180 
1181 	/* Optionally capture errored message object(s) */
1182 	err = msg ? msg_errcode(msg) : 0;
1183 	if (unlikely(err)) {
1184 		anc_data[0] = err;
1185 		anc_data[1] = msg_data_sz(msg);
1186 		res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1187 		if (res)
1188 			return res;
1189 		if (anc_data[1]) {
1190 			res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1191 				       msg_data(msg));
1192 			if (res)
1193 				return res;
1194 		}
1195 	}
1196 
1197 	/* Optionally capture message destination object */
1198 	dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1199 	switch (dest_type) {
1200 	case TIPC_NAMED_MSG:
1201 		has_name = 1;
1202 		anc_data[0] = msg_nametype(msg);
1203 		anc_data[1] = msg_namelower(msg);
1204 		anc_data[2] = msg_namelower(msg);
1205 		break;
1206 	case TIPC_MCAST_MSG:
1207 		has_name = 1;
1208 		anc_data[0] = msg_nametype(msg);
1209 		anc_data[1] = msg_namelower(msg);
1210 		anc_data[2] = msg_nameupper(msg);
1211 		break;
1212 	case TIPC_CONN_MSG:
1213 		has_name = (tsk->conn_type != 0);
1214 		anc_data[0] = tsk->conn_type;
1215 		anc_data[1] = tsk->conn_instance;
1216 		anc_data[2] = tsk->conn_instance;
1217 		break;
1218 	default:
1219 		has_name = 0;
1220 	}
1221 	if (has_name) {
1222 		res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1223 		if (res)
1224 			return res;
1225 	}
1226 
1227 	return 0;
1228 }
1229 
tipc_sk_send_ack(struct tipc_sock * tsk,uint ack)1230 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1231 {
1232 	struct net *net = sock_net(&tsk->sk);
1233 	struct sk_buff *skb = NULL;
1234 	struct tipc_msg *msg;
1235 	u32 peer_port = tsk_peer_port(tsk);
1236 	u32 dnode = tsk_peer_node(tsk);
1237 
1238 	if (!tsk->connected)
1239 		return;
1240 	skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1241 			      dnode, tsk_own_node(tsk), peer_port,
1242 			      tsk->portid, TIPC_OK);
1243 	if (!skb)
1244 		return;
1245 	msg = buf_msg(skb);
1246 	msg_set_msgcnt(msg, ack);
1247 	tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1248 }
1249 
tipc_wait_for_rcvmsg(struct socket * sock,long * timeop)1250 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1251 {
1252 	struct sock *sk = sock->sk;
1253 	DEFINE_WAIT(wait);
1254 	long timeo = *timeop;
1255 	int err;
1256 
1257 	for (;;) {
1258 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1259 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1260 			if (sock->state == SS_DISCONNECTING) {
1261 				err = -ENOTCONN;
1262 				break;
1263 			}
1264 			release_sock(sk);
1265 			timeo = schedule_timeout(timeo);
1266 			lock_sock(sk);
1267 		}
1268 		err = 0;
1269 		if (!skb_queue_empty(&sk->sk_receive_queue))
1270 			break;
1271 		err = -EAGAIN;
1272 		if (!timeo)
1273 			break;
1274 		err = sock_intr_errno(timeo);
1275 		if (signal_pending(current))
1276 			break;
1277 	}
1278 	finish_wait(sk_sleep(sk), &wait);
1279 	*timeop = timeo;
1280 	return err;
1281 }
1282 
1283 /**
1284  * tipc_recvmsg - receive packet-oriented message
1285  * @m: descriptor for message info
1286  * @buf_len: total size of user buffer area
1287  * @flags: receive flags
1288  *
1289  * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1290  * If the complete message doesn't fit in user area, truncate it.
1291  *
1292  * Returns size of returned message data, errno otherwise
1293  */
tipc_recvmsg(struct socket * sock,struct msghdr * m,size_t buf_len,int flags)1294 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1295 			int flags)
1296 {
1297 	struct sock *sk = sock->sk;
1298 	struct tipc_sock *tsk = tipc_sk(sk);
1299 	struct sk_buff *buf;
1300 	struct tipc_msg *msg;
1301 	long timeo;
1302 	unsigned int sz;
1303 	u32 err;
1304 	int res;
1305 
1306 	/* Catch invalid receive requests */
1307 	if (unlikely(!buf_len))
1308 		return -EINVAL;
1309 
1310 	lock_sock(sk);
1311 
1312 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1313 		res = -ENOTCONN;
1314 		goto exit;
1315 	}
1316 
1317 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1318 restart:
1319 
1320 	/* Look for a message in receive queue; wait if necessary */
1321 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1322 	if (res)
1323 		goto exit;
1324 
1325 	/* Look at first message in receive queue */
1326 	buf = skb_peek(&sk->sk_receive_queue);
1327 	msg = buf_msg(buf);
1328 	sz = msg_data_sz(msg);
1329 	err = msg_errcode(msg);
1330 
1331 	/* Discard an empty non-errored message & try again */
1332 	if ((!sz) && (!err)) {
1333 		tsk_advance_rx_queue(sk);
1334 		goto restart;
1335 	}
1336 
1337 	/* Capture sender's address (optional) */
1338 	set_orig_addr(m, msg);
1339 
1340 	/* Capture ancillary data (optional) */
1341 	res = tipc_sk_anc_data_recv(m, msg, tsk);
1342 	if (res)
1343 		goto exit;
1344 
1345 	/* Capture message data (if valid) & compute return value (always) */
1346 	if (!err) {
1347 		if (unlikely(buf_len < sz)) {
1348 			sz = buf_len;
1349 			m->msg_flags |= MSG_TRUNC;
1350 		}
1351 		res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1352 		if (res)
1353 			goto exit;
1354 		res = sz;
1355 	} else {
1356 		if ((sock->state == SS_READY) ||
1357 		    ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1358 			res = 0;
1359 		else
1360 			res = -ECONNRESET;
1361 	}
1362 
1363 	/* Consume received message (optional) */
1364 	if (likely(!(flags & MSG_PEEK))) {
1365 		if ((sock->state != SS_READY) &&
1366 		    (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1367 			tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1368 			tsk->rcv_unacked = 0;
1369 		}
1370 		tsk_advance_rx_queue(sk);
1371 	}
1372 exit:
1373 	release_sock(sk);
1374 	return res;
1375 }
1376 
1377 /**
1378  * tipc_recv_stream - receive stream-oriented data
1379  * @m: descriptor for message info
1380  * @buf_len: total size of user buffer area
1381  * @flags: receive flags
1382  *
1383  * Used for SOCK_STREAM messages only.  If not enough data is available
1384  * will optionally wait for more; never truncates data.
1385  *
1386  * Returns size of returned message data, errno otherwise
1387  */
tipc_recv_stream(struct socket * sock,struct msghdr * m,size_t buf_len,int flags)1388 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1389 			    size_t buf_len, int flags)
1390 {
1391 	struct sock *sk = sock->sk;
1392 	struct tipc_sock *tsk = tipc_sk(sk);
1393 	struct sk_buff *buf;
1394 	struct tipc_msg *msg;
1395 	long timeo;
1396 	unsigned int sz;
1397 	int sz_to_copy, target, needed;
1398 	int sz_copied = 0;
1399 	u32 err;
1400 	int res = 0;
1401 
1402 	/* Catch invalid receive attempts */
1403 	if (unlikely(!buf_len))
1404 		return -EINVAL;
1405 
1406 	lock_sock(sk);
1407 
1408 	if (unlikely(sock->state == SS_UNCONNECTED)) {
1409 		res = -ENOTCONN;
1410 		goto exit;
1411 	}
1412 
1413 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1414 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1415 
1416 restart:
1417 	/* Look for a message in receive queue; wait if necessary */
1418 	res = tipc_wait_for_rcvmsg(sock, &timeo);
1419 	if (res)
1420 		goto exit;
1421 
1422 	/* Look at first message in receive queue */
1423 	buf = skb_peek(&sk->sk_receive_queue);
1424 	msg = buf_msg(buf);
1425 	sz = msg_data_sz(msg);
1426 	err = msg_errcode(msg);
1427 
1428 	/* Discard an empty non-errored message & try again */
1429 	if ((!sz) && (!err)) {
1430 		tsk_advance_rx_queue(sk);
1431 		goto restart;
1432 	}
1433 
1434 	/* Optionally capture sender's address & ancillary data of first msg */
1435 	if (sz_copied == 0) {
1436 		set_orig_addr(m, msg);
1437 		res = tipc_sk_anc_data_recv(m, msg, tsk);
1438 		if (res)
1439 			goto exit;
1440 	}
1441 
1442 	/* Capture message data (if valid) & compute return value (always) */
1443 	if (!err) {
1444 		u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1445 
1446 		sz -= offset;
1447 		needed = (buf_len - sz_copied);
1448 		sz_to_copy = (sz <= needed) ? sz : needed;
1449 
1450 		res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1451 					    m, sz_to_copy);
1452 		if (res)
1453 			goto exit;
1454 
1455 		sz_copied += sz_to_copy;
1456 
1457 		if (sz_to_copy < sz) {
1458 			if (!(flags & MSG_PEEK))
1459 				TIPC_SKB_CB(buf)->handle =
1460 				(void *)(unsigned long)(offset + sz_to_copy);
1461 			goto exit;
1462 		}
1463 	} else {
1464 		if (sz_copied != 0)
1465 			goto exit; /* can't add error msg to valid data */
1466 
1467 		if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1468 			res = 0;
1469 		else
1470 			res = -ECONNRESET;
1471 	}
1472 
1473 	/* Consume received message (optional) */
1474 	if (likely(!(flags & MSG_PEEK))) {
1475 		if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1476 			tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1477 			tsk->rcv_unacked = 0;
1478 		}
1479 		tsk_advance_rx_queue(sk);
1480 	}
1481 
1482 	/* Loop around if more data is required */
1483 	if ((sz_copied < buf_len) &&	/* didn't get all requested data */
1484 	    (!skb_queue_empty(&sk->sk_receive_queue) ||
1485 	    (sz_copied < target)) &&	/* and more is ready or required */
1486 	    (!(flags & MSG_PEEK)) &&	/* and aren't just peeking at data */
1487 	    (!err))			/* and haven't reached a FIN */
1488 		goto restart;
1489 
1490 exit:
1491 	release_sock(sk);
1492 	return sz_copied ? sz_copied : res;
1493 }
1494 
1495 /**
1496  * tipc_write_space - wake up thread if port congestion is released
1497  * @sk: socket
1498  */
tipc_write_space(struct sock * sk)1499 static void tipc_write_space(struct sock *sk)
1500 {
1501 	struct socket_wq *wq;
1502 
1503 	rcu_read_lock();
1504 	wq = rcu_dereference(sk->sk_wq);
1505 	if (wq_has_sleeper(wq))
1506 		wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1507 						POLLWRNORM | POLLWRBAND);
1508 	rcu_read_unlock();
1509 }
1510 
1511 /**
1512  * tipc_data_ready - wake up threads to indicate messages have been received
1513  * @sk: socket
1514  * @len: the length of messages
1515  */
tipc_data_ready(struct sock * sk)1516 static void tipc_data_ready(struct sock *sk)
1517 {
1518 	struct socket_wq *wq;
1519 
1520 	rcu_read_lock();
1521 	wq = rcu_dereference(sk->sk_wq);
1522 	if (wq_has_sleeper(wq))
1523 		wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1524 						POLLRDNORM | POLLRDBAND);
1525 	rcu_read_unlock();
1526 }
1527 
tipc_sock_destruct(struct sock * sk)1528 static void tipc_sock_destruct(struct sock *sk)
1529 {
1530 	__skb_queue_purge(&sk->sk_receive_queue);
1531 }
1532 
1533 /**
1534  * filter_connect - Handle all incoming messages for a connection-based socket
1535  * @tsk: TIPC socket
1536  * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1537  *
1538  * Returns true if everything ok, false otherwise
1539  */
filter_connect(struct tipc_sock * tsk,struct sk_buff * skb)1540 static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1541 {
1542 	struct sock *sk = &tsk->sk;
1543 	struct net *net = sock_net(sk);
1544 	struct socket *sock = sk->sk_socket;
1545 	struct tipc_msg *hdr = buf_msg(skb);
1546 
1547 	if (unlikely(msg_mcast(hdr)))
1548 		return false;
1549 
1550 	switch ((int)sock->state) {
1551 	case SS_CONNECTED:
1552 
1553 		/* Accept only connection-based messages sent by peer */
1554 		if (unlikely(!tsk_peer_msg(tsk, hdr)))
1555 			return false;
1556 
1557 		if (unlikely(msg_errcode(hdr))) {
1558 			sock->state = SS_DISCONNECTING;
1559 			tsk->connected = 0;
1560 			/* Let timer expire on it's own */
1561 			tipc_node_remove_conn(net, tsk_peer_node(tsk),
1562 					      tsk->portid);
1563 		}
1564 		return true;
1565 
1566 	case SS_CONNECTING:
1567 
1568 		/* Accept only ACK or NACK message */
1569 		if (unlikely(!msg_connected(hdr)))
1570 			return false;
1571 
1572 		if (unlikely(msg_errcode(hdr))) {
1573 			sock->state = SS_DISCONNECTING;
1574 			sk->sk_err = ECONNREFUSED;
1575 			return true;
1576 		}
1577 
1578 		if (unlikely(!msg_isdata(hdr))) {
1579 			sock->state = SS_DISCONNECTING;
1580 			sk->sk_err = EINVAL;
1581 			return true;
1582 		}
1583 
1584 		tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1585 		msg_set_importance(&tsk->phdr, msg_importance(hdr));
1586 		sock->state = SS_CONNECTED;
1587 
1588 		/* If 'ACK+' message, add to socket receive queue */
1589 		if (msg_data_sz(hdr))
1590 			return true;
1591 
1592 		/* If empty 'ACK-' message, wake up sleeping connect() */
1593 		if (waitqueue_active(sk_sleep(sk)))
1594 			wake_up_interruptible(sk_sleep(sk));
1595 
1596 		/* 'ACK-' message is neither accepted nor rejected: */
1597 		msg_set_dest_droppable(hdr, 1);
1598 		return false;
1599 
1600 	case SS_LISTENING:
1601 	case SS_UNCONNECTED:
1602 
1603 		/* Accept only SYN message */
1604 		if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1605 			return true;
1606 		break;
1607 	case SS_DISCONNECTING:
1608 		break;
1609 	default:
1610 		pr_err("Unknown socket state %u\n", sock->state);
1611 	}
1612 	return false;
1613 }
1614 
1615 /**
1616  * rcvbuf_limit - get proper overload limit of socket receive queue
1617  * @sk: socket
1618  * @buf: message
1619  *
1620  * For all connection oriented messages, irrespective of importance,
1621  * the default overload value (i.e. 67MB) is set as limit.
1622  *
1623  * For all connectionless messages, by default new queue limits are
1624  * as belows:
1625  *
1626  * TIPC_LOW_IMPORTANCE       (4 MB)
1627  * TIPC_MEDIUM_IMPORTANCE    (8 MB)
1628  * TIPC_HIGH_IMPORTANCE      (16 MB)
1629  * TIPC_CRITICAL_IMPORTANCE  (32 MB)
1630  *
1631  * Returns overload limit according to corresponding message importance
1632  */
rcvbuf_limit(struct sock * sk,struct sk_buff * buf)1633 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1634 {
1635 	struct tipc_msg *msg = buf_msg(buf);
1636 
1637 	if (msg_connected(msg))
1638 		return sysctl_tipc_rmem[2];
1639 
1640 	return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1641 		msg_importance(msg);
1642 }
1643 
1644 /**
1645  * filter_rcv - validate incoming message
1646  * @sk: socket
1647  * @skb: pointer to message.
1648  *
1649  * Enqueues message on receive queue if acceptable; optionally handles
1650  * disconnect indication for a connected socket.
1651  *
1652  * Called with socket lock already taken
1653  *
1654  * Returns true if message was added to socket receive queue, otherwise false
1655  */
filter_rcv(struct sock * sk,struct sk_buff * skb,struct sk_buff_head * xmitq)1656 static bool filter_rcv(struct sock *sk, struct sk_buff *skb,
1657 		       struct sk_buff_head *xmitq)
1658 {
1659 	struct socket *sock = sk->sk_socket;
1660 	struct tipc_sock *tsk = tipc_sk(sk);
1661 	struct tipc_msg *hdr = buf_msg(skb);
1662 	unsigned int limit = rcvbuf_limit(sk, skb);
1663 	int err = TIPC_OK;
1664 	int usr = msg_user(hdr);
1665 
1666 	if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1667 		tipc_sk_proto_rcv(tsk, skb, xmitq);
1668 		return false;
1669 	}
1670 
1671 	if (unlikely(usr == SOCK_WAKEUP)) {
1672 		kfree_skb(skb);
1673 		tsk->link_cong = 0;
1674 		sk->sk_write_space(sk);
1675 		return false;
1676 	}
1677 
1678 	/* Drop if illegal message type */
1679 	if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1680 		kfree_skb(skb);
1681 		return false;
1682 	}
1683 
1684 	/* Reject if wrong message type for current socket state */
1685 	if (unlikely(sock->state == SS_READY)) {
1686 		if (msg_connected(hdr)) {
1687 			err = TIPC_ERR_NO_PORT;
1688 			goto reject;
1689 		}
1690 	} else if (unlikely(!filter_connect(tsk, skb))) {
1691 		err = TIPC_ERR_NO_PORT;
1692 		goto reject;
1693 	}
1694 
1695 	/* Reject message if there isn't room to queue it */
1696 	if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1697 		err = TIPC_ERR_OVERLOAD;
1698 		goto reject;
1699 	}
1700 
1701 	/* Enqueue message */
1702 	TIPC_SKB_CB(skb)->handle = NULL;
1703 	__skb_queue_tail(&sk->sk_receive_queue, skb);
1704 	skb_set_owner_r(skb, sk);
1705 
1706 	sk->sk_data_ready(sk);
1707 	return true;
1708 
1709 reject:
1710 	if (tipc_msg_reverse(tsk_own_node(tsk), &skb, err))
1711 		__skb_queue_tail(xmitq, skb);
1712 	return false;
1713 }
1714 
1715 /**
1716  * tipc_backlog_rcv - handle incoming message from backlog queue
1717  * @sk: socket
1718  * @skb: message
1719  *
1720  * Caller must hold socket lock
1721  *
1722  * Returns 0
1723  */
tipc_backlog_rcv(struct sock * sk,struct sk_buff * skb)1724 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1725 {
1726 	unsigned int truesize = skb->truesize;
1727 	struct sk_buff_head xmitq;
1728 	u32 dnode, selector;
1729 
1730 	__skb_queue_head_init(&xmitq);
1731 
1732 	if (likely(filter_rcv(sk, skb, &xmitq))) {
1733 		atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1734 		return 0;
1735 	}
1736 
1737 	if (skb_queue_empty(&xmitq))
1738 		return 0;
1739 
1740 	/* Send response/rejected message */
1741 	skb = __skb_dequeue(&xmitq);
1742 	dnode = msg_destnode(buf_msg(skb));
1743 	selector = msg_origport(buf_msg(skb));
1744 	tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
1745 	return 0;
1746 }
1747 
1748 /**
1749  * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1750  *                   inputq and try adding them to socket or backlog queue
1751  * @inputq: list of incoming buffers with potentially different destinations
1752  * @sk: socket where the buffers should be enqueued
1753  * @dport: port number for the socket
1754  *
1755  * Caller must hold socket lock
1756  */
tipc_sk_enqueue(struct sk_buff_head * inputq,struct sock * sk,u32 dport,struct sk_buff_head * xmitq)1757 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1758 			    u32 dport, struct sk_buff_head *xmitq)
1759 {
1760 	unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
1761 	struct sk_buff *skb;
1762 	unsigned int lim;
1763 	atomic_t *dcnt;
1764 	u32 onode;
1765 
1766 	while (skb_queue_len(inputq)) {
1767 		if (unlikely(time_after_eq(jiffies, time_limit)))
1768 			return;
1769 
1770 		skb = tipc_skb_dequeue(inputq, dport);
1771 		if (unlikely(!skb))
1772 			return;
1773 
1774 		/* Add message directly to receive queue if possible */
1775 		if (!sock_owned_by_user(sk)) {
1776 			filter_rcv(sk, skb, xmitq);
1777 			continue;
1778 		}
1779 
1780 		/* Try backlog, compensating for double-counted bytes */
1781 		dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1782 		if (!sk->sk_backlog.len)
1783 			atomic_set(dcnt, 0);
1784 		lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1785 		if (likely(!sk_add_backlog(sk, skb, lim)))
1786 			continue;
1787 
1788 		/* Overload => reject message back to sender */
1789 		onode = tipc_own_addr(sock_net(sk));
1790 		if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
1791 			__skb_queue_tail(xmitq, skb);
1792 		break;
1793 	}
1794 }
1795 
1796 /**
1797  * tipc_sk_rcv - handle a chain of incoming buffers
1798  * @inputq: buffer list containing the buffers
1799  * Consumes all buffers in list until inputq is empty
1800  * Note: may be called in multiple threads referring to the same queue
1801  */
tipc_sk_rcv(struct net * net,struct sk_buff_head * inputq)1802 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1803 {
1804 	struct sk_buff_head xmitq;
1805 	u32 dnode, dport = 0;
1806 	int err;
1807 	struct tipc_sock *tsk;
1808 	struct sock *sk;
1809 	struct sk_buff *skb;
1810 
1811 	__skb_queue_head_init(&xmitq);
1812 	while (skb_queue_len(inputq)) {
1813 		dport = tipc_skb_peek_port(inputq, dport);
1814 		tsk = tipc_sk_lookup(net, dport);
1815 
1816 		if (likely(tsk)) {
1817 			sk = &tsk->sk;
1818 			if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1819 				tipc_sk_enqueue(inputq, sk, dport, &xmitq);
1820 				spin_unlock_bh(&sk->sk_lock.slock);
1821 			}
1822 			/* Send pending response/rejected messages, if any */
1823 			while ((skb = __skb_dequeue(&xmitq))) {
1824 				dnode = msg_destnode(buf_msg(skb));
1825 				tipc_node_xmit_skb(net, skb, dnode, dport);
1826 			}
1827 			sock_put(sk);
1828 			continue;
1829 		}
1830 
1831 		/* No destination socket => dequeue skb if still there */
1832 		skb = tipc_skb_dequeue(inputq, dport);
1833 		if (!skb)
1834 			return;
1835 
1836 		/* Try secondary lookup if unresolved named message */
1837 		err = TIPC_ERR_NO_PORT;
1838 		if (tipc_msg_lookup_dest(net, skb, &err))
1839 			goto xmit;
1840 
1841 		/* Prepare for message rejection */
1842 		if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1843 			continue;
1844 xmit:
1845 		dnode = msg_destnode(buf_msg(skb));
1846 		tipc_node_xmit_skb(net, skb, dnode, dport);
1847 	}
1848 }
1849 
tipc_wait_for_connect(struct socket * sock,long * timeo_p)1850 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1851 {
1852 	struct sock *sk = sock->sk;
1853 	DEFINE_WAIT(wait);
1854 	int done;
1855 
1856 	do {
1857 		int err = sock_error(sk);
1858 		if (err)
1859 			return err;
1860 		if (!*timeo_p)
1861 			return -ETIMEDOUT;
1862 		if (signal_pending(current))
1863 			return sock_intr_errno(*timeo_p);
1864 
1865 		prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1866 		done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1867 		finish_wait(sk_sleep(sk), &wait);
1868 	} while (!done);
1869 	return 0;
1870 }
1871 
1872 /**
1873  * tipc_connect - establish a connection to another TIPC port
1874  * @sock: socket structure
1875  * @dest: socket address for destination port
1876  * @destlen: size of socket address data structure
1877  * @flags: file-related flags associated with socket
1878  *
1879  * Returns 0 on success, errno otherwise
1880  */
tipc_connect(struct socket * sock,struct sockaddr * dest,int destlen,int flags)1881 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1882 			int destlen, int flags)
1883 {
1884 	struct sock *sk = sock->sk;
1885 	struct tipc_sock *tsk = tipc_sk(sk);
1886 	struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1887 	struct msghdr m = {NULL,};
1888 	long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1889 	socket_state previous;
1890 	int res = 0;
1891 
1892 	lock_sock(sk);
1893 
1894 	/* DGRAM/RDM connect(), just save the destaddr */
1895 	if (sock->state == SS_READY) {
1896 		if (dst->family == AF_UNSPEC) {
1897 			memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1898 			tsk->connected = 0;
1899 		} else if (destlen != sizeof(struct sockaddr_tipc)) {
1900 			res = -EINVAL;
1901 		} else {
1902 			memcpy(&tsk->remote, dest, destlen);
1903 			tsk->connected = 1;
1904 		}
1905 		goto exit;
1906 	}
1907 
1908 	/*
1909 	 * Reject connection attempt using multicast address
1910 	 *
1911 	 * Note: send_msg() validates the rest of the address fields,
1912 	 *       so there's no need to do it here
1913 	 */
1914 	if (dst->addrtype == TIPC_ADDR_MCAST) {
1915 		res = -EINVAL;
1916 		goto exit;
1917 	}
1918 
1919 	previous = sock->state;
1920 	switch (sock->state) {
1921 	case SS_UNCONNECTED:
1922 		/* Send a 'SYN-' to destination */
1923 		m.msg_name = dest;
1924 		m.msg_namelen = destlen;
1925 
1926 		/* If connect is in non-blocking case, set MSG_DONTWAIT to
1927 		 * indicate send_msg() is never blocked.
1928 		 */
1929 		if (!timeout)
1930 			m.msg_flags = MSG_DONTWAIT;
1931 
1932 		res = __tipc_sendmsg(sock, &m, 0);
1933 		if ((res < 0) && (res != -EWOULDBLOCK))
1934 			goto exit;
1935 
1936 		/* Just entered SS_CONNECTING state; the only
1937 		 * difference is that return value in non-blocking
1938 		 * case is EINPROGRESS, rather than EALREADY.
1939 		 */
1940 		res = -EINPROGRESS;
1941 	case SS_CONNECTING:
1942 		if (previous == SS_CONNECTING)
1943 			res = -EALREADY;
1944 		if (!timeout)
1945 			goto exit;
1946 		timeout = msecs_to_jiffies(timeout);
1947 		/* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1948 		res = tipc_wait_for_connect(sock, &timeout);
1949 		break;
1950 	case SS_CONNECTED:
1951 		res = -EISCONN;
1952 		break;
1953 	default:
1954 		res = -EINVAL;
1955 		break;
1956 	}
1957 exit:
1958 	release_sock(sk);
1959 	return res;
1960 }
1961 
1962 /**
1963  * tipc_listen - allow socket to listen for incoming connections
1964  * @sock: socket structure
1965  * @len: (unused)
1966  *
1967  * Returns 0 on success, errno otherwise
1968  */
tipc_listen(struct socket * sock,int len)1969 static int tipc_listen(struct socket *sock, int len)
1970 {
1971 	struct sock *sk = sock->sk;
1972 	int res;
1973 
1974 	lock_sock(sk);
1975 
1976 	if (sock->state != SS_UNCONNECTED)
1977 		res = -EINVAL;
1978 	else {
1979 		sock->state = SS_LISTENING;
1980 		res = 0;
1981 	}
1982 
1983 	release_sock(sk);
1984 	return res;
1985 }
1986 
tipc_wait_for_accept(struct socket * sock,long timeo)1987 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1988 {
1989 	struct sock *sk = sock->sk;
1990 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
1991 	int err;
1992 
1993 	/* True wake-one mechanism for incoming connections: only
1994 	 * one process gets woken up, not the 'whole herd'.
1995 	 * Since we do not 'race & poll' for established sockets
1996 	 * anymore, the common case will execute the loop only once.
1997 	*/
1998 	for (;;) {
1999 		if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2000 			add_wait_queue(sk_sleep(sk), &wait);
2001 			release_sock(sk);
2002 			timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2003 			lock_sock(sk);
2004 			remove_wait_queue(sk_sleep(sk), &wait);
2005 		}
2006 		err = 0;
2007 		if (!skb_queue_empty(&sk->sk_receive_queue))
2008 			break;
2009 		err = -EINVAL;
2010 		if (sock->state != SS_LISTENING)
2011 			break;
2012 		err = -EAGAIN;
2013 		if (!timeo)
2014 			break;
2015 		err = sock_intr_errno(timeo);
2016 		if (signal_pending(current))
2017 			break;
2018 	}
2019 	return err;
2020 }
2021 
2022 /**
2023  * tipc_accept - wait for connection request
2024  * @sock: listening socket
2025  * @newsock: new socket that is to be connected
2026  * @flags: file-related flags associated with socket
2027  *
2028  * Returns 0 on success, errno otherwise
2029  */
tipc_accept(struct socket * sock,struct socket * new_sock,int flags)2030 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2031 {
2032 	struct sock *new_sk, *sk = sock->sk;
2033 	struct sk_buff *buf;
2034 	struct tipc_sock *new_tsock;
2035 	struct tipc_msg *msg;
2036 	long timeo;
2037 	int res;
2038 
2039 	lock_sock(sk);
2040 
2041 	if (sock->state != SS_LISTENING) {
2042 		res = -EINVAL;
2043 		goto exit;
2044 	}
2045 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2046 	res = tipc_wait_for_accept(sock, timeo);
2047 	if (res)
2048 		goto exit;
2049 
2050 	buf = skb_peek(&sk->sk_receive_queue);
2051 
2052 	res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2053 	if (res)
2054 		goto exit;
2055 	security_sk_clone(sock->sk, new_sock->sk);
2056 
2057 	new_sk = new_sock->sk;
2058 	new_tsock = tipc_sk(new_sk);
2059 	msg = buf_msg(buf);
2060 
2061 	/* we lock on new_sk; but lockdep sees the lock on sk */
2062 	lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2063 
2064 	/*
2065 	 * Reject any stray messages received by new socket
2066 	 * before the socket lock was taken (very, very unlikely)
2067 	 */
2068 	tsk_rej_rx_queue(new_sk);
2069 
2070 	/* Connect new socket to it's peer */
2071 	tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2072 	new_sock->state = SS_CONNECTED;
2073 
2074 	tsk_set_importance(new_tsock, msg_importance(msg));
2075 	if (msg_named(msg)) {
2076 		new_tsock->conn_type = msg_nametype(msg);
2077 		new_tsock->conn_instance = msg_nameinst(msg);
2078 	}
2079 
2080 	/*
2081 	 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2082 	 * Respond to 'SYN+' by queuing it on new socket.
2083 	 */
2084 	if (!msg_data_sz(msg)) {
2085 		struct msghdr m = {NULL,};
2086 
2087 		tsk_advance_rx_queue(sk);
2088 		__tipc_send_stream(new_sock, &m, 0);
2089 	} else {
2090 		__skb_dequeue(&sk->sk_receive_queue);
2091 		__skb_queue_head(&new_sk->sk_receive_queue, buf);
2092 		skb_set_owner_r(buf, new_sk);
2093 	}
2094 	release_sock(new_sk);
2095 exit:
2096 	release_sock(sk);
2097 	return res;
2098 }
2099 
2100 /**
2101  * tipc_shutdown - shutdown socket connection
2102  * @sock: socket structure
2103  * @how: direction to close (must be SHUT_RDWR)
2104  *
2105  * Terminates connection (if necessary), then purges socket's receive queue.
2106  *
2107  * Returns 0 on success, errno otherwise
2108  */
tipc_shutdown(struct socket * sock,int how)2109 static int tipc_shutdown(struct socket *sock, int how)
2110 {
2111 	struct sock *sk = sock->sk;
2112 	struct net *net = sock_net(sk);
2113 	struct tipc_sock *tsk = tipc_sk(sk);
2114 	struct sk_buff *skb;
2115 	u32 dnode = tsk_peer_node(tsk);
2116 	u32 dport = tsk_peer_port(tsk);
2117 	u32 onode = tipc_own_addr(net);
2118 	u32 oport = tsk->portid;
2119 	int res;
2120 
2121 	if (how != SHUT_RDWR)
2122 		return -EINVAL;
2123 
2124 	lock_sock(sk);
2125 
2126 	switch (sock->state) {
2127 	case SS_CONNECTING:
2128 	case SS_CONNECTED:
2129 
2130 restart:
2131 		dnode = tsk_peer_node(tsk);
2132 
2133 		/* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2134 		skb = __skb_dequeue(&sk->sk_receive_queue);
2135 		if (skb) {
2136 			if (TIPC_SKB_CB(skb)->handle != NULL) {
2137 				kfree_skb(skb);
2138 				goto restart;
2139 			}
2140 			tipc_sk_respond(sk, skb, TIPC_CONN_SHUTDOWN);
2141 		} else {
2142 			skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2143 					      TIPC_CONN_MSG, SHORT_H_SIZE,
2144 					      0, dnode, onode, dport, oport,
2145 					      TIPC_CONN_SHUTDOWN);
2146 			if (skb)
2147 				tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
2148 		}
2149 		tsk->connected = 0;
2150 		sock->state = SS_DISCONNECTING;
2151 		tipc_node_remove_conn(net, dnode, tsk->portid);
2152 		/* fall through */
2153 
2154 	case SS_DISCONNECTING:
2155 
2156 		/* Discard any unreceived messages */
2157 		__skb_queue_purge(&sk->sk_receive_queue);
2158 
2159 		/* Wake up anyone sleeping in poll */
2160 		sk->sk_state_change(sk);
2161 		res = 0;
2162 		break;
2163 
2164 	default:
2165 		res = -ENOTCONN;
2166 	}
2167 
2168 	release_sock(sk);
2169 	return res;
2170 }
2171 
tipc_sk_timeout(unsigned long data)2172 static void tipc_sk_timeout(unsigned long data)
2173 {
2174 	struct tipc_sock *tsk = (struct tipc_sock *)data;
2175 	struct sock *sk = &tsk->sk;
2176 	struct sk_buff *skb = NULL;
2177 	u32 peer_port, peer_node;
2178 	u32 own_node = tsk_own_node(tsk);
2179 
2180 	bh_lock_sock(sk);
2181 	if (!tsk->connected) {
2182 		bh_unlock_sock(sk);
2183 		goto exit;
2184 	}
2185 	peer_port = tsk_peer_port(tsk);
2186 	peer_node = tsk_peer_node(tsk);
2187 
2188 	if (tsk->probing_state == TIPC_CONN_PROBING) {
2189 		if (!sock_owned_by_user(sk)) {
2190 			sk->sk_socket->state = SS_DISCONNECTING;
2191 			tsk->connected = 0;
2192 			tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2193 					      tsk_peer_port(tsk));
2194 			sk->sk_state_change(sk);
2195 		} else {
2196 			/* Try again later */
2197 			sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2198 		}
2199 
2200 	} else {
2201 		skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2202 				      INT_H_SIZE, 0, peer_node, own_node,
2203 				      peer_port, tsk->portid, TIPC_OK);
2204 		tsk->probing_state = TIPC_CONN_PROBING;
2205 		sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2206 	}
2207 	bh_unlock_sock(sk);
2208 	if (skb)
2209 		tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2210 exit:
2211 	sock_put(sk);
2212 }
2213 
tipc_sk_publish(struct tipc_sock * tsk,uint scope,struct tipc_name_seq const * seq)2214 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2215 			   struct tipc_name_seq const *seq)
2216 {
2217 	struct net *net = sock_net(&tsk->sk);
2218 	struct publication *publ;
2219 	u32 key;
2220 
2221 	if (tsk->connected)
2222 		return -EINVAL;
2223 	key = tsk->portid + tsk->pub_count + 1;
2224 	if (key == tsk->portid)
2225 		return -EADDRINUSE;
2226 
2227 	publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2228 				    scope, tsk->portid, key);
2229 	if (unlikely(!publ))
2230 		return -EINVAL;
2231 
2232 	list_add(&publ->pport_list, &tsk->publications);
2233 	tsk->pub_count++;
2234 	tsk->published = 1;
2235 	return 0;
2236 }
2237 
tipc_sk_withdraw(struct tipc_sock * tsk,uint scope,struct tipc_name_seq const * seq)2238 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2239 			    struct tipc_name_seq const *seq)
2240 {
2241 	struct net *net = sock_net(&tsk->sk);
2242 	struct publication *publ;
2243 	struct publication *safe;
2244 	int rc = -EINVAL;
2245 
2246 	list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2247 		if (seq) {
2248 			if (publ->scope != scope)
2249 				continue;
2250 			if (publ->type != seq->type)
2251 				continue;
2252 			if (publ->lower != seq->lower)
2253 				continue;
2254 			if (publ->upper != seq->upper)
2255 				break;
2256 			tipc_nametbl_withdraw(net, publ->type, publ->lower,
2257 					      publ->ref, publ->key);
2258 			rc = 0;
2259 			break;
2260 		}
2261 		tipc_nametbl_withdraw(net, publ->type, publ->lower,
2262 				      publ->ref, publ->key);
2263 		rc = 0;
2264 	}
2265 	if (list_empty(&tsk->publications))
2266 		tsk->published = 0;
2267 	return rc;
2268 }
2269 
2270 /* tipc_sk_reinit: set non-zero address in all existing sockets
2271  *                 when we go from standalone to network mode.
2272  */
tipc_sk_reinit(struct net * net)2273 void tipc_sk_reinit(struct net *net)
2274 {
2275 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2276 	const struct bucket_table *tbl;
2277 	struct rhash_head *pos;
2278 	struct tipc_sock *tsk;
2279 	struct tipc_msg *msg;
2280 	int i;
2281 
2282 	rcu_read_lock();
2283 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2284 	for (i = 0; i < tbl->size; i++) {
2285 		rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2286 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2287 			msg = &tsk->phdr;
2288 			msg_set_prevnode(msg, tn->own_addr);
2289 			msg_set_orignode(msg, tn->own_addr);
2290 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2291 		}
2292 	}
2293 	rcu_read_unlock();
2294 }
2295 
tipc_sk_lookup(struct net * net,u32 portid)2296 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2297 {
2298 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2299 	struct tipc_sock *tsk;
2300 
2301 	rcu_read_lock();
2302 	tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2303 	if (tsk)
2304 		sock_hold(&tsk->sk);
2305 	rcu_read_unlock();
2306 
2307 	return tsk;
2308 }
2309 
tipc_sk_insert(struct tipc_sock * tsk)2310 static int tipc_sk_insert(struct tipc_sock *tsk)
2311 {
2312 	struct sock *sk = &tsk->sk;
2313 	struct net *net = sock_net(sk);
2314 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2315 	u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2316 	u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2317 
2318 	while (remaining--) {
2319 		portid++;
2320 		if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2321 			portid = TIPC_MIN_PORT;
2322 		tsk->portid = portid;
2323 		sock_hold(&tsk->sk);
2324 		if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2325 						   tsk_rht_params))
2326 			return 0;
2327 		sock_put(&tsk->sk);
2328 	}
2329 
2330 	return -1;
2331 }
2332 
tipc_sk_remove(struct tipc_sock * tsk)2333 static void tipc_sk_remove(struct tipc_sock *tsk)
2334 {
2335 	struct sock *sk = &tsk->sk;
2336 	struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2337 
2338 	if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2339 		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2340 		__sock_put(sk);
2341 	}
2342 }
2343 
2344 static const struct rhashtable_params tsk_rht_params = {
2345 	.nelem_hint = 192,
2346 	.head_offset = offsetof(struct tipc_sock, node),
2347 	.key_offset = offsetof(struct tipc_sock, portid),
2348 	.key_len = sizeof(u32), /* portid */
2349 	.max_size = 1048576,
2350 	.min_size = 256,
2351 	.automatic_shrinking = true,
2352 };
2353 
tipc_sk_rht_init(struct net * net)2354 int tipc_sk_rht_init(struct net *net)
2355 {
2356 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2357 
2358 	return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2359 }
2360 
tipc_sk_rht_destroy(struct net * net)2361 void tipc_sk_rht_destroy(struct net *net)
2362 {
2363 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2364 
2365 	/* Wait for socket readers to complete */
2366 	synchronize_net();
2367 
2368 	rhashtable_destroy(&tn->sk_rht);
2369 }
2370 
2371 /**
2372  * tipc_setsockopt - set socket option
2373  * @sock: socket structure
2374  * @lvl: option level
2375  * @opt: option identifier
2376  * @ov: pointer to new option value
2377  * @ol: length of option value
2378  *
2379  * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2380  * (to ease compatibility).
2381  *
2382  * Returns 0 on success, errno otherwise
2383  */
tipc_setsockopt(struct socket * sock,int lvl,int opt,char __user * ov,unsigned int ol)2384 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2385 			   char __user *ov, unsigned int ol)
2386 {
2387 	struct sock *sk = sock->sk;
2388 	struct tipc_sock *tsk = tipc_sk(sk);
2389 	u32 value;
2390 	int res;
2391 
2392 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2393 		return 0;
2394 	if (lvl != SOL_TIPC)
2395 		return -ENOPROTOOPT;
2396 	if (ol < sizeof(value))
2397 		return -EINVAL;
2398 	res = get_user(value, (u32 __user *)ov);
2399 	if (res)
2400 		return res;
2401 
2402 	lock_sock(sk);
2403 
2404 	switch (opt) {
2405 	case TIPC_IMPORTANCE:
2406 		res = tsk_set_importance(tsk, value);
2407 		break;
2408 	case TIPC_SRC_DROPPABLE:
2409 		if (sock->type != SOCK_STREAM)
2410 			tsk_set_unreliable(tsk, value);
2411 		else
2412 			res = -ENOPROTOOPT;
2413 		break;
2414 	case TIPC_DEST_DROPPABLE:
2415 		tsk_set_unreturnable(tsk, value);
2416 		break;
2417 	case TIPC_CONN_TIMEOUT:
2418 		tipc_sk(sk)->conn_timeout = value;
2419 		/* no need to set "res", since already 0 at this point */
2420 		break;
2421 	default:
2422 		res = -EINVAL;
2423 	}
2424 
2425 	release_sock(sk);
2426 
2427 	return res;
2428 }
2429 
2430 /**
2431  * tipc_getsockopt - get socket option
2432  * @sock: socket structure
2433  * @lvl: option level
2434  * @opt: option identifier
2435  * @ov: receptacle for option value
2436  * @ol: receptacle for length of option value
2437  *
2438  * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2439  * (to ease compatibility).
2440  *
2441  * Returns 0 on success, errno otherwise
2442  */
tipc_getsockopt(struct socket * sock,int lvl,int opt,char __user * ov,int __user * ol)2443 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2444 			   char __user *ov, int __user *ol)
2445 {
2446 	struct sock *sk = sock->sk;
2447 	struct tipc_sock *tsk = tipc_sk(sk);
2448 	int len;
2449 	u32 value;
2450 	int res;
2451 
2452 	if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2453 		return put_user(0, ol);
2454 	if (lvl != SOL_TIPC)
2455 		return -ENOPROTOOPT;
2456 	res = get_user(len, ol);
2457 	if (res)
2458 		return res;
2459 
2460 	lock_sock(sk);
2461 
2462 	switch (opt) {
2463 	case TIPC_IMPORTANCE:
2464 		value = tsk_importance(tsk);
2465 		break;
2466 	case TIPC_SRC_DROPPABLE:
2467 		value = tsk_unreliable(tsk);
2468 		break;
2469 	case TIPC_DEST_DROPPABLE:
2470 		value = tsk_unreturnable(tsk);
2471 		break;
2472 	case TIPC_CONN_TIMEOUT:
2473 		value = tsk->conn_timeout;
2474 		/* no need to set "res", since already 0 at this point */
2475 		break;
2476 	case TIPC_NODE_RECVQ_DEPTH:
2477 		value = 0; /* was tipc_queue_size, now obsolete */
2478 		break;
2479 	case TIPC_SOCK_RECVQ_DEPTH:
2480 		value = skb_queue_len(&sk->sk_receive_queue);
2481 		break;
2482 	default:
2483 		res = -EINVAL;
2484 	}
2485 
2486 	release_sock(sk);
2487 
2488 	if (res)
2489 		return res;	/* "get" failed */
2490 
2491 	if (len < sizeof(value))
2492 		return -EINVAL;
2493 
2494 	if (copy_to_user(ov, &value, sizeof(value)))
2495 		return -EFAULT;
2496 
2497 	return put_user(sizeof(value), ol);
2498 }
2499 
tipc_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)2500 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2501 {
2502 	struct sock *sk = sock->sk;
2503 	struct tipc_sioc_ln_req lnr;
2504 	void __user *argp = (void __user *)arg;
2505 
2506 	switch (cmd) {
2507 	case SIOCGETLINKNAME:
2508 		if (copy_from_user(&lnr, argp, sizeof(lnr)))
2509 			return -EFAULT;
2510 		if (!tipc_node_get_linkname(sock_net(sk),
2511 					    lnr.bearer_id & 0xffff, lnr.peer,
2512 					    lnr.linkname, TIPC_MAX_LINK_NAME)) {
2513 			if (copy_to_user(argp, &lnr, sizeof(lnr)))
2514 				return -EFAULT;
2515 			return 0;
2516 		}
2517 		return -EADDRNOTAVAIL;
2518 	default:
2519 		return -ENOIOCTLCMD;
2520 	}
2521 }
2522 
2523 /* Protocol switches for the various types of TIPC sockets */
2524 
2525 static const struct proto_ops msg_ops = {
2526 	.owner		= THIS_MODULE,
2527 	.family		= AF_TIPC,
2528 	.release	= tipc_release,
2529 	.bind		= tipc_bind,
2530 	.connect	= tipc_connect,
2531 	.socketpair	= sock_no_socketpair,
2532 	.accept		= sock_no_accept,
2533 	.getname	= tipc_getname,
2534 	.poll		= tipc_poll,
2535 	.ioctl		= tipc_ioctl,
2536 	.listen		= sock_no_listen,
2537 	.shutdown	= tipc_shutdown,
2538 	.setsockopt	= tipc_setsockopt,
2539 	.getsockopt	= tipc_getsockopt,
2540 	.sendmsg	= tipc_sendmsg,
2541 	.recvmsg	= tipc_recvmsg,
2542 	.mmap		= sock_no_mmap,
2543 	.sendpage	= sock_no_sendpage
2544 };
2545 
2546 static const struct proto_ops packet_ops = {
2547 	.owner		= THIS_MODULE,
2548 	.family		= AF_TIPC,
2549 	.release	= tipc_release,
2550 	.bind		= tipc_bind,
2551 	.connect	= tipc_connect,
2552 	.socketpair	= sock_no_socketpair,
2553 	.accept		= tipc_accept,
2554 	.getname	= tipc_getname,
2555 	.poll		= tipc_poll,
2556 	.ioctl		= tipc_ioctl,
2557 	.listen		= tipc_listen,
2558 	.shutdown	= tipc_shutdown,
2559 	.setsockopt	= tipc_setsockopt,
2560 	.getsockopt	= tipc_getsockopt,
2561 	.sendmsg	= tipc_send_packet,
2562 	.recvmsg	= tipc_recvmsg,
2563 	.mmap		= sock_no_mmap,
2564 	.sendpage	= sock_no_sendpage
2565 };
2566 
2567 static const struct proto_ops stream_ops = {
2568 	.owner		= THIS_MODULE,
2569 	.family		= AF_TIPC,
2570 	.release	= tipc_release,
2571 	.bind		= tipc_bind,
2572 	.connect	= tipc_connect,
2573 	.socketpair	= sock_no_socketpair,
2574 	.accept		= tipc_accept,
2575 	.getname	= tipc_getname,
2576 	.poll		= tipc_poll,
2577 	.ioctl		= tipc_ioctl,
2578 	.listen		= tipc_listen,
2579 	.shutdown	= tipc_shutdown,
2580 	.setsockopt	= tipc_setsockopt,
2581 	.getsockopt	= tipc_getsockopt,
2582 	.sendmsg	= tipc_send_stream,
2583 	.recvmsg	= tipc_recv_stream,
2584 	.mmap		= sock_no_mmap,
2585 	.sendpage	= sock_no_sendpage
2586 };
2587 
2588 static const struct net_proto_family tipc_family_ops = {
2589 	.owner		= THIS_MODULE,
2590 	.family		= AF_TIPC,
2591 	.create		= tipc_sk_create
2592 };
2593 
2594 static struct proto tipc_proto = {
2595 	.name		= "TIPC",
2596 	.owner		= THIS_MODULE,
2597 	.obj_size	= sizeof(struct tipc_sock),
2598 	.sysctl_rmem	= sysctl_tipc_rmem
2599 };
2600 
2601 /**
2602  * tipc_socket_init - initialize TIPC socket interface
2603  *
2604  * Returns 0 on success, errno otherwise
2605  */
tipc_socket_init(void)2606 int tipc_socket_init(void)
2607 {
2608 	int res;
2609 
2610 	res = proto_register(&tipc_proto, 1);
2611 	if (res) {
2612 		pr_err("Failed to register TIPC protocol type\n");
2613 		goto out;
2614 	}
2615 
2616 	res = sock_register(&tipc_family_ops);
2617 	if (res) {
2618 		pr_err("Failed to register TIPC socket type\n");
2619 		proto_unregister(&tipc_proto);
2620 		goto out;
2621 	}
2622  out:
2623 	return res;
2624 }
2625 
2626 /**
2627  * tipc_socket_stop - stop TIPC socket interface
2628  */
tipc_socket_stop(void)2629 void tipc_socket_stop(void)
2630 {
2631 	sock_unregister(tipc_family_ops.family);
2632 	proto_unregister(&tipc_proto);
2633 }
2634 
2635 /* Caller should hold socket lock for the passed tipc socket. */
__tipc_nl_add_sk_con(struct sk_buff * skb,struct tipc_sock * tsk)2636 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2637 {
2638 	u32 peer_node;
2639 	u32 peer_port;
2640 	struct nlattr *nest;
2641 
2642 	peer_node = tsk_peer_node(tsk);
2643 	peer_port = tsk_peer_port(tsk);
2644 
2645 	nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2646 
2647 	if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2648 		goto msg_full;
2649 	if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2650 		goto msg_full;
2651 
2652 	if (tsk->conn_type != 0) {
2653 		if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2654 			goto msg_full;
2655 		if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2656 			goto msg_full;
2657 		if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2658 			goto msg_full;
2659 	}
2660 	nla_nest_end(skb, nest);
2661 
2662 	return 0;
2663 
2664 msg_full:
2665 	nla_nest_cancel(skb, nest);
2666 
2667 	return -EMSGSIZE;
2668 }
2669 
2670 /* Caller should hold socket lock for the passed tipc socket. */
__tipc_nl_add_sk(struct sk_buff * skb,struct netlink_callback * cb,struct tipc_sock * tsk)2671 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2672 			    struct tipc_sock *tsk)
2673 {
2674 	int err;
2675 	void *hdr;
2676 	struct nlattr *attrs;
2677 	struct net *net = sock_net(skb->sk);
2678 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2679 
2680 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2681 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2682 	if (!hdr)
2683 		goto msg_cancel;
2684 
2685 	attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2686 	if (!attrs)
2687 		goto genlmsg_cancel;
2688 	if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2689 		goto attr_msg_cancel;
2690 	if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2691 		goto attr_msg_cancel;
2692 
2693 	if (tsk->connected) {
2694 		err = __tipc_nl_add_sk_con(skb, tsk);
2695 		if (err)
2696 			goto attr_msg_cancel;
2697 	} else if (!list_empty(&tsk->publications)) {
2698 		if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2699 			goto attr_msg_cancel;
2700 	}
2701 	nla_nest_end(skb, attrs);
2702 	genlmsg_end(skb, hdr);
2703 
2704 	return 0;
2705 
2706 attr_msg_cancel:
2707 	nla_nest_cancel(skb, attrs);
2708 genlmsg_cancel:
2709 	genlmsg_cancel(skb, hdr);
2710 msg_cancel:
2711 	return -EMSGSIZE;
2712 }
2713 
tipc_nl_sk_dump(struct sk_buff * skb,struct netlink_callback * cb)2714 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2715 {
2716 	int err;
2717 	struct tipc_sock *tsk;
2718 	const struct bucket_table *tbl;
2719 	struct rhash_head *pos;
2720 	struct net *net = sock_net(skb->sk);
2721 	struct tipc_net *tn = net_generic(net, tipc_net_id);
2722 	u32 tbl_id = cb->args[0];
2723 	u32 prev_portid = cb->args[1];
2724 
2725 	rcu_read_lock();
2726 	tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2727 	for (; tbl_id < tbl->size; tbl_id++) {
2728 		rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2729 			spin_lock_bh(&tsk->sk.sk_lock.slock);
2730 			if (prev_portid && prev_portid != tsk->portid) {
2731 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2732 				continue;
2733 			}
2734 
2735 			err = __tipc_nl_add_sk(skb, cb, tsk);
2736 			if (err) {
2737 				prev_portid = tsk->portid;
2738 				spin_unlock_bh(&tsk->sk.sk_lock.slock);
2739 				goto out;
2740 			}
2741 			prev_portid = 0;
2742 			spin_unlock_bh(&tsk->sk.sk_lock.slock);
2743 		}
2744 	}
2745 out:
2746 	rcu_read_unlock();
2747 	cb->args[0] = tbl_id;
2748 	cb->args[1] = prev_portid;
2749 
2750 	return skb->len;
2751 }
2752 
2753 /* Caller should hold socket lock for the passed tipc socket. */
__tipc_nl_add_sk_publ(struct sk_buff * skb,struct netlink_callback * cb,struct publication * publ)2754 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2755 				 struct netlink_callback *cb,
2756 				 struct publication *publ)
2757 {
2758 	void *hdr;
2759 	struct nlattr *attrs;
2760 
2761 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2762 			  &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2763 	if (!hdr)
2764 		goto msg_cancel;
2765 
2766 	attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2767 	if (!attrs)
2768 		goto genlmsg_cancel;
2769 
2770 	if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2771 		goto attr_msg_cancel;
2772 	if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2773 		goto attr_msg_cancel;
2774 	if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2775 		goto attr_msg_cancel;
2776 	if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2777 		goto attr_msg_cancel;
2778 
2779 	nla_nest_end(skb, attrs);
2780 	genlmsg_end(skb, hdr);
2781 
2782 	return 0;
2783 
2784 attr_msg_cancel:
2785 	nla_nest_cancel(skb, attrs);
2786 genlmsg_cancel:
2787 	genlmsg_cancel(skb, hdr);
2788 msg_cancel:
2789 	return -EMSGSIZE;
2790 }
2791 
2792 /* Caller should hold socket lock for the passed tipc socket. */
__tipc_nl_list_sk_publ(struct sk_buff * skb,struct netlink_callback * cb,struct tipc_sock * tsk,u32 * last_publ)2793 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2794 				  struct netlink_callback *cb,
2795 				  struct tipc_sock *tsk, u32 *last_publ)
2796 {
2797 	int err;
2798 	struct publication *p;
2799 
2800 	if (*last_publ) {
2801 		list_for_each_entry(p, &tsk->publications, pport_list) {
2802 			if (p->key == *last_publ)
2803 				break;
2804 		}
2805 		if (p->key != *last_publ) {
2806 			/* We never set seq or call nl_dump_check_consistent()
2807 			 * this means that setting prev_seq here will cause the
2808 			 * consistence check to fail in the netlink callback
2809 			 * handler. Resulting in the last NLMSG_DONE message
2810 			 * having the NLM_F_DUMP_INTR flag set.
2811 			 */
2812 			cb->prev_seq = 1;
2813 			*last_publ = 0;
2814 			return -EPIPE;
2815 		}
2816 	} else {
2817 		p = list_first_entry(&tsk->publications, struct publication,
2818 				     pport_list);
2819 	}
2820 
2821 	list_for_each_entry_from(p, &tsk->publications, pport_list) {
2822 		err = __tipc_nl_add_sk_publ(skb, cb, p);
2823 		if (err) {
2824 			*last_publ = p->key;
2825 			return err;
2826 		}
2827 	}
2828 	*last_publ = 0;
2829 
2830 	return 0;
2831 }
2832 
tipc_nl_publ_dump(struct sk_buff * skb,struct netlink_callback * cb)2833 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2834 {
2835 	int err;
2836 	u32 tsk_portid = cb->args[0];
2837 	u32 last_publ = cb->args[1];
2838 	u32 done = cb->args[2];
2839 	struct net *net = sock_net(skb->sk);
2840 	struct tipc_sock *tsk;
2841 
2842 	if (!tsk_portid) {
2843 		struct nlattr **attrs;
2844 		struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2845 
2846 		err = tipc_nlmsg_parse(cb->nlh, &attrs);
2847 		if (err)
2848 			return err;
2849 
2850 		if (!attrs[TIPC_NLA_SOCK])
2851 			return -EINVAL;
2852 
2853 		err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2854 				       attrs[TIPC_NLA_SOCK],
2855 				       tipc_nl_sock_policy);
2856 		if (err)
2857 			return err;
2858 
2859 		if (!sock[TIPC_NLA_SOCK_REF])
2860 			return -EINVAL;
2861 
2862 		tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2863 	}
2864 
2865 	if (done)
2866 		return 0;
2867 
2868 	tsk = tipc_sk_lookup(net, tsk_portid);
2869 	if (!tsk)
2870 		return -EINVAL;
2871 
2872 	lock_sock(&tsk->sk);
2873 	err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2874 	if (!err)
2875 		done = 1;
2876 	release_sock(&tsk->sk);
2877 	sock_put(&tsk->sk);
2878 
2879 	cb->args[0] = tsk_portid;
2880 	cb->args[1] = last_publ;
2881 	cb->args[2] = done;
2882 
2883 	return skb->len;
2884 }
2885