Loading...
1/*
2 * net/tipc/socket.c: TIPC socket API
3 *
4 * Copyright (c) 2001-2007, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2011, 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 <net/sock.h>
38
39#include "core.h"
40#include "port.h"
41
42#define SS_LISTENING -1 /* socket is listening */
43#define SS_READY -2 /* socket is connectionless */
44
45#define OVERLOAD_LIMIT_BASE 5000
46#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
47
48struct tipc_sock {
49 struct sock sk;
50 struct tipc_port *p;
51 struct tipc_portid peer_name;
52 long conn_timeout;
53};
54
55#define tipc_sk(sk) ((struct tipc_sock *)(sk))
56#define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
57
58#define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
59 (sock->state == SS_DISCONNECTING))
60
61static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
62static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
63static void wakeupdispatch(struct tipc_port *tport);
64
65static const struct proto_ops packet_ops;
66static const struct proto_ops stream_ops;
67static const struct proto_ops msg_ops;
68
69static struct proto tipc_proto;
70
71static int sockets_enabled;
72
73static atomic_t tipc_queue_size = ATOMIC_INIT(0);
74
75/*
76 * Revised TIPC socket locking policy:
77 *
78 * Most socket operations take the standard socket lock when they start
79 * and hold it until they finish (or until they need to sleep). Acquiring
80 * this lock grants the owner exclusive access to the fields of the socket
81 * data structures, with the exception of the backlog queue. A few socket
82 * operations can be done without taking the socket lock because they only
83 * read socket information that never changes during the life of the socket.
84 *
85 * Socket operations may acquire the lock for the associated TIPC port if they
86 * need to perform an operation on the port. If any routine needs to acquire
87 * both the socket lock and the port lock it must take the socket lock first
88 * to avoid the risk of deadlock.
89 *
90 * The dispatcher handling incoming messages cannot grab the socket lock in
91 * the standard fashion, since invoked it runs at the BH level and cannot block.
92 * Instead, it checks to see if the socket lock is currently owned by someone,
93 * and either handles the message itself or adds it to the socket's backlog
94 * queue; in the latter case the queued message is processed once the process
95 * owning the socket lock releases it.
96 *
97 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
98 * the problem of a blocked socket operation preventing any other operations
99 * from occurring. However, applications must be careful if they have
100 * multiple threads trying to send (or receive) on the same socket, as these
101 * operations might interfere with each other. For example, doing a connect
102 * and a receive at the same time might allow the receive to consume the
103 * ACK message meant for the connect. While additional work could be done
104 * to try and overcome this, it doesn't seem to be worthwhile at the present.
105 *
106 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
107 * that another operation that must be performed in a non-blocking manner is
108 * not delayed for very long because the lock has already been taken.
109 *
110 * NOTE: This code assumes that certain fields of a port/socket pair are
111 * constant over its lifetime; such fields can be examined without taking
112 * the socket lock and/or port lock, and do not need to be re-read even
113 * after resuming processing after waiting. These fields include:
114 * - socket type
115 * - pointer to socket sk structure (aka tipc_sock structure)
116 * - pointer to port structure
117 * - port reference
118 */
119
120/**
121 * advance_rx_queue - discard first buffer in socket receive queue
122 *
123 * Caller must hold socket lock
124 */
125
126static void advance_rx_queue(struct sock *sk)
127{
128 buf_discard(__skb_dequeue(&sk->sk_receive_queue));
129 atomic_dec(&tipc_queue_size);
130}
131
132/**
133 * discard_rx_queue - discard all buffers in socket receive queue
134 *
135 * Caller must hold socket lock
136 */
137
138static void discard_rx_queue(struct sock *sk)
139{
140 struct sk_buff *buf;
141
142 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
143 atomic_dec(&tipc_queue_size);
144 buf_discard(buf);
145 }
146}
147
148/**
149 * reject_rx_queue - reject all buffers in socket receive queue
150 *
151 * Caller must hold socket lock
152 */
153
154static void reject_rx_queue(struct sock *sk)
155{
156 struct sk_buff *buf;
157
158 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
159 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
160 atomic_dec(&tipc_queue_size);
161 }
162}
163
164/**
165 * tipc_create - create a TIPC socket
166 * @net: network namespace (must be default network)
167 * @sock: pre-allocated socket structure
168 * @protocol: protocol indicator (must be 0)
169 * @kern: caused by kernel or by userspace?
170 *
171 * This routine creates additional data structures used by the TIPC socket,
172 * initializes them, and links them together.
173 *
174 * Returns 0 on success, errno otherwise
175 */
176
177static int tipc_create(struct net *net, struct socket *sock, int protocol,
178 int kern)
179{
180 const struct proto_ops *ops;
181 socket_state state;
182 struct sock *sk;
183 struct tipc_port *tp_ptr;
184
185 /* Validate arguments */
186
187 if (!net_eq(net, &init_net))
188 return -EAFNOSUPPORT;
189
190 if (unlikely(protocol != 0))
191 return -EPROTONOSUPPORT;
192
193 switch (sock->type) {
194 case SOCK_STREAM:
195 ops = &stream_ops;
196 state = SS_UNCONNECTED;
197 break;
198 case SOCK_SEQPACKET:
199 ops = &packet_ops;
200 state = SS_UNCONNECTED;
201 break;
202 case SOCK_DGRAM:
203 case SOCK_RDM:
204 ops = &msg_ops;
205 state = SS_READY;
206 break;
207 default:
208 return -EPROTOTYPE;
209 }
210
211 /* Allocate socket's protocol area */
212
213 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
214 if (sk == NULL)
215 return -ENOMEM;
216
217 /* Allocate TIPC port for socket to use */
218
219 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
220 TIPC_LOW_IMPORTANCE);
221 if (unlikely(!tp_ptr)) {
222 sk_free(sk);
223 return -ENOMEM;
224 }
225
226 /* Finish initializing socket data structures */
227
228 sock->ops = ops;
229 sock->state = state;
230
231 sock_init_data(sock, sk);
232 sk->sk_backlog_rcv = backlog_rcv;
233 tipc_sk(sk)->p = tp_ptr;
234 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
235
236 spin_unlock_bh(tp_ptr->lock);
237
238 if (sock->state == SS_READY) {
239 tipc_set_portunreturnable(tp_ptr->ref, 1);
240 if (sock->type == SOCK_DGRAM)
241 tipc_set_portunreliable(tp_ptr->ref, 1);
242 }
243
244 return 0;
245}
246
247/**
248 * release - destroy a TIPC socket
249 * @sock: socket to destroy
250 *
251 * This routine cleans up any messages that are still queued on the socket.
252 * For DGRAM and RDM socket types, all queued messages are rejected.
253 * For SEQPACKET and STREAM socket types, the first message is rejected
254 * and any others are discarded. (If the first message on a STREAM socket
255 * is partially-read, it is discarded and the next one is rejected instead.)
256 *
257 * NOTE: Rejected messages are not necessarily returned to the sender! They
258 * are returned or discarded according to the "destination droppable" setting
259 * specified for the message by the sender.
260 *
261 * Returns 0 on success, errno otherwise
262 */
263
264static int release(struct socket *sock)
265{
266 struct sock *sk = sock->sk;
267 struct tipc_port *tport;
268 struct sk_buff *buf;
269 int res;
270
271 /*
272 * Exit if socket isn't fully initialized (occurs when a failed accept()
273 * releases a pre-allocated child socket that was never used)
274 */
275
276 if (sk == NULL)
277 return 0;
278
279 tport = tipc_sk_port(sk);
280 lock_sock(sk);
281
282 /*
283 * Reject all unreceived messages, except on an active connection
284 * (which disconnects locally & sends a 'FIN+' to peer)
285 */
286
287 while (sock->state != SS_DISCONNECTING) {
288 buf = __skb_dequeue(&sk->sk_receive_queue);
289 if (buf == NULL)
290 break;
291 atomic_dec(&tipc_queue_size);
292 if (TIPC_SKB_CB(buf)->handle != 0)
293 buf_discard(buf);
294 else {
295 if ((sock->state == SS_CONNECTING) ||
296 (sock->state == SS_CONNECTED)) {
297 sock->state = SS_DISCONNECTING;
298 tipc_disconnect(tport->ref);
299 }
300 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
301 }
302 }
303
304 /*
305 * Delete TIPC port; this ensures no more messages are queued
306 * (also disconnects an active connection & sends a 'FIN-' to peer)
307 */
308
309 res = tipc_deleteport(tport->ref);
310
311 /* Discard any remaining (connection-based) messages in receive queue */
312
313 discard_rx_queue(sk);
314
315 /* Reject any messages that accumulated in backlog queue */
316
317 sock->state = SS_DISCONNECTING;
318 release_sock(sk);
319
320 sock_put(sk);
321 sock->sk = NULL;
322
323 return res;
324}
325
326/**
327 * bind - associate or disassocate TIPC name(s) with a socket
328 * @sock: socket structure
329 * @uaddr: socket address describing name(s) and desired operation
330 * @uaddr_len: size of socket address data structure
331 *
332 * Name and name sequence binding is indicated using a positive scope value;
333 * a negative scope value unbinds the specified name. Specifying no name
334 * (i.e. a socket address length of 0) unbinds all names from the socket.
335 *
336 * Returns 0 on success, errno otherwise
337 *
338 * NOTE: This routine doesn't need to take the socket lock since it doesn't
339 * access any non-constant socket information.
340 */
341
342static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
343{
344 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
345 u32 portref = tipc_sk_port(sock->sk)->ref;
346
347 if (unlikely(!uaddr_len))
348 return tipc_withdraw(portref, 0, NULL);
349
350 if (uaddr_len < sizeof(struct sockaddr_tipc))
351 return -EINVAL;
352 if (addr->family != AF_TIPC)
353 return -EAFNOSUPPORT;
354
355 if (addr->addrtype == TIPC_ADDR_NAME)
356 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
357 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
358 return -EAFNOSUPPORT;
359
360 return (addr->scope > 0) ?
361 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
362 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
363}
364
365/**
366 * get_name - get port ID of socket or peer socket
367 * @sock: socket structure
368 * @uaddr: area for returned socket address
369 * @uaddr_len: area for returned length of socket address
370 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
371 *
372 * Returns 0 on success, errno otherwise
373 *
374 * NOTE: This routine doesn't need to take the socket lock since it only
375 * accesses socket information that is unchanging (or which changes in
376 * a completely predictable manner).
377 */
378
379static int get_name(struct socket *sock, struct sockaddr *uaddr,
380 int *uaddr_len, int peer)
381{
382 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
383 struct tipc_sock *tsock = tipc_sk(sock->sk);
384
385 memset(addr, 0, sizeof(*addr));
386 if (peer) {
387 if ((sock->state != SS_CONNECTED) &&
388 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
389 return -ENOTCONN;
390 addr->addr.id.ref = tsock->peer_name.ref;
391 addr->addr.id.node = tsock->peer_name.node;
392 } else {
393 addr->addr.id.ref = tsock->p->ref;
394 addr->addr.id.node = tipc_own_addr;
395 }
396
397 *uaddr_len = sizeof(*addr);
398 addr->addrtype = TIPC_ADDR_ID;
399 addr->family = AF_TIPC;
400 addr->scope = 0;
401 addr->addr.name.domain = 0;
402
403 return 0;
404}
405
406/**
407 * poll - read and possibly block on pollmask
408 * @file: file structure associated with the socket
409 * @sock: socket for which to calculate the poll bits
410 * @wait: ???
411 *
412 * Returns pollmask value
413 *
414 * COMMENTARY:
415 * It appears that the usual socket locking mechanisms are not useful here
416 * since the pollmask info is potentially out-of-date the moment this routine
417 * exits. TCP and other protocols seem to rely on higher level poll routines
418 * to handle any preventable race conditions, so TIPC will do the same ...
419 *
420 * TIPC sets the returned events as follows:
421 *
422 * socket state flags set
423 * ------------ ---------
424 * unconnected no read flags
425 * no write flags
426 *
427 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
428 * no write flags
429 *
430 * connected POLLIN/POLLRDNORM if data in rx queue
431 * POLLOUT if port is not congested
432 *
433 * disconnecting POLLIN/POLLRDNORM/POLLHUP
434 * no write flags
435 *
436 * listening POLLIN if SYN in rx queue
437 * no write flags
438 *
439 * ready POLLIN/POLLRDNORM if data in rx queue
440 * [connectionless] POLLOUT (since port cannot be congested)
441 *
442 * IMPORTANT: The fact that a read or write operation is indicated does NOT
443 * imply that the operation will succeed, merely that it should be performed
444 * and will not block.
445 */
446
447static unsigned int poll(struct file *file, struct socket *sock,
448 poll_table *wait)
449{
450 struct sock *sk = sock->sk;
451 u32 mask = 0;
452
453 poll_wait(file, sk_sleep(sk), wait);
454
455 switch ((int)sock->state) {
456 case SS_READY:
457 case SS_CONNECTED:
458 if (!tipc_sk_port(sk)->congested)
459 mask |= POLLOUT;
460 /* fall thru' */
461 case SS_CONNECTING:
462 case SS_LISTENING:
463 if (!skb_queue_empty(&sk->sk_receive_queue))
464 mask |= (POLLIN | POLLRDNORM);
465 break;
466 case SS_DISCONNECTING:
467 mask = (POLLIN | POLLRDNORM | POLLHUP);
468 break;
469 }
470
471 return mask;
472}
473
474/**
475 * dest_name_check - verify user is permitted to send to specified port name
476 * @dest: destination address
477 * @m: descriptor for message to be sent
478 *
479 * Prevents restricted configuration commands from being issued by
480 * unauthorized users.
481 *
482 * Returns 0 if permission is granted, otherwise errno
483 */
484
485static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
486{
487 struct tipc_cfg_msg_hdr hdr;
488
489 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
490 return 0;
491 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
492 return 0;
493 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
494 return -EACCES;
495
496 if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
497 return -EMSGSIZE;
498 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
499 return -EFAULT;
500 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
501 return -EACCES;
502
503 return 0;
504}
505
506/**
507 * send_msg - send message in connectionless manner
508 * @iocb: if NULL, indicates that socket lock is already held
509 * @sock: socket structure
510 * @m: message to send
511 * @total_len: length of message
512 *
513 * Message must have an destination specified explicitly.
514 * Used for SOCK_RDM and SOCK_DGRAM messages,
515 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
516 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
517 *
518 * Returns the number of bytes sent on success, or errno otherwise
519 */
520
521static int send_msg(struct kiocb *iocb, struct socket *sock,
522 struct msghdr *m, size_t total_len)
523{
524 struct sock *sk = sock->sk;
525 struct tipc_port *tport = tipc_sk_port(sk);
526 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
527 int needs_conn;
528 int res = -EINVAL;
529
530 if (unlikely(!dest))
531 return -EDESTADDRREQ;
532 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
533 (dest->family != AF_TIPC)))
534 return -EINVAL;
535 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
536 (m->msg_iovlen > (unsigned)INT_MAX))
537 return -EMSGSIZE;
538
539 if (iocb)
540 lock_sock(sk);
541
542 needs_conn = (sock->state != SS_READY);
543 if (unlikely(needs_conn)) {
544 if (sock->state == SS_LISTENING) {
545 res = -EPIPE;
546 goto exit;
547 }
548 if (sock->state != SS_UNCONNECTED) {
549 res = -EISCONN;
550 goto exit;
551 }
552 if ((tport->published) ||
553 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
554 res = -EOPNOTSUPP;
555 goto exit;
556 }
557 if (dest->addrtype == TIPC_ADDR_NAME) {
558 tport->conn_type = dest->addr.name.name.type;
559 tport->conn_instance = dest->addr.name.name.instance;
560 }
561
562 /* Abort any pending connection attempts (very unlikely) */
563
564 reject_rx_queue(sk);
565 }
566
567 do {
568 if (dest->addrtype == TIPC_ADDR_NAME) {
569 res = dest_name_check(dest, m);
570 if (res)
571 break;
572 res = tipc_send2name(tport->ref,
573 &dest->addr.name.name,
574 dest->addr.name.domain,
575 m->msg_iovlen,
576 m->msg_iov,
577 total_len);
578 } else if (dest->addrtype == TIPC_ADDR_ID) {
579 res = tipc_send2port(tport->ref,
580 &dest->addr.id,
581 m->msg_iovlen,
582 m->msg_iov,
583 total_len);
584 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
585 if (needs_conn) {
586 res = -EOPNOTSUPP;
587 break;
588 }
589 res = dest_name_check(dest, m);
590 if (res)
591 break;
592 res = tipc_multicast(tport->ref,
593 &dest->addr.nameseq,
594 m->msg_iovlen,
595 m->msg_iov,
596 total_len);
597 }
598 if (likely(res != -ELINKCONG)) {
599 if (needs_conn && (res >= 0))
600 sock->state = SS_CONNECTING;
601 break;
602 }
603 if (m->msg_flags & MSG_DONTWAIT) {
604 res = -EWOULDBLOCK;
605 break;
606 }
607 release_sock(sk);
608 res = wait_event_interruptible(*sk_sleep(sk),
609 !tport->congested);
610 lock_sock(sk);
611 if (res)
612 break;
613 } while (1);
614
615exit:
616 if (iocb)
617 release_sock(sk);
618 return res;
619}
620
621/**
622 * send_packet - send a connection-oriented message
623 * @iocb: if NULL, indicates that socket lock is already held
624 * @sock: socket structure
625 * @m: message to send
626 * @total_len: length of message
627 *
628 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
629 *
630 * Returns the number of bytes sent on success, or errno otherwise
631 */
632
633static int send_packet(struct kiocb *iocb, struct socket *sock,
634 struct msghdr *m, size_t total_len)
635{
636 struct sock *sk = sock->sk;
637 struct tipc_port *tport = tipc_sk_port(sk);
638 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
639 int res;
640
641 /* Handle implied connection establishment */
642
643 if (unlikely(dest))
644 return send_msg(iocb, sock, m, total_len);
645
646 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
647 (m->msg_iovlen > (unsigned)INT_MAX))
648 return -EMSGSIZE;
649
650 if (iocb)
651 lock_sock(sk);
652
653 do {
654 if (unlikely(sock->state != SS_CONNECTED)) {
655 if (sock->state == SS_DISCONNECTING)
656 res = -EPIPE;
657 else
658 res = -ENOTCONN;
659 break;
660 }
661
662 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
663 total_len);
664 if (likely(res != -ELINKCONG))
665 break;
666 if (m->msg_flags & MSG_DONTWAIT) {
667 res = -EWOULDBLOCK;
668 break;
669 }
670 release_sock(sk);
671 res = wait_event_interruptible(*sk_sleep(sk),
672 (!tport->congested || !tport->connected));
673 lock_sock(sk);
674 if (res)
675 break;
676 } while (1);
677
678 if (iocb)
679 release_sock(sk);
680 return res;
681}
682
683/**
684 * send_stream - send stream-oriented data
685 * @iocb: (unused)
686 * @sock: socket structure
687 * @m: data to send
688 * @total_len: total length of data to be sent
689 *
690 * Used for SOCK_STREAM data.
691 *
692 * Returns the number of bytes sent on success (or partial success),
693 * or errno if no data sent
694 */
695
696static int send_stream(struct kiocb *iocb, struct socket *sock,
697 struct msghdr *m, size_t total_len)
698{
699 struct sock *sk = sock->sk;
700 struct tipc_port *tport = tipc_sk_port(sk);
701 struct msghdr my_msg;
702 struct iovec my_iov;
703 struct iovec *curr_iov;
704 int curr_iovlen;
705 char __user *curr_start;
706 u32 hdr_size;
707 int curr_left;
708 int bytes_to_send;
709 int bytes_sent;
710 int res;
711
712 lock_sock(sk);
713
714 /* Handle special cases where there is no connection */
715
716 if (unlikely(sock->state != SS_CONNECTED)) {
717 if (sock->state == SS_UNCONNECTED) {
718 res = send_packet(NULL, sock, m, total_len);
719 goto exit;
720 } else if (sock->state == SS_DISCONNECTING) {
721 res = -EPIPE;
722 goto exit;
723 } else {
724 res = -ENOTCONN;
725 goto exit;
726 }
727 }
728
729 if (unlikely(m->msg_name)) {
730 res = -EISCONN;
731 goto exit;
732 }
733
734 if ((total_len > (unsigned)INT_MAX) ||
735 (m->msg_iovlen > (unsigned)INT_MAX)) {
736 res = -EMSGSIZE;
737 goto exit;
738 }
739
740 /*
741 * Send each iovec entry using one or more messages
742 *
743 * Note: This algorithm is good for the most likely case
744 * (i.e. one large iovec entry), but could be improved to pass sets
745 * of small iovec entries into send_packet().
746 */
747
748 curr_iov = m->msg_iov;
749 curr_iovlen = m->msg_iovlen;
750 my_msg.msg_iov = &my_iov;
751 my_msg.msg_iovlen = 1;
752 my_msg.msg_flags = m->msg_flags;
753 my_msg.msg_name = NULL;
754 bytes_sent = 0;
755
756 hdr_size = msg_hdr_sz(&tport->phdr);
757
758 while (curr_iovlen--) {
759 curr_start = curr_iov->iov_base;
760 curr_left = curr_iov->iov_len;
761
762 while (curr_left) {
763 bytes_to_send = tport->max_pkt - hdr_size;
764 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
765 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
766 if (curr_left < bytes_to_send)
767 bytes_to_send = curr_left;
768 my_iov.iov_base = curr_start;
769 my_iov.iov_len = bytes_to_send;
770 res = send_packet(NULL, sock, &my_msg, bytes_to_send);
771 if (res < 0) {
772 if (bytes_sent)
773 res = bytes_sent;
774 goto exit;
775 }
776 curr_left -= bytes_to_send;
777 curr_start += bytes_to_send;
778 bytes_sent += bytes_to_send;
779 }
780
781 curr_iov++;
782 }
783 res = bytes_sent;
784exit:
785 release_sock(sk);
786 return res;
787}
788
789/**
790 * auto_connect - complete connection setup to a remote port
791 * @sock: socket structure
792 * @msg: peer's response message
793 *
794 * Returns 0 on success, errno otherwise
795 */
796
797static int auto_connect(struct socket *sock, struct tipc_msg *msg)
798{
799 struct tipc_sock *tsock = tipc_sk(sock->sk);
800
801 if (msg_errcode(msg)) {
802 sock->state = SS_DISCONNECTING;
803 return -ECONNREFUSED;
804 }
805
806 tsock->peer_name.ref = msg_origport(msg);
807 tsock->peer_name.node = msg_orignode(msg);
808 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
809 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
810 sock->state = SS_CONNECTED;
811 return 0;
812}
813
814/**
815 * set_orig_addr - capture sender's address for received message
816 * @m: descriptor for message info
817 * @msg: received message header
818 *
819 * Note: Address is not captured if not requested by receiver.
820 */
821
822static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
823{
824 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
825
826 if (addr) {
827 addr->family = AF_TIPC;
828 addr->addrtype = TIPC_ADDR_ID;
829 addr->addr.id.ref = msg_origport(msg);
830 addr->addr.id.node = msg_orignode(msg);
831 addr->addr.name.domain = 0; /* could leave uninitialized */
832 addr->scope = 0; /* could leave uninitialized */
833 m->msg_namelen = sizeof(struct sockaddr_tipc);
834 }
835}
836
837/**
838 * anc_data_recv - optionally capture ancillary data for received message
839 * @m: descriptor for message info
840 * @msg: received message header
841 * @tport: TIPC port associated with message
842 *
843 * Note: Ancillary data is not captured if not requested by receiver.
844 *
845 * Returns 0 if successful, otherwise errno
846 */
847
848static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
849 struct tipc_port *tport)
850{
851 u32 anc_data[3];
852 u32 err;
853 u32 dest_type;
854 int has_name;
855 int res;
856
857 if (likely(m->msg_controllen == 0))
858 return 0;
859
860 /* Optionally capture errored message object(s) */
861
862 err = msg ? msg_errcode(msg) : 0;
863 if (unlikely(err)) {
864 anc_data[0] = err;
865 anc_data[1] = msg_data_sz(msg);
866 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
867 if (res)
868 return res;
869 if (anc_data[1]) {
870 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
871 msg_data(msg));
872 if (res)
873 return res;
874 }
875 }
876
877 /* Optionally capture message destination object */
878
879 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
880 switch (dest_type) {
881 case TIPC_NAMED_MSG:
882 has_name = 1;
883 anc_data[0] = msg_nametype(msg);
884 anc_data[1] = msg_namelower(msg);
885 anc_data[2] = msg_namelower(msg);
886 break;
887 case TIPC_MCAST_MSG:
888 has_name = 1;
889 anc_data[0] = msg_nametype(msg);
890 anc_data[1] = msg_namelower(msg);
891 anc_data[2] = msg_nameupper(msg);
892 break;
893 case TIPC_CONN_MSG:
894 has_name = (tport->conn_type != 0);
895 anc_data[0] = tport->conn_type;
896 anc_data[1] = tport->conn_instance;
897 anc_data[2] = tport->conn_instance;
898 break;
899 default:
900 has_name = 0;
901 }
902 if (has_name) {
903 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
904 if (res)
905 return res;
906 }
907
908 return 0;
909}
910
911/**
912 * recv_msg - receive packet-oriented message
913 * @iocb: (unused)
914 * @m: descriptor for message info
915 * @buf_len: total size of user buffer area
916 * @flags: receive flags
917 *
918 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
919 * If the complete message doesn't fit in user area, truncate it.
920 *
921 * Returns size of returned message data, errno otherwise
922 */
923
924static int recv_msg(struct kiocb *iocb, struct socket *sock,
925 struct msghdr *m, size_t buf_len, int flags)
926{
927 struct sock *sk = sock->sk;
928 struct tipc_port *tport = tipc_sk_port(sk);
929 struct sk_buff *buf;
930 struct tipc_msg *msg;
931 long timeout;
932 unsigned int sz;
933 u32 err;
934 int res;
935
936 /* Catch invalid receive requests */
937
938 if (unlikely(!buf_len))
939 return -EINVAL;
940
941 lock_sock(sk);
942
943 if (unlikely(sock->state == SS_UNCONNECTED)) {
944 res = -ENOTCONN;
945 goto exit;
946 }
947
948 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
949restart:
950
951 /* Look for a message in receive queue; wait if necessary */
952
953 while (skb_queue_empty(&sk->sk_receive_queue)) {
954 if (sock->state == SS_DISCONNECTING) {
955 res = -ENOTCONN;
956 goto exit;
957 }
958 if (timeout <= 0L) {
959 res = timeout ? timeout : -EWOULDBLOCK;
960 goto exit;
961 }
962 release_sock(sk);
963 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
964 tipc_rx_ready(sock),
965 timeout);
966 lock_sock(sk);
967 }
968
969 /* Look at first message in receive queue */
970
971 buf = skb_peek(&sk->sk_receive_queue);
972 msg = buf_msg(buf);
973 sz = msg_data_sz(msg);
974 err = msg_errcode(msg);
975
976 /* Complete connection setup for an implied connect */
977
978 if (unlikely(sock->state == SS_CONNECTING)) {
979 res = auto_connect(sock, msg);
980 if (res)
981 goto exit;
982 }
983
984 /* Discard an empty non-errored message & try again */
985
986 if ((!sz) && (!err)) {
987 advance_rx_queue(sk);
988 goto restart;
989 }
990
991 /* Capture sender's address (optional) */
992
993 set_orig_addr(m, msg);
994
995 /* Capture ancillary data (optional) */
996
997 res = anc_data_recv(m, msg, tport);
998 if (res)
999 goto exit;
1000
1001 /* Capture message data (if valid) & compute return value (always) */
1002
1003 if (!err) {
1004 if (unlikely(buf_len < sz)) {
1005 sz = buf_len;
1006 m->msg_flags |= MSG_TRUNC;
1007 }
1008 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1009 m->msg_iov, sz);
1010 if (res)
1011 goto exit;
1012 res = sz;
1013 } else {
1014 if ((sock->state == SS_READY) ||
1015 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1016 res = 0;
1017 else
1018 res = -ECONNRESET;
1019 }
1020
1021 /* Consume received message (optional) */
1022
1023 if (likely(!(flags & MSG_PEEK))) {
1024 if ((sock->state != SS_READY) &&
1025 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1026 tipc_acknowledge(tport->ref, tport->conn_unacked);
1027 advance_rx_queue(sk);
1028 }
1029exit:
1030 release_sock(sk);
1031 return res;
1032}
1033
1034/**
1035 * recv_stream - receive stream-oriented data
1036 * @iocb: (unused)
1037 * @m: descriptor for message info
1038 * @buf_len: total size of user buffer area
1039 * @flags: receive flags
1040 *
1041 * Used for SOCK_STREAM messages only. If not enough data is available
1042 * will optionally wait for more; never truncates data.
1043 *
1044 * Returns size of returned message data, errno otherwise
1045 */
1046
1047static int recv_stream(struct kiocb *iocb, struct socket *sock,
1048 struct msghdr *m, size_t buf_len, int flags)
1049{
1050 struct sock *sk = sock->sk;
1051 struct tipc_port *tport = tipc_sk_port(sk);
1052 struct sk_buff *buf;
1053 struct tipc_msg *msg;
1054 long timeout;
1055 unsigned int sz;
1056 int sz_to_copy, target, needed;
1057 int sz_copied = 0;
1058 u32 err;
1059 int res = 0;
1060
1061 /* Catch invalid receive attempts */
1062
1063 if (unlikely(!buf_len))
1064 return -EINVAL;
1065
1066 lock_sock(sk);
1067
1068 if (unlikely((sock->state == SS_UNCONNECTED) ||
1069 (sock->state == SS_CONNECTING))) {
1070 res = -ENOTCONN;
1071 goto exit;
1072 }
1073
1074 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1075 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1076restart:
1077
1078 /* Look for a message in receive queue; wait if necessary */
1079
1080 while (skb_queue_empty(&sk->sk_receive_queue)) {
1081 if (sock->state == SS_DISCONNECTING) {
1082 res = -ENOTCONN;
1083 goto exit;
1084 }
1085 if (timeout <= 0L) {
1086 res = timeout ? timeout : -EWOULDBLOCK;
1087 goto exit;
1088 }
1089 release_sock(sk);
1090 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1091 tipc_rx_ready(sock),
1092 timeout);
1093 lock_sock(sk);
1094 }
1095
1096 /* Look at first message in receive queue */
1097
1098 buf = skb_peek(&sk->sk_receive_queue);
1099 msg = buf_msg(buf);
1100 sz = msg_data_sz(msg);
1101 err = msg_errcode(msg);
1102
1103 /* Discard an empty non-errored message & try again */
1104
1105 if ((!sz) && (!err)) {
1106 advance_rx_queue(sk);
1107 goto restart;
1108 }
1109
1110 /* Optionally capture sender's address & ancillary data of first msg */
1111
1112 if (sz_copied == 0) {
1113 set_orig_addr(m, msg);
1114 res = anc_data_recv(m, msg, tport);
1115 if (res)
1116 goto exit;
1117 }
1118
1119 /* Capture message data (if valid) & compute return value (always) */
1120
1121 if (!err) {
1122 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1123
1124 sz -= offset;
1125 needed = (buf_len - sz_copied);
1126 sz_to_copy = (sz <= needed) ? sz : needed;
1127
1128 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1129 m->msg_iov, sz_to_copy);
1130 if (res)
1131 goto exit;
1132
1133 sz_copied += sz_to_copy;
1134
1135 if (sz_to_copy < sz) {
1136 if (!(flags & MSG_PEEK))
1137 TIPC_SKB_CB(buf)->handle =
1138 (void *)(unsigned long)(offset + sz_to_copy);
1139 goto exit;
1140 }
1141 } else {
1142 if (sz_copied != 0)
1143 goto exit; /* can't add error msg to valid data */
1144
1145 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1146 res = 0;
1147 else
1148 res = -ECONNRESET;
1149 }
1150
1151 /* Consume received message (optional) */
1152
1153 if (likely(!(flags & MSG_PEEK))) {
1154 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1155 tipc_acknowledge(tport->ref, tport->conn_unacked);
1156 advance_rx_queue(sk);
1157 }
1158
1159 /* Loop around if more data is required */
1160
1161 if ((sz_copied < buf_len) && /* didn't get all requested data */
1162 (!skb_queue_empty(&sk->sk_receive_queue) ||
1163 (sz_copied < target)) && /* and more is ready or required */
1164 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1165 (!err)) /* and haven't reached a FIN */
1166 goto restart;
1167
1168exit:
1169 release_sock(sk);
1170 return sz_copied ? sz_copied : res;
1171}
1172
1173/**
1174 * rx_queue_full - determine if receive queue can accept another message
1175 * @msg: message to be added to queue
1176 * @queue_size: current size of queue
1177 * @base: nominal maximum size of queue
1178 *
1179 * Returns 1 if queue is unable to accept message, 0 otherwise
1180 */
1181
1182static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1183{
1184 u32 threshold;
1185 u32 imp = msg_importance(msg);
1186
1187 if (imp == TIPC_LOW_IMPORTANCE)
1188 threshold = base;
1189 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1190 threshold = base * 2;
1191 else if (imp == TIPC_HIGH_IMPORTANCE)
1192 threshold = base * 100;
1193 else
1194 return 0;
1195
1196 if (msg_connected(msg))
1197 threshold *= 4;
1198
1199 return queue_size >= threshold;
1200}
1201
1202/**
1203 * filter_rcv - validate incoming message
1204 * @sk: socket
1205 * @buf: message
1206 *
1207 * Enqueues message on receive queue if acceptable; optionally handles
1208 * disconnect indication for a connected socket.
1209 *
1210 * Called with socket lock already taken; port lock may also be taken.
1211 *
1212 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1213 */
1214
1215static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1216{
1217 struct socket *sock = sk->sk_socket;
1218 struct tipc_msg *msg = buf_msg(buf);
1219 u32 recv_q_len;
1220
1221 /* Reject message if it is wrong sort of message for socket */
1222
1223 /*
1224 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1225 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1226 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1227 */
1228
1229 if (sock->state == SS_READY) {
1230 if (msg_connected(msg))
1231 return TIPC_ERR_NO_PORT;
1232 } else {
1233 if (msg_mcast(msg))
1234 return TIPC_ERR_NO_PORT;
1235 if (sock->state == SS_CONNECTED) {
1236 if (!msg_connected(msg))
1237 return TIPC_ERR_NO_PORT;
1238 } else if (sock->state == SS_CONNECTING) {
1239 if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1240 return TIPC_ERR_NO_PORT;
1241 } else if (sock->state == SS_LISTENING) {
1242 if (msg_connected(msg) || msg_errcode(msg))
1243 return TIPC_ERR_NO_PORT;
1244 } else if (sock->state == SS_DISCONNECTING) {
1245 return TIPC_ERR_NO_PORT;
1246 } else /* (sock->state == SS_UNCONNECTED) */ {
1247 if (msg_connected(msg) || msg_errcode(msg))
1248 return TIPC_ERR_NO_PORT;
1249 }
1250 }
1251
1252 /* Reject message if there isn't room to queue it */
1253
1254 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1255 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1256 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1257 return TIPC_ERR_OVERLOAD;
1258 }
1259 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1260 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1261 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1262 return TIPC_ERR_OVERLOAD;
1263 }
1264
1265 /* Enqueue message (finally!) */
1266
1267 TIPC_SKB_CB(buf)->handle = 0;
1268 atomic_inc(&tipc_queue_size);
1269 __skb_queue_tail(&sk->sk_receive_queue, buf);
1270
1271 /* Initiate connection termination for an incoming 'FIN' */
1272
1273 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1274 sock->state = SS_DISCONNECTING;
1275 tipc_disconnect_port(tipc_sk_port(sk));
1276 }
1277
1278 if (waitqueue_active(sk_sleep(sk)))
1279 wake_up_interruptible(sk_sleep(sk));
1280 return TIPC_OK;
1281}
1282
1283/**
1284 * backlog_rcv - handle incoming message from backlog queue
1285 * @sk: socket
1286 * @buf: message
1287 *
1288 * Caller must hold socket lock, but not port lock.
1289 *
1290 * Returns 0
1291 */
1292
1293static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1294{
1295 u32 res;
1296
1297 res = filter_rcv(sk, buf);
1298 if (res)
1299 tipc_reject_msg(buf, res);
1300 return 0;
1301}
1302
1303/**
1304 * dispatch - handle incoming message
1305 * @tport: TIPC port that received message
1306 * @buf: message
1307 *
1308 * Called with port lock already taken.
1309 *
1310 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1311 */
1312
1313static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1314{
1315 struct sock *sk = (struct sock *)tport->usr_handle;
1316 u32 res;
1317
1318 /*
1319 * Process message if socket is unlocked; otherwise add to backlog queue
1320 *
1321 * This code is based on sk_receive_skb(), but must be distinct from it
1322 * since a TIPC-specific filter/reject mechanism is utilized
1323 */
1324
1325 bh_lock_sock(sk);
1326 if (!sock_owned_by_user(sk)) {
1327 res = filter_rcv(sk, buf);
1328 } else {
1329 if (sk_add_backlog(sk, buf))
1330 res = TIPC_ERR_OVERLOAD;
1331 else
1332 res = TIPC_OK;
1333 }
1334 bh_unlock_sock(sk);
1335
1336 return res;
1337}
1338
1339/**
1340 * wakeupdispatch - wake up port after congestion
1341 * @tport: port to wakeup
1342 *
1343 * Called with port lock already taken.
1344 */
1345
1346static void wakeupdispatch(struct tipc_port *tport)
1347{
1348 struct sock *sk = (struct sock *)tport->usr_handle;
1349
1350 if (waitqueue_active(sk_sleep(sk)))
1351 wake_up_interruptible(sk_sleep(sk));
1352}
1353
1354/**
1355 * connect - establish a connection to another TIPC port
1356 * @sock: socket structure
1357 * @dest: socket address for destination port
1358 * @destlen: size of socket address data structure
1359 * @flags: file-related flags associated with socket
1360 *
1361 * Returns 0 on success, errno otherwise
1362 */
1363
1364static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1365 int flags)
1366{
1367 struct sock *sk = sock->sk;
1368 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1369 struct msghdr m = {NULL,};
1370 struct sk_buff *buf;
1371 struct tipc_msg *msg;
1372 long timeout;
1373 int res;
1374
1375 lock_sock(sk);
1376
1377 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1378
1379 if (sock->state == SS_READY) {
1380 res = -EOPNOTSUPP;
1381 goto exit;
1382 }
1383
1384 /* For now, TIPC does not support the non-blocking form of connect() */
1385
1386 if (flags & O_NONBLOCK) {
1387 res = -EOPNOTSUPP;
1388 goto exit;
1389 }
1390
1391 /* Issue Posix-compliant error code if socket is in the wrong state */
1392
1393 if (sock->state == SS_LISTENING) {
1394 res = -EOPNOTSUPP;
1395 goto exit;
1396 }
1397 if (sock->state == SS_CONNECTING) {
1398 res = -EALREADY;
1399 goto exit;
1400 }
1401 if (sock->state != SS_UNCONNECTED) {
1402 res = -EISCONN;
1403 goto exit;
1404 }
1405
1406 /*
1407 * Reject connection attempt using multicast address
1408 *
1409 * Note: send_msg() validates the rest of the address fields,
1410 * so there's no need to do it here
1411 */
1412
1413 if (dst->addrtype == TIPC_ADDR_MCAST) {
1414 res = -EINVAL;
1415 goto exit;
1416 }
1417
1418 /* Reject any messages already in receive queue (very unlikely) */
1419
1420 reject_rx_queue(sk);
1421
1422 /* Send a 'SYN-' to destination */
1423
1424 m.msg_name = dest;
1425 m.msg_namelen = destlen;
1426 res = send_msg(NULL, sock, &m, 0);
1427 if (res < 0)
1428 goto exit;
1429
1430 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1431
1432 timeout = tipc_sk(sk)->conn_timeout;
1433 release_sock(sk);
1434 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1435 (!skb_queue_empty(&sk->sk_receive_queue) ||
1436 (sock->state != SS_CONNECTING)),
1437 timeout ? timeout : MAX_SCHEDULE_TIMEOUT);
1438 lock_sock(sk);
1439
1440 if (res > 0) {
1441 buf = skb_peek(&sk->sk_receive_queue);
1442 if (buf != NULL) {
1443 msg = buf_msg(buf);
1444 res = auto_connect(sock, msg);
1445 if (!res) {
1446 if (!msg_data_sz(msg))
1447 advance_rx_queue(sk);
1448 }
1449 } else {
1450 if (sock->state == SS_CONNECTED)
1451 res = -EISCONN;
1452 else
1453 res = -ECONNREFUSED;
1454 }
1455 } else {
1456 if (res == 0)
1457 res = -ETIMEDOUT;
1458 else
1459 ; /* leave "res" unchanged */
1460 sock->state = SS_DISCONNECTING;
1461 }
1462
1463exit:
1464 release_sock(sk);
1465 return res;
1466}
1467
1468/**
1469 * listen - allow socket to listen for incoming connections
1470 * @sock: socket structure
1471 * @len: (unused)
1472 *
1473 * Returns 0 on success, errno otherwise
1474 */
1475
1476static int listen(struct socket *sock, int len)
1477{
1478 struct sock *sk = sock->sk;
1479 int res;
1480
1481 lock_sock(sk);
1482
1483 if (sock->state == SS_READY)
1484 res = -EOPNOTSUPP;
1485 else if (sock->state != SS_UNCONNECTED)
1486 res = -EINVAL;
1487 else {
1488 sock->state = SS_LISTENING;
1489 res = 0;
1490 }
1491
1492 release_sock(sk);
1493 return res;
1494}
1495
1496/**
1497 * accept - wait for connection request
1498 * @sock: listening socket
1499 * @newsock: new socket that is to be connected
1500 * @flags: file-related flags associated with socket
1501 *
1502 * Returns 0 on success, errno otherwise
1503 */
1504
1505static int accept(struct socket *sock, struct socket *new_sock, int flags)
1506{
1507 struct sock *sk = sock->sk;
1508 struct sk_buff *buf;
1509 int res;
1510
1511 lock_sock(sk);
1512
1513 if (sock->state == SS_READY) {
1514 res = -EOPNOTSUPP;
1515 goto exit;
1516 }
1517 if (sock->state != SS_LISTENING) {
1518 res = -EINVAL;
1519 goto exit;
1520 }
1521
1522 while (skb_queue_empty(&sk->sk_receive_queue)) {
1523 if (flags & O_NONBLOCK) {
1524 res = -EWOULDBLOCK;
1525 goto exit;
1526 }
1527 release_sock(sk);
1528 res = wait_event_interruptible(*sk_sleep(sk),
1529 (!skb_queue_empty(&sk->sk_receive_queue)));
1530 lock_sock(sk);
1531 if (res)
1532 goto exit;
1533 }
1534
1535 buf = skb_peek(&sk->sk_receive_queue);
1536
1537 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1538 if (!res) {
1539 struct sock *new_sk = new_sock->sk;
1540 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1541 struct tipc_port *new_tport = new_tsock->p;
1542 u32 new_ref = new_tport->ref;
1543 struct tipc_msg *msg = buf_msg(buf);
1544
1545 lock_sock(new_sk);
1546
1547 /*
1548 * Reject any stray messages received by new socket
1549 * before the socket lock was taken (very, very unlikely)
1550 */
1551
1552 reject_rx_queue(new_sk);
1553
1554 /* Connect new socket to it's peer */
1555
1556 new_tsock->peer_name.ref = msg_origport(msg);
1557 new_tsock->peer_name.node = msg_orignode(msg);
1558 tipc_connect2port(new_ref, &new_tsock->peer_name);
1559 new_sock->state = SS_CONNECTED;
1560
1561 tipc_set_portimportance(new_ref, msg_importance(msg));
1562 if (msg_named(msg)) {
1563 new_tport->conn_type = msg_nametype(msg);
1564 new_tport->conn_instance = msg_nameinst(msg);
1565 }
1566
1567 /*
1568 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1569 * Respond to 'SYN+' by queuing it on new socket.
1570 */
1571
1572 if (!msg_data_sz(msg)) {
1573 struct msghdr m = {NULL,};
1574
1575 advance_rx_queue(sk);
1576 send_packet(NULL, new_sock, &m, 0);
1577 } else {
1578 __skb_dequeue(&sk->sk_receive_queue);
1579 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1580 }
1581 release_sock(new_sk);
1582 }
1583exit:
1584 release_sock(sk);
1585 return res;
1586}
1587
1588/**
1589 * shutdown - shutdown socket connection
1590 * @sock: socket structure
1591 * @how: direction to close (must be SHUT_RDWR)
1592 *
1593 * Terminates connection (if necessary), then purges socket's receive queue.
1594 *
1595 * Returns 0 on success, errno otherwise
1596 */
1597
1598static int shutdown(struct socket *sock, int how)
1599{
1600 struct sock *sk = sock->sk;
1601 struct tipc_port *tport = tipc_sk_port(sk);
1602 struct sk_buff *buf;
1603 int res;
1604
1605 if (how != SHUT_RDWR)
1606 return -EINVAL;
1607
1608 lock_sock(sk);
1609
1610 switch (sock->state) {
1611 case SS_CONNECTING:
1612 case SS_CONNECTED:
1613
1614 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1615restart:
1616 buf = __skb_dequeue(&sk->sk_receive_queue);
1617 if (buf) {
1618 atomic_dec(&tipc_queue_size);
1619 if (TIPC_SKB_CB(buf)->handle != 0) {
1620 buf_discard(buf);
1621 goto restart;
1622 }
1623 tipc_disconnect(tport->ref);
1624 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1625 } else {
1626 tipc_shutdown(tport->ref);
1627 }
1628
1629 sock->state = SS_DISCONNECTING;
1630
1631 /* fall through */
1632
1633 case SS_DISCONNECTING:
1634
1635 /* Discard any unreceived messages; wake up sleeping tasks */
1636
1637 discard_rx_queue(sk);
1638 if (waitqueue_active(sk_sleep(sk)))
1639 wake_up_interruptible(sk_sleep(sk));
1640 res = 0;
1641 break;
1642
1643 default:
1644 res = -ENOTCONN;
1645 }
1646
1647 release_sock(sk);
1648 return res;
1649}
1650
1651/**
1652 * setsockopt - set socket option
1653 * @sock: socket structure
1654 * @lvl: option level
1655 * @opt: option identifier
1656 * @ov: pointer to new option value
1657 * @ol: length of option value
1658 *
1659 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1660 * (to ease compatibility).
1661 *
1662 * Returns 0 on success, errno otherwise
1663 */
1664
1665static int setsockopt(struct socket *sock,
1666 int lvl, int opt, char __user *ov, unsigned int ol)
1667{
1668 struct sock *sk = sock->sk;
1669 struct tipc_port *tport = tipc_sk_port(sk);
1670 u32 value;
1671 int res;
1672
1673 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1674 return 0;
1675 if (lvl != SOL_TIPC)
1676 return -ENOPROTOOPT;
1677 if (ol < sizeof(value))
1678 return -EINVAL;
1679 res = get_user(value, (u32 __user *)ov);
1680 if (res)
1681 return res;
1682
1683 lock_sock(sk);
1684
1685 switch (opt) {
1686 case TIPC_IMPORTANCE:
1687 res = tipc_set_portimportance(tport->ref, value);
1688 break;
1689 case TIPC_SRC_DROPPABLE:
1690 if (sock->type != SOCK_STREAM)
1691 res = tipc_set_portunreliable(tport->ref, value);
1692 else
1693 res = -ENOPROTOOPT;
1694 break;
1695 case TIPC_DEST_DROPPABLE:
1696 res = tipc_set_portunreturnable(tport->ref, value);
1697 break;
1698 case TIPC_CONN_TIMEOUT:
1699 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value);
1700 /* no need to set "res", since already 0 at this point */
1701 break;
1702 default:
1703 res = -EINVAL;
1704 }
1705
1706 release_sock(sk);
1707
1708 return res;
1709}
1710
1711/**
1712 * getsockopt - get socket option
1713 * @sock: socket structure
1714 * @lvl: option level
1715 * @opt: option identifier
1716 * @ov: receptacle for option value
1717 * @ol: receptacle for length of option value
1718 *
1719 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1720 * (to ease compatibility).
1721 *
1722 * Returns 0 on success, errno otherwise
1723 */
1724
1725static int getsockopt(struct socket *sock,
1726 int lvl, int opt, char __user *ov, int __user *ol)
1727{
1728 struct sock *sk = sock->sk;
1729 struct tipc_port *tport = tipc_sk_port(sk);
1730 int len;
1731 u32 value;
1732 int res;
1733
1734 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1735 return put_user(0, ol);
1736 if (lvl != SOL_TIPC)
1737 return -ENOPROTOOPT;
1738 res = get_user(len, ol);
1739 if (res)
1740 return res;
1741
1742 lock_sock(sk);
1743
1744 switch (opt) {
1745 case TIPC_IMPORTANCE:
1746 res = tipc_portimportance(tport->ref, &value);
1747 break;
1748 case TIPC_SRC_DROPPABLE:
1749 res = tipc_portunreliable(tport->ref, &value);
1750 break;
1751 case TIPC_DEST_DROPPABLE:
1752 res = tipc_portunreturnable(tport->ref, &value);
1753 break;
1754 case TIPC_CONN_TIMEOUT:
1755 value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout);
1756 /* no need to set "res", since already 0 at this point */
1757 break;
1758 case TIPC_NODE_RECVQ_DEPTH:
1759 value = (u32)atomic_read(&tipc_queue_size);
1760 break;
1761 case TIPC_SOCK_RECVQ_DEPTH:
1762 value = skb_queue_len(&sk->sk_receive_queue);
1763 break;
1764 default:
1765 res = -EINVAL;
1766 }
1767
1768 release_sock(sk);
1769
1770 if (res)
1771 return res; /* "get" failed */
1772
1773 if (len < sizeof(value))
1774 return -EINVAL;
1775
1776 if (copy_to_user(ov, &value, sizeof(value)))
1777 return -EFAULT;
1778
1779 return put_user(sizeof(value), ol);
1780}
1781
1782/**
1783 * Protocol switches for the various types of TIPC sockets
1784 */
1785
1786static const struct proto_ops msg_ops = {
1787 .owner = THIS_MODULE,
1788 .family = AF_TIPC,
1789 .release = release,
1790 .bind = bind,
1791 .connect = connect,
1792 .socketpair = sock_no_socketpair,
1793 .accept = accept,
1794 .getname = get_name,
1795 .poll = poll,
1796 .ioctl = sock_no_ioctl,
1797 .listen = listen,
1798 .shutdown = shutdown,
1799 .setsockopt = setsockopt,
1800 .getsockopt = getsockopt,
1801 .sendmsg = send_msg,
1802 .recvmsg = recv_msg,
1803 .mmap = sock_no_mmap,
1804 .sendpage = sock_no_sendpage
1805};
1806
1807static const struct proto_ops packet_ops = {
1808 .owner = THIS_MODULE,
1809 .family = AF_TIPC,
1810 .release = release,
1811 .bind = bind,
1812 .connect = connect,
1813 .socketpair = sock_no_socketpair,
1814 .accept = accept,
1815 .getname = get_name,
1816 .poll = poll,
1817 .ioctl = sock_no_ioctl,
1818 .listen = listen,
1819 .shutdown = shutdown,
1820 .setsockopt = setsockopt,
1821 .getsockopt = getsockopt,
1822 .sendmsg = send_packet,
1823 .recvmsg = recv_msg,
1824 .mmap = sock_no_mmap,
1825 .sendpage = sock_no_sendpage
1826};
1827
1828static const struct proto_ops stream_ops = {
1829 .owner = THIS_MODULE,
1830 .family = AF_TIPC,
1831 .release = release,
1832 .bind = bind,
1833 .connect = connect,
1834 .socketpair = sock_no_socketpair,
1835 .accept = accept,
1836 .getname = get_name,
1837 .poll = poll,
1838 .ioctl = sock_no_ioctl,
1839 .listen = listen,
1840 .shutdown = shutdown,
1841 .setsockopt = setsockopt,
1842 .getsockopt = getsockopt,
1843 .sendmsg = send_stream,
1844 .recvmsg = recv_stream,
1845 .mmap = sock_no_mmap,
1846 .sendpage = sock_no_sendpage
1847};
1848
1849static const struct net_proto_family tipc_family_ops = {
1850 .owner = THIS_MODULE,
1851 .family = AF_TIPC,
1852 .create = tipc_create
1853};
1854
1855static struct proto tipc_proto = {
1856 .name = "TIPC",
1857 .owner = THIS_MODULE,
1858 .obj_size = sizeof(struct tipc_sock)
1859};
1860
1861/**
1862 * tipc_socket_init - initialize TIPC socket interface
1863 *
1864 * Returns 0 on success, errno otherwise
1865 */
1866int tipc_socket_init(void)
1867{
1868 int res;
1869
1870 res = proto_register(&tipc_proto, 1);
1871 if (res) {
1872 err("Failed to register TIPC protocol type\n");
1873 goto out;
1874 }
1875
1876 res = sock_register(&tipc_family_ops);
1877 if (res) {
1878 err("Failed to register TIPC socket type\n");
1879 proto_unregister(&tipc_proto);
1880 goto out;
1881 }
1882
1883 sockets_enabled = 1;
1884 out:
1885 return res;
1886}
1887
1888/**
1889 * tipc_socket_stop - stop TIPC socket interface
1890 */
1891
1892void tipc_socket_stop(void)
1893{
1894 if (!sockets_enabled)
1895 return;
1896
1897 sockets_enabled = 0;
1898 sock_unregister(tipc_family_ops.family);
1899 proto_unregister(&tipc_proto);
1900}
1901
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#include "netlink.h"
46
47#define SS_LISTENING -1 /* socket is listening */
48#define SS_READY -2 /* socket is connectionless */
49
50#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51#define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
52#define TIPC_FWD_MSG 1
53#define TIPC_CONN_OK 0
54#define TIPC_CONN_PROBING 1
55#define TIPC_MAX_PORT 0xffffffff
56#define TIPC_MIN_PORT 1
57
58/**
59 * struct tipc_sock - TIPC socket structure
60 * @sk: socket - interacts with 'port' and with user via the socket API
61 * @connected: non-zero if port is currently connected to a peer port
62 * @conn_type: TIPC type used when connection was established
63 * @conn_instance: TIPC instance used when connection was established
64 * @published: non-zero if port has one or more associated names
65 * @max_pkt: maximum packet size "hint" used when building messages sent by port
66 * @portid: unique port identity in TIPC socket hash table
67 * @phdr: preformatted message header used when sending messages
68 * @port_list: adjacent ports in TIPC's global list of ports
69 * @publications: list of publications for port
70 * @pub_count: total # of publications port has made during its lifetime
71 * @probing_state:
72 * @probing_intv:
73 * @conn_timeout: the time we can wait for an unresponded setup request
74 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
75 * @link_cong: non-zero if owner must sleep because of link congestion
76 * @sent_unacked: # messages sent by socket, and not yet acked by peer
77 * @rcv_unacked: # messages read by user, but not yet acked back to peer
78 * @remote: 'connected' peer for dgram/rdm
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
81 */
82struct tipc_sock {
83 struct sock sk;
84 int connected;
85 u32 conn_type;
86 u32 conn_instance;
87 int published;
88 u32 max_pkt;
89 u32 portid;
90 struct tipc_msg phdr;
91 struct list_head sock_list;
92 struct list_head publications;
93 u32 pub_count;
94 u32 probing_state;
95 unsigned long probing_intv;
96 uint conn_timeout;
97 atomic_t dupl_rcvcnt;
98 bool link_cong;
99 uint sent_unacked;
100 uint rcv_unacked;
101 struct sockaddr_tipc remote;
102 struct rhash_head node;
103 struct rcu_head rcu;
104};
105
106static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
107static void tipc_data_ready(struct sock *sk);
108static void tipc_write_space(struct sock *sk);
109static void tipc_sock_destruct(struct sock *sk);
110static int tipc_release(struct socket *sock);
111static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
112static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
113static void tipc_sk_timeout(unsigned long data);
114static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
115 struct tipc_name_seq const *seq);
116static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
117 struct tipc_name_seq const *seq);
118static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
119static int tipc_sk_insert(struct tipc_sock *tsk);
120static void tipc_sk_remove(struct tipc_sock *tsk);
121static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
122 size_t dsz);
123static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
124
125static const struct proto_ops packet_ops;
126static const struct proto_ops stream_ops;
127static const struct proto_ops msg_ops;
128static struct proto tipc_proto;
129
130static const struct rhashtable_params tsk_rht_params;
131
132/*
133 * Revised TIPC socket locking policy:
134 *
135 * Most socket operations take the standard socket lock when they start
136 * and hold it until they finish (or until they need to sleep). Acquiring
137 * this lock grants the owner exclusive access to the fields of the socket
138 * data structures, with the exception of the backlog queue. A few socket
139 * operations can be done without taking the socket lock because they only
140 * read socket information that never changes during the life of the socket.
141 *
142 * Socket operations may acquire the lock for the associated TIPC port if they
143 * need to perform an operation on the port. If any routine needs to acquire
144 * both the socket lock and the port lock it must take the socket lock first
145 * to avoid the risk of deadlock.
146 *
147 * The dispatcher handling incoming messages cannot grab the socket lock in
148 * the standard fashion, since invoked it runs at the BH level and cannot block.
149 * Instead, it checks to see if the socket lock is currently owned by someone,
150 * and either handles the message itself or adds it to the socket's backlog
151 * queue; in the latter case the queued message is processed once the process
152 * owning the socket lock releases it.
153 *
154 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
155 * the problem of a blocked socket operation preventing any other operations
156 * from occurring. However, applications must be careful if they have
157 * multiple threads trying to send (or receive) on the same socket, as these
158 * operations might interfere with each other. For example, doing a connect
159 * and a receive at the same time might allow the receive to consume the
160 * ACK message meant for the connect. While additional work could be done
161 * to try and overcome this, it doesn't seem to be worthwhile at the present.
162 *
163 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
164 * that another operation that must be performed in a non-blocking manner is
165 * not delayed for very long because the lock has already been taken.
166 *
167 * NOTE: This code assumes that certain fields of a port/socket pair are
168 * constant over its lifetime; such fields can be examined without taking
169 * the socket lock and/or port lock, and do not need to be re-read even
170 * after resuming processing after waiting. These fields include:
171 * - socket type
172 * - pointer to socket sk structure (aka tipc_sock structure)
173 * - pointer to port structure
174 * - port reference
175 */
176
177static u32 tsk_own_node(struct tipc_sock *tsk)
178{
179 return msg_prevnode(&tsk->phdr);
180}
181
182static u32 tsk_peer_node(struct tipc_sock *tsk)
183{
184 return msg_destnode(&tsk->phdr);
185}
186
187static u32 tsk_peer_port(struct tipc_sock *tsk)
188{
189 return msg_destport(&tsk->phdr);
190}
191
192static bool tsk_unreliable(struct tipc_sock *tsk)
193{
194 return msg_src_droppable(&tsk->phdr) != 0;
195}
196
197static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
198{
199 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
200}
201
202static bool tsk_unreturnable(struct tipc_sock *tsk)
203{
204 return msg_dest_droppable(&tsk->phdr) != 0;
205}
206
207static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
208{
209 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
210}
211
212static int tsk_importance(struct tipc_sock *tsk)
213{
214 return msg_importance(&tsk->phdr);
215}
216
217static int tsk_set_importance(struct tipc_sock *tsk, int imp)
218{
219 if (imp > TIPC_CRITICAL_IMPORTANCE)
220 return -EINVAL;
221 msg_set_importance(&tsk->phdr, (u32)imp);
222 return 0;
223}
224
225static struct tipc_sock *tipc_sk(const struct sock *sk)
226{
227 return container_of(sk, struct tipc_sock, sk);
228}
229
230static int tsk_conn_cong(struct tipc_sock *tsk)
231{
232 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
233}
234
235/**
236 * tsk_advance_rx_queue - discard first buffer in socket receive queue
237 *
238 * Caller must hold socket lock
239 */
240static void tsk_advance_rx_queue(struct sock *sk)
241{
242 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
243}
244
245/* tipc_sk_respond() : send response message back to sender
246 */
247static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
248{
249 u32 selector;
250 u32 dnode;
251 u32 onode = tipc_own_addr(sock_net(sk));
252
253 if (!tipc_msg_reverse(onode, &skb, err))
254 return;
255
256 dnode = msg_destnode(buf_msg(skb));
257 selector = msg_origport(buf_msg(skb));
258 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
259}
260
261/**
262 * tsk_rej_rx_queue - reject all buffers in socket receive queue
263 *
264 * Caller must hold socket lock
265 */
266static void tsk_rej_rx_queue(struct sock *sk)
267{
268 struct sk_buff *skb;
269
270 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
271 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
272}
273
274/* tsk_peer_msg - verify if message was sent by connected port's peer
275 *
276 * Handles cases where the node's network address has changed from
277 * the default of <0.0.0> to its configured setting.
278 */
279static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
280{
281 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
282 u32 peer_port = tsk_peer_port(tsk);
283 u32 orig_node;
284 u32 peer_node;
285
286 if (unlikely(!tsk->connected))
287 return false;
288
289 if (unlikely(msg_origport(msg) != peer_port))
290 return false;
291
292 orig_node = msg_orignode(msg);
293 peer_node = tsk_peer_node(tsk);
294
295 if (likely(orig_node == peer_node))
296 return true;
297
298 if (!orig_node && (peer_node == tn->own_addr))
299 return true;
300
301 if (!peer_node && (orig_node == tn->own_addr))
302 return true;
303
304 return false;
305}
306
307/**
308 * tipc_sk_create - create a TIPC socket
309 * @net: network namespace (must be default network)
310 * @sock: pre-allocated socket structure
311 * @protocol: protocol indicator (must be 0)
312 * @kern: caused by kernel or by userspace?
313 *
314 * This routine creates additional data structures used by the TIPC socket,
315 * initializes them, and links them together.
316 *
317 * Returns 0 on success, errno otherwise
318 */
319static int tipc_sk_create(struct net *net, struct socket *sock,
320 int protocol, int kern)
321{
322 struct tipc_net *tn;
323 const struct proto_ops *ops;
324 socket_state state;
325 struct sock *sk;
326 struct tipc_sock *tsk;
327 struct tipc_msg *msg;
328
329 /* Validate arguments */
330 if (unlikely(protocol != 0))
331 return -EPROTONOSUPPORT;
332
333 switch (sock->type) {
334 case SOCK_STREAM:
335 ops = &stream_ops;
336 state = SS_UNCONNECTED;
337 break;
338 case SOCK_SEQPACKET:
339 ops = &packet_ops;
340 state = SS_UNCONNECTED;
341 break;
342 case SOCK_DGRAM:
343 case SOCK_RDM:
344 ops = &msg_ops;
345 state = SS_READY;
346 break;
347 default:
348 return -EPROTOTYPE;
349 }
350
351 /* Allocate socket's protocol area */
352 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
353 if (sk == NULL)
354 return -ENOMEM;
355
356 tsk = tipc_sk(sk);
357 tsk->max_pkt = MAX_PKT_DEFAULT;
358 INIT_LIST_HEAD(&tsk->publications);
359 msg = &tsk->phdr;
360 tn = net_generic(sock_net(sk), tipc_net_id);
361 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
362 NAMED_H_SIZE, 0);
363
364 /* Finish initializing socket data structures */
365 sock->ops = ops;
366 sock->state = state;
367 sock_init_data(sock, sk);
368 if (tipc_sk_insert(tsk)) {
369 pr_warn("Socket create failed; port numbrer exhausted\n");
370 return -EINVAL;
371 }
372 msg_set_origport(msg, tsk->portid);
373 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
374 sk->sk_backlog_rcv = tipc_backlog_rcv;
375 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
376 sk->sk_data_ready = tipc_data_ready;
377 sk->sk_write_space = tipc_write_space;
378 sk->sk_destruct = tipc_sock_destruct;
379 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
380 tsk->sent_unacked = 0;
381 atomic_set(&tsk->dupl_rcvcnt, 0);
382
383 if (sock->state == SS_READY) {
384 tsk_set_unreturnable(tsk, true);
385 if (sock->type == SOCK_DGRAM)
386 tsk_set_unreliable(tsk, true);
387 }
388 return 0;
389}
390
391static void tipc_sk_callback(struct rcu_head *head)
392{
393 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
394
395 sock_put(&tsk->sk);
396}
397
398/**
399 * tipc_release - destroy a TIPC socket
400 * @sock: socket to destroy
401 *
402 * This routine cleans up any messages that are still queued on the socket.
403 * For DGRAM and RDM socket types, all queued messages are rejected.
404 * For SEQPACKET and STREAM socket types, the first message is rejected
405 * and any others are discarded. (If the first message on a STREAM socket
406 * is partially-read, it is discarded and the next one is rejected instead.)
407 *
408 * NOTE: Rejected messages are not necessarily returned to the sender! They
409 * are returned or discarded according to the "destination droppable" setting
410 * specified for the message by the sender.
411 *
412 * Returns 0 on success, errno otherwise
413 */
414static int tipc_release(struct socket *sock)
415{
416 struct sock *sk = sock->sk;
417 struct net *net;
418 struct tipc_sock *tsk;
419 struct sk_buff *skb;
420 u32 dnode;
421
422 /*
423 * Exit if socket isn't fully initialized (occurs when a failed accept()
424 * releases a pre-allocated child socket that was never used)
425 */
426 if (sk == NULL)
427 return 0;
428
429 net = sock_net(sk);
430 tsk = tipc_sk(sk);
431 lock_sock(sk);
432
433 /*
434 * Reject all unreceived messages, except on an active connection
435 * (which disconnects locally & sends a 'FIN+' to peer)
436 */
437 dnode = tsk_peer_node(tsk);
438 while (sock->state != SS_DISCONNECTING) {
439 skb = __skb_dequeue(&sk->sk_receive_queue);
440 if (skb == NULL)
441 break;
442 if (TIPC_SKB_CB(skb)->handle != NULL)
443 kfree_skb(skb);
444 else {
445 if ((sock->state == SS_CONNECTING) ||
446 (sock->state == SS_CONNECTED)) {
447 sock->state = SS_DISCONNECTING;
448 tsk->connected = 0;
449 tipc_node_remove_conn(net, dnode, tsk->portid);
450 }
451 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
452 }
453 }
454
455 tipc_sk_withdraw(tsk, 0, NULL);
456 sk_stop_timer(sk, &sk->sk_timer);
457 tipc_sk_remove(tsk);
458 if (tsk->connected) {
459 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
460 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
461 tsk_own_node(tsk), tsk_peer_port(tsk),
462 tsk->portid, TIPC_ERR_NO_PORT);
463 if (skb)
464 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
465 tipc_node_remove_conn(net, dnode, tsk->portid);
466 }
467
468 /* Reject any messages that accumulated in backlog queue */
469 sock->state = SS_DISCONNECTING;
470 release_sock(sk);
471
472 call_rcu(&tsk->rcu, tipc_sk_callback);
473 sock->sk = NULL;
474
475 return 0;
476}
477
478/**
479 * tipc_bind - associate or disassocate TIPC name(s) with a socket
480 * @sock: socket structure
481 * @uaddr: socket address describing name(s) and desired operation
482 * @uaddr_len: size of socket address data structure
483 *
484 * Name and name sequence binding is indicated using a positive scope value;
485 * a negative scope value unbinds the specified name. Specifying no name
486 * (i.e. a socket address length of 0) unbinds all names from the socket.
487 *
488 * Returns 0 on success, errno otherwise
489 *
490 * NOTE: This routine doesn't need to take the socket lock since it doesn't
491 * access any non-constant socket information.
492 */
493static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
494 int uaddr_len)
495{
496 struct sock *sk = sock->sk;
497 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
498 struct tipc_sock *tsk = tipc_sk(sk);
499 int res = -EINVAL;
500
501 lock_sock(sk);
502 if (unlikely(!uaddr_len)) {
503 res = tipc_sk_withdraw(tsk, 0, NULL);
504 goto exit;
505 }
506
507 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
508 res = -EINVAL;
509 goto exit;
510 }
511 if (addr->family != AF_TIPC) {
512 res = -EAFNOSUPPORT;
513 goto exit;
514 }
515
516 if (addr->addrtype == TIPC_ADDR_NAME)
517 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
518 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
519 res = -EAFNOSUPPORT;
520 goto exit;
521 }
522
523 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
524 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
525 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
526 res = -EACCES;
527 goto exit;
528 }
529
530 res = (addr->scope > 0) ?
531 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
532 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
533exit:
534 release_sock(sk);
535 return res;
536}
537
538/**
539 * tipc_getname - get port ID of socket or peer socket
540 * @sock: socket structure
541 * @uaddr: area for returned socket address
542 * @uaddr_len: area for returned length of socket address
543 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
544 *
545 * Returns 0 on success, errno otherwise
546 *
547 * NOTE: This routine doesn't need to take the socket lock since it only
548 * accesses socket information that is unchanging (or which changes in
549 * a completely predictable manner).
550 */
551static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
552 int *uaddr_len, int peer)
553{
554 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
555 struct tipc_sock *tsk = tipc_sk(sock->sk);
556 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
557
558 memset(addr, 0, sizeof(*addr));
559 if (peer) {
560 if ((sock->state != SS_CONNECTED) &&
561 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
562 return -ENOTCONN;
563 addr->addr.id.ref = tsk_peer_port(tsk);
564 addr->addr.id.node = tsk_peer_node(tsk);
565 } else {
566 addr->addr.id.ref = tsk->portid;
567 addr->addr.id.node = tn->own_addr;
568 }
569
570 *uaddr_len = sizeof(*addr);
571 addr->addrtype = TIPC_ADDR_ID;
572 addr->family = AF_TIPC;
573 addr->scope = 0;
574 addr->addr.name.domain = 0;
575
576 return 0;
577}
578
579/**
580 * tipc_poll - read and possibly block on pollmask
581 * @file: file structure associated with the socket
582 * @sock: socket for which to calculate the poll bits
583 * @wait: ???
584 *
585 * Returns pollmask value
586 *
587 * COMMENTARY:
588 * It appears that the usual socket locking mechanisms are not useful here
589 * since the pollmask info is potentially out-of-date the moment this routine
590 * exits. TCP and other protocols seem to rely on higher level poll routines
591 * to handle any preventable race conditions, so TIPC will do the same ...
592 *
593 * TIPC sets the returned events as follows:
594 *
595 * socket state flags set
596 * ------------ ---------
597 * unconnected no read flags
598 * POLLOUT if port is not congested
599 *
600 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
601 * no write flags
602 *
603 * connected POLLIN/POLLRDNORM if data in rx queue
604 * POLLOUT if port is not congested
605 *
606 * disconnecting POLLIN/POLLRDNORM/POLLHUP
607 * no write flags
608 *
609 * listening POLLIN if SYN in rx queue
610 * no write flags
611 *
612 * ready POLLIN/POLLRDNORM if data in rx queue
613 * [connectionless] POLLOUT (since port cannot be congested)
614 *
615 * IMPORTANT: The fact that a read or write operation is indicated does NOT
616 * imply that the operation will succeed, merely that it should be performed
617 * and will not block.
618 */
619static unsigned int tipc_poll(struct file *file, struct socket *sock,
620 poll_table *wait)
621{
622 struct sock *sk = sock->sk;
623 struct tipc_sock *tsk = tipc_sk(sk);
624 u32 mask = 0;
625
626 sock_poll_wait(file, sk_sleep(sk), wait);
627
628 switch ((int)sock->state) {
629 case SS_UNCONNECTED:
630 if (!tsk->link_cong)
631 mask |= POLLOUT;
632 break;
633 case SS_READY:
634 case SS_CONNECTED:
635 if (!tsk->link_cong && !tsk_conn_cong(tsk))
636 mask |= POLLOUT;
637 /* fall thru' */
638 case SS_CONNECTING:
639 case SS_LISTENING:
640 if (!skb_queue_empty(&sk->sk_receive_queue))
641 mask |= (POLLIN | POLLRDNORM);
642 break;
643 case SS_DISCONNECTING:
644 mask = (POLLIN | POLLRDNORM | POLLHUP);
645 break;
646 }
647
648 return mask;
649}
650
651/**
652 * tipc_sendmcast - send multicast message
653 * @sock: socket structure
654 * @seq: destination address
655 * @msg: message to send
656 * @dsz: total length of message data
657 * @timeo: timeout to wait for wakeup
658 *
659 * Called from function tipc_sendmsg(), which has done all sanity checks
660 * Returns the number of bytes sent on success, or errno
661 */
662static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
663 struct msghdr *msg, size_t dsz, long timeo)
664{
665 struct sock *sk = sock->sk;
666 struct tipc_sock *tsk = tipc_sk(sk);
667 struct net *net = sock_net(sk);
668 struct tipc_msg *mhdr = &tsk->phdr;
669 struct sk_buff_head pktchain;
670 struct iov_iter save = msg->msg_iter;
671 uint mtu;
672 int rc;
673
674 msg_set_type(mhdr, TIPC_MCAST_MSG);
675 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
676 msg_set_destport(mhdr, 0);
677 msg_set_destnode(mhdr, 0);
678 msg_set_nametype(mhdr, seq->type);
679 msg_set_namelower(mhdr, seq->lower);
680 msg_set_nameupper(mhdr, seq->upper);
681 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
682
683 skb_queue_head_init(&pktchain);
684
685new_mtu:
686 mtu = tipc_bcast_get_mtu(net);
687 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
688 if (unlikely(rc < 0))
689 return rc;
690
691 do {
692 rc = tipc_bcast_xmit(net, &pktchain);
693 if (likely(!rc))
694 return dsz;
695
696 if (rc == -ELINKCONG) {
697 tsk->link_cong = 1;
698 rc = tipc_wait_for_sndmsg(sock, &timeo);
699 if (!rc)
700 continue;
701 }
702 __skb_queue_purge(&pktchain);
703 if (rc == -EMSGSIZE) {
704 msg->msg_iter = save;
705 goto new_mtu;
706 }
707 break;
708 } while (1);
709 return rc;
710}
711
712/**
713 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
714 * @arrvq: queue with arriving messages, to be cloned after destination lookup
715 * @inputq: queue with cloned messages, delivered to socket after dest lookup
716 *
717 * Multi-threaded: parallel calls with reference to same queues may occur
718 */
719void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
720 struct sk_buff_head *inputq)
721{
722 struct tipc_msg *msg;
723 struct tipc_plist dports;
724 u32 portid;
725 u32 scope = TIPC_CLUSTER_SCOPE;
726 struct sk_buff_head tmpq;
727 uint hsz;
728 struct sk_buff *skb, *_skb;
729
730 __skb_queue_head_init(&tmpq);
731 tipc_plist_init(&dports);
732
733 skb = tipc_skb_peek(arrvq, &inputq->lock);
734 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
735 msg = buf_msg(skb);
736 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
737
738 if (in_own_node(net, msg_orignode(msg)))
739 scope = TIPC_NODE_SCOPE;
740
741 /* Create destination port list and message clones: */
742 tipc_nametbl_mc_translate(net,
743 msg_nametype(msg), msg_namelower(msg),
744 msg_nameupper(msg), scope, &dports);
745 portid = tipc_plist_pop(&dports);
746 for (; portid; portid = tipc_plist_pop(&dports)) {
747 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
748 if (_skb) {
749 msg_set_destport(buf_msg(_skb), portid);
750 __skb_queue_tail(&tmpq, _skb);
751 continue;
752 }
753 pr_warn("Failed to clone mcast rcv buffer\n");
754 }
755 /* Append to inputq if not already done by other thread */
756 spin_lock_bh(&inputq->lock);
757 if (skb_peek(arrvq) == skb) {
758 skb_queue_splice_tail_init(&tmpq, inputq);
759 kfree_skb(__skb_dequeue(arrvq));
760 }
761 spin_unlock_bh(&inputq->lock);
762 __skb_queue_purge(&tmpq);
763 kfree_skb(skb);
764 }
765 tipc_sk_rcv(net, inputq);
766}
767
768/**
769 * tipc_sk_proto_rcv - receive a connection mng protocol message
770 * @tsk: receiving socket
771 * @skb: pointer to message buffer.
772 */
773static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb)
774{
775 struct sock *sk = &tsk->sk;
776 struct tipc_msg *hdr = buf_msg(skb);
777 int mtyp = msg_type(hdr);
778 int conn_cong;
779
780 /* Ignore if connection cannot be validated: */
781 if (!tsk_peer_msg(tsk, hdr))
782 goto exit;
783
784 tsk->probing_state = TIPC_CONN_OK;
785
786 if (mtyp == CONN_PROBE) {
787 msg_set_type(hdr, CONN_PROBE_REPLY);
788 tipc_sk_respond(sk, skb, TIPC_OK);
789 return;
790 } else if (mtyp == CONN_ACK) {
791 conn_cong = tsk_conn_cong(tsk);
792 tsk->sent_unacked -= msg_msgcnt(hdr);
793 if (conn_cong)
794 sk->sk_write_space(sk);
795 } else if (mtyp != CONN_PROBE_REPLY) {
796 pr_warn("Received unknown CONN_PROTO msg\n");
797 }
798exit:
799 kfree_skb(skb);
800}
801
802static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
803{
804 struct sock *sk = sock->sk;
805 struct tipc_sock *tsk = tipc_sk(sk);
806 DEFINE_WAIT(wait);
807 int done;
808
809 do {
810 int err = sock_error(sk);
811 if (err)
812 return err;
813 if (sock->state == SS_DISCONNECTING)
814 return -EPIPE;
815 if (!*timeo_p)
816 return -EAGAIN;
817 if (signal_pending(current))
818 return sock_intr_errno(*timeo_p);
819
820 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
821 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
822 finish_wait(sk_sleep(sk), &wait);
823 } while (!done);
824 return 0;
825}
826
827/**
828 * tipc_sendmsg - send message in connectionless manner
829 * @sock: socket structure
830 * @m: message to send
831 * @dsz: amount of user data to be sent
832 *
833 * Message must have an destination specified explicitly.
834 * Used for SOCK_RDM and SOCK_DGRAM messages,
835 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
836 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
837 *
838 * Returns the number of bytes sent on success, or errno otherwise
839 */
840static int tipc_sendmsg(struct socket *sock,
841 struct msghdr *m, size_t dsz)
842{
843 struct sock *sk = sock->sk;
844 int ret;
845
846 lock_sock(sk);
847 ret = __tipc_sendmsg(sock, m, dsz);
848 release_sock(sk);
849
850 return ret;
851}
852
853static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
854{
855 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
856 struct sock *sk = sock->sk;
857 struct tipc_sock *tsk = tipc_sk(sk);
858 struct net *net = sock_net(sk);
859 struct tipc_msg *mhdr = &tsk->phdr;
860 u32 dnode, dport;
861 struct sk_buff_head pktchain;
862 struct sk_buff *skb;
863 struct tipc_name_seq *seq;
864 struct iov_iter save;
865 u32 mtu;
866 long timeo;
867 int rc;
868
869 if (dsz > TIPC_MAX_USER_MSG_SIZE)
870 return -EMSGSIZE;
871 if (unlikely(!dest)) {
872 if (tsk->connected && sock->state == SS_READY)
873 dest = &tsk->remote;
874 else
875 return -EDESTADDRREQ;
876 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
877 dest->family != AF_TIPC) {
878 return -EINVAL;
879 }
880 if (unlikely(sock->state != SS_READY)) {
881 if (sock->state == SS_LISTENING)
882 return -EPIPE;
883 if (sock->state != SS_UNCONNECTED)
884 return -EISCONN;
885 if (tsk->published)
886 return -EOPNOTSUPP;
887 if (dest->addrtype == TIPC_ADDR_NAME) {
888 tsk->conn_type = dest->addr.name.name.type;
889 tsk->conn_instance = dest->addr.name.name.instance;
890 }
891 }
892 seq = &dest->addr.nameseq;
893 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
894
895 if (dest->addrtype == TIPC_ADDR_MCAST) {
896 return tipc_sendmcast(sock, seq, m, dsz, timeo);
897 } else if (dest->addrtype == TIPC_ADDR_NAME) {
898 u32 type = dest->addr.name.name.type;
899 u32 inst = dest->addr.name.name.instance;
900 u32 domain = dest->addr.name.domain;
901
902 dnode = domain;
903 msg_set_type(mhdr, TIPC_NAMED_MSG);
904 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
905 msg_set_nametype(mhdr, type);
906 msg_set_nameinst(mhdr, inst);
907 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
908 dport = tipc_nametbl_translate(net, type, inst, &dnode);
909 msg_set_destnode(mhdr, dnode);
910 msg_set_destport(mhdr, dport);
911 if (unlikely(!dport && !dnode))
912 return -EHOSTUNREACH;
913 } else if (dest->addrtype == TIPC_ADDR_ID) {
914 dnode = dest->addr.id.node;
915 msg_set_type(mhdr, TIPC_DIRECT_MSG);
916 msg_set_lookup_scope(mhdr, 0);
917 msg_set_destnode(mhdr, dnode);
918 msg_set_destport(mhdr, dest->addr.id.ref);
919 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
920 }
921
922 skb_queue_head_init(&pktchain);
923 save = m->msg_iter;
924new_mtu:
925 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
926 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
927 if (rc < 0)
928 return rc;
929
930 do {
931 skb = skb_peek(&pktchain);
932 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
933 rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
934 if (likely(!rc)) {
935 if (sock->state != SS_READY)
936 sock->state = SS_CONNECTING;
937 return dsz;
938 }
939 if (rc == -ELINKCONG) {
940 tsk->link_cong = 1;
941 rc = tipc_wait_for_sndmsg(sock, &timeo);
942 if (!rc)
943 continue;
944 }
945 __skb_queue_purge(&pktchain);
946 if (rc == -EMSGSIZE) {
947 m->msg_iter = save;
948 goto new_mtu;
949 }
950 break;
951 } while (1);
952
953 return rc;
954}
955
956static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
957{
958 struct sock *sk = sock->sk;
959 struct tipc_sock *tsk = tipc_sk(sk);
960 DEFINE_WAIT(wait);
961 int done;
962
963 do {
964 int err = sock_error(sk);
965 if (err)
966 return err;
967 if (sock->state == SS_DISCONNECTING)
968 return -EPIPE;
969 else if (sock->state != SS_CONNECTED)
970 return -ENOTCONN;
971 if (!*timeo_p)
972 return -EAGAIN;
973 if (signal_pending(current))
974 return sock_intr_errno(*timeo_p);
975
976 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
977 done = sk_wait_event(sk, timeo_p,
978 (!tsk->link_cong &&
979 !tsk_conn_cong(tsk)) ||
980 !tsk->connected);
981 finish_wait(sk_sleep(sk), &wait);
982 } while (!done);
983 return 0;
984}
985
986/**
987 * tipc_send_stream - send stream-oriented data
988 * @sock: socket structure
989 * @m: data to send
990 * @dsz: total length of data to be transmitted
991 *
992 * Used for SOCK_STREAM data.
993 *
994 * Returns the number of bytes sent on success (or partial success),
995 * or errno if no data sent
996 */
997static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
998{
999 struct sock *sk = sock->sk;
1000 int ret;
1001
1002 lock_sock(sk);
1003 ret = __tipc_send_stream(sock, m, dsz);
1004 release_sock(sk);
1005
1006 return ret;
1007}
1008
1009static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1010{
1011 struct sock *sk = sock->sk;
1012 struct net *net = sock_net(sk);
1013 struct tipc_sock *tsk = tipc_sk(sk);
1014 struct tipc_msg *mhdr = &tsk->phdr;
1015 struct sk_buff_head pktchain;
1016 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1017 u32 portid = tsk->portid;
1018 int rc = -EINVAL;
1019 long timeo;
1020 u32 dnode;
1021 uint mtu, send, sent = 0;
1022 struct iov_iter save;
1023
1024 /* Handle implied connection establishment */
1025 if (unlikely(dest)) {
1026 rc = __tipc_sendmsg(sock, m, dsz);
1027 if (dsz && (dsz == rc))
1028 tsk->sent_unacked = 1;
1029 return rc;
1030 }
1031 if (dsz > (uint)INT_MAX)
1032 return -EMSGSIZE;
1033
1034 if (unlikely(sock->state != SS_CONNECTED)) {
1035 if (sock->state == SS_DISCONNECTING)
1036 return -EPIPE;
1037 else
1038 return -ENOTCONN;
1039 }
1040
1041 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1042 dnode = tsk_peer_node(tsk);
1043 skb_queue_head_init(&pktchain);
1044
1045next:
1046 save = m->msg_iter;
1047 mtu = tsk->max_pkt;
1048 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1049 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1050 if (unlikely(rc < 0))
1051 return rc;
1052
1053 do {
1054 if (likely(!tsk_conn_cong(tsk))) {
1055 rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1056 if (likely(!rc)) {
1057 tsk->sent_unacked++;
1058 sent += send;
1059 if (sent == dsz)
1060 return dsz;
1061 goto next;
1062 }
1063 if (rc == -EMSGSIZE) {
1064 __skb_queue_purge(&pktchain);
1065 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1066 portid);
1067 m->msg_iter = save;
1068 goto next;
1069 }
1070 if (rc != -ELINKCONG)
1071 break;
1072
1073 tsk->link_cong = 1;
1074 }
1075 rc = tipc_wait_for_sndpkt(sock, &timeo);
1076 } while (!rc);
1077
1078 __skb_queue_purge(&pktchain);
1079 return sent ? sent : rc;
1080}
1081
1082/**
1083 * tipc_send_packet - send a connection-oriented message
1084 * @sock: socket structure
1085 * @m: message to send
1086 * @dsz: length of data to be transmitted
1087 *
1088 * Used for SOCK_SEQPACKET messages.
1089 *
1090 * Returns the number of bytes sent on success, or errno otherwise
1091 */
1092static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1093{
1094 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1095 return -EMSGSIZE;
1096
1097 return tipc_send_stream(sock, m, dsz);
1098}
1099
1100/* tipc_sk_finish_conn - complete the setup of a connection
1101 */
1102static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1103 u32 peer_node)
1104{
1105 struct sock *sk = &tsk->sk;
1106 struct net *net = sock_net(sk);
1107 struct tipc_msg *msg = &tsk->phdr;
1108
1109 msg_set_destnode(msg, peer_node);
1110 msg_set_destport(msg, peer_port);
1111 msg_set_type(msg, TIPC_CONN_MSG);
1112 msg_set_lookup_scope(msg, 0);
1113 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1114
1115 tsk->probing_intv = CONN_PROBING_INTERVAL;
1116 tsk->probing_state = TIPC_CONN_OK;
1117 tsk->connected = 1;
1118 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1119 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1120 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1121}
1122
1123/**
1124 * set_orig_addr - capture sender's address for received message
1125 * @m: descriptor for message info
1126 * @msg: received message header
1127 *
1128 * Note: Address is not captured if not requested by receiver.
1129 */
1130static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1131{
1132 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1133
1134 if (addr) {
1135 addr->family = AF_TIPC;
1136 addr->addrtype = TIPC_ADDR_ID;
1137 memset(&addr->addr, 0, sizeof(addr->addr));
1138 addr->addr.id.ref = msg_origport(msg);
1139 addr->addr.id.node = msg_orignode(msg);
1140 addr->addr.name.domain = 0; /* could leave uninitialized */
1141 addr->scope = 0; /* could leave uninitialized */
1142 m->msg_namelen = sizeof(struct sockaddr_tipc);
1143 }
1144}
1145
1146/**
1147 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1148 * @m: descriptor for message info
1149 * @msg: received message header
1150 * @tsk: TIPC port associated with message
1151 *
1152 * Note: Ancillary data is not captured if not requested by receiver.
1153 *
1154 * Returns 0 if successful, otherwise errno
1155 */
1156static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1157 struct tipc_sock *tsk)
1158{
1159 u32 anc_data[3];
1160 u32 err;
1161 u32 dest_type;
1162 int has_name;
1163 int res;
1164
1165 if (likely(m->msg_controllen == 0))
1166 return 0;
1167
1168 /* Optionally capture errored message object(s) */
1169 err = msg ? msg_errcode(msg) : 0;
1170 if (unlikely(err)) {
1171 anc_data[0] = err;
1172 anc_data[1] = msg_data_sz(msg);
1173 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1174 if (res)
1175 return res;
1176 if (anc_data[1]) {
1177 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1178 msg_data(msg));
1179 if (res)
1180 return res;
1181 }
1182 }
1183
1184 /* Optionally capture message destination object */
1185 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1186 switch (dest_type) {
1187 case TIPC_NAMED_MSG:
1188 has_name = 1;
1189 anc_data[0] = msg_nametype(msg);
1190 anc_data[1] = msg_namelower(msg);
1191 anc_data[2] = msg_namelower(msg);
1192 break;
1193 case TIPC_MCAST_MSG:
1194 has_name = 1;
1195 anc_data[0] = msg_nametype(msg);
1196 anc_data[1] = msg_namelower(msg);
1197 anc_data[2] = msg_nameupper(msg);
1198 break;
1199 case TIPC_CONN_MSG:
1200 has_name = (tsk->conn_type != 0);
1201 anc_data[0] = tsk->conn_type;
1202 anc_data[1] = tsk->conn_instance;
1203 anc_data[2] = tsk->conn_instance;
1204 break;
1205 default:
1206 has_name = 0;
1207 }
1208 if (has_name) {
1209 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1210 if (res)
1211 return res;
1212 }
1213
1214 return 0;
1215}
1216
1217static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1218{
1219 struct net *net = sock_net(&tsk->sk);
1220 struct sk_buff *skb = NULL;
1221 struct tipc_msg *msg;
1222 u32 peer_port = tsk_peer_port(tsk);
1223 u32 dnode = tsk_peer_node(tsk);
1224
1225 if (!tsk->connected)
1226 return;
1227 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1228 dnode, tsk_own_node(tsk), peer_port,
1229 tsk->portid, TIPC_OK);
1230 if (!skb)
1231 return;
1232 msg = buf_msg(skb);
1233 msg_set_msgcnt(msg, ack);
1234 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1235}
1236
1237static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1238{
1239 struct sock *sk = sock->sk;
1240 DEFINE_WAIT(wait);
1241 long timeo = *timeop;
1242 int err;
1243
1244 for (;;) {
1245 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1246 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1247 if (sock->state == SS_DISCONNECTING) {
1248 err = -ENOTCONN;
1249 break;
1250 }
1251 release_sock(sk);
1252 timeo = schedule_timeout(timeo);
1253 lock_sock(sk);
1254 }
1255 err = 0;
1256 if (!skb_queue_empty(&sk->sk_receive_queue))
1257 break;
1258 err = -EAGAIN;
1259 if (!timeo)
1260 break;
1261 err = sock_intr_errno(timeo);
1262 if (signal_pending(current))
1263 break;
1264 }
1265 finish_wait(sk_sleep(sk), &wait);
1266 *timeop = timeo;
1267 return err;
1268}
1269
1270/**
1271 * tipc_recvmsg - receive packet-oriented message
1272 * @m: descriptor for message info
1273 * @buf_len: total size of user buffer area
1274 * @flags: receive flags
1275 *
1276 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1277 * If the complete message doesn't fit in user area, truncate it.
1278 *
1279 * Returns size of returned message data, errno otherwise
1280 */
1281static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1282 int flags)
1283{
1284 struct sock *sk = sock->sk;
1285 struct tipc_sock *tsk = tipc_sk(sk);
1286 struct sk_buff *buf;
1287 struct tipc_msg *msg;
1288 long timeo;
1289 unsigned int sz;
1290 u32 err;
1291 int res;
1292
1293 /* Catch invalid receive requests */
1294 if (unlikely(!buf_len))
1295 return -EINVAL;
1296
1297 lock_sock(sk);
1298
1299 if (unlikely(sock->state == SS_UNCONNECTED)) {
1300 res = -ENOTCONN;
1301 goto exit;
1302 }
1303
1304 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1305restart:
1306
1307 /* Look for a message in receive queue; wait if necessary */
1308 res = tipc_wait_for_rcvmsg(sock, &timeo);
1309 if (res)
1310 goto exit;
1311
1312 /* Look at first message in receive queue */
1313 buf = skb_peek(&sk->sk_receive_queue);
1314 msg = buf_msg(buf);
1315 sz = msg_data_sz(msg);
1316 err = msg_errcode(msg);
1317
1318 /* Discard an empty non-errored message & try again */
1319 if ((!sz) && (!err)) {
1320 tsk_advance_rx_queue(sk);
1321 goto restart;
1322 }
1323
1324 /* Capture sender's address (optional) */
1325 set_orig_addr(m, msg);
1326
1327 /* Capture ancillary data (optional) */
1328 res = tipc_sk_anc_data_recv(m, msg, tsk);
1329 if (res)
1330 goto exit;
1331
1332 /* Capture message data (if valid) & compute return value (always) */
1333 if (!err) {
1334 if (unlikely(buf_len < sz)) {
1335 sz = buf_len;
1336 m->msg_flags |= MSG_TRUNC;
1337 }
1338 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1339 if (res)
1340 goto exit;
1341 res = sz;
1342 } else {
1343 if ((sock->state == SS_READY) ||
1344 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1345 res = 0;
1346 else
1347 res = -ECONNRESET;
1348 }
1349
1350 /* Consume received message (optional) */
1351 if (likely(!(flags & MSG_PEEK))) {
1352 if ((sock->state != SS_READY) &&
1353 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1354 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1355 tsk->rcv_unacked = 0;
1356 }
1357 tsk_advance_rx_queue(sk);
1358 }
1359exit:
1360 release_sock(sk);
1361 return res;
1362}
1363
1364/**
1365 * tipc_recv_stream - receive stream-oriented data
1366 * @m: descriptor for message info
1367 * @buf_len: total size of user buffer area
1368 * @flags: receive flags
1369 *
1370 * Used for SOCK_STREAM messages only. If not enough data is available
1371 * will optionally wait for more; never truncates data.
1372 *
1373 * Returns size of returned message data, errno otherwise
1374 */
1375static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1376 size_t buf_len, int flags)
1377{
1378 struct sock *sk = sock->sk;
1379 struct tipc_sock *tsk = tipc_sk(sk);
1380 struct sk_buff *buf;
1381 struct tipc_msg *msg;
1382 long timeo;
1383 unsigned int sz;
1384 int sz_to_copy, target, needed;
1385 int sz_copied = 0;
1386 u32 err;
1387 int res = 0;
1388
1389 /* Catch invalid receive attempts */
1390 if (unlikely(!buf_len))
1391 return -EINVAL;
1392
1393 lock_sock(sk);
1394
1395 if (unlikely(sock->state == SS_UNCONNECTED)) {
1396 res = -ENOTCONN;
1397 goto exit;
1398 }
1399
1400 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1401 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1402
1403restart:
1404 /* Look for a message in receive queue; wait if necessary */
1405 res = tipc_wait_for_rcvmsg(sock, &timeo);
1406 if (res)
1407 goto exit;
1408
1409 /* Look at first message in receive queue */
1410 buf = skb_peek(&sk->sk_receive_queue);
1411 msg = buf_msg(buf);
1412 sz = msg_data_sz(msg);
1413 err = msg_errcode(msg);
1414
1415 /* Discard an empty non-errored message & try again */
1416 if ((!sz) && (!err)) {
1417 tsk_advance_rx_queue(sk);
1418 goto restart;
1419 }
1420
1421 /* Optionally capture sender's address & ancillary data of first msg */
1422 if (sz_copied == 0) {
1423 set_orig_addr(m, msg);
1424 res = tipc_sk_anc_data_recv(m, msg, tsk);
1425 if (res)
1426 goto exit;
1427 }
1428
1429 /* Capture message data (if valid) & compute return value (always) */
1430 if (!err) {
1431 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1432
1433 sz -= offset;
1434 needed = (buf_len - sz_copied);
1435 sz_to_copy = (sz <= needed) ? sz : needed;
1436
1437 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1438 m, sz_to_copy);
1439 if (res)
1440 goto exit;
1441
1442 sz_copied += sz_to_copy;
1443
1444 if (sz_to_copy < sz) {
1445 if (!(flags & MSG_PEEK))
1446 TIPC_SKB_CB(buf)->handle =
1447 (void *)(unsigned long)(offset + sz_to_copy);
1448 goto exit;
1449 }
1450 } else {
1451 if (sz_copied != 0)
1452 goto exit; /* can't add error msg to valid data */
1453
1454 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1455 res = 0;
1456 else
1457 res = -ECONNRESET;
1458 }
1459
1460 /* Consume received message (optional) */
1461 if (likely(!(flags & MSG_PEEK))) {
1462 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1463 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1464 tsk->rcv_unacked = 0;
1465 }
1466 tsk_advance_rx_queue(sk);
1467 }
1468
1469 /* Loop around if more data is required */
1470 if ((sz_copied < buf_len) && /* didn't get all requested data */
1471 (!skb_queue_empty(&sk->sk_receive_queue) ||
1472 (sz_copied < target)) && /* and more is ready or required */
1473 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1474 (!err)) /* and haven't reached a FIN */
1475 goto restart;
1476
1477exit:
1478 release_sock(sk);
1479 return sz_copied ? sz_copied : res;
1480}
1481
1482/**
1483 * tipc_write_space - wake up thread if port congestion is released
1484 * @sk: socket
1485 */
1486static void tipc_write_space(struct sock *sk)
1487{
1488 struct socket_wq *wq;
1489
1490 rcu_read_lock();
1491 wq = rcu_dereference(sk->sk_wq);
1492 if (skwq_has_sleeper(wq))
1493 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1494 POLLWRNORM | POLLWRBAND);
1495 rcu_read_unlock();
1496}
1497
1498/**
1499 * tipc_data_ready - wake up threads to indicate messages have been received
1500 * @sk: socket
1501 * @len: the length of messages
1502 */
1503static void tipc_data_ready(struct sock *sk)
1504{
1505 struct socket_wq *wq;
1506
1507 rcu_read_lock();
1508 wq = rcu_dereference(sk->sk_wq);
1509 if (skwq_has_sleeper(wq))
1510 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1511 POLLRDNORM | POLLRDBAND);
1512 rcu_read_unlock();
1513}
1514
1515static void tipc_sock_destruct(struct sock *sk)
1516{
1517 __skb_queue_purge(&sk->sk_receive_queue);
1518}
1519
1520/**
1521 * filter_connect - Handle all incoming messages for a connection-based socket
1522 * @tsk: TIPC socket
1523 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1524 *
1525 * Returns true if everything ok, false otherwise
1526 */
1527static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1528{
1529 struct sock *sk = &tsk->sk;
1530 struct net *net = sock_net(sk);
1531 struct socket *sock = sk->sk_socket;
1532 struct tipc_msg *hdr = buf_msg(skb);
1533
1534 if (unlikely(msg_mcast(hdr)))
1535 return false;
1536
1537 switch ((int)sock->state) {
1538 case SS_CONNECTED:
1539
1540 /* Accept only connection-based messages sent by peer */
1541 if (unlikely(!tsk_peer_msg(tsk, hdr)))
1542 return false;
1543
1544 if (unlikely(msg_errcode(hdr))) {
1545 sock->state = SS_DISCONNECTING;
1546 tsk->connected = 0;
1547 /* Let timer expire on it's own */
1548 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1549 tsk->portid);
1550 }
1551 return true;
1552
1553 case SS_CONNECTING:
1554
1555 /* Accept only ACK or NACK message */
1556 if (unlikely(!msg_connected(hdr)))
1557 return false;
1558
1559 if (unlikely(msg_errcode(hdr))) {
1560 sock->state = SS_DISCONNECTING;
1561 sk->sk_err = ECONNREFUSED;
1562 return true;
1563 }
1564
1565 if (unlikely(!msg_isdata(hdr))) {
1566 sock->state = SS_DISCONNECTING;
1567 sk->sk_err = EINVAL;
1568 return true;
1569 }
1570
1571 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1572 msg_set_importance(&tsk->phdr, msg_importance(hdr));
1573 sock->state = SS_CONNECTED;
1574
1575 /* If 'ACK+' message, add to socket receive queue */
1576 if (msg_data_sz(hdr))
1577 return true;
1578
1579 /* If empty 'ACK-' message, wake up sleeping connect() */
1580 if (waitqueue_active(sk_sleep(sk)))
1581 wake_up_interruptible(sk_sleep(sk));
1582
1583 /* 'ACK-' message is neither accepted nor rejected: */
1584 msg_set_dest_droppable(hdr, 1);
1585 return false;
1586
1587 case SS_LISTENING:
1588 case SS_UNCONNECTED:
1589
1590 /* Accept only SYN message */
1591 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1592 return true;
1593 break;
1594 case SS_DISCONNECTING:
1595 break;
1596 default:
1597 pr_err("Unknown socket state %u\n", sock->state);
1598 }
1599 return false;
1600}
1601
1602/**
1603 * rcvbuf_limit - get proper overload limit of socket receive queue
1604 * @sk: socket
1605 * @buf: message
1606 *
1607 * For all connection oriented messages, irrespective of importance,
1608 * the default overload value (i.e. 67MB) is set as limit.
1609 *
1610 * For all connectionless messages, by default new queue limits are
1611 * as belows:
1612 *
1613 * TIPC_LOW_IMPORTANCE (4 MB)
1614 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1615 * TIPC_HIGH_IMPORTANCE (16 MB)
1616 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1617 *
1618 * Returns overload limit according to corresponding message importance
1619 */
1620static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1621{
1622 struct tipc_msg *msg = buf_msg(buf);
1623
1624 if (msg_connected(msg))
1625 return sysctl_tipc_rmem[2];
1626
1627 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1628 msg_importance(msg);
1629}
1630
1631/**
1632 * filter_rcv - validate incoming message
1633 * @sk: socket
1634 * @skb: pointer to message.
1635 *
1636 * Enqueues message on receive queue if acceptable; optionally handles
1637 * disconnect indication for a connected socket.
1638 *
1639 * Called with socket lock already taken
1640 *
1641 * Returns true if message was added to socket receive queue, otherwise false
1642 */
1643static bool filter_rcv(struct sock *sk, struct sk_buff *skb)
1644{
1645 struct socket *sock = sk->sk_socket;
1646 struct tipc_sock *tsk = tipc_sk(sk);
1647 struct tipc_msg *hdr = buf_msg(skb);
1648 unsigned int limit = rcvbuf_limit(sk, skb);
1649 int err = TIPC_OK;
1650 int usr = msg_user(hdr);
1651
1652 if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1653 tipc_sk_proto_rcv(tsk, skb);
1654 return false;
1655 }
1656
1657 if (unlikely(usr == SOCK_WAKEUP)) {
1658 kfree_skb(skb);
1659 tsk->link_cong = 0;
1660 sk->sk_write_space(sk);
1661 return false;
1662 }
1663
1664 /* Drop if illegal message type */
1665 if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1666 kfree_skb(skb);
1667 return false;
1668 }
1669
1670 /* Reject if wrong message type for current socket state */
1671 if (unlikely(sock->state == SS_READY)) {
1672 if (msg_connected(hdr)) {
1673 err = TIPC_ERR_NO_PORT;
1674 goto reject;
1675 }
1676 } else if (unlikely(!filter_connect(tsk, skb))) {
1677 err = TIPC_ERR_NO_PORT;
1678 goto reject;
1679 }
1680
1681 /* Reject message if there isn't room to queue it */
1682 if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1683 err = TIPC_ERR_OVERLOAD;
1684 goto reject;
1685 }
1686
1687 /* Enqueue message */
1688 TIPC_SKB_CB(skb)->handle = NULL;
1689 __skb_queue_tail(&sk->sk_receive_queue, skb);
1690 skb_set_owner_r(skb, sk);
1691
1692 sk->sk_data_ready(sk);
1693 return true;
1694
1695reject:
1696 tipc_sk_respond(sk, skb, err);
1697 return false;
1698}
1699
1700/**
1701 * tipc_backlog_rcv - handle incoming message from backlog queue
1702 * @sk: socket
1703 * @skb: message
1704 *
1705 * Caller must hold socket lock
1706 *
1707 * Returns 0
1708 */
1709static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1710{
1711 unsigned int truesize = skb->truesize;
1712
1713 if (likely(filter_rcv(sk, skb)))
1714 atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1715 return 0;
1716}
1717
1718/**
1719 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1720 * inputq and try adding them to socket or backlog queue
1721 * @inputq: list of incoming buffers with potentially different destinations
1722 * @sk: socket where the buffers should be enqueued
1723 * @dport: port number for the socket
1724 *
1725 * Caller must hold socket lock
1726 */
1727static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1728 u32 dport)
1729{
1730 unsigned int lim;
1731 atomic_t *dcnt;
1732 struct sk_buff *skb;
1733 unsigned long time_limit = jiffies + 2;
1734
1735 while (skb_queue_len(inputq)) {
1736 if (unlikely(time_after_eq(jiffies, time_limit)))
1737 return;
1738
1739 skb = tipc_skb_dequeue(inputq, dport);
1740 if (unlikely(!skb))
1741 return;
1742
1743 /* Add message directly to receive queue if possible */
1744 if (!sock_owned_by_user(sk)) {
1745 filter_rcv(sk, skb);
1746 continue;
1747 }
1748
1749 /* Try backlog, compensating for double-counted bytes */
1750 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1751 if (sk->sk_backlog.len)
1752 atomic_set(dcnt, 0);
1753 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1754 if (likely(!sk_add_backlog(sk, skb, lim)))
1755 continue;
1756
1757 /* Overload => reject message back to sender */
1758 tipc_sk_respond(sk, skb, TIPC_ERR_OVERLOAD);
1759 break;
1760 }
1761}
1762
1763/**
1764 * tipc_sk_rcv - handle a chain of incoming buffers
1765 * @inputq: buffer list containing the buffers
1766 * Consumes all buffers in list until inputq is empty
1767 * Note: may be called in multiple threads referring to the same queue
1768 */
1769void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1770{
1771 u32 dnode, dport = 0;
1772 int err;
1773 struct tipc_sock *tsk;
1774 struct sock *sk;
1775 struct sk_buff *skb;
1776
1777 while (skb_queue_len(inputq)) {
1778 dport = tipc_skb_peek_port(inputq, dport);
1779 tsk = tipc_sk_lookup(net, dport);
1780
1781 if (likely(tsk)) {
1782 sk = &tsk->sk;
1783 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1784 tipc_sk_enqueue(inputq, sk, dport);
1785 spin_unlock_bh(&sk->sk_lock.slock);
1786 }
1787 sock_put(sk);
1788 continue;
1789 }
1790
1791 /* No destination socket => dequeue skb if still there */
1792 skb = tipc_skb_dequeue(inputq, dport);
1793 if (!skb)
1794 return;
1795
1796 /* Try secondary lookup if unresolved named message */
1797 err = TIPC_ERR_NO_PORT;
1798 if (tipc_msg_lookup_dest(net, skb, &err))
1799 goto xmit;
1800
1801 /* Prepare for message rejection */
1802 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1803 continue;
1804xmit:
1805 dnode = msg_destnode(buf_msg(skb));
1806 tipc_node_xmit_skb(net, skb, dnode, dport);
1807 }
1808}
1809
1810static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1811{
1812 struct sock *sk = sock->sk;
1813 DEFINE_WAIT(wait);
1814 int done;
1815
1816 do {
1817 int err = sock_error(sk);
1818 if (err)
1819 return err;
1820 if (!*timeo_p)
1821 return -ETIMEDOUT;
1822 if (signal_pending(current))
1823 return sock_intr_errno(*timeo_p);
1824
1825 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1826 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1827 finish_wait(sk_sleep(sk), &wait);
1828 } while (!done);
1829 return 0;
1830}
1831
1832/**
1833 * tipc_connect - establish a connection to another TIPC port
1834 * @sock: socket structure
1835 * @dest: socket address for destination port
1836 * @destlen: size of socket address data structure
1837 * @flags: file-related flags associated with socket
1838 *
1839 * Returns 0 on success, errno otherwise
1840 */
1841static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1842 int destlen, int flags)
1843{
1844 struct sock *sk = sock->sk;
1845 struct tipc_sock *tsk = tipc_sk(sk);
1846 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1847 struct msghdr m = {NULL,};
1848 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1849 socket_state previous;
1850 int res = 0;
1851
1852 lock_sock(sk);
1853
1854 /* DGRAM/RDM connect(), just save the destaddr */
1855 if (sock->state == SS_READY) {
1856 if (dst->family == AF_UNSPEC) {
1857 memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1858 tsk->connected = 0;
1859 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1860 res = -EINVAL;
1861 } else {
1862 memcpy(&tsk->remote, dest, destlen);
1863 tsk->connected = 1;
1864 }
1865 goto exit;
1866 }
1867
1868 /*
1869 * Reject connection attempt using multicast address
1870 *
1871 * Note: send_msg() validates the rest of the address fields,
1872 * so there's no need to do it here
1873 */
1874 if (dst->addrtype == TIPC_ADDR_MCAST) {
1875 res = -EINVAL;
1876 goto exit;
1877 }
1878
1879 previous = sock->state;
1880 switch (sock->state) {
1881 case SS_UNCONNECTED:
1882 /* Send a 'SYN-' to destination */
1883 m.msg_name = dest;
1884 m.msg_namelen = destlen;
1885
1886 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1887 * indicate send_msg() is never blocked.
1888 */
1889 if (!timeout)
1890 m.msg_flags = MSG_DONTWAIT;
1891
1892 res = __tipc_sendmsg(sock, &m, 0);
1893 if ((res < 0) && (res != -EWOULDBLOCK))
1894 goto exit;
1895
1896 /* Just entered SS_CONNECTING state; the only
1897 * difference is that return value in non-blocking
1898 * case is EINPROGRESS, rather than EALREADY.
1899 */
1900 res = -EINPROGRESS;
1901 case SS_CONNECTING:
1902 if (previous == SS_CONNECTING)
1903 res = -EALREADY;
1904 if (!timeout)
1905 goto exit;
1906 timeout = msecs_to_jiffies(timeout);
1907 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1908 res = tipc_wait_for_connect(sock, &timeout);
1909 break;
1910 case SS_CONNECTED:
1911 res = -EISCONN;
1912 break;
1913 default:
1914 res = -EINVAL;
1915 break;
1916 }
1917exit:
1918 release_sock(sk);
1919 return res;
1920}
1921
1922/**
1923 * tipc_listen - allow socket to listen for incoming connections
1924 * @sock: socket structure
1925 * @len: (unused)
1926 *
1927 * Returns 0 on success, errno otherwise
1928 */
1929static int tipc_listen(struct socket *sock, int len)
1930{
1931 struct sock *sk = sock->sk;
1932 int res;
1933
1934 lock_sock(sk);
1935
1936 if (sock->state != SS_UNCONNECTED)
1937 res = -EINVAL;
1938 else {
1939 sock->state = SS_LISTENING;
1940 res = 0;
1941 }
1942
1943 release_sock(sk);
1944 return res;
1945}
1946
1947static int tipc_wait_for_accept(struct socket *sock, long timeo)
1948{
1949 struct sock *sk = sock->sk;
1950 DEFINE_WAIT(wait);
1951 int err;
1952
1953 /* True wake-one mechanism for incoming connections: only
1954 * one process gets woken up, not the 'whole herd'.
1955 * Since we do not 'race & poll' for established sockets
1956 * anymore, the common case will execute the loop only once.
1957 */
1958 for (;;) {
1959 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1960 TASK_INTERRUPTIBLE);
1961 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1962 release_sock(sk);
1963 timeo = schedule_timeout(timeo);
1964 lock_sock(sk);
1965 }
1966 err = 0;
1967 if (!skb_queue_empty(&sk->sk_receive_queue))
1968 break;
1969 err = -EINVAL;
1970 if (sock->state != SS_LISTENING)
1971 break;
1972 err = -EAGAIN;
1973 if (!timeo)
1974 break;
1975 err = sock_intr_errno(timeo);
1976 if (signal_pending(current))
1977 break;
1978 }
1979 finish_wait(sk_sleep(sk), &wait);
1980 return err;
1981}
1982
1983/**
1984 * tipc_accept - wait for connection request
1985 * @sock: listening socket
1986 * @newsock: new socket that is to be connected
1987 * @flags: file-related flags associated with socket
1988 *
1989 * Returns 0 on success, errno otherwise
1990 */
1991static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1992{
1993 struct sock *new_sk, *sk = sock->sk;
1994 struct sk_buff *buf;
1995 struct tipc_sock *new_tsock;
1996 struct tipc_msg *msg;
1997 long timeo;
1998 int res;
1999
2000 lock_sock(sk);
2001
2002 if (sock->state != SS_LISTENING) {
2003 res = -EINVAL;
2004 goto exit;
2005 }
2006 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2007 res = tipc_wait_for_accept(sock, timeo);
2008 if (res)
2009 goto exit;
2010
2011 buf = skb_peek(&sk->sk_receive_queue);
2012
2013 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2014 if (res)
2015 goto exit;
2016 security_sk_clone(sock->sk, new_sock->sk);
2017
2018 new_sk = new_sock->sk;
2019 new_tsock = tipc_sk(new_sk);
2020 msg = buf_msg(buf);
2021
2022 /* we lock on new_sk; but lockdep sees the lock on sk */
2023 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2024
2025 /*
2026 * Reject any stray messages received by new socket
2027 * before the socket lock was taken (very, very unlikely)
2028 */
2029 tsk_rej_rx_queue(new_sk);
2030
2031 /* Connect new socket to it's peer */
2032 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2033 new_sock->state = SS_CONNECTED;
2034
2035 tsk_set_importance(new_tsock, msg_importance(msg));
2036 if (msg_named(msg)) {
2037 new_tsock->conn_type = msg_nametype(msg);
2038 new_tsock->conn_instance = msg_nameinst(msg);
2039 }
2040
2041 /*
2042 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2043 * Respond to 'SYN+' by queuing it on new socket.
2044 */
2045 if (!msg_data_sz(msg)) {
2046 struct msghdr m = {NULL,};
2047
2048 tsk_advance_rx_queue(sk);
2049 __tipc_send_stream(new_sock, &m, 0);
2050 } else {
2051 __skb_dequeue(&sk->sk_receive_queue);
2052 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2053 skb_set_owner_r(buf, new_sk);
2054 }
2055 release_sock(new_sk);
2056exit:
2057 release_sock(sk);
2058 return res;
2059}
2060
2061/**
2062 * tipc_shutdown - shutdown socket connection
2063 * @sock: socket structure
2064 * @how: direction to close (must be SHUT_RDWR)
2065 *
2066 * Terminates connection (if necessary), then purges socket's receive queue.
2067 *
2068 * Returns 0 on success, errno otherwise
2069 */
2070static int tipc_shutdown(struct socket *sock, int how)
2071{
2072 struct sock *sk = sock->sk;
2073 struct net *net = sock_net(sk);
2074 struct tipc_sock *tsk = tipc_sk(sk);
2075 struct sk_buff *skb;
2076 u32 dnode = tsk_peer_node(tsk);
2077 u32 dport = tsk_peer_port(tsk);
2078 u32 onode = tipc_own_addr(net);
2079 u32 oport = tsk->portid;
2080 int res;
2081
2082 if (how != SHUT_RDWR)
2083 return -EINVAL;
2084
2085 lock_sock(sk);
2086
2087 switch (sock->state) {
2088 case SS_CONNECTING:
2089 case SS_CONNECTED:
2090
2091restart:
2092 dnode = tsk_peer_node(tsk);
2093
2094 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2095 skb = __skb_dequeue(&sk->sk_receive_queue);
2096 if (skb) {
2097 if (TIPC_SKB_CB(skb)->handle != NULL) {
2098 kfree_skb(skb);
2099 goto restart;
2100 }
2101 tipc_sk_respond(sk, skb, TIPC_CONN_SHUTDOWN);
2102 } else {
2103 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2104 TIPC_CONN_MSG, SHORT_H_SIZE,
2105 0, dnode, onode, dport, oport,
2106 TIPC_CONN_SHUTDOWN);
2107 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
2108 }
2109 tsk->connected = 0;
2110 sock->state = SS_DISCONNECTING;
2111 tipc_node_remove_conn(net, dnode, tsk->portid);
2112 /* fall through */
2113
2114 case SS_DISCONNECTING:
2115
2116 /* Discard any unreceived messages */
2117 __skb_queue_purge(&sk->sk_receive_queue);
2118
2119 /* Wake up anyone sleeping in poll */
2120 sk->sk_state_change(sk);
2121 res = 0;
2122 break;
2123
2124 default:
2125 res = -ENOTCONN;
2126 }
2127
2128 release_sock(sk);
2129 return res;
2130}
2131
2132static void tipc_sk_timeout(unsigned long data)
2133{
2134 struct tipc_sock *tsk = (struct tipc_sock *)data;
2135 struct sock *sk = &tsk->sk;
2136 struct sk_buff *skb = NULL;
2137 u32 peer_port, peer_node;
2138 u32 own_node = tsk_own_node(tsk);
2139
2140 bh_lock_sock(sk);
2141 if (!tsk->connected) {
2142 bh_unlock_sock(sk);
2143 goto exit;
2144 }
2145 peer_port = tsk_peer_port(tsk);
2146 peer_node = tsk_peer_node(tsk);
2147
2148 if (tsk->probing_state == TIPC_CONN_PROBING) {
2149 if (!sock_owned_by_user(sk)) {
2150 sk->sk_socket->state = SS_DISCONNECTING;
2151 tsk->connected = 0;
2152 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2153 tsk_peer_port(tsk));
2154 sk->sk_state_change(sk);
2155 } else {
2156 /* Try again later */
2157 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2158 }
2159
2160 } else {
2161 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2162 INT_H_SIZE, 0, peer_node, own_node,
2163 peer_port, tsk->portid, TIPC_OK);
2164 tsk->probing_state = TIPC_CONN_PROBING;
2165 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2166 }
2167 bh_unlock_sock(sk);
2168 if (skb)
2169 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2170exit:
2171 sock_put(sk);
2172}
2173
2174static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2175 struct tipc_name_seq const *seq)
2176{
2177 struct net *net = sock_net(&tsk->sk);
2178 struct publication *publ;
2179 u32 key;
2180
2181 if (tsk->connected)
2182 return -EINVAL;
2183 key = tsk->portid + tsk->pub_count + 1;
2184 if (key == tsk->portid)
2185 return -EADDRINUSE;
2186
2187 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2188 scope, tsk->portid, key);
2189 if (unlikely(!publ))
2190 return -EINVAL;
2191
2192 list_add(&publ->pport_list, &tsk->publications);
2193 tsk->pub_count++;
2194 tsk->published = 1;
2195 return 0;
2196}
2197
2198static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2199 struct tipc_name_seq const *seq)
2200{
2201 struct net *net = sock_net(&tsk->sk);
2202 struct publication *publ;
2203 struct publication *safe;
2204 int rc = -EINVAL;
2205
2206 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2207 if (seq) {
2208 if (publ->scope != scope)
2209 continue;
2210 if (publ->type != seq->type)
2211 continue;
2212 if (publ->lower != seq->lower)
2213 continue;
2214 if (publ->upper != seq->upper)
2215 break;
2216 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2217 publ->ref, publ->key);
2218 rc = 0;
2219 break;
2220 }
2221 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2222 publ->ref, publ->key);
2223 rc = 0;
2224 }
2225 if (list_empty(&tsk->publications))
2226 tsk->published = 0;
2227 return rc;
2228}
2229
2230/* tipc_sk_reinit: set non-zero address in all existing sockets
2231 * when we go from standalone to network mode.
2232 */
2233void tipc_sk_reinit(struct net *net)
2234{
2235 struct tipc_net *tn = net_generic(net, tipc_net_id);
2236 const struct bucket_table *tbl;
2237 struct rhash_head *pos;
2238 struct tipc_sock *tsk;
2239 struct tipc_msg *msg;
2240 int i;
2241
2242 rcu_read_lock();
2243 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2244 for (i = 0; i < tbl->size; i++) {
2245 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2246 spin_lock_bh(&tsk->sk.sk_lock.slock);
2247 msg = &tsk->phdr;
2248 msg_set_prevnode(msg, tn->own_addr);
2249 msg_set_orignode(msg, tn->own_addr);
2250 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2251 }
2252 }
2253 rcu_read_unlock();
2254}
2255
2256static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2257{
2258 struct tipc_net *tn = net_generic(net, tipc_net_id);
2259 struct tipc_sock *tsk;
2260
2261 rcu_read_lock();
2262 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2263 if (tsk)
2264 sock_hold(&tsk->sk);
2265 rcu_read_unlock();
2266
2267 return tsk;
2268}
2269
2270static int tipc_sk_insert(struct tipc_sock *tsk)
2271{
2272 struct sock *sk = &tsk->sk;
2273 struct net *net = sock_net(sk);
2274 struct tipc_net *tn = net_generic(net, tipc_net_id);
2275 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2276 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2277
2278 while (remaining--) {
2279 portid++;
2280 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2281 portid = TIPC_MIN_PORT;
2282 tsk->portid = portid;
2283 sock_hold(&tsk->sk);
2284 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2285 tsk_rht_params))
2286 return 0;
2287 sock_put(&tsk->sk);
2288 }
2289
2290 return -1;
2291}
2292
2293static void tipc_sk_remove(struct tipc_sock *tsk)
2294{
2295 struct sock *sk = &tsk->sk;
2296 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2297
2298 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2299 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2300 __sock_put(sk);
2301 }
2302}
2303
2304static const struct rhashtable_params tsk_rht_params = {
2305 .nelem_hint = 192,
2306 .head_offset = offsetof(struct tipc_sock, node),
2307 .key_offset = offsetof(struct tipc_sock, portid),
2308 .key_len = sizeof(u32), /* portid */
2309 .max_size = 1048576,
2310 .min_size = 256,
2311 .automatic_shrinking = true,
2312};
2313
2314int tipc_sk_rht_init(struct net *net)
2315{
2316 struct tipc_net *tn = net_generic(net, tipc_net_id);
2317
2318 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2319}
2320
2321void tipc_sk_rht_destroy(struct net *net)
2322{
2323 struct tipc_net *tn = net_generic(net, tipc_net_id);
2324
2325 /* Wait for socket readers to complete */
2326 synchronize_net();
2327
2328 rhashtable_destroy(&tn->sk_rht);
2329}
2330
2331/**
2332 * tipc_setsockopt - set socket option
2333 * @sock: socket structure
2334 * @lvl: option level
2335 * @opt: option identifier
2336 * @ov: pointer to new option value
2337 * @ol: length of option value
2338 *
2339 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2340 * (to ease compatibility).
2341 *
2342 * Returns 0 on success, errno otherwise
2343 */
2344static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2345 char __user *ov, unsigned int ol)
2346{
2347 struct sock *sk = sock->sk;
2348 struct tipc_sock *tsk = tipc_sk(sk);
2349 u32 value;
2350 int res;
2351
2352 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2353 return 0;
2354 if (lvl != SOL_TIPC)
2355 return -ENOPROTOOPT;
2356 if (ol < sizeof(value))
2357 return -EINVAL;
2358 res = get_user(value, (u32 __user *)ov);
2359 if (res)
2360 return res;
2361
2362 lock_sock(sk);
2363
2364 switch (opt) {
2365 case TIPC_IMPORTANCE:
2366 res = tsk_set_importance(tsk, value);
2367 break;
2368 case TIPC_SRC_DROPPABLE:
2369 if (sock->type != SOCK_STREAM)
2370 tsk_set_unreliable(tsk, value);
2371 else
2372 res = -ENOPROTOOPT;
2373 break;
2374 case TIPC_DEST_DROPPABLE:
2375 tsk_set_unreturnable(tsk, value);
2376 break;
2377 case TIPC_CONN_TIMEOUT:
2378 tipc_sk(sk)->conn_timeout = value;
2379 /* no need to set "res", since already 0 at this point */
2380 break;
2381 default:
2382 res = -EINVAL;
2383 }
2384
2385 release_sock(sk);
2386
2387 return res;
2388}
2389
2390/**
2391 * tipc_getsockopt - get socket option
2392 * @sock: socket structure
2393 * @lvl: option level
2394 * @opt: option identifier
2395 * @ov: receptacle for option value
2396 * @ol: receptacle for length of option value
2397 *
2398 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2399 * (to ease compatibility).
2400 *
2401 * Returns 0 on success, errno otherwise
2402 */
2403static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2404 char __user *ov, int __user *ol)
2405{
2406 struct sock *sk = sock->sk;
2407 struct tipc_sock *tsk = tipc_sk(sk);
2408 int len;
2409 u32 value;
2410 int res;
2411
2412 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2413 return put_user(0, ol);
2414 if (lvl != SOL_TIPC)
2415 return -ENOPROTOOPT;
2416 res = get_user(len, ol);
2417 if (res)
2418 return res;
2419
2420 lock_sock(sk);
2421
2422 switch (opt) {
2423 case TIPC_IMPORTANCE:
2424 value = tsk_importance(tsk);
2425 break;
2426 case TIPC_SRC_DROPPABLE:
2427 value = tsk_unreliable(tsk);
2428 break;
2429 case TIPC_DEST_DROPPABLE:
2430 value = tsk_unreturnable(tsk);
2431 break;
2432 case TIPC_CONN_TIMEOUT:
2433 value = tsk->conn_timeout;
2434 /* no need to set "res", since already 0 at this point */
2435 break;
2436 case TIPC_NODE_RECVQ_DEPTH:
2437 value = 0; /* was tipc_queue_size, now obsolete */
2438 break;
2439 case TIPC_SOCK_RECVQ_DEPTH:
2440 value = skb_queue_len(&sk->sk_receive_queue);
2441 break;
2442 default:
2443 res = -EINVAL;
2444 }
2445
2446 release_sock(sk);
2447
2448 if (res)
2449 return res; /* "get" failed */
2450
2451 if (len < sizeof(value))
2452 return -EINVAL;
2453
2454 if (copy_to_user(ov, &value, sizeof(value)))
2455 return -EFAULT;
2456
2457 return put_user(sizeof(value), ol);
2458}
2459
2460static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2461{
2462 struct sock *sk = sock->sk;
2463 struct tipc_sioc_ln_req lnr;
2464 void __user *argp = (void __user *)arg;
2465
2466 switch (cmd) {
2467 case SIOCGETLINKNAME:
2468 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2469 return -EFAULT;
2470 if (!tipc_node_get_linkname(sock_net(sk),
2471 lnr.bearer_id & 0xffff, lnr.peer,
2472 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2473 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2474 return -EFAULT;
2475 return 0;
2476 }
2477 return -EADDRNOTAVAIL;
2478 default:
2479 return -ENOIOCTLCMD;
2480 }
2481}
2482
2483/* Protocol switches for the various types of TIPC sockets */
2484
2485static const struct proto_ops msg_ops = {
2486 .owner = THIS_MODULE,
2487 .family = AF_TIPC,
2488 .release = tipc_release,
2489 .bind = tipc_bind,
2490 .connect = tipc_connect,
2491 .socketpair = sock_no_socketpair,
2492 .accept = sock_no_accept,
2493 .getname = tipc_getname,
2494 .poll = tipc_poll,
2495 .ioctl = tipc_ioctl,
2496 .listen = sock_no_listen,
2497 .shutdown = tipc_shutdown,
2498 .setsockopt = tipc_setsockopt,
2499 .getsockopt = tipc_getsockopt,
2500 .sendmsg = tipc_sendmsg,
2501 .recvmsg = tipc_recvmsg,
2502 .mmap = sock_no_mmap,
2503 .sendpage = sock_no_sendpage
2504};
2505
2506static const struct proto_ops packet_ops = {
2507 .owner = THIS_MODULE,
2508 .family = AF_TIPC,
2509 .release = tipc_release,
2510 .bind = tipc_bind,
2511 .connect = tipc_connect,
2512 .socketpair = sock_no_socketpair,
2513 .accept = tipc_accept,
2514 .getname = tipc_getname,
2515 .poll = tipc_poll,
2516 .ioctl = tipc_ioctl,
2517 .listen = tipc_listen,
2518 .shutdown = tipc_shutdown,
2519 .setsockopt = tipc_setsockopt,
2520 .getsockopt = tipc_getsockopt,
2521 .sendmsg = tipc_send_packet,
2522 .recvmsg = tipc_recvmsg,
2523 .mmap = sock_no_mmap,
2524 .sendpage = sock_no_sendpage
2525};
2526
2527static const struct proto_ops stream_ops = {
2528 .owner = THIS_MODULE,
2529 .family = AF_TIPC,
2530 .release = tipc_release,
2531 .bind = tipc_bind,
2532 .connect = tipc_connect,
2533 .socketpair = sock_no_socketpair,
2534 .accept = tipc_accept,
2535 .getname = tipc_getname,
2536 .poll = tipc_poll,
2537 .ioctl = tipc_ioctl,
2538 .listen = tipc_listen,
2539 .shutdown = tipc_shutdown,
2540 .setsockopt = tipc_setsockopt,
2541 .getsockopt = tipc_getsockopt,
2542 .sendmsg = tipc_send_stream,
2543 .recvmsg = tipc_recv_stream,
2544 .mmap = sock_no_mmap,
2545 .sendpage = sock_no_sendpage
2546};
2547
2548static const struct net_proto_family tipc_family_ops = {
2549 .owner = THIS_MODULE,
2550 .family = AF_TIPC,
2551 .create = tipc_sk_create
2552};
2553
2554static struct proto tipc_proto = {
2555 .name = "TIPC",
2556 .owner = THIS_MODULE,
2557 .obj_size = sizeof(struct tipc_sock),
2558 .sysctl_rmem = sysctl_tipc_rmem
2559};
2560
2561/**
2562 * tipc_socket_init - initialize TIPC socket interface
2563 *
2564 * Returns 0 on success, errno otherwise
2565 */
2566int tipc_socket_init(void)
2567{
2568 int res;
2569
2570 res = proto_register(&tipc_proto, 1);
2571 if (res) {
2572 pr_err("Failed to register TIPC protocol type\n");
2573 goto out;
2574 }
2575
2576 res = sock_register(&tipc_family_ops);
2577 if (res) {
2578 pr_err("Failed to register TIPC socket type\n");
2579 proto_unregister(&tipc_proto);
2580 goto out;
2581 }
2582 out:
2583 return res;
2584}
2585
2586/**
2587 * tipc_socket_stop - stop TIPC socket interface
2588 */
2589void tipc_socket_stop(void)
2590{
2591 sock_unregister(tipc_family_ops.family);
2592 proto_unregister(&tipc_proto);
2593}
2594
2595/* Caller should hold socket lock for the passed tipc socket. */
2596static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2597{
2598 u32 peer_node;
2599 u32 peer_port;
2600 struct nlattr *nest;
2601
2602 peer_node = tsk_peer_node(tsk);
2603 peer_port = tsk_peer_port(tsk);
2604
2605 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2606
2607 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2608 goto msg_full;
2609 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2610 goto msg_full;
2611
2612 if (tsk->conn_type != 0) {
2613 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2614 goto msg_full;
2615 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2616 goto msg_full;
2617 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2618 goto msg_full;
2619 }
2620 nla_nest_end(skb, nest);
2621
2622 return 0;
2623
2624msg_full:
2625 nla_nest_cancel(skb, nest);
2626
2627 return -EMSGSIZE;
2628}
2629
2630/* Caller should hold socket lock for the passed tipc socket. */
2631static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2632 struct tipc_sock *tsk)
2633{
2634 int err;
2635 void *hdr;
2636 struct nlattr *attrs;
2637 struct net *net = sock_net(skb->sk);
2638 struct tipc_net *tn = net_generic(net, tipc_net_id);
2639
2640 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2641 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2642 if (!hdr)
2643 goto msg_cancel;
2644
2645 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2646 if (!attrs)
2647 goto genlmsg_cancel;
2648 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2649 goto attr_msg_cancel;
2650 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2651 goto attr_msg_cancel;
2652
2653 if (tsk->connected) {
2654 err = __tipc_nl_add_sk_con(skb, tsk);
2655 if (err)
2656 goto attr_msg_cancel;
2657 } else if (!list_empty(&tsk->publications)) {
2658 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2659 goto attr_msg_cancel;
2660 }
2661 nla_nest_end(skb, attrs);
2662 genlmsg_end(skb, hdr);
2663
2664 return 0;
2665
2666attr_msg_cancel:
2667 nla_nest_cancel(skb, attrs);
2668genlmsg_cancel:
2669 genlmsg_cancel(skb, hdr);
2670msg_cancel:
2671 return -EMSGSIZE;
2672}
2673
2674int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2675{
2676 int err;
2677 struct tipc_sock *tsk;
2678 const struct bucket_table *tbl;
2679 struct rhash_head *pos;
2680 struct net *net = sock_net(skb->sk);
2681 struct tipc_net *tn = net_generic(net, tipc_net_id);
2682 u32 tbl_id = cb->args[0];
2683 u32 prev_portid = cb->args[1];
2684
2685 rcu_read_lock();
2686 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2687 for (; tbl_id < tbl->size; tbl_id++) {
2688 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2689 spin_lock_bh(&tsk->sk.sk_lock.slock);
2690 if (prev_portid && prev_portid != tsk->portid) {
2691 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2692 continue;
2693 }
2694
2695 err = __tipc_nl_add_sk(skb, cb, tsk);
2696 if (err) {
2697 prev_portid = tsk->portid;
2698 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2699 goto out;
2700 }
2701 prev_portid = 0;
2702 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2703 }
2704 }
2705out:
2706 rcu_read_unlock();
2707 cb->args[0] = tbl_id;
2708 cb->args[1] = prev_portid;
2709
2710 return skb->len;
2711}
2712
2713/* Caller should hold socket lock for the passed tipc socket. */
2714static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2715 struct netlink_callback *cb,
2716 struct publication *publ)
2717{
2718 void *hdr;
2719 struct nlattr *attrs;
2720
2721 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2722 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2723 if (!hdr)
2724 goto msg_cancel;
2725
2726 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2727 if (!attrs)
2728 goto genlmsg_cancel;
2729
2730 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2731 goto attr_msg_cancel;
2732 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2733 goto attr_msg_cancel;
2734 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2735 goto attr_msg_cancel;
2736 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2737 goto attr_msg_cancel;
2738
2739 nla_nest_end(skb, attrs);
2740 genlmsg_end(skb, hdr);
2741
2742 return 0;
2743
2744attr_msg_cancel:
2745 nla_nest_cancel(skb, attrs);
2746genlmsg_cancel:
2747 genlmsg_cancel(skb, hdr);
2748msg_cancel:
2749 return -EMSGSIZE;
2750}
2751
2752/* Caller should hold socket lock for the passed tipc socket. */
2753static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2754 struct netlink_callback *cb,
2755 struct tipc_sock *tsk, u32 *last_publ)
2756{
2757 int err;
2758 struct publication *p;
2759
2760 if (*last_publ) {
2761 list_for_each_entry(p, &tsk->publications, pport_list) {
2762 if (p->key == *last_publ)
2763 break;
2764 }
2765 if (p->key != *last_publ) {
2766 /* We never set seq or call nl_dump_check_consistent()
2767 * this means that setting prev_seq here will cause the
2768 * consistence check to fail in the netlink callback
2769 * handler. Resulting in the last NLMSG_DONE message
2770 * having the NLM_F_DUMP_INTR flag set.
2771 */
2772 cb->prev_seq = 1;
2773 *last_publ = 0;
2774 return -EPIPE;
2775 }
2776 } else {
2777 p = list_first_entry(&tsk->publications, struct publication,
2778 pport_list);
2779 }
2780
2781 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2782 err = __tipc_nl_add_sk_publ(skb, cb, p);
2783 if (err) {
2784 *last_publ = p->key;
2785 return err;
2786 }
2787 }
2788 *last_publ = 0;
2789
2790 return 0;
2791}
2792
2793int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2794{
2795 int err;
2796 u32 tsk_portid = cb->args[0];
2797 u32 last_publ = cb->args[1];
2798 u32 done = cb->args[2];
2799 struct net *net = sock_net(skb->sk);
2800 struct tipc_sock *tsk;
2801
2802 if (!tsk_portid) {
2803 struct nlattr **attrs;
2804 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2805
2806 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2807 if (err)
2808 return err;
2809
2810 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2811 attrs[TIPC_NLA_SOCK],
2812 tipc_nl_sock_policy);
2813 if (err)
2814 return err;
2815
2816 if (!sock[TIPC_NLA_SOCK_REF])
2817 return -EINVAL;
2818
2819 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2820 }
2821
2822 if (done)
2823 return 0;
2824
2825 tsk = tipc_sk_lookup(net, tsk_portid);
2826 if (!tsk)
2827 return -EINVAL;
2828
2829 lock_sock(&tsk->sk);
2830 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2831 if (!err)
2832 done = 1;
2833 release_sock(&tsk->sk);
2834 sock_put(&tsk->sk);
2835
2836 cb->args[0] = tsk_portid;
2837 cb->args[1] = last_publ;
2838 cb->args[2] = done;
2839
2840 return skb->len;
2841}