blob: 648a734e604418ee38eff4da3a95c35e7cdddea1 [file] [log] [blame]
Per Lidenb97bf3f2006-01-02 19:04:38 +01001/*
2 * net/tipc/socket.c: TIPC socket API
3 *
Per Liden593a5f22006-01-11 19:14:19 +01004 * Copyright (c) 2001-2006, Ericsson AB
Per Lidenb97bf3f2006-01-02 19:04:38 +01005 * Copyright (c) 2004-2005, Wind River Systems
Per Lidenb97bf3f2006-01-02 19:04:38 +01006 * All rights reserved.
7 *
Per Liden9ea1fd32006-01-11 13:30:43 +01008 * Redistribution and use in source and binary forms, with or without
Per Lidenb97bf3f2006-01-02 19:04:38 +01009 * modification, are permitted provided that the following conditions are met:
10 *
Per Liden9ea1fd32006-01-11 13:30:43 +010011 * 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.
Per Lidenb97bf3f2006-01-02 19:04:38 +010019 *
Per Liden9ea1fd32006-01-11 13:30:43 +010020 * 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
Per Lidenb97bf3f2006-01-02 19:04:38 +010034 * POSSIBILITY OF SUCH DAMAGE.
35 */
36
37#include <linux/module.h>
38#include <linux/types.h>
39#include <linux/net.h>
40#include <linux/socket.h>
41#include <linux/errno.h>
42#include <linux/mm.h>
43#include <linux/slab.h>
44#include <linux/poll.h>
Per Lidenb97bf3f2006-01-02 19:04:38 +010045#include <linux/fcntl.h>
Per Lidenb97bf3f2006-01-02 19:04:38 +010046#include <asm/semaphore.h>
47#include <asm/string.h>
48#include <asm/atomic.h>
49#include <net/sock.h>
50
51#include <linux/tipc.h>
Per Lidenea714cc2006-01-11 12:28:47 +010052#include <linux/tipc_config.h>
Per Lidenb97bf3f2006-01-02 19:04:38 +010053#include <net/tipc/tipc_msg.h>
54#include <net/tipc/tipc_port.h>
55
56#include "core.h"
57
58#define SS_LISTENING -1 /* socket is listening */
59#define SS_READY -2 /* socket is connectionless */
60
61#define OVERLOAD_LIMIT_BASE 5000
62
63struct tipc_sock {
64 struct sock sk;
65 struct tipc_port *p;
66 struct semaphore sem;
67};
68
69#define tipc_sk(sk) ((struct tipc_sock*)sk)
70
71static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
72static void wakeupdispatch(struct tipc_port *tport);
73
74static struct proto_ops packet_ops;
75static struct proto_ops stream_ops;
76static struct proto_ops msg_ops;
77
78static struct proto tipc_proto;
79
80static int sockets_enabled = 0;
81
82static atomic_t tipc_queue_size = ATOMIC_INIT(0);
83
84
85/*
86 * sock_lock(): Lock a port/socket pair. lock_sock() can
87 * not be used here, since the same lock must protect ports
88 * with non-socket interfaces.
89 * See net.c for description of locking policy.
90 */
Sam Ravnborg05790c62006-03-20 22:37:04 -080091static void sock_lock(struct tipc_sock* tsock)
Per Lidenb97bf3f2006-01-02 19:04:38 +010092{
93 spin_lock_bh(tsock->p->lock);
94}
95
96/*
97 * sock_unlock(): Unlock a port/socket pair
98 */
Sam Ravnborg05790c62006-03-20 22:37:04 -080099static void sock_unlock(struct tipc_sock* tsock)
Per Lidenb97bf3f2006-01-02 19:04:38 +0100100{
101 spin_unlock_bh(tsock->p->lock);
102}
103
104/**
105 * pollmask - determine the current set of poll() events for a socket
106 * @sock: socket structure
107 *
108 * TIPC sets the returned events as follows:
109 * a) POLLRDNORM and POLLIN are set if the socket's receive queue is non-empty
110 * or if a connection-oriented socket is does not have an active connection
111 * (i.e. a read operation will not block).
112 * b) POLLOUT is set except when a socket's connection has been terminated
113 * (i.e. a write operation will not block).
114 * c) POLLHUP is set when a socket's connection has been terminated.
115 *
116 * IMPORTANT: The fact that a read or write operation will not block does NOT
117 * imply that the operation will succeed!
118 *
119 * Returns pollmask value
120 */
121
Sam Ravnborg05790c62006-03-20 22:37:04 -0800122static u32 pollmask(struct socket *sock)
Per Lidenb97bf3f2006-01-02 19:04:38 +0100123{
124 u32 mask;
125
126 if ((skb_queue_len(&sock->sk->sk_receive_queue) != 0) ||
127 (sock->state == SS_UNCONNECTED) ||
128 (sock->state == SS_DISCONNECTING))
129 mask = (POLLRDNORM | POLLIN);
130 else
131 mask = 0;
132
133 if (sock->state == SS_DISCONNECTING)
134 mask |= POLLHUP;
135 else
136 mask |= POLLOUT;
137
138 return mask;
139}
140
141
142/**
143 * advance_queue - discard first buffer in queue
144 * @tsock: TIPC socket
145 */
146
Sam Ravnborg05790c62006-03-20 22:37:04 -0800147static void advance_queue(struct tipc_sock *tsock)
Per Lidenb97bf3f2006-01-02 19:04:38 +0100148{
149 sock_lock(tsock);
150 buf_discard(skb_dequeue(&tsock->sk.sk_receive_queue));
151 sock_unlock(tsock);
152 atomic_dec(&tipc_queue_size);
153}
154
155/**
156 * tipc_create - create a TIPC socket
157 * @sock: pre-allocated socket structure
158 * @protocol: protocol indicator (must be 0)
159 *
160 * This routine creates and attaches a 'struct sock' to the 'struct socket',
161 * then create and attaches a TIPC port to the 'struct sock' part.
162 *
163 * Returns 0 on success, errno otherwise
164 */
165static int tipc_create(struct socket *sock, int protocol)
166{
167 struct tipc_sock *tsock;
168 struct tipc_port *port;
169 struct sock *sk;
170 u32 ref;
171
172 if ((sock->type != SOCK_STREAM) &&
173 (sock->type != SOCK_SEQPACKET) &&
174 (sock->type != SOCK_DGRAM) &&
175 (sock->type != SOCK_RDM))
176 return -EPROTOTYPE;
177
178 if (unlikely(protocol != 0))
179 return -EPROTONOSUPPORT;
180
Sam Ravnborg1fc54d82006-03-20 22:36:47 -0800181 ref = tipc_createport_raw(NULL, &dispatch, &wakeupdispatch, TIPC_LOW_IMPORTANCE);
Per Lidenb97bf3f2006-01-02 19:04:38 +0100182 if (unlikely(!ref))
183 return -ENOMEM;
184
185 sock->state = SS_UNCONNECTED;
186
187 switch (sock->type) {
188 case SOCK_STREAM:
189 sock->ops = &stream_ops;
190 break;
191 case SOCK_SEQPACKET:
192 sock->ops = &packet_ops;
193 break;
194 case SOCK_DGRAM:
195 tipc_set_portunreliable(ref, 1);
196 /* fall through */
197 case SOCK_RDM:
198 tipc_set_portunreturnable(ref, 1);
199 sock->ops = &msg_ops;
200 sock->state = SS_READY;
201 break;
202 }
203
204 sk = sk_alloc(AF_TIPC, GFP_KERNEL, &tipc_proto, 1);
205 if (!sk) {
206 tipc_deleteport(ref);
207 return -ENOMEM;
208 }
209
210 sock_init_data(sock, sk);
211 init_waitqueue_head(sk->sk_sleep);
212 sk->sk_rcvtimeo = 8 * HZ; /* default connect timeout = 8s */
213
214 tsock = tipc_sk(sk);
215 port = tipc_get_port(ref);
216
217 tsock->p = port;
218 port->usr_handle = tsock;
219
220 init_MUTEX(&tsock->sem);
221
222 dbg("sock_create: %x\n",tsock);
223
224 atomic_inc(&tipc_user_count);
225
226 return 0;
227}
228
229/**
230 * release - destroy a TIPC socket
231 * @sock: socket to destroy
232 *
233 * This routine cleans up any messages that are still queued on the socket.
234 * For DGRAM and RDM socket types, all queued messages are rejected.
235 * For SEQPACKET and STREAM socket types, the first message is rejected
236 * and any others are discarded. (If the first message on a STREAM socket
237 * is partially-read, it is discarded and the next one is rejected instead.)
238 *
239 * NOTE: Rejected messages are not necessarily returned to the sender! They
240 * are returned or discarded according to the "destination droppable" setting
241 * specified for the message by the sender.
242 *
243 * Returns 0 on success, errno otherwise
244 */
245
246static int release(struct socket *sock)
247{
248 struct tipc_sock *tsock = tipc_sk(sock->sk);
249 struct sock *sk = sock->sk;
250 int res = TIPC_OK;
251 struct sk_buff *buf;
252
253 dbg("sock_delete: %x\n",tsock);
254 if (!tsock)
255 return 0;
256 down_interruptible(&tsock->sem);
257 if (!sock->sk) {
258 up(&tsock->sem);
259 return 0;
260 }
261
262 /* Reject unreceived messages, unless no longer connected */
263
264 while (sock->state != SS_DISCONNECTING) {
265 sock_lock(tsock);
266 buf = skb_dequeue(&sk->sk_receive_queue);
267 if (!buf)
Sam Ravnborg1fc54d82006-03-20 22:36:47 -0800268 tsock->p->usr_handle = NULL;
Per Lidenb97bf3f2006-01-02 19:04:38 +0100269 sock_unlock(tsock);
270 if (!buf)
271 break;
272 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
273 buf_discard(buf);
274 else
275 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
276 atomic_dec(&tipc_queue_size);
277 }
278
279 /* Delete TIPC port */
280
281 res = tipc_deleteport(tsock->p->ref);
282 sock->sk = NULL;
283
284 /* Discard any remaining messages */
285
286 while ((buf = skb_dequeue(&sk->sk_receive_queue))) {
287 buf_discard(buf);
288 atomic_dec(&tipc_queue_size);
289 }
290
291 up(&tsock->sem);
292
293 sock_put(sk);
294
295 atomic_dec(&tipc_user_count);
296 return res;
297}
298
299/**
300 * bind - associate or disassocate TIPC name(s) with a socket
301 * @sock: socket structure
302 * @uaddr: socket address describing name(s) and desired operation
303 * @uaddr_len: size of socket address data structure
304 *
305 * Name and name sequence binding is indicated using a positive scope value;
306 * a negative scope value unbinds the specified name. Specifying no name
307 * (i.e. a socket address length of 0) unbinds all names from the socket.
308 *
309 * Returns 0 on success, errno otherwise
310 */
311
312static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
313{
314 struct tipc_sock *tsock = tipc_sk(sock->sk);
315 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
316 int res;
317
318 if (down_interruptible(&tsock->sem))
319 return -ERESTARTSYS;
320
321 if (unlikely(!uaddr_len)) {
Sam Ravnborg1fc54d82006-03-20 22:36:47 -0800322 res = tipc_withdraw(tsock->p->ref, 0, NULL);
Per Lidenb97bf3f2006-01-02 19:04:38 +0100323 goto exit;
324 }
325
326 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
327 res = -EINVAL;
328 goto exit;
329 }
330
331 if (addr->family != AF_TIPC) {
332 res = -EAFNOSUPPORT;
333 goto exit;
334 }
335 if (addr->addrtype == TIPC_ADDR_NAME)
336 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
337 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
338 res = -EAFNOSUPPORT;
339 goto exit;
340 }
341
342 if (addr->scope > 0)
343 res = tipc_publish(tsock->p->ref, addr->scope,
344 &addr->addr.nameseq);
345 else
346 res = tipc_withdraw(tsock->p->ref, -addr->scope,
347 &addr->addr.nameseq);
348exit:
349 up(&tsock->sem);
350 return res;
351}
352
353/**
354 * get_name - get port ID of socket or peer socket
355 * @sock: socket structure
356 * @uaddr: area for returned socket address
357 * @uaddr_len: area for returned length of socket address
358 * @peer: 0 to obtain socket name, 1 to obtain peer socket name
359 *
360 * Returns 0 on success, errno otherwise
361 */
362
363static int get_name(struct socket *sock, struct sockaddr *uaddr,
364 int *uaddr_len, int peer)
365{
366 struct tipc_sock *tsock = tipc_sk(sock->sk);
367 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
368 u32 res;
369
370 if (down_interruptible(&tsock->sem))
371 return -ERESTARTSYS;
372
373 *uaddr_len = sizeof(*addr);
374 addr->addrtype = TIPC_ADDR_ID;
375 addr->family = AF_TIPC;
376 addr->scope = 0;
377 if (peer)
378 res = tipc_peer(tsock->p->ref, &addr->addr.id);
379 else
380 res = tipc_ownidentity(tsock->p->ref, &addr->addr.id);
381 addr->addr.name.domain = 0;
382
383 up(&tsock->sem);
384 return res;
385}
386
387/**
388 * poll - read and possibly block on pollmask
389 * @file: file structure associated with the socket
390 * @sock: socket for which to calculate the poll bits
391 * @wait: ???
392 *
393 * Returns the pollmask
394 */
395
396static unsigned int poll(struct file *file, struct socket *sock,
397 poll_table *wait)
398{
399 poll_wait(file, sock->sk->sk_sleep, wait);
400 /* NEED LOCK HERE? */
401 return pollmask(sock);
402}
403
404/**
405 * dest_name_check - verify user is permitted to send to specified port name
406 * @dest: destination address
407 * @m: descriptor for message to be sent
408 *
409 * Prevents restricted configuration commands from being issued by
410 * unauthorized users.
411 *
412 * Returns 0 if permission is granted, otherwise errno
413 */
414
Sam Ravnborg05790c62006-03-20 22:37:04 -0800415static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
Per Lidenb97bf3f2006-01-02 19:04:38 +0100416{
417 struct tipc_cfg_msg_hdr hdr;
418
419 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
420 return 0;
421 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
422 return 0;
423
424 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
425 return -EACCES;
426
427 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
428 return -EFAULT;
429 if ((ntohs(hdr.tcm_type) & 0xC000) & (!capable(CAP_NET_ADMIN)))
430 return -EACCES;
431
432 return 0;
433}
434
435/**
436 * send_msg - send message in connectionless manner
437 * @iocb: (unused)
438 * @sock: socket structure
439 * @m: message to send
440 * @total_len: (unused)
441 *
442 * Message must have an destination specified explicitly.
443 * Used for SOCK_RDM and SOCK_DGRAM messages,
444 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
445 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
446 *
447 * Returns the number of bytes sent on success, or errno otherwise
448 */
449
450static int send_msg(struct kiocb *iocb, struct socket *sock,
451 struct msghdr *m, size_t total_len)
452{
453 struct tipc_sock *tsock = tipc_sk(sock->sk);
454 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
455 struct sk_buff *buf;
456 int needs_conn;
457 int res = -EINVAL;
458
459 if (unlikely(!dest))
460 return -EDESTADDRREQ;
461 if (unlikely(dest->family != AF_TIPC))
462 return -EINVAL;
463
464 needs_conn = (sock->state != SS_READY);
465 if (unlikely(needs_conn)) {
466 if (sock->state == SS_LISTENING)
467 return -EPIPE;
468 if (sock->state != SS_UNCONNECTED)
469 return -EISCONN;
470 if ((tsock->p->published) ||
471 ((sock->type == SOCK_STREAM) && (total_len != 0)))
472 return -EOPNOTSUPP;
473 }
474
475 if (down_interruptible(&tsock->sem))
476 return -ERESTARTSYS;
477
478 if (needs_conn) {
479
480 /* Abort any pending connection attempts (very unlikely) */
481
482 while ((buf = skb_dequeue(&sock->sk->sk_receive_queue))) {
483 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
484 atomic_dec(&tipc_queue_size);
485 }
486
487 sock->state = SS_CONNECTING;
488 }
489
490 do {
491 if (dest->addrtype == TIPC_ADDR_NAME) {
492 if ((res = dest_name_check(dest, m)))
493 goto exit;
494 res = tipc_send2name(tsock->p->ref,
495 &dest->addr.name.name,
496 dest->addr.name.domain,
497 m->msg_iovlen,
498 m->msg_iov);
499 }
500 else if (dest->addrtype == TIPC_ADDR_ID) {
501 res = tipc_send2port(tsock->p->ref,
502 &dest->addr.id,
503 m->msg_iovlen,
504 m->msg_iov);
505 }
506 else if (dest->addrtype == TIPC_ADDR_MCAST) {
507 if (needs_conn) {
508 res = -EOPNOTSUPP;
509 goto exit;
510 }
511 if ((res = dest_name_check(dest, m)))
512 goto exit;
513 res = tipc_multicast(tsock->p->ref,
514 &dest->addr.nameseq,
515 0,
516 m->msg_iovlen,
517 m->msg_iov);
518 }
519 if (likely(res != -ELINKCONG)) {
520exit:
521 up(&tsock->sem);
522 return res;
523 }
524 if (m->msg_flags & MSG_DONTWAIT) {
525 res = -EWOULDBLOCK;
526 goto exit;
527 }
528 if (wait_event_interruptible(*sock->sk->sk_sleep,
529 !tsock->p->congested)) {
530 res = -ERESTARTSYS;
531 goto exit;
532 }
533 } while (1);
534}
535
536/**
537 * send_packet - send a connection-oriented message
538 * @iocb: (unused)
539 * @sock: socket structure
540 * @m: message to send
541 * @total_len: (unused)
542 *
543 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
544 *
545 * Returns the number of bytes sent on success, or errno otherwise
546 */
547
548static int send_packet(struct kiocb *iocb, struct socket *sock,
549 struct msghdr *m, size_t total_len)
550{
551 struct tipc_sock *tsock = tipc_sk(sock->sk);
552 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
553 int res;
554
555 /* Handle implied connection establishment */
556
557 if (unlikely(dest))
558 return send_msg(iocb, sock, m, total_len);
559
560 if (down_interruptible(&tsock->sem)) {
561 return -ERESTARTSYS;
562 }
563
564 if (unlikely(sock->state != SS_CONNECTED)) {
565 if (sock->state == SS_DISCONNECTING)
566 res = -EPIPE;
567 else
568 res = -ENOTCONN;
569 goto exit;
570 }
571
572 do {
573 res = tipc_send(tsock->p->ref, m->msg_iovlen, m->msg_iov);
574 if (likely(res != -ELINKCONG)) {
575exit:
576 up(&tsock->sem);
577 return res;
578 }
579 if (m->msg_flags & MSG_DONTWAIT) {
580 res = -EWOULDBLOCK;
581 goto exit;
582 }
583 if (wait_event_interruptible(*sock->sk->sk_sleep,
584 !tsock->p->congested)) {
585 res = -ERESTARTSYS;
586 goto exit;
587 }
588 } while (1);
589}
590
591/**
592 * send_stream - send stream-oriented data
593 * @iocb: (unused)
594 * @sock: socket structure
595 * @m: data to send
596 * @total_len: total length of data to be sent
597 *
598 * Used for SOCK_STREAM data.
599 *
600 * Returns the number of bytes sent on success, or errno otherwise
601 */
602
603
604static int send_stream(struct kiocb *iocb, struct socket *sock,
605 struct msghdr *m, size_t total_len)
606{
607 struct msghdr my_msg;
608 struct iovec my_iov;
609 struct iovec *curr_iov;
610 int curr_iovlen;
611 char __user *curr_start;
612 int curr_left;
613 int bytes_to_send;
614 int res;
615
616 if (likely(total_len <= TIPC_MAX_USER_MSG_SIZE))
617 return send_packet(iocb, sock, m, total_len);
618
619 /* Can only send large data streams if already connected */
620
621 if (unlikely(sock->state != SS_CONNECTED)) {
622 if (sock->state == SS_DISCONNECTING)
623 return -EPIPE;
624 else
625 return -ENOTCONN;
626 }
627
628 /*
629 * Send each iovec entry using one or more messages
630 *
631 * Note: This algorithm is good for the most likely case
632 * (i.e. one large iovec entry), but could be improved to pass sets
633 * of small iovec entries into send_packet().
634 */
635
636 my_msg = *m;
637 curr_iov = my_msg.msg_iov;
638 curr_iovlen = my_msg.msg_iovlen;
639 my_msg.msg_iov = &my_iov;
640 my_msg.msg_iovlen = 1;
641
642 while (curr_iovlen--) {
643 curr_start = curr_iov->iov_base;
644 curr_left = curr_iov->iov_len;
645
646 while (curr_left) {
647 bytes_to_send = (curr_left < TIPC_MAX_USER_MSG_SIZE)
648 ? curr_left : TIPC_MAX_USER_MSG_SIZE;
649 my_iov.iov_base = curr_start;
650 my_iov.iov_len = bytes_to_send;
651 if ((res = send_packet(iocb, sock, &my_msg, 0)) < 0)
652 return res;
653 curr_left -= bytes_to_send;
654 curr_start += bytes_to_send;
655 }
656
657 curr_iov++;
658 }
659
660 return total_len;
661}
662
663/**
664 * auto_connect - complete connection setup to a remote port
665 * @sock: socket structure
666 * @tsock: TIPC-specific socket structure
667 * @msg: peer's response message
668 *
669 * Returns 0 on success, errno otherwise
670 */
671
672static int auto_connect(struct socket *sock, struct tipc_sock *tsock,
673 struct tipc_msg *msg)
674{
675 struct tipc_portid peer;
676
677 if (msg_errcode(msg)) {
678 sock->state = SS_DISCONNECTING;
679 return -ECONNREFUSED;
680 }
681
682 peer.ref = msg_origport(msg);
683 peer.node = msg_orignode(msg);
684 tipc_connect2port(tsock->p->ref, &peer);
685 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
686 sock->state = SS_CONNECTED;
687 return 0;
688}
689
690/**
691 * set_orig_addr - capture sender's address for received message
692 * @m: descriptor for message info
693 * @msg: received message header
694 *
695 * Note: Address is not captured if not requested by receiver.
696 */
697
Sam Ravnborg05790c62006-03-20 22:37:04 -0800698static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
Per Lidenb97bf3f2006-01-02 19:04:38 +0100699{
700 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
701
702 if (addr) {
703 addr->family = AF_TIPC;
704 addr->addrtype = TIPC_ADDR_ID;
705 addr->addr.id.ref = msg_origport(msg);
706 addr->addr.id.node = msg_orignode(msg);
707 addr->addr.name.domain = 0; /* could leave uninitialized */
708 addr->scope = 0; /* could leave uninitialized */
709 m->msg_namelen = sizeof(struct sockaddr_tipc);
710 }
711}
712
713/**
714 * anc_data_recv - optionally capture ancillary data for received message
715 * @m: descriptor for message info
716 * @msg: received message header
717 * @tport: TIPC port associated with message
718 *
719 * Note: Ancillary data is not captured if not requested by receiver.
720 *
721 * Returns 0 if successful, otherwise errno
722 */
723
Sam Ravnborg05790c62006-03-20 22:37:04 -0800724static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
Per Lidenb97bf3f2006-01-02 19:04:38 +0100725 struct tipc_port *tport)
726{
727 u32 anc_data[3];
728 u32 err;
729 u32 dest_type;
730 int res;
731
732 if (likely(m->msg_controllen == 0))
733 return 0;
734
735 /* Optionally capture errored message object(s) */
736
737 err = msg ? msg_errcode(msg) : 0;
738 if (unlikely(err)) {
739 anc_data[0] = err;
740 anc_data[1] = msg_data_sz(msg);
741 if ((res = put_cmsg(m, SOL_SOCKET, TIPC_ERRINFO, 8, anc_data)))
742 return res;
743 if (anc_data[1] &&
744 (res = put_cmsg(m, SOL_SOCKET, TIPC_RETDATA, anc_data[1],
745 msg_data(msg))))
746 return res;
747 }
748
749 /* Optionally capture message destination object */
750
751 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
752 switch (dest_type) {
753 case TIPC_NAMED_MSG:
754 anc_data[0] = msg_nametype(msg);
755 anc_data[1] = msg_namelower(msg);
756 anc_data[2] = msg_namelower(msg);
757 break;
758 case TIPC_MCAST_MSG:
759 anc_data[0] = msg_nametype(msg);
760 anc_data[1] = msg_namelower(msg);
761 anc_data[2] = msg_nameupper(msg);
762 break;
763 case TIPC_CONN_MSG:
764 anc_data[0] = tport->conn_type;
765 anc_data[1] = tport->conn_instance;
766 anc_data[2] = tport->conn_instance;
767 break;
768 default:
769 anc_data[0] = 0;
770 }
771 if (anc_data[0] &&
772 (res = put_cmsg(m, SOL_SOCKET, TIPC_DESTNAME, 12, anc_data)))
773 return res;
774
775 return 0;
776}
777
778/**
779 * recv_msg - receive packet-oriented message
780 * @iocb: (unused)
781 * @m: descriptor for message info
782 * @buf_len: total size of user buffer area
783 * @flags: receive flags
784 *
785 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
786 * If the complete message doesn't fit in user area, truncate it.
787 *
788 * Returns size of returned message data, errno otherwise
789 */
790
791static int recv_msg(struct kiocb *iocb, struct socket *sock,
792 struct msghdr *m, size_t buf_len, int flags)
793{
794 struct tipc_sock *tsock = tipc_sk(sock->sk);
795 struct sk_buff *buf;
796 struct tipc_msg *msg;
797 unsigned int q_len;
798 unsigned int sz;
799 u32 err;
800 int res;
801
802 /* Currently doesn't support receiving into multiple iovec entries */
803
804 if (m->msg_iovlen != 1)
805 return -EOPNOTSUPP;
806
807 /* Catch invalid receive attempts */
808
809 if (unlikely(!buf_len))
810 return -EINVAL;
811
812 if (sock->type == SOCK_SEQPACKET) {
813 if (unlikely(sock->state == SS_UNCONNECTED))
814 return -ENOTCONN;
815 if (unlikely((sock->state == SS_DISCONNECTING) &&
816 (skb_queue_len(&sock->sk->sk_receive_queue) == 0)))
817 return -ENOTCONN;
818 }
819
820 /* Look for a message in receive queue; wait if necessary */
821
822 if (unlikely(down_interruptible(&tsock->sem)))
823 return -ERESTARTSYS;
824
825restart:
826 if (unlikely((skb_queue_len(&sock->sk->sk_receive_queue) == 0) &&
827 (flags & MSG_DONTWAIT))) {
828 res = -EWOULDBLOCK;
829 goto exit;
830 }
831
832 if ((res = wait_event_interruptible(
833 *sock->sk->sk_sleep,
834 ((q_len = skb_queue_len(&sock->sk->sk_receive_queue)) ||
835 (sock->state == SS_DISCONNECTING))) )) {
836 goto exit;
837 }
838
839 /* Catch attempt to receive on an already terminated connection */
840 /* [THIS CHECK MAY OVERLAP WITH AN EARLIER CHECK] */
841
842 if (!q_len) {
843 res = -ENOTCONN;
844 goto exit;
845 }
846
847 /* Get access to first message in receive queue */
848
849 buf = skb_peek(&sock->sk->sk_receive_queue);
850 msg = buf_msg(buf);
851 sz = msg_data_sz(msg);
852 err = msg_errcode(msg);
853
854 /* Complete connection setup for an implied connect */
855
856 if (unlikely(sock->state == SS_CONNECTING)) {
857 if ((res = auto_connect(sock, tsock, msg)))
858 goto exit;
859 }
860
861 /* Discard an empty non-errored message & try again */
862
863 if ((!sz) && (!err)) {
864 advance_queue(tsock);
865 goto restart;
866 }
867
868 /* Capture sender's address (optional) */
869
870 set_orig_addr(m, msg);
871
872 /* Capture ancillary data (optional) */
873
874 if ((res = anc_data_recv(m, msg, tsock->p)))
875 goto exit;
876
877 /* Capture message data (if valid) & compute return value (always) */
878
879 if (!err) {
880 if (unlikely(buf_len < sz)) {
881 sz = buf_len;
882 m->msg_flags |= MSG_TRUNC;
883 }
884 if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
885 sz))) {
886 res = -EFAULT;
887 goto exit;
888 }
889 res = sz;
890 } else {
891 if ((sock->state == SS_READY) ||
892 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
893 res = 0;
894 else
895 res = -ECONNRESET;
896 }
897
898 /* Consume received message (optional) */
899
900 if (likely(!(flags & MSG_PEEK))) {
901 if (unlikely(++tsock->p->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
902 tipc_acknowledge(tsock->p->ref, tsock->p->conn_unacked);
903 advance_queue(tsock);
904 }
905exit:
906 up(&tsock->sem);
907 return res;
908}
909
910/**
911 * recv_stream - receive stream-oriented data
912 * @iocb: (unused)
913 * @m: descriptor for message info
914 * @buf_len: total size of user buffer area
915 * @flags: receive flags
916 *
917 * Used for SOCK_STREAM messages only. If not enough data is available
918 * will optionally wait for more; never truncates data.
919 *
920 * Returns size of returned message data, errno otherwise
921 */
922
923static int recv_stream(struct kiocb *iocb, struct socket *sock,
924 struct msghdr *m, size_t buf_len, int flags)
925{
926 struct tipc_sock *tsock = tipc_sk(sock->sk);
927 struct sk_buff *buf;
928 struct tipc_msg *msg;
929 unsigned int q_len;
930 unsigned int sz;
931 int sz_to_copy;
932 int sz_copied = 0;
933 int needed;
934 char *crs = m->msg_iov->iov_base;
935 unsigned char *buf_crs;
936 u32 err;
937 int res;
938
939 /* Currently doesn't support receiving into multiple iovec entries */
940
941 if (m->msg_iovlen != 1)
942 return -EOPNOTSUPP;
943
944 /* Catch invalid receive attempts */
945
946 if (unlikely(!buf_len))
947 return -EINVAL;
948
949 if (unlikely(sock->state == SS_DISCONNECTING)) {
950 if (skb_queue_len(&sock->sk->sk_receive_queue) == 0)
951 return -ENOTCONN;
952 } else if (unlikely(sock->state != SS_CONNECTED))
953 return -ENOTCONN;
954
955 /* Look for a message in receive queue; wait if necessary */
956
957 if (unlikely(down_interruptible(&tsock->sem)))
958 return -ERESTARTSYS;
959
960restart:
961 if (unlikely((skb_queue_len(&sock->sk->sk_receive_queue) == 0) &&
962 (flags & MSG_DONTWAIT))) {
963 res = (sz_copied == 0) ? -EWOULDBLOCK : 0;
964 goto exit;
965 }
966
967 if ((res = wait_event_interruptible(
968 *sock->sk->sk_sleep,
969 ((q_len = skb_queue_len(&sock->sk->sk_receive_queue)) ||
970 (sock->state == SS_DISCONNECTING))) )) {
971 goto exit;
972 }
973
974 /* Catch attempt to receive on an already terminated connection */
975 /* [THIS CHECK MAY OVERLAP WITH AN EARLIER CHECK] */
976
977 if (!q_len) {
978 res = -ENOTCONN;
979 goto exit;
980 }
981
982 /* Get access to first message in receive queue */
983
984 buf = skb_peek(&sock->sk->sk_receive_queue);
985 msg = buf_msg(buf);
986 sz = msg_data_sz(msg);
987 err = msg_errcode(msg);
988
989 /* Discard an empty non-errored message & try again */
990
991 if ((!sz) && (!err)) {
992 advance_queue(tsock);
993 goto restart;
994 }
995
996 /* Optionally capture sender's address & ancillary data of first msg */
997
998 if (sz_copied == 0) {
999 set_orig_addr(m, msg);
1000 if ((res = anc_data_recv(m, msg, tsock->p)))
1001 goto exit;
1002 }
1003
1004 /* Capture message data (if valid) & compute return value (always) */
1005
1006 if (!err) {
1007 buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1008 sz = buf->tail - buf_crs;
1009
1010 needed = (buf_len - sz_copied);
1011 sz_to_copy = (sz <= needed) ? sz : needed;
1012 if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1013 res = -EFAULT;
1014 goto exit;
1015 }
1016 sz_copied += sz_to_copy;
1017
1018 if (sz_to_copy < sz) {
1019 if (!(flags & MSG_PEEK))
1020 TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1021 goto exit;
1022 }
1023
1024 crs += sz_to_copy;
1025 } else {
1026 if (sz_copied != 0)
1027 goto exit; /* can't add error msg to valid data */
1028
1029 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1030 res = 0;
1031 else
1032 res = -ECONNRESET;
1033 }
1034
1035 /* Consume received message (optional) */
1036
1037 if (likely(!(flags & MSG_PEEK))) {
1038 if (unlikely(++tsock->p->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1039 tipc_acknowledge(tsock->p->ref, tsock->p->conn_unacked);
1040 advance_queue(tsock);
1041 }
1042
1043 /* Loop around if more data is required */
1044
1045 if ((sz_copied < buf_len) /* didn't get all requested data */
1046 && (flags & MSG_WAITALL) /* ... and need to wait for more */
1047 && (!(flags & MSG_PEEK)) /* ... and aren't just peeking at data */
1048 && (!err) /* ... and haven't reached a FIN */
1049 )
1050 goto restart;
1051
1052exit:
1053 up(&tsock->sem);
1054 return res ? res : sz_copied;
1055}
1056
1057/**
1058 * queue_overloaded - test if queue overload condition exists
1059 * @queue_size: current size of queue
1060 * @base: nominal maximum size of queue
1061 * @msg: message to be added to queue
1062 *
1063 * Returns 1 if queue is currently overloaded, 0 otherwise
1064 */
1065
1066static int queue_overloaded(u32 queue_size, u32 base, struct tipc_msg *msg)
1067{
1068 u32 threshold;
1069 u32 imp = msg_importance(msg);
1070
1071 if (imp == TIPC_LOW_IMPORTANCE)
1072 threshold = base;
1073 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1074 threshold = base * 2;
1075 else if (imp == TIPC_HIGH_IMPORTANCE)
1076 threshold = base * 100;
1077 else
1078 return 0;
1079
1080 if (msg_connected(msg))
1081 threshold *= 4;
1082
1083 return (queue_size > threshold);
1084}
1085
1086/**
1087 * async_disconnect - wrapper function used to disconnect port
1088 * @portref: TIPC port reference (passed as pointer-sized value)
1089 */
1090
1091static void async_disconnect(unsigned long portref)
1092{
1093 tipc_disconnect((u32)portref);
1094}
1095
1096/**
1097 * dispatch - handle arriving message
1098 * @tport: TIPC port that received message
1099 * @buf: message
1100 *
1101 * Called with port locked. Must not take socket lock to avoid deadlock risk.
1102 *
1103 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1104 */
1105
1106static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1107{
1108 struct tipc_msg *msg = buf_msg(buf);
1109 struct tipc_sock *tsock = (struct tipc_sock *)tport->usr_handle;
1110 struct socket *sock;
1111 u32 recv_q_len;
1112
1113 /* Reject message if socket is closing */
1114
1115 if (!tsock)
1116 return TIPC_ERR_NO_PORT;
1117
1118 /* Reject message if it is wrong sort of message for socket */
1119
1120 /*
1121 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1122 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1123 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1124 */
1125 sock = tsock->sk.sk_socket;
1126 if (sock->state == SS_READY) {
1127 if (msg_connected(msg)) {
1128 msg_dbg(msg, "dispatch filter 1\n");
1129 return TIPC_ERR_NO_PORT;
1130 }
1131 } else {
1132 if (msg_mcast(msg)) {
1133 msg_dbg(msg, "dispatch filter 2\n");
1134 return TIPC_ERR_NO_PORT;
1135 }
1136 if (sock->state == SS_CONNECTED) {
1137 if (!msg_connected(msg)) {
1138 msg_dbg(msg, "dispatch filter 3\n");
1139 return TIPC_ERR_NO_PORT;
1140 }
1141 }
1142 else if (sock->state == SS_CONNECTING) {
1143 if (!msg_connected(msg) && (msg_errcode(msg) == 0)) {
1144 msg_dbg(msg, "dispatch filter 4\n");
1145 return TIPC_ERR_NO_PORT;
1146 }
1147 }
1148 else if (sock->state == SS_LISTENING) {
1149 if (msg_connected(msg) || msg_errcode(msg)) {
1150 msg_dbg(msg, "dispatch filter 5\n");
1151 return TIPC_ERR_NO_PORT;
1152 }
1153 }
1154 else if (sock->state == SS_DISCONNECTING) {
1155 msg_dbg(msg, "dispatch filter 6\n");
1156 return TIPC_ERR_NO_PORT;
1157 }
1158 else /* (sock->state == SS_UNCONNECTED) */ {
1159 if (msg_connected(msg) || msg_errcode(msg)) {
1160 msg_dbg(msg, "dispatch filter 7\n");
1161 return TIPC_ERR_NO_PORT;
1162 }
1163 }
1164 }
1165
1166 /* Reject message if there isn't room to queue it */
1167
1168 if (unlikely((u32)atomic_read(&tipc_queue_size) >
1169 OVERLOAD_LIMIT_BASE)) {
1170 if (queue_overloaded(atomic_read(&tipc_queue_size),
1171 OVERLOAD_LIMIT_BASE, msg))
1172 return TIPC_ERR_OVERLOAD;
1173 }
1174 recv_q_len = skb_queue_len(&tsock->sk.sk_receive_queue);
1175 if (unlikely(recv_q_len > (OVERLOAD_LIMIT_BASE / 2))) {
1176 if (queue_overloaded(recv_q_len,
1177 OVERLOAD_LIMIT_BASE / 2, msg))
1178 return TIPC_ERR_OVERLOAD;
1179 }
1180
1181 /* Initiate connection termination for an incoming 'FIN' */
1182
1183 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1184 sock->state = SS_DISCONNECTING;
1185 /* Note: Use signal since port lock is already taken! */
Per Liden4323add2006-01-18 00:38:21 +01001186 tipc_k_signal((Handler)async_disconnect, tport->ref);
Per Lidenb97bf3f2006-01-02 19:04:38 +01001187 }
1188
1189 /* Enqueue message (finally!) */
1190
1191 msg_dbg(msg,"<DISP<: ");
1192 TIPC_SKB_CB(buf)->handle = msg_data(msg);
1193 atomic_inc(&tipc_queue_size);
1194 skb_queue_tail(&sock->sk->sk_receive_queue, buf);
1195
1196 wake_up_interruptible(sock->sk->sk_sleep);
1197 return TIPC_OK;
1198}
1199
1200/**
1201 * wakeupdispatch - wake up port after congestion
1202 * @tport: port to wakeup
1203 *
1204 * Called with port lock on.
1205 */
1206
1207static void wakeupdispatch(struct tipc_port *tport)
1208{
1209 struct tipc_sock *tsock = (struct tipc_sock *)tport->usr_handle;
1210
1211 wake_up_interruptible(tsock->sk.sk_sleep);
1212}
1213
1214/**
1215 * connect - establish a connection to another TIPC port
1216 * @sock: socket structure
1217 * @dest: socket address for destination port
1218 * @destlen: size of socket address data structure
1219 * @flags: (unused)
1220 *
1221 * Returns 0 on success, errno otherwise
1222 */
1223
1224static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1225 int flags)
1226{
1227 struct tipc_sock *tsock = tipc_sk(sock->sk);
1228 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
Sam Ravnborg1fc54d82006-03-20 22:36:47 -08001229 struct msghdr m = {NULL,};
Per Lidenb97bf3f2006-01-02 19:04:38 +01001230 struct sk_buff *buf;
1231 struct tipc_msg *msg;
1232 int res;
1233
1234 /* For now, TIPC does not allow use of connect() with DGRAM or RDM types */
1235
1236 if (sock->state == SS_READY)
1237 return -EOPNOTSUPP;
1238
1239 /* MOVE THE REST OF THIS ERROR CHECKING TO send_msg()? */
1240 if (sock->state == SS_LISTENING)
1241 return -EOPNOTSUPP;
1242 if (sock->state == SS_CONNECTING)
1243 return -EALREADY;
1244 if (sock->state != SS_UNCONNECTED)
1245 return -EISCONN;
1246
1247 if ((dst->family != AF_TIPC) ||
1248 ((dst->addrtype != TIPC_ADDR_NAME) && (dst->addrtype != TIPC_ADDR_ID)))
1249 return -EINVAL;
1250
1251 /* Send a 'SYN-' to destination */
1252
1253 m.msg_name = dest;
Sam Ravnborg1fc54d82006-03-20 22:36:47 -08001254 if ((res = send_msg(NULL, sock, &m, 0)) < 0) {
Per Lidenb97bf3f2006-01-02 19:04:38 +01001255 sock->state = SS_DISCONNECTING;
1256 return res;
1257 }
1258
1259 if (down_interruptible(&tsock->sem))
1260 return -ERESTARTSYS;
1261
1262 /* Wait for destination's 'ACK' response */
1263
1264 res = wait_event_interruptible_timeout(*sock->sk->sk_sleep,
1265 skb_queue_len(&sock->sk->sk_receive_queue),
1266 sock->sk->sk_rcvtimeo);
1267 buf = skb_peek(&sock->sk->sk_receive_queue);
1268 if (res > 0) {
1269 msg = buf_msg(buf);
1270 res = auto_connect(sock, tsock, msg);
1271 if (!res) {
1272 if (dst->addrtype == TIPC_ADDR_NAME) {
1273 tsock->p->conn_type = dst->addr.name.name.type;
1274 tsock->p->conn_instance = dst->addr.name.name.instance;
1275 }
1276 if (!msg_data_sz(msg))
1277 advance_queue(tsock);
1278 }
1279 } else {
1280 if (res == 0) {
1281 res = -ETIMEDOUT;
1282 } else
1283 { /* leave "res" unchanged */ }
1284 sock->state = SS_DISCONNECTING;
1285 }
1286
1287 up(&tsock->sem);
1288 return res;
1289}
1290
1291/**
1292 * listen - allow socket to listen for incoming connections
1293 * @sock: socket structure
1294 * @len: (unused)
1295 *
1296 * Returns 0 on success, errno otherwise
1297 */
1298
1299static int listen(struct socket *sock, int len)
1300{
1301 /* REQUIRES SOCKET LOCKING OF SOME SORT? */
1302
1303 if (sock->state == SS_READY)
1304 return -EOPNOTSUPP;
1305 if (sock->state != SS_UNCONNECTED)
1306 return -EINVAL;
1307 sock->state = SS_LISTENING;
1308 return 0;
1309}
1310
1311/**
1312 * accept - wait for connection request
1313 * @sock: listening socket
1314 * @newsock: new socket that is to be connected
1315 * @flags: file-related flags associated with socket
1316 *
1317 * Returns 0 on success, errno otherwise
1318 */
1319
1320static int accept(struct socket *sock, struct socket *newsock, int flags)
1321{
1322 struct tipc_sock *tsock = tipc_sk(sock->sk);
1323 struct sk_buff *buf;
1324 int res = -EFAULT;
1325
1326 if (sock->state == SS_READY)
1327 return -EOPNOTSUPP;
1328 if (sock->state != SS_LISTENING)
1329 return -EINVAL;
1330
1331 if (unlikely((skb_queue_len(&sock->sk->sk_receive_queue) == 0) &&
1332 (flags & O_NONBLOCK)))
1333 return -EWOULDBLOCK;
1334
1335 if (down_interruptible(&tsock->sem))
1336 return -ERESTARTSYS;
1337
1338 if (wait_event_interruptible(*sock->sk->sk_sleep,
1339 skb_queue_len(&sock->sk->sk_receive_queue))) {
1340 res = -ERESTARTSYS;
1341 goto exit;
1342 }
1343 buf = skb_peek(&sock->sk->sk_receive_queue);
1344
1345 res = tipc_create(newsock, 0);
1346 if (!res) {
1347 struct tipc_sock *new_tsock = tipc_sk(newsock->sk);
1348 struct tipc_portid id;
1349 struct tipc_msg *msg = buf_msg(buf);
1350 u32 new_ref = new_tsock->p->ref;
1351
1352 id.ref = msg_origport(msg);
1353 id.node = msg_orignode(msg);
1354 tipc_connect2port(new_ref, &id);
1355 newsock->state = SS_CONNECTED;
1356
1357 tipc_set_portimportance(new_ref, msg_importance(msg));
1358 if (msg_named(msg)) {
1359 new_tsock->p->conn_type = msg_nametype(msg);
1360 new_tsock->p->conn_instance = msg_nameinst(msg);
1361 }
1362
1363 /*
1364 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1365 * Respond to 'SYN+' by queuing it on new socket.
1366 */
1367
1368 msg_dbg(msg,"<ACC<: ");
1369 if (!msg_data_sz(msg)) {
Sam Ravnborg1fc54d82006-03-20 22:36:47 -08001370 struct msghdr m = {NULL,};
Per Lidenb97bf3f2006-01-02 19:04:38 +01001371
Sam Ravnborg1fc54d82006-03-20 22:36:47 -08001372 send_packet(NULL, newsock, &m, 0);
Per Lidenb97bf3f2006-01-02 19:04:38 +01001373 advance_queue(tsock);
1374 } else {
1375 sock_lock(tsock);
1376 skb_dequeue(&sock->sk->sk_receive_queue);
1377 sock_unlock(tsock);
1378 skb_queue_head(&newsock->sk->sk_receive_queue, buf);
1379 }
1380 }
1381exit:
1382 up(&tsock->sem);
1383 return res;
1384}
1385
1386/**
1387 * shutdown - shutdown socket connection
1388 * @sock: socket structure
1389 * @how: direction to close (always treated as read + write)
1390 *
1391 * Terminates connection (if necessary), then purges socket's receive queue.
1392 *
1393 * Returns 0 on success, errno otherwise
1394 */
1395
1396static int shutdown(struct socket *sock, int how)
1397{
1398 struct tipc_sock* tsock = tipc_sk(sock->sk);
1399 struct sk_buff *buf;
1400 int res;
1401
1402 /* Could return -EINVAL for an invalid "how", but why bother? */
1403
1404 if (down_interruptible(&tsock->sem))
1405 return -ERESTARTSYS;
1406
1407 sock_lock(tsock);
1408
1409 switch (sock->state) {
1410 case SS_CONNECTED:
1411
1412 /* Send 'FIN+' or 'FIN-' message to peer */
1413
1414 sock_unlock(tsock);
1415restart:
1416 if ((buf = skb_dequeue(&sock->sk->sk_receive_queue))) {
1417 atomic_dec(&tipc_queue_size);
1418 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1419 buf_discard(buf);
1420 goto restart;
1421 }
1422 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1423 }
1424 else {
1425 tipc_shutdown(tsock->p->ref);
1426 }
1427 sock_lock(tsock);
1428
1429 /* fall through */
1430
1431 case SS_DISCONNECTING:
1432
1433 /* Discard any unreceived messages */
1434
1435 while ((buf = skb_dequeue(&sock->sk->sk_receive_queue))) {
1436 atomic_dec(&tipc_queue_size);
1437 buf_discard(buf);
1438 }
1439 tsock->p->conn_unacked = 0;
1440
1441 /* fall through */
1442
1443 case SS_CONNECTING:
1444 sock->state = SS_DISCONNECTING;
1445 res = 0;
1446 break;
1447
1448 default:
1449 res = -ENOTCONN;
1450 }
1451
1452 sock_unlock(tsock);
1453
1454 up(&tsock->sem);
1455 return res;
1456}
1457
1458/**
1459 * setsockopt - set socket option
1460 * @sock: socket structure
1461 * @lvl: option level
1462 * @opt: option identifier
1463 * @ov: pointer to new option value
1464 * @ol: length of option value
1465 *
1466 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1467 * (to ease compatibility).
1468 *
1469 * Returns 0 on success, errno otherwise
1470 */
1471
1472static int setsockopt(struct socket *sock, int lvl, int opt, char *ov, int ol)
1473{
1474 struct tipc_sock *tsock = tipc_sk(sock->sk);
1475 u32 value;
1476 int res;
1477
1478 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1479 return 0;
1480 if (lvl != SOL_TIPC)
1481 return -ENOPROTOOPT;
1482 if (ol < sizeof(value))
1483 return -EINVAL;
1484 if ((res = get_user(value, (u32 *)ov)))
1485 return res;
1486
1487 if (down_interruptible(&tsock->sem))
1488 return -ERESTARTSYS;
1489
1490 switch (opt) {
1491 case TIPC_IMPORTANCE:
1492 res = tipc_set_portimportance(tsock->p->ref, value);
1493 break;
1494 case TIPC_SRC_DROPPABLE:
1495 if (sock->type != SOCK_STREAM)
1496 res = tipc_set_portunreliable(tsock->p->ref, value);
1497 else
1498 res = -ENOPROTOOPT;
1499 break;
1500 case TIPC_DEST_DROPPABLE:
1501 res = tipc_set_portunreturnable(tsock->p->ref, value);
1502 break;
1503 case TIPC_CONN_TIMEOUT:
1504 sock->sk->sk_rcvtimeo = (value * HZ / 1000);
1505 break;
1506 default:
1507 res = -EINVAL;
1508 }
1509
1510 up(&tsock->sem);
1511 return res;
1512}
1513
1514/**
1515 * getsockopt - get socket option
1516 * @sock: socket structure
1517 * @lvl: option level
1518 * @opt: option identifier
1519 * @ov: receptacle for option value
1520 * @ol: receptacle for length of option value
1521 *
1522 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1523 * (to ease compatibility).
1524 *
1525 * Returns 0 on success, errno otherwise
1526 */
1527
1528static int getsockopt(struct socket *sock, int lvl, int opt, char *ov, int *ol)
1529{
1530 struct tipc_sock *tsock = tipc_sk(sock->sk);
1531 int len;
1532 u32 value;
1533 int res;
1534
1535 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1536 return put_user(0, ol);
1537 if (lvl != SOL_TIPC)
1538 return -ENOPROTOOPT;
1539 if ((res = get_user(len, ol)))
1540 return res;
1541
1542 if (down_interruptible(&tsock->sem))
1543 return -ERESTARTSYS;
1544
1545 switch (opt) {
1546 case TIPC_IMPORTANCE:
1547 res = tipc_portimportance(tsock->p->ref, &value);
1548 break;
1549 case TIPC_SRC_DROPPABLE:
1550 res = tipc_portunreliable(tsock->p->ref, &value);
1551 break;
1552 case TIPC_DEST_DROPPABLE:
1553 res = tipc_portunreturnable(tsock->p->ref, &value);
1554 break;
1555 case TIPC_CONN_TIMEOUT:
1556 value = (sock->sk->sk_rcvtimeo * 1000) / HZ;
1557 break;
1558 default:
1559 res = -EINVAL;
1560 }
1561
1562 if (res) {
1563 /* "get" failed */
1564 }
1565 else if (len < sizeof(value)) {
1566 res = -EINVAL;
1567 }
1568 else if ((res = copy_to_user(ov, &value, sizeof(value)))) {
1569 /* couldn't return value */
1570 }
1571 else {
1572 res = put_user(sizeof(value), ol);
1573 }
1574
1575 up(&tsock->sem);
1576 return res;
1577}
1578
1579/**
1580 * Placeholders for non-implemented functionality
1581 *
1582 * Returns error code (POSIX-compliant where defined)
1583 */
1584
1585static int ioctl(struct socket *s, u32 cmd, unsigned long arg)
1586{
1587 return -EINVAL;
1588}
1589
1590static int no_mmap(struct file *file, struct socket *sock,
1591 struct vm_area_struct *vma)
1592{
1593 return -EINVAL;
1594}
1595static ssize_t no_sendpage(struct socket *sock, struct page *page,
1596 int offset, size_t size, int flags)
1597{
1598 return -EINVAL;
1599}
1600
1601static int no_skpair(struct socket *s1, struct socket *s2)
1602{
1603 return -EOPNOTSUPP;
1604}
1605
1606/**
1607 * Protocol switches for the various types of TIPC sockets
1608 */
1609
1610static struct proto_ops msg_ops = {
1611 .owner = THIS_MODULE,
1612 .family = AF_TIPC,
1613 .release = release,
1614 .bind = bind,
1615 .connect = connect,
1616 .socketpair = no_skpair,
1617 .accept = accept,
1618 .getname = get_name,
1619 .poll = poll,
1620 .ioctl = ioctl,
1621 .listen = listen,
1622 .shutdown = shutdown,
1623 .setsockopt = setsockopt,
1624 .getsockopt = getsockopt,
1625 .sendmsg = send_msg,
1626 .recvmsg = recv_msg,
1627 .mmap = no_mmap,
1628 .sendpage = no_sendpage
1629};
1630
1631static struct proto_ops packet_ops = {
1632 .owner = THIS_MODULE,
1633 .family = AF_TIPC,
1634 .release = release,
1635 .bind = bind,
1636 .connect = connect,
1637 .socketpair = no_skpair,
1638 .accept = accept,
1639 .getname = get_name,
1640 .poll = poll,
1641 .ioctl = ioctl,
1642 .listen = listen,
1643 .shutdown = shutdown,
1644 .setsockopt = setsockopt,
1645 .getsockopt = getsockopt,
1646 .sendmsg = send_packet,
1647 .recvmsg = recv_msg,
1648 .mmap = no_mmap,
1649 .sendpage = no_sendpage
1650};
1651
1652static struct proto_ops stream_ops = {
1653 .owner = THIS_MODULE,
1654 .family = AF_TIPC,
1655 .release = release,
1656 .bind = bind,
1657 .connect = connect,
1658 .socketpair = no_skpair,
1659 .accept = accept,
1660 .getname = get_name,
1661 .poll = poll,
1662 .ioctl = ioctl,
1663 .listen = listen,
1664 .shutdown = shutdown,
1665 .setsockopt = setsockopt,
1666 .getsockopt = getsockopt,
1667 .sendmsg = send_stream,
1668 .recvmsg = recv_stream,
1669 .mmap = no_mmap,
1670 .sendpage = no_sendpage
1671};
1672
1673static struct net_proto_family tipc_family_ops = {
1674 .owner = THIS_MODULE,
1675 .family = AF_TIPC,
1676 .create = tipc_create
1677};
1678
1679static struct proto tipc_proto = {
1680 .name = "TIPC",
1681 .owner = THIS_MODULE,
1682 .obj_size = sizeof(struct tipc_sock)
1683};
1684
1685/**
Per Liden4323add2006-01-18 00:38:21 +01001686 * tipc_socket_init - initialize TIPC socket interface
Per Lidenb97bf3f2006-01-02 19:04:38 +01001687 *
1688 * Returns 0 on success, errno otherwise
1689 */
Per Liden4323add2006-01-18 00:38:21 +01001690int tipc_socket_init(void)
Per Lidenb97bf3f2006-01-02 19:04:38 +01001691{
1692 int res;
1693
1694 res = proto_register(&tipc_proto, 1);
1695 if (res) {
Per Lidend0a14a92006-01-11 13:52:51 +01001696 err("Failed to register TIPC protocol type\n");
Per Lidenb97bf3f2006-01-02 19:04:38 +01001697 goto out;
1698 }
1699
1700 res = sock_register(&tipc_family_ops);
1701 if (res) {
Per Lidend0a14a92006-01-11 13:52:51 +01001702 err("Failed to register TIPC socket type\n");
Per Lidenb97bf3f2006-01-02 19:04:38 +01001703 proto_unregister(&tipc_proto);
1704 goto out;
1705 }
1706
1707 sockets_enabled = 1;
1708 out:
1709 return res;
1710}
1711
1712/**
Per Liden4323add2006-01-18 00:38:21 +01001713 * tipc_socket_stop - stop TIPC socket interface
Per Lidenb97bf3f2006-01-02 19:04:38 +01001714 */
Per Liden4323add2006-01-18 00:38:21 +01001715void tipc_socket_stop(void)
Per Lidenb97bf3f2006-01-02 19:04:38 +01001716{
1717 if (!sockets_enabled)
1718 return;
1719
1720 sockets_enabled = 0;
1721 sock_unregister(tipc_family_ops.family);
1722 proto_unregister(&tipc_proto);
1723}
1724