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, 2012-2017, 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 <linux/sched/signal.h>
39
40#include "core.h"
41#include "name_table.h"
42#include "node.h"
43#include "link.h"
44#include "name_distr.h"
45#include "socket.h"
46#include "bcast.h"
47#include "netlink.h"
48#include "group.h"
49#include "trace.h"
50
51#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
52#define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
53#define TIPC_FWD_MSG 1
54#define TIPC_MAX_PORT 0xffffffff
55#define TIPC_MIN_PORT 1
56#define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
57
58enum {
59 TIPC_LISTEN = TCP_LISTEN,
60 TIPC_ESTABLISHED = TCP_ESTABLISHED,
61 TIPC_OPEN = TCP_CLOSE,
62 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
63 TIPC_CONNECTING = TCP_SYN_SENT,
64};
65
66struct sockaddr_pair {
67 struct sockaddr_tipc sock;
68 struct sockaddr_tipc member;
69};
70
71/**
72 * struct tipc_sock - TIPC socket structure
73 * @sk: socket - interacts with 'port' and with user via the socket API
74 * @conn_type: TIPC type used when connection was established
75 * @conn_instance: TIPC instance used when connection was established
76 * @published: non-zero if port has one or more associated names
77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
78 * @portid: unique port identity in TIPC socket hash table
79 * @phdr: preformatted message header used when sending messages
80 * #cong_links: list of congested links
81 * @publications: list of publications for port
82 * @blocking_link: address of the congested link we are currently sleeping on
83 * @pub_count: total # of publications port has made during its lifetime
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86 * @cong_link_cnt: number of congested links
87 * @snt_unacked: # messages sent by socket, and not yet acked by peer
88 * @rcv_unacked: # messages read by user, but not yet acked back to peer
89 * @peer: 'connected' peer for dgram/rdm
90 * @node: hash table node
91 * @mc_method: cookie for use between socket and broadcast layer
92 * @rcu: rcu struct for tipc_sock
93 */
94struct tipc_sock {
95 struct sock sk;
96 u32 conn_type;
97 u32 conn_instance;
98 int published;
99 u32 max_pkt;
100 u32 portid;
101 struct tipc_msg phdr;
102 struct list_head cong_links;
103 struct list_head publications;
104 u32 pub_count;
105 atomic_t dupl_rcvcnt;
106 u16 conn_timeout;
107 bool probe_unacked;
108 u16 cong_link_cnt;
109 u16 snt_unacked;
110 u16 snd_win;
111 u16 peer_caps;
112 u16 rcv_unacked;
113 u16 rcv_win;
114 struct sockaddr_tipc peer;
115 struct rhash_head node;
116 struct tipc_mc_method mc_method;
117 struct rcu_head rcu;
118 struct tipc_group *group;
119 bool group_is_open;
120};
121
122static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123static void tipc_data_ready(struct sock *sk);
124static void tipc_write_space(struct sock *sk);
125static void tipc_sock_destruct(struct sock *sk);
126static int tipc_release(struct socket *sock);
127static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
128 bool kern);
129static void tipc_sk_timeout(struct timer_list *t);
130static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 struct tipc_name_seq const *seq);
132static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 struct tipc_name_seq const *seq);
134static int tipc_sk_leave(struct tipc_sock *tsk);
135static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136static int tipc_sk_insert(struct tipc_sock *tsk);
137static void tipc_sk_remove(struct tipc_sock *tsk);
138static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
140
141static const struct proto_ops packet_ops;
142static const struct proto_ops stream_ops;
143static const struct proto_ops msg_ops;
144static struct proto tipc_proto;
145static const struct rhashtable_params tsk_rht_params;
146
147static u32 tsk_own_node(struct tipc_sock *tsk)
148{
149 return msg_prevnode(&tsk->phdr);
150}
151
152static u32 tsk_peer_node(struct tipc_sock *tsk)
153{
154 return msg_destnode(&tsk->phdr);
155}
156
157static u32 tsk_peer_port(struct tipc_sock *tsk)
158{
159 return msg_destport(&tsk->phdr);
160}
161
162static bool tsk_unreliable(struct tipc_sock *tsk)
163{
164 return msg_src_droppable(&tsk->phdr) != 0;
165}
166
167static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
168{
169 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
170}
171
172static bool tsk_unreturnable(struct tipc_sock *tsk)
173{
174 return msg_dest_droppable(&tsk->phdr) != 0;
175}
176
177static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
178{
179 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
180}
181
182static int tsk_importance(struct tipc_sock *tsk)
183{
184 return msg_importance(&tsk->phdr);
185}
186
187static int tsk_set_importance(struct tipc_sock *tsk, int imp)
188{
189 if (imp > TIPC_CRITICAL_IMPORTANCE)
190 return -EINVAL;
191 msg_set_importance(&tsk->phdr, (u32)imp);
192 return 0;
193}
194
195static struct tipc_sock *tipc_sk(const struct sock *sk)
196{
197 return container_of(sk, struct tipc_sock, sk);
198}
199
200static bool tsk_conn_cong(struct tipc_sock *tsk)
201{
202 return tsk->snt_unacked > tsk->snd_win;
203}
204
205static u16 tsk_blocks(int len)
206{
207 return ((len / FLOWCTL_BLK_SZ) + 1);
208}
209
210/* tsk_blocks(): translate a buffer size in bytes to number of
211 * advertisable blocks, taking into account the ratio truesize(len)/len
212 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
213 */
214static u16 tsk_adv_blocks(int len)
215{
216 return len / FLOWCTL_BLK_SZ / 4;
217}
218
219/* tsk_inc(): increment counter for sent or received data
220 * - If block based flow control is not supported by peer we
221 * fall back to message based ditto, incrementing the counter
222 */
223static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
224{
225 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
226 return ((msglen / FLOWCTL_BLK_SZ) + 1);
227 return 1;
228}
229
230/**
231 * tsk_advance_rx_queue - discard first buffer in socket receive queue
232 *
233 * Caller must hold socket lock
234 */
235static void tsk_advance_rx_queue(struct sock *sk)
236{
237 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
238 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
239}
240
241/* tipc_sk_respond() : send response message back to sender
242 */
243static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
244{
245 u32 selector;
246 u32 dnode;
247 u32 onode = tipc_own_addr(sock_net(sk));
248
249 if (!tipc_msg_reverse(onode, &skb, err))
250 return;
251
252 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
253 dnode = msg_destnode(buf_msg(skb));
254 selector = msg_origport(buf_msg(skb));
255 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
256}
257
258/**
259 * tsk_rej_rx_queue - reject all buffers in socket receive queue
260 *
261 * Caller must hold socket lock
262 */
263static void tsk_rej_rx_queue(struct sock *sk)
264{
265 struct sk_buff *skb;
266
267 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
268 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
269}
270
271static bool tipc_sk_connected(struct sock *sk)
272{
273 return sk->sk_state == TIPC_ESTABLISHED;
274}
275
276/* tipc_sk_type_connectionless - check if the socket is datagram socket
277 * @sk: socket
278 *
279 * Returns true if connection less, false otherwise
280 */
281static bool tipc_sk_type_connectionless(struct sock *sk)
282{
283 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
284}
285
286/* tsk_peer_msg - verify if message was sent by connected port's peer
287 *
288 * Handles cases where the node's network address has changed from
289 * the default of <0.0.0> to its configured setting.
290 */
291static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
292{
293 struct sock *sk = &tsk->sk;
294 u32 self = tipc_own_addr(sock_net(sk));
295 u32 peer_port = tsk_peer_port(tsk);
296 u32 orig_node, peer_node;
297
298 if (unlikely(!tipc_sk_connected(sk)))
299 return false;
300
301 if (unlikely(msg_origport(msg) != peer_port))
302 return false;
303
304 orig_node = msg_orignode(msg);
305 peer_node = tsk_peer_node(tsk);
306
307 if (likely(orig_node == peer_node))
308 return true;
309
310 if (!orig_node && peer_node == self)
311 return true;
312
313 if (!peer_node && orig_node == self)
314 return true;
315
316 return false;
317}
318
319/* tipc_set_sk_state - set the sk_state of the socket
320 * @sk: socket
321 *
322 * Caller must hold socket lock
323 *
324 * Returns 0 on success, errno otherwise
325 */
326static int tipc_set_sk_state(struct sock *sk, int state)
327{
328 int oldsk_state = sk->sk_state;
329 int res = -EINVAL;
330
331 switch (state) {
332 case TIPC_OPEN:
333 res = 0;
334 break;
335 case TIPC_LISTEN:
336 case TIPC_CONNECTING:
337 if (oldsk_state == TIPC_OPEN)
338 res = 0;
339 break;
340 case TIPC_ESTABLISHED:
341 if (oldsk_state == TIPC_CONNECTING ||
342 oldsk_state == TIPC_OPEN)
343 res = 0;
344 break;
345 case TIPC_DISCONNECTING:
346 if (oldsk_state == TIPC_CONNECTING ||
347 oldsk_state == TIPC_ESTABLISHED)
348 res = 0;
349 break;
350 }
351
352 if (!res)
353 sk->sk_state = state;
354
355 return res;
356}
357
358static int tipc_sk_sock_err(struct socket *sock, long *timeout)
359{
360 struct sock *sk = sock->sk;
361 int err = sock_error(sk);
362 int typ = sock->type;
363
364 if (err)
365 return err;
366 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
367 if (sk->sk_state == TIPC_DISCONNECTING)
368 return -EPIPE;
369 else if (!tipc_sk_connected(sk))
370 return -ENOTCONN;
371 }
372 if (!*timeout)
373 return -EAGAIN;
374 if (signal_pending(current))
375 return sock_intr_errno(*timeout);
376
377 return 0;
378}
379
380#define tipc_wait_for_cond(sock_, timeo_, condition_) \
381({ \
382 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
383 struct sock *sk_; \
384 int rc_; \
385 \
386 while ((rc_ = !(condition_))) { \
387 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
388 smp_rmb(); \
389 sk_ = (sock_)->sk; \
390 rc_ = tipc_sk_sock_err((sock_), timeo_); \
391 if (rc_) \
392 break; \
393 add_wait_queue(sk_sleep(sk_), &wait_); \
394 release_sock(sk_); \
395 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
396 sched_annotate_sleep(); \
397 lock_sock(sk_); \
398 remove_wait_queue(sk_sleep(sk_), &wait_); \
399 } \
400 rc_; \
401})
402
403/**
404 * tipc_sk_create - create a TIPC socket
405 * @net: network namespace (must be default network)
406 * @sock: pre-allocated socket structure
407 * @protocol: protocol indicator (must be 0)
408 * @kern: caused by kernel or by userspace?
409 *
410 * This routine creates additional data structures used by the TIPC socket,
411 * initializes them, and links them together.
412 *
413 * Returns 0 on success, errno otherwise
414 */
415static int tipc_sk_create(struct net *net, struct socket *sock,
416 int protocol, int kern)
417{
418 const struct proto_ops *ops;
419 struct sock *sk;
420 struct tipc_sock *tsk;
421 struct tipc_msg *msg;
422
423 /* Validate arguments */
424 if (unlikely(protocol != 0))
425 return -EPROTONOSUPPORT;
426
427 switch (sock->type) {
428 case SOCK_STREAM:
429 ops = &stream_ops;
430 break;
431 case SOCK_SEQPACKET:
432 ops = &packet_ops;
433 break;
434 case SOCK_DGRAM:
435 case SOCK_RDM:
436 ops = &msg_ops;
437 break;
438 default:
439 return -EPROTOTYPE;
440 }
441
442 /* Allocate socket's protocol area */
443 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
444 if (sk == NULL)
445 return -ENOMEM;
446
447 tsk = tipc_sk(sk);
448 tsk->max_pkt = MAX_PKT_DEFAULT;
449 INIT_LIST_HEAD(&tsk->publications);
450 INIT_LIST_HEAD(&tsk->cong_links);
451 msg = &tsk->phdr;
452
453 /* Finish initializing socket data structures */
454 sock->ops = ops;
455 sock_init_data(sock, sk);
456 tipc_set_sk_state(sk, TIPC_OPEN);
457 if (tipc_sk_insert(tsk)) {
458 pr_warn("Socket create failed; port number exhausted\n");
459 return -EINVAL;
460 }
461
462 /* Ensure tsk is visible before we read own_addr. */
463 smp_mb();
464
465 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
466 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
467
468 msg_set_origport(msg, tsk->portid);
469 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
470 sk->sk_shutdown = 0;
471 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
472 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
473 sk->sk_data_ready = tipc_data_ready;
474 sk->sk_write_space = tipc_write_space;
475 sk->sk_destruct = tipc_sock_destruct;
476 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
477 tsk->group_is_open = true;
478 atomic_set(&tsk->dupl_rcvcnt, 0);
479
480 /* Start out with safe limits until we receive an advertised window */
481 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
482 tsk->rcv_win = tsk->snd_win;
483
484 if (tipc_sk_type_connectionless(sk)) {
485 tsk_set_unreturnable(tsk, true);
486 if (sock->type == SOCK_DGRAM)
487 tsk_set_unreliable(tsk, true);
488 }
489 __skb_queue_head_init(&tsk->mc_method.deferredq);
490 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
491 return 0;
492}
493
494static void tipc_sk_callback(struct rcu_head *head)
495{
496 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
497
498 sock_put(&tsk->sk);
499}
500
501/* Caller should hold socket lock for the socket. */
502static void __tipc_shutdown(struct socket *sock, int error)
503{
504 struct sock *sk = sock->sk;
505 struct tipc_sock *tsk = tipc_sk(sk);
506 struct net *net = sock_net(sk);
507 long timeout = CONN_TIMEOUT_DEFAULT;
508 u32 dnode = tsk_peer_node(tsk);
509 struct sk_buff *skb;
510
511 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
512 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
513 !tsk_conn_cong(tsk)));
514
515 /* Remove any pending SYN message */
516 __skb_queue_purge(&sk->sk_write_queue);
517
518 /* Reject all unreceived messages, except on an active connection
519 * (which disconnects locally & sends a 'FIN+' to peer).
520 */
521 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
522 if (TIPC_SKB_CB(skb)->bytes_read) {
523 kfree_skb(skb);
524 continue;
525 }
526 if (!tipc_sk_type_connectionless(sk) &&
527 sk->sk_state != TIPC_DISCONNECTING) {
528 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
529 tipc_node_remove_conn(net, dnode, tsk->portid);
530 }
531 tipc_sk_respond(sk, skb, error);
532 }
533
534 if (tipc_sk_type_connectionless(sk))
535 return;
536
537 if (sk->sk_state != TIPC_DISCONNECTING) {
538 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
539 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
540 tsk_own_node(tsk), tsk_peer_port(tsk),
541 tsk->portid, error);
542 if (skb)
543 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
544 tipc_node_remove_conn(net, dnode, tsk->portid);
545 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
546 }
547}
548
549/**
550 * tipc_release - destroy a TIPC socket
551 * @sock: socket to destroy
552 *
553 * This routine cleans up any messages that are still queued on the socket.
554 * For DGRAM and RDM socket types, all queued messages are rejected.
555 * For SEQPACKET and STREAM socket types, the first message is rejected
556 * and any others are discarded. (If the first message on a STREAM socket
557 * is partially-read, it is discarded and the next one is rejected instead.)
558 *
559 * NOTE: Rejected messages are not necessarily returned to the sender! They
560 * are returned or discarded according to the "destination droppable" setting
561 * specified for the message by the sender.
562 *
563 * Returns 0 on success, errno otherwise
564 */
565static int tipc_release(struct socket *sock)
566{
567 struct sock *sk = sock->sk;
568 struct tipc_sock *tsk;
569
570 /*
571 * Exit if socket isn't fully initialized (occurs when a failed accept()
572 * releases a pre-allocated child socket that was never used)
573 */
574 if (sk == NULL)
575 return 0;
576
577 tsk = tipc_sk(sk);
578 lock_sock(sk);
579
580 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
581 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
582 sk->sk_shutdown = SHUTDOWN_MASK;
583 tipc_sk_leave(tsk);
584 tipc_sk_withdraw(tsk, 0, NULL);
585 __skb_queue_purge(&tsk->mc_method.deferredq);
586 sk_stop_timer(sk, &sk->sk_timer);
587 tipc_sk_remove(tsk);
588
589 sock_orphan(sk);
590 /* Reject any messages that accumulated in backlog queue */
591 release_sock(sk);
592 tipc_dest_list_purge(&tsk->cong_links);
593 tsk->cong_link_cnt = 0;
594 call_rcu(&tsk->rcu, tipc_sk_callback);
595 sock->sk = NULL;
596
597 return 0;
598}
599
600/**
601 * tipc_bind - associate or disassocate TIPC name(s) with a socket
602 * @sock: socket structure
603 * @uaddr: socket address describing name(s) and desired operation
604 * @uaddr_len: size of socket address data structure
605 *
606 * Name and name sequence binding is indicated using a positive scope value;
607 * a negative scope value unbinds the specified name. Specifying no name
608 * (i.e. a socket address length of 0) unbinds all names from the socket.
609 *
610 * Returns 0 on success, errno otherwise
611 *
612 * NOTE: This routine doesn't need to take the socket lock since it doesn't
613 * access any non-constant socket information.
614 */
615static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
616 int uaddr_len)
617{
618 struct sock *sk = sock->sk;
619 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
620 struct tipc_sock *tsk = tipc_sk(sk);
621 int res = -EINVAL;
622
623 lock_sock(sk);
624 if (unlikely(!uaddr_len)) {
625 res = tipc_sk_withdraw(tsk, 0, NULL);
626 goto exit;
627 }
628 if (tsk->group) {
629 res = -EACCES;
630 goto exit;
631 }
632 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
633 res = -EINVAL;
634 goto exit;
635 }
636 if (addr->family != AF_TIPC) {
637 res = -EAFNOSUPPORT;
638 goto exit;
639 }
640
641 if (addr->addrtype == TIPC_ADDR_NAME)
642 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
643 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
644 res = -EAFNOSUPPORT;
645 goto exit;
646 }
647
648 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
649 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
650 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
651 res = -EACCES;
652 goto exit;
653 }
654
655 res = (addr->scope >= 0) ?
656 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
657 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
658exit:
659 release_sock(sk);
660 return res;
661}
662
663/**
664 * tipc_getname - get port ID of socket or peer socket
665 * @sock: socket structure
666 * @uaddr: area for returned socket address
667 * @uaddr_len: area for returned length of socket address
668 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
669 *
670 * Returns 0 on success, errno otherwise
671 *
672 * NOTE: This routine doesn't need to take the socket lock since it only
673 * accesses socket information that is unchanging (or which changes in
674 * a completely predictable manner).
675 */
676static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
677 int peer)
678{
679 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
680 struct sock *sk = sock->sk;
681 struct tipc_sock *tsk = tipc_sk(sk);
682
683 memset(addr, 0, sizeof(*addr));
684 if (peer) {
685 if ((!tipc_sk_connected(sk)) &&
686 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
687 return -ENOTCONN;
688 addr->addr.id.ref = tsk_peer_port(tsk);
689 addr->addr.id.node = tsk_peer_node(tsk);
690 } else {
691 addr->addr.id.ref = tsk->portid;
692 addr->addr.id.node = tipc_own_addr(sock_net(sk));
693 }
694
695 addr->addrtype = TIPC_ADDR_ID;
696 addr->family = AF_TIPC;
697 addr->scope = 0;
698 addr->addr.name.domain = 0;
699
700 return sizeof(*addr);
701}
702
703/**
704 * tipc_poll - read and possibly block on pollmask
705 * @file: file structure associated with the socket
706 * @sock: socket for which to calculate the poll bits
707 * @wait: ???
708 *
709 * Returns pollmask value
710 *
711 * COMMENTARY:
712 * It appears that the usual socket locking mechanisms are not useful here
713 * since the pollmask info is potentially out-of-date the moment this routine
714 * exits. TCP and other protocols seem to rely on higher level poll routines
715 * to handle any preventable race conditions, so TIPC will do the same ...
716 *
717 * IMPORTANT: The fact that a read or write operation is indicated does NOT
718 * imply that the operation will succeed, merely that it should be performed
719 * and will not block.
720 */
721static __poll_t tipc_poll(struct file *file, struct socket *sock,
722 poll_table *wait)
723{
724 struct sock *sk = sock->sk;
725 struct tipc_sock *tsk = tipc_sk(sk);
726 __poll_t revents = 0;
727
728 sock_poll_wait(file, sock, wait);
729 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
730
731 if (sk->sk_shutdown & RCV_SHUTDOWN)
732 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
733 if (sk->sk_shutdown == SHUTDOWN_MASK)
734 revents |= EPOLLHUP;
735
736 switch (sk->sk_state) {
737 case TIPC_ESTABLISHED:
738 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
739 revents |= EPOLLOUT;
740 /* fall through */
741 case TIPC_LISTEN:
742 case TIPC_CONNECTING:
743 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
744 revents |= EPOLLIN | EPOLLRDNORM;
745 break;
746 case TIPC_OPEN:
747 if (tsk->group_is_open && !tsk->cong_link_cnt)
748 revents |= EPOLLOUT;
749 if (!tipc_sk_type_connectionless(sk))
750 break;
751 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
752 break;
753 revents |= EPOLLIN | EPOLLRDNORM;
754 break;
755 case TIPC_DISCONNECTING:
756 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
757 break;
758 }
759 return revents;
760}
761
762/**
763 * tipc_sendmcast - send multicast message
764 * @sock: socket structure
765 * @seq: destination address
766 * @msg: message to send
767 * @dlen: length of data to send
768 * @timeout: timeout to wait for wakeup
769 *
770 * Called from function tipc_sendmsg(), which has done all sanity checks
771 * Returns the number of bytes sent on success, or errno
772 */
773static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
774 struct msghdr *msg, size_t dlen, long timeout)
775{
776 struct sock *sk = sock->sk;
777 struct tipc_sock *tsk = tipc_sk(sk);
778 struct tipc_msg *hdr = &tsk->phdr;
779 struct net *net = sock_net(sk);
780 int mtu = tipc_bcast_get_mtu(net);
781 struct tipc_mc_method *method = &tsk->mc_method;
782 struct sk_buff_head pkts;
783 struct tipc_nlist dsts;
784 int rc;
785
786 if (tsk->group)
787 return -EACCES;
788
789 /* Block or return if any destination link is congested */
790 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
791 if (unlikely(rc))
792 return rc;
793
794 /* Lookup destination nodes */
795 tipc_nlist_init(&dsts, tipc_own_addr(net));
796 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
797 seq->upper, &dsts);
798 if (!dsts.local && !dsts.remote)
799 return -EHOSTUNREACH;
800
801 /* Build message header */
802 msg_set_type(hdr, TIPC_MCAST_MSG);
803 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
804 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
805 msg_set_destport(hdr, 0);
806 msg_set_destnode(hdr, 0);
807 msg_set_nametype(hdr, seq->type);
808 msg_set_namelower(hdr, seq->lower);
809 msg_set_nameupper(hdr, seq->upper);
810
811 /* Build message as chain of buffers */
812 __skb_queue_head_init(&pkts);
813 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
814
815 /* Send message if build was successful */
816 if (unlikely(rc == dlen)) {
817 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
818 TIPC_DUMP_SK_SNDQ, " ");
819 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
820 &tsk->cong_link_cnt);
821 }
822
823 tipc_nlist_purge(&dsts);
824
825 return rc ? rc : dlen;
826}
827
828/**
829 * tipc_send_group_msg - send a message to a member in the group
830 * @net: network namespace
831 * @m: message to send
832 * @mb: group member
833 * @dnode: destination node
834 * @dport: destination port
835 * @dlen: total length of message data
836 */
837static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
838 struct msghdr *m, struct tipc_member *mb,
839 u32 dnode, u32 dport, int dlen)
840{
841 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
842 struct tipc_mc_method *method = &tsk->mc_method;
843 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
844 struct tipc_msg *hdr = &tsk->phdr;
845 struct sk_buff_head pkts;
846 int mtu, rc;
847
848 /* Complete message header */
849 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
850 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
851 msg_set_destport(hdr, dport);
852 msg_set_destnode(hdr, dnode);
853 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
854
855 /* Build message as chain of buffers */
856 __skb_queue_head_init(&pkts);
857 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
858 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
859 if (unlikely(rc != dlen))
860 return rc;
861
862 /* Send message */
863 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
864 if (unlikely(rc == -ELINKCONG)) {
865 tipc_dest_push(&tsk->cong_links, dnode, 0);
866 tsk->cong_link_cnt++;
867 }
868
869 /* Update send window */
870 tipc_group_update_member(mb, blks);
871
872 /* A broadcast sent within next EXPIRE period must follow same path */
873 method->rcast = true;
874 method->mandatory = true;
875 return dlen;
876}
877
878/**
879 * tipc_send_group_unicast - send message to a member in the group
880 * @sock: socket structure
881 * @m: message to send
882 * @dlen: total length of message data
883 * @timeout: timeout to wait for wakeup
884 *
885 * Called from function tipc_sendmsg(), which has done all sanity checks
886 * Returns the number of bytes sent on success, or errno
887 */
888static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
889 int dlen, long timeout)
890{
891 struct sock *sk = sock->sk;
892 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
893 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
894 struct tipc_sock *tsk = tipc_sk(sk);
895 struct net *net = sock_net(sk);
896 struct tipc_member *mb = NULL;
897 u32 node, port;
898 int rc;
899
900 node = dest->addr.id.node;
901 port = dest->addr.id.ref;
902 if (!port && !node)
903 return -EHOSTUNREACH;
904
905 /* Block or return if destination link or member is congested */
906 rc = tipc_wait_for_cond(sock, &timeout,
907 !tipc_dest_find(&tsk->cong_links, node, 0) &&
908 tsk->group &&
909 !tipc_group_cong(tsk->group, node, port, blks,
910 &mb));
911 if (unlikely(rc))
912 return rc;
913
914 if (unlikely(!mb))
915 return -EHOSTUNREACH;
916
917 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
918
919 return rc ? rc : dlen;
920}
921
922/**
923 * tipc_send_group_anycast - send message to any member with given identity
924 * @sock: socket structure
925 * @m: message to send
926 * @dlen: total length of message data
927 * @timeout: timeout to wait for wakeup
928 *
929 * Called from function tipc_sendmsg(), which has done all sanity checks
930 * Returns the number of bytes sent on success, or errno
931 */
932static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
933 int dlen, long timeout)
934{
935 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
936 struct sock *sk = sock->sk;
937 struct tipc_sock *tsk = tipc_sk(sk);
938 struct list_head *cong_links = &tsk->cong_links;
939 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
940 struct tipc_msg *hdr = &tsk->phdr;
941 struct tipc_member *first = NULL;
942 struct tipc_member *mbr = NULL;
943 struct net *net = sock_net(sk);
944 u32 node, port, exclude;
945 struct list_head dsts;
946 u32 type, inst, scope;
947 int lookups = 0;
948 int dstcnt, rc;
949 bool cong;
950
951 INIT_LIST_HEAD(&dsts);
952
953 type = msg_nametype(hdr);
954 inst = dest->addr.name.name.instance;
955 scope = msg_lookup_scope(hdr);
956
957 while (++lookups < 4) {
958 exclude = tipc_group_exclude(tsk->group);
959
960 first = NULL;
961
962 /* Look for a non-congested destination member, if any */
963 while (1) {
964 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
965 &dstcnt, exclude, false))
966 return -EHOSTUNREACH;
967 tipc_dest_pop(&dsts, &node, &port);
968 cong = tipc_group_cong(tsk->group, node, port, blks,
969 &mbr);
970 if (!cong)
971 break;
972 if (mbr == first)
973 break;
974 if (!first)
975 first = mbr;
976 }
977
978 /* Start over if destination was not in member list */
979 if (unlikely(!mbr))
980 continue;
981
982 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
983 break;
984
985 /* Block or return if destination link or member is congested */
986 rc = tipc_wait_for_cond(sock, &timeout,
987 !tipc_dest_find(cong_links, node, 0) &&
988 tsk->group &&
989 !tipc_group_cong(tsk->group, node, port,
990 blks, &mbr));
991 if (unlikely(rc))
992 return rc;
993
994 /* Send, unless destination disappeared while waiting */
995 if (likely(mbr))
996 break;
997 }
998
999 if (unlikely(lookups >= 4))
1000 return -EHOSTUNREACH;
1001
1002 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1003
1004 return rc ? rc : dlen;
1005}
1006
1007/**
1008 * tipc_send_group_bcast - send message to all members in communication group
1009 * @sk: socket structure
1010 * @m: message to send
1011 * @dlen: total length of message data
1012 * @timeout: timeout to wait for wakeup
1013 *
1014 * Called from function tipc_sendmsg(), which has done all sanity checks
1015 * Returns the number of bytes sent on success, or errno
1016 */
1017static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1018 int dlen, long timeout)
1019{
1020 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1021 struct sock *sk = sock->sk;
1022 struct net *net = sock_net(sk);
1023 struct tipc_sock *tsk = tipc_sk(sk);
1024 struct tipc_nlist *dsts;
1025 struct tipc_mc_method *method = &tsk->mc_method;
1026 bool ack = method->mandatory && method->rcast;
1027 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1028 struct tipc_msg *hdr = &tsk->phdr;
1029 int mtu = tipc_bcast_get_mtu(net);
1030 struct sk_buff_head pkts;
1031 int rc = -EHOSTUNREACH;
1032
1033 /* Block or return if any destination link or member is congested */
1034 rc = tipc_wait_for_cond(sock, &timeout,
1035 !tsk->cong_link_cnt && tsk->group &&
1036 !tipc_group_bc_cong(tsk->group, blks));
1037 if (unlikely(rc))
1038 return rc;
1039
1040 dsts = tipc_group_dests(tsk->group);
1041 if (!dsts->local && !dsts->remote)
1042 return -EHOSTUNREACH;
1043
1044 /* Complete message header */
1045 if (dest) {
1046 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1047 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1048 } else {
1049 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1050 msg_set_nameinst(hdr, 0);
1051 }
1052 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1053 msg_set_destport(hdr, 0);
1054 msg_set_destnode(hdr, 0);
1055 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1056
1057 /* Avoid getting stuck with repeated forced replicasts */
1058 msg_set_grp_bc_ack_req(hdr, ack);
1059
1060 /* Build message as chain of buffers */
1061 __skb_queue_head_init(&pkts);
1062 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1063 if (unlikely(rc != dlen))
1064 return rc;
1065
1066 /* Send message */
1067 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1068 if (unlikely(rc))
1069 return rc;
1070
1071 /* Update broadcast sequence number and send windows */
1072 tipc_group_update_bc_members(tsk->group, blks, ack);
1073
1074 /* Broadcast link is now free to choose method for next broadcast */
1075 method->mandatory = false;
1076 method->expires = jiffies;
1077
1078 return dlen;
1079}
1080
1081/**
1082 * tipc_send_group_mcast - send message to all members with given identity
1083 * @sock: socket structure
1084 * @m: message to send
1085 * @dlen: total length of message data
1086 * @timeout: timeout to wait for wakeup
1087 *
1088 * Called from function tipc_sendmsg(), which has done all sanity checks
1089 * Returns the number of bytes sent on success, or errno
1090 */
1091static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1092 int dlen, long timeout)
1093{
1094 struct sock *sk = sock->sk;
1095 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1096 struct tipc_sock *tsk = tipc_sk(sk);
1097 struct tipc_group *grp = tsk->group;
1098 struct tipc_msg *hdr = &tsk->phdr;
1099 struct net *net = sock_net(sk);
1100 u32 type, inst, scope, exclude;
1101 struct list_head dsts;
1102 u32 dstcnt;
1103
1104 INIT_LIST_HEAD(&dsts);
1105
1106 type = msg_nametype(hdr);
1107 inst = dest->addr.name.name.instance;
1108 scope = msg_lookup_scope(hdr);
1109 exclude = tipc_group_exclude(grp);
1110
1111 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1112 &dstcnt, exclude, true))
1113 return -EHOSTUNREACH;
1114
1115 if (dstcnt == 1) {
1116 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1117 return tipc_send_group_unicast(sock, m, dlen, timeout);
1118 }
1119
1120 tipc_dest_list_purge(&dsts);
1121 return tipc_send_group_bcast(sock, m, dlen, timeout);
1122}
1123
1124/**
1125 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1126 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1127 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1128 *
1129 * Multi-threaded: parallel calls with reference to same queues may occur
1130 */
1131void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1132 struct sk_buff_head *inputq)
1133{
1134 u32 self = tipc_own_addr(net);
1135 u32 type, lower, upper, scope;
1136 struct sk_buff *skb, *_skb;
1137 u32 portid, onode;
1138 struct sk_buff_head tmpq;
1139 struct list_head dports;
1140 struct tipc_msg *hdr;
1141 int user, mtyp, hlen;
1142 bool exact;
1143
1144 __skb_queue_head_init(&tmpq);
1145 INIT_LIST_HEAD(&dports);
1146
1147 skb = tipc_skb_peek(arrvq, &inputq->lock);
1148 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1149 hdr = buf_msg(skb);
1150 user = msg_user(hdr);
1151 mtyp = msg_type(hdr);
1152 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1153 onode = msg_orignode(hdr);
1154 type = msg_nametype(hdr);
1155
1156 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1157 spin_lock_bh(&inputq->lock);
1158 if (skb_peek(arrvq) == skb) {
1159 __skb_dequeue(arrvq);
1160 __skb_queue_tail(inputq, skb);
1161 }
1162 kfree_skb(skb);
1163 spin_unlock_bh(&inputq->lock);
1164 continue;
1165 }
1166
1167 /* Group messages require exact scope match */
1168 if (msg_in_group(hdr)) {
1169 lower = 0;
1170 upper = ~0;
1171 scope = msg_lookup_scope(hdr);
1172 exact = true;
1173 } else {
1174 /* TIPC_NODE_SCOPE means "any scope" in this context */
1175 if (onode == self)
1176 scope = TIPC_NODE_SCOPE;
1177 else
1178 scope = TIPC_CLUSTER_SCOPE;
1179 exact = false;
1180 lower = msg_namelower(hdr);
1181 upper = msg_nameupper(hdr);
1182 }
1183
1184 /* Create destination port list: */
1185 tipc_nametbl_mc_lookup(net, type, lower, upper,
1186 scope, exact, &dports);
1187
1188 /* Clone message per destination */
1189 while (tipc_dest_pop(&dports, NULL, &portid)) {
1190 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1191 if (_skb) {
1192 msg_set_destport(buf_msg(_skb), portid);
1193 __skb_queue_tail(&tmpq, _skb);
1194 continue;
1195 }
1196 pr_warn("Failed to clone mcast rcv buffer\n");
1197 }
1198 /* Append to inputq if not already done by other thread */
1199 spin_lock_bh(&inputq->lock);
1200 if (skb_peek(arrvq) == skb) {
1201 skb_queue_splice_tail_init(&tmpq, inputq);
1202 kfree_skb(__skb_dequeue(arrvq));
1203 }
1204 spin_unlock_bh(&inputq->lock);
1205 __skb_queue_purge(&tmpq);
1206 kfree_skb(skb);
1207 }
1208 tipc_sk_rcv(net, inputq);
1209}
1210
1211/**
1212 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1213 * @tsk: receiving socket
1214 * @skb: pointer to message buffer.
1215 */
1216static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1217 struct sk_buff_head *inputq,
1218 struct sk_buff_head *xmitq)
1219{
1220 struct tipc_msg *hdr = buf_msg(skb);
1221 u32 onode = tsk_own_node(tsk);
1222 struct sock *sk = &tsk->sk;
1223 int mtyp = msg_type(hdr);
1224 bool conn_cong;
1225
1226 /* Ignore if connection cannot be validated: */
1227 if (!tsk_peer_msg(tsk, hdr)) {
1228 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1229 goto exit;
1230 }
1231
1232 if (unlikely(msg_errcode(hdr))) {
1233 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1234 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1235 tsk_peer_port(tsk));
1236 sk->sk_state_change(sk);
1237
1238 /* State change is ignored if socket already awake,
1239 * - convert msg to abort msg and add to inqueue
1240 */
1241 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1242 msg_set_type(hdr, TIPC_CONN_MSG);
1243 msg_set_size(hdr, BASIC_H_SIZE);
1244 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1245 __skb_queue_tail(inputq, skb);
1246 return;
1247 }
1248
1249 tsk->probe_unacked = false;
1250
1251 if (mtyp == CONN_PROBE) {
1252 msg_set_type(hdr, CONN_PROBE_REPLY);
1253 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1254 __skb_queue_tail(xmitq, skb);
1255 return;
1256 } else if (mtyp == CONN_ACK) {
1257 conn_cong = tsk_conn_cong(tsk);
1258 tsk->snt_unacked -= msg_conn_ack(hdr);
1259 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1260 tsk->snd_win = msg_adv_win(hdr);
1261 if (conn_cong)
1262 sk->sk_write_space(sk);
1263 } else if (mtyp != CONN_PROBE_REPLY) {
1264 pr_warn("Received unknown CONN_PROTO msg\n");
1265 }
1266exit:
1267 kfree_skb(skb);
1268}
1269
1270/**
1271 * tipc_sendmsg - send message in connectionless manner
1272 * @sock: socket structure
1273 * @m: message to send
1274 * @dsz: amount of user data to be sent
1275 *
1276 * Message must have an destination specified explicitly.
1277 * Used for SOCK_RDM and SOCK_DGRAM messages,
1278 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1279 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1280 *
1281 * Returns the number of bytes sent on success, or errno otherwise
1282 */
1283static int tipc_sendmsg(struct socket *sock,
1284 struct msghdr *m, size_t dsz)
1285{
1286 struct sock *sk = sock->sk;
1287 int ret;
1288
1289 lock_sock(sk);
1290 ret = __tipc_sendmsg(sock, m, dsz);
1291 release_sock(sk);
1292
1293 return ret;
1294}
1295
1296static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1297{
1298 struct sock *sk = sock->sk;
1299 struct net *net = sock_net(sk);
1300 struct tipc_sock *tsk = tipc_sk(sk);
1301 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1302 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1303 struct list_head *clinks = &tsk->cong_links;
1304 bool syn = !tipc_sk_type_connectionless(sk);
1305 struct tipc_group *grp = tsk->group;
1306 struct tipc_msg *hdr = &tsk->phdr;
1307 struct tipc_name_seq *seq;
1308 struct sk_buff_head pkts;
1309 u32 dport, dnode = 0;
1310 u32 type, inst;
1311 int mtu, rc;
1312
1313 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1314 return -EMSGSIZE;
1315
1316 if (likely(dest)) {
1317 if (unlikely(m->msg_namelen < sizeof(*dest)))
1318 return -EINVAL;
1319 if (unlikely(dest->family != AF_TIPC))
1320 return -EINVAL;
1321 }
1322
1323 if (grp) {
1324 if (!dest)
1325 return tipc_send_group_bcast(sock, m, dlen, timeout);
1326 if (dest->addrtype == TIPC_ADDR_NAME)
1327 return tipc_send_group_anycast(sock, m, dlen, timeout);
1328 if (dest->addrtype == TIPC_ADDR_ID)
1329 return tipc_send_group_unicast(sock, m, dlen, timeout);
1330 if (dest->addrtype == TIPC_ADDR_MCAST)
1331 return tipc_send_group_mcast(sock, m, dlen, timeout);
1332 return -EINVAL;
1333 }
1334
1335 if (unlikely(!dest)) {
1336 dest = &tsk->peer;
1337 if (!syn && dest->family != AF_TIPC)
1338 return -EDESTADDRREQ;
1339 }
1340
1341 if (unlikely(syn)) {
1342 if (sk->sk_state == TIPC_LISTEN)
1343 return -EPIPE;
1344 if (sk->sk_state != TIPC_OPEN)
1345 return -EISCONN;
1346 if (tsk->published)
1347 return -EOPNOTSUPP;
1348 if (dest->addrtype == TIPC_ADDR_NAME) {
1349 tsk->conn_type = dest->addr.name.name.type;
1350 tsk->conn_instance = dest->addr.name.name.instance;
1351 }
1352 msg_set_syn(hdr, 1);
1353 }
1354
1355 seq = &dest->addr.nameseq;
1356 if (dest->addrtype == TIPC_ADDR_MCAST)
1357 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1358
1359 if (dest->addrtype == TIPC_ADDR_NAME) {
1360 type = dest->addr.name.name.type;
1361 inst = dest->addr.name.name.instance;
1362 dnode = dest->addr.name.domain;
1363 msg_set_type(hdr, TIPC_NAMED_MSG);
1364 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1365 msg_set_nametype(hdr, type);
1366 msg_set_nameinst(hdr, inst);
1367 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1368 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1369 msg_set_destnode(hdr, dnode);
1370 msg_set_destport(hdr, dport);
1371 if (unlikely(!dport && !dnode))
1372 return -EHOSTUNREACH;
1373 } else if (dest->addrtype == TIPC_ADDR_ID) {
1374 dnode = dest->addr.id.node;
1375 msg_set_type(hdr, TIPC_DIRECT_MSG);
1376 msg_set_lookup_scope(hdr, 0);
1377 msg_set_destnode(hdr, dnode);
1378 msg_set_destport(hdr, dest->addr.id.ref);
1379 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1380 } else {
1381 return -EINVAL;
1382 }
1383
1384 /* Block or return if destination link is congested */
1385 rc = tipc_wait_for_cond(sock, &timeout,
1386 !tipc_dest_find(clinks, dnode, 0));
1387 if (unlikely(rc))
1388 return rc;
1389
1390 __skb_queue_head_init(&pkts);
1391 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1392 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1393 if (unlikely(rc != dlen))
1394 return rc;
1395 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1396 return -ENOMEM;
1397
1398 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1399 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1400 if (unlikely(rc == -ELINKCONG)) {
1401 tipc_dest_push(clinks, dnode, 0);
1402 tsk->cong_link_cnt++;
1403 rc = 0;
1404 }
1405
1406 if (unlikely(syn && !rc))
1407 tipc_set_sk_state(sk, TIPC_CONNECTING);
1408
1409 return rc ? rc : dlen;
1410}
1411
1412/**
1413 * tipc_sendstream - send stream-oriented data
1414 * @sock: socket structure
1415 * @m: data to send
1416 * @dsz: total length of data to be transmitted
1417 *
1418 * Used for SOCK_STREAM data.
1419 *
1420 * Returns the number of bytes sent on success (or partial success),
1421 * or errno if no data sent
1422 */
1423static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1424{
1425 struct sock *sk = sock->sk;
1426 int ret;
1427
1428 lock_sock(sk);
1429 ret = __tipc_sendstream(sock, m, dsz);
1430 release_sock(sk);
1431
1432 return ret;
1433}
1434
1435static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1436{
1437 struct sock *sk = sock->sk;
1438 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1439 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1440 struct tipc_sock *tsk = tipc_sk(sk);
1441 struct tipc_msg *hdr = &tsk->phdr;
1442 struct net *net = sock_net(sk);
1443 struct sk_buff_head pkts;
1444 u32 dnode = tsk_peer_node(tsk);
1445 int send, sent = 0;
1446 int rc = 0;
1447
1448 __skb_queue_head_init(&pkts);
1449
1450 if (unlikely(dlen > INT_MAX))
1451 return -EMSGSIZE;
1452
1453 /* Handle implicit connection setup */
1454 if (unlikely(dest)) {
1455 rc = __tipc_sendmsg(sock, m, dlen);
1456 if (dlen && dlen == rc) {
1457 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1458 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1459 }
1460 return rc;
1461 }
1462
1463 do {
1464 rc = tipc_wait_for_cond(sock, &timeout,
1465 (!tsk->cong_link_cnt &&
1466 !tsk_conn_cong(tsk) &&
1467 tipc_sk_connected(sk)));
1468 if (unlikely(rc))
1469 break;
1470
1471 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1472 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1473 if (unlikely(rc != send))
1474 break;
1475
1476 trace_tipc_sk_sendstream(sk, skb_peek(&pkts),
1477 TIPC_DUMP_SK_SNDQ, " ");
1478 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1479 if (unlikely(rc == -ELINKCONG)) {
1480 tsk->cong_link_cnt = 1;
1481 rc = 0;
1482 }
1483 if (likely(!rc)) {
1484 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1485 sent += send;
1486 }
1487 } while (sent < dlen && !rc);
1488
1489 return sent ? sent : rc;
1490}
1491
1492/**
1493 * tipc_send_packet - send a connection-oriented message
1494 * @sock: socket structure
1495 * @m: message to send
1496 * @dsz: length of data to be transmitted
1497 *
1498 * Used for SOCK_SEQPACKET messages.
1499 *
1500 * Returns the number of bytes sent on success, or errno otherwise
1501 */
1502static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1503{
1504 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1505 return -EMSGSIZE;
1506
1507 return tipc_sendstream(sock, m, dsz);
1508}
1509
1510/* tipc_sk_finish_conn - complete the setup of a connection
1511 */
1512static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1513 u32 peer_node)
1514{
1515 struct sock *sk = &tsk->sk;
1516 struct net *net = sock_net(sk);
1517 struct tipc_msg *msg = &tsk->phdr;
1518
1519 msg_set_syn(msg, 0);
1520 msg_set_destnode(msg, peer_node);
1521 msg_set_destport(msg, peer_port);
1522 msg_set_type(msg, TIPC_CONN_MSG);
1523 msg_set_lookup_scope(msg, 0);
1524 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1525
1526 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1527 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1528 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1529 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1530 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1531 __skb_queue_purge(&sk->sk_write_queue);
1532 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1533 return;
1534
1535 /* Fall back to message based flow control */
1536 tsk->rcv_win = FLOWCTL_MSG_WIN;
1537 tsk->snd_win = FLOWCTL_MSG_WIN;
1538}
1539
1540/**
1541 * tipc_sk_set_orig_addr - capture sender's address for received message
1542 * @m: descriptor for message info
1543 * @hdr: received message header
1544 *
1545 * Note: Address is not captured if not requested by receiver.
1546 */
1547static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1548{
1549 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1550 struct tipc_msg *hdr = buf_msg(skb);
1551
1552 if (!srcaddr)
1553 return;
1554
1555 srcaddr->sock.family = AF_TIPC;
1556 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1557 srcaddr->sock.scope = 0;
1558 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1559 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1560 srcaddr->sock.addr.name.domain = 0;
1561 m->msg_namelen = sizeof(struct sockaddr_tipc);
1562
1563 if (!msg_in_group(hdr))
1564 return;
1565
1566 /* Group message users may also want to know sending member's id */
1567 srcaddr->member.family = AF_TIPC;
1568 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1569 srcaddr->member.scope = 0;
1570 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1571 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1572 srcaddr->member.addr.name.domain = 0;
1573 m->msg_namelen = sizeof(*srcaddr);
1574}
1575
1576/**
1577 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1578 * @m: descriptor for message info
1579 * @skb: received message buffer
1580 * @tsk: TIPC port associated with message
1581 *
1582 * Note: Ancillary data is not captured if not requested by receiver.
1583 *
1584 * Returns 0 if successful, otherwise errno
1585 */
1586static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1587 struct tipc_sock *tsk)
1588{
1589 struct tipc_msg *msg;
1590 u32 anc_data[3];
1591 u32 err;
1592 u32 dest_type;
1593 int has_name;
1594 int res;
1595
1596 if (likely(m->msg_controllen == 0))
1597 return 0;
1598 msg = buf_msg(skb);
1599
1600 /* Optionally capture errored message object(s) */
1601 err = msg ? msg_errcode(msg) : 0;
1602 if (unlikely(err)) {
1603 anc_data[0] = err;
1604 anc_data[1] = msg_data_sz(msg);
1605 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1606 if (res)
1607 return res;
1608 if (anc_data[1]) {
1609 if (skb_linearize(skb))
1610 return -ENOMEM;
1611 msg = buf_msg(skb);
1612 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1613 msg_data(msg));
1614 if (res)
1615 return res;
1616 }
1617 }
1618
1619 /* Optionally capture message destination object */
1620 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1621 switch (dest_type) {
1622 case TIPC_NAMED_MSG:
1623 has_name = 1;
1624 anc_data[0] = msg_nametype(msg);
1625 anc_data[1] = msg_namelower(msg);
1626 anc_data[2] = msg_namelower(msg);
1627 break;
1628 case TIPC_MCAST_MSG:
1629 has_name = 1;
1630 anc_data[0] = msg_nametype(msg);
1631 anc_data[1] = msg_namelower(msg);
1632 anc_data[2] = msg_nameupper(msg);
1633 break;
1634 case TIPC_CONN_MSG:
1635 has_name = (tsk->conn_type != 0);
1636 anc_data[0] = tsk->conn_type;
1637 anc_data[1] = tsk->conn_instance;
1638 anc_data[2] = tsk->conn_instance;
1639 break;
1640 default:
1641 has_name = 0;
1642 }
1643 if (has_name) {
1644 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1645 if (res)
1646 return res;
1647 }
1648
1649 return 0;
1650}
1651
1652static void tipc_sk_send_ack(struct tipc_sock *tsk)
1653{
1654 struct sock *sk = &tsk->sk;
1655 struct net *net = sock_net(sk);
1656 struct sk_buff *skb = NULL;
1657 struct tipc_msg *msg;
1658 u32 peer_port = tsk_peer_port(tsk);
1659 u32 dnode = tsk_peer_node(tsk);
1660
1661 if (!tipc_sk_connected(sk))
1662 return;
1663 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1664 dnode, tsk_own_node(tsk), peer_port,
1665 tsk->portid, TIPC_OK);
1666 if (!skb)
1667 return;
1668 msg = buf_msg(skb);
1669 msg_set_conn_ack(msg, tsk->rcv_unacked);
1670 tsk->rcv_unacked = 0;
1671
1672 /* Adjust to and advertize the correct window limit */
1673 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1674 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1675 msg_set_adv_win(msg, tsk->rcv_win);
1676 }
1677 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1678}
1679
1680static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1681{
1682 struct sock *sk = sock->sk;
1683 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1684 long timeo = *timeop;
1685 int err = sock_error(sk);
1686
1687 if (err)
1688 return err;
1689
1690 for (;;) {
1691 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1692 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1693 err = -ENOTCONN;
1694 break;
1695 }
1696 add_wait_queue(sk_sleep(sk), &wait);
1697 release_sock(sk);
1698 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1699 sched_annotate_sleep();
1700 lock_sock(sk);
1701 remove_wait_queue(sk_sleep(sk), &wait);
1702 }
1703 err = 0;
1704 if (!skb_queue_empty(&sk->sk_receive_queue))
1705 break;
1706 err = -EAGAIN;
1707 if (!timeo)
1708 break;
1709 err = sock_intr_errno(timeo);
1710 if (signal_pending(current))
1711 break;
1712
1713 err = sock_error(sk);
1714 if (err)
1715 break;
1716 }
1717 *timeop = timeo;
1718 return err;
1719}
1720
1721/**
1722 * tipc_recvmsg - receive packet-oriented message
1723 * @m: descriptor for message info
1724 * @buflen: length of user buffer area
1725 * @flags: receive flags
1726 *
1727 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1728 * If the complete message doesn't fit in user area, truncate it.
1729 *
1730 * Returns size of returned message data, errno otherwise
1731 */
1732static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1733 size_t buflen, int flags)
1734{
1735 struct sock *sk = sock->sk;
1736 bool connected = !tipc_sk_type_connectionless(sk);
1737 struct tipc_sock *tsk = tipc_sk(sk);
1738 int rc, err, hlen, dlen, copy;
1739 struct sk_buff_head xmitq;
1740 struct tipc_msg *hdr;
1741 struct sk_buff *skb;
1742 bool grp_evt;
1743 long timeout;
1744
1745 /* Catch invalid receive requests */
1746 if (unlikely(!buflen))
1747 return -EINVAL;
1748
1749 lock_sock(sk);
1750 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1751 rc = -ENOTCONN;
1752 goto exit;
1753 }
1754 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1755
1756 /* Step rcv queue to first msg with data or error; wait if necessary */
1757 do {
1758 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1759 if (unlikely(rc))
1760 goto exit;
1761 skb = skb_peek(&sk->sk_receive_queue);
1762 hdr = buf_msg(skb);
1763 dlen = msg_data_sz(hdr);
1764 hlen = msg_hdr_sz(hdr);
1765 err = msg_errcode(hdr);
1766 grp_evt = msg_is_grp_evt(hdr);
1767 if (likely(dlen || err))
1768 break;
1769 tsk_advance_rx_queue(sk);
1770 } while (1);
1771
1772 /* Collect msg meta data, including error code and rejected data */
1773 tipc_sk_set_orig_addr(m, skb);
1774 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1775 if (unlikely(rc))
1776 goto exit;
1777 hdr = buf_msg(skb);
1778
1779 /* Capture data if non-error msg, otherwise just set return value */
1780 if (likely(!err)) {
1781 copy = min_t(int, dlen, buflen);
1782 if (unlikely(copy != dlen))
1783 m->msg_flags |= MSG_TRUNC;
1784 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1785 } else {
1786 copy = 0;
1787 rc = 0;
1788 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1789 rc = -ECONNRESET;
1790 }
1791 if (unlikely(rc))
1792 goto exit;
1793
1794 /* Mark message as group event if applicable */
1795 if (unlikely(grp_evt)) {
1796 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1797 m->msg_flags |= MSG_EOR;
1798 m->msg_flags |= MSG_OOB;
1799 copy = 0;
1800 }
1801
1802 /* Caption of data or error code/rejected data was successful */
1803 if (unlikely(flags & MSG_PEEK))
1804 goto exit;
1805
1806 /* Send group flow control advertisement when applicable */
1807 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1808 __skb_queue_head_init(&xmitq);
1809 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1810 msg_orignode(hdr), msg_origport(hdr),
1811 &xmitq);
1812 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1813 }
1814
1815 tsk_advance_rx_queue(sk);
1816
1817 if (likely(!connected))
1818 goto exit;
1819
1820 /* Send connection flow control advertisement when applicable */
1821 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1822 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1823 tipc_sk_send_ack(tsk);
1824exit:
1825 release_sock(sk);
1826 return rc ? rc : copy;
1827}
1828
1829/**
1830 * tipc_recvstream - receive stream-oriented data
1831 * @m: descriptor for message info
1832 * @buflen: total size of user buffer area
1833 * @flags: receive flags
1834 *
1835 * Used for SOCK_STREAM messages only. If not enough data is available
1836 * will optionally wait for more; never truncates data.
1837 *
1838 * Returns size of returned message data, errno otherwise
1839 */
1840static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1841 size_t buflen, int flags)
1842{
1843 struct sock *sk = sock->sk;
1844 struct tipc_sock *tsk = tipc_sk(sk);
1845 struct sk_buff *skb;
1846 struct tipc_msg *hdr;
1847 struct tipc_skb_cb *skb_cb;
1848 bool peek = flags & MSG_PEEK;
1849 int offset, required, copy, copied = 0;
1850 int hlen, dlen, err, rc;
1851 long timeout;
1852
1853 /* Catch invalid receive attempts */
1854 if (unlikely(!buflen))
1855 return -EINVAL;
1856
1857 lock_sock(sk);
1858
1859 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1860 rc = -ENOTCONN;
1861 goto exit;
1862 }
1863 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1864 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1865
1866 do {
1867 /* Look at first msg in receive queue; wait if necessary */
1868 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1869 if (unlikely(rc))
1870 break;
1871 skb = skb_peek(&sk->sk_receive_queue);
1872 skb_cb = TIPC_SKB_CB(skb);
1873 hdr = buf_msg(skb);
1874 dlen = msg_data_sz(hdr);
1875 hlen = msg_hdr_sz(hdr);
1876 err = msg_errcode(hdr);
1877
1878 /* Discard any empty non-errored (SYN-) message */
1879 if (unlikely(!dlen && !err)) {
1880 tsk_advance_rx_queue(sk);
1881 continue;
1882 }
1883
1884 /* Collect msg meta data, incl. error code and rejected data */
1885 if (!copied) {
1886 tipc_sk_set_orig_addr(m, skb);
1887 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1888 if (rc)
1889 break;
1890 hdr = buf_msg(skb);
1891 }
1892
1893 /* Copy data if msg ok, otherwise return error/partial data */
1894 if (likely(!err)) {
1895 offset = skb_cb->bytes_read;
1896 copy = min_t(int, dlen - offset, buflen - copied);
1897 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1898 if (unlikely(rc))
1899 break;
1900 copied += copy;
1901 offset += copy;
1902 if (unlikely(offset < dlen)) {
1903 if (!peek)
1904 skb_cb->bytes_read = offset;
1905 break;
1906 }
1907 } else {
1908 rc = 0;
1909 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1910 rc = -ECONNRESET;
1911 if (copied || rc)
1912 break;
1913 }
1914
1915 if (unlikely(peek))
1916 break;
1917
1918 tsk_advance_rx_queue(sk);
1919
1920 /* Send connection flow control advertisement when applicable */
1921 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1922 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1923 tipc_sk_send_ack(tsk);
1924
1925 /* Exit if all requested data or FIN/error received */
1926 if (copied == buflen || err)
1927 break;
1928
1929 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1930exit:
1931 release_sock(sk);
1932 return copied ? copied : rc;
1933}
1934
1935/**
1936 * tipc_write_space - wake up thread if port congestion is released
1937 * @sk: socket
1938 */
1939static void tipc_write_space(struct sock *sk)
1940{
1941 struct socket_wq *wq;
1942
1943 rcu_read_lock();
1944 wq = rcu_dereference(sk->sk_wq);
1945 if (skwq_has_sleeper(wq))
1946 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1947 EPOLLWRNORM | EPOLLWRBAND);
1948 rcu_read_unlock();
1949}
1950
1951/**
1952 * tipc_data_ready - wake up threads to indicate messages have been received
1953 * @sk: socket
1954 * @len: the length of messages
1955 */
1956static void tipc_data_ready(struct sock *sk)
1957{
1958 struct socket_wq *wq;
1959
1960 rcu_read_lock();
1961 wq = rcu_dereference(sk->sk_wq);
1962 if (skwq_has_sleeper(wq))
1963 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1964 EPOLLRDNORM | EPOLLRDBAND);
1965 rcu_read_unlock();
1966}
1967
1968static void tipc_sock_destruct(struct sock *sk)
1969{
1970 __skb_queue_purge(&sk->sk_receive_queue);
1971}
1972
1973static void tipc_sk_proto_rcv(struct sock *sk,
1974 struct sk_buff_head *inputq,
1975 struct sk_buff_head *xmitq)
1976{
1977 struct sk_buff *skb = __skb_dequeue(inputq);
1978 struct tipc_sock *tsk = tipc_sk(sk);
1979 struct tipc_msg *hdr = buf_msg(skb);
1980 struct tipc_group *grp = tsk->group;
1981 bool wakeup = false;
1982
1983 switch (msg_user(hdr)) {
1984 case CONN_MANAGER:
1985 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1986 return;
1987 case SOCK_WAKEUP:
1988 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1989 /* coupled with smp_rmb() in tipc_wait_for_cond() */
1990 smp_wmb();
1991 tsk->cong_link_cnt--;
1992 wakeup = true;
1993 break;
1994 case GROUP_PROTOCOL:
1995 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1996 break;
1997 case TOP_SRV:
1998 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1999 hdr, inputq, xmitq);
2000 break;
2001 default:
2002 break;
2003 }
2004
2005 if (wakeup)
2006 sk->sk_write_space(sk);
2007
2008 kfree_skb(skb);
2009}
2010
2011/**
2012 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2013 * @tsk: TIPC socket
2014 * @skb: pointer to message buffer.
2015 * Returns true if message should be added to receive queue, false otherwise
2016 */
2017static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2018{
2019 struct sock *sk = &tsk->sk;
2020 struct net *net = sock_net(sk);
2021 struct tipc_msg *hdr = buf_msg(skb);
2022 bool con_msg = msg_connected(hdr);
2023 u32 pport = tsk_peer_port(tsk);
2024 u32 pnode = tsk_peer_node(tsk);
2025 u32 oport = msg_origport(hdr);
2026 u32 onode = msg_orignode(hdr);
2027 int err = msg_errcode(hdr);
2028 unsigned long delay;
2029
2030 if (unlikely(msg_mcast(hdr)))
2031 return false;
2032
2033 switch (sk->sk_state) {
2034 case TIPC_CONNECTING:
2035 /* Setup ACK */
2036 if (likely(con_msg)) {
2037 if (err)
2038 break;
2039 tipc_sk_finish_conn(tsk, oport, onode);
2040 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2041 /* ACK+ message with data is added to receive queue */
2042 if (msg_data_sz(hdr))
2043 return true;
2044 /* Empty ACK-, - wake up sleeping connect() and drop */
2045 sk->sk_state_change(sk);
2046 msg_set_dest_droppable(hdr, 1);
2047 return false;
2048 }
2049 /* Ignore connectionless message if not from listening socket */
2050 if (oport != pport || onode != pnode)
2051 return false;
2052
2053 /* Rejected SYN */
2054 if (err != TIPC_ERR_OVERLOAD)
2055 break;
2056
2057 /* Prepare for new setup attempt if we have a SYN clone */
2058 if (skb_queue_empty(&sk->sk_write_queue))
2059 break;
2060 get_random_bytes(&delay, 2);
2061 delay %= (tsk->conn_timeout / 4);
2062 delay = msecs_to_jiffies(delay + 100);
2063 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2064 return false;
2065 case TIPC_OPEN:
2066 case TIPC_DISCONNECTING:
2067 return false;
2068 case TIPC_LISTEN:
2069 /* Accept only SYN message */
2070 if (!msg_is_syn(hdr) &&
2071 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2072 return false;
2073 if (!con_msg && !err)
2074 return true;
2075 return false;
2076 case TIPC_ESTABLISHED:
2077 /* Accept only connection-based messages sent by peer */
2078 if (likely(con_msg && !err && pport == oport && pnode == onode))
2079 return true;
2080 if (!tsk_peer_msg(tsk, hdr))
2081 return false;
2082 if (!err)
2083 return true;
2084 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2085 tipc_node_remove_conn(net, pnode, tsk->portid);
2086 sk->sk_state_change(sk);
2087 return true;
2088 default:
2089 pr_err("Unknown sk_state %u\n", sk->sk_state);
2090 }
2091 /* Abort connection setup attempt */
2092 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2093 sk->sk_err = ECONNREFUSED;
2094 sk->sk_state_change(sk);
2095 return true;
2096}
2097
2098/**
2099 * rcvbuf_limit - get proper overload limit of socket receive queue
2100 * @sk: socket
2101 * @skb: message
2102 *
2103 * For connection oriented messages, irrespective of importance,
2104 * default queue limit is 2 MB.
2105 *
2106 * For connectionless messages, queue limits are based on message
2107 * importance as follows:
2108 *
2109 * TIPC_LOW_IMPORTANCE (2 MB)
2110 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2111 * TIPC_HIGH_IMPORTANCE (8 MB)
2112 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2113 *
2114 * Returns overload limit according to corresponding message importance
2115 */
2116static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2117{
2118 struct tipc_sock *tsk = tipc_sk(sk);
2119 struct tipc_msg *hdr = buf_msg(skb);
2120
2121 if (unlikely(msg_in_group(hdr)))
2122 return READ_ONCE(sk->sk_rcvbuf);
2123
2124 if (unlikely(!msg_connected(hdr)))
2125 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2126
2127 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2128 return READ_ONCE(sk->sk_rcvbuf);
2129
2130 return FLOWCTL_MSG_LIM;
2131}
2132
2133/**
2134 * tipc_sk_filter_rcv - validate incoming message
2135 * @sk: socket
2136 * @skb: pointer to message.
2137 *
2138 * Enqueues message on receive queue if acceptable; optionally handles
2139 * disconnect indication for a connected socket.
2140 *
2141 * Called with socket lock already taken
2142 *
2143 */
2144static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2145 struct sk_buff_head *xmitq)
2146{
2147 bool sk_conn = !tipc_sk_type_connectionless(sk);
2148 struct tipc_sock *tsk = tipc_sk(sk);
2149 struct tipc_group *grp = tsk->group;
2150 struct tipc_msg *hdr = buf_msg(skb);
2151 struct net *net = sock_net(sk);
2152 struct sk_buff_head inputq;
2153 int mtyp = msg_type(hdr);
2154 int limit, err = TIPC_OK;
2155
2156 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2157 TIPC_SKB_CB(skb)->bytes_read = 0;
2158 __skb_queue_head_init(&inputq);
2159 __skb_queue_tail(&inputq, skb);
2160
2161 if (unlikely(!msg_isdata(hdr)))
2162 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2163
2164 if (unlikely(grp))
2165 tipc_group_filter_msg(grp, &inputq, xmitq);
2166
2167 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2168 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2169
2170 /* Validate and add to receive buffer if there is space */
2171 while ((skb = __skb_dequeue(&inputq))) {
2172 hdr = buf_msg(skb);
2173 limit = rcvbuf_limit(sk, skb);
2174 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2175 (!sk_conn && msg_connected(hdr)) ||
2176 (!grp && msg_in_group(hdr)))
2177 err = TIPC_ERR_NO_PORT;
2178 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2179 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2180 "err_overload2!");
2181 atomic_inc(&sk->sk_drops);
2182 err = TIPC_ERR_OVERLOAD;
2183 }
2184
2185 if (unlikely(err)) {
2186 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2187 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2188 "@filter_rcv!");
2189 __skb_queue_tail(xmitq, skb);
2190 }
2191 err = TIPC_OK;
2192 continue;
2193 }
2194 __skb_queue_tail(&sk->sk_receive_queue, skb);
2195 skb_set_owner_r(skb, sk);
2196 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2197 "rcvq >90% allocated!");
2198 sk->sk_data_ready(sk);
2199 }
2200}
2201
2202/**
2203 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2204 * @sk: socket
2205 * @skb: message
2206 *
2207 * Caller must hold socket lock
2208 */
2209static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2210{
2211 unsigned int before = sk_rmem_alloc_get(sk);
2212 struct sk_buff_head xmitq;
2213 unsigned int added;
2214
2215 __skb_queue_head_init(&xmitq);
2216
2217 tipc_sk_filter_rcv(sk, skb, &xmitq);
2218 added = sk_rmem_alloc_get(sk) - before;
2219 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2220
2221 /* Send pending response/rejected messages, if any */
2222 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2223 return 0;
2224}
2225
2226/**
2227 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2228 * inputq and try adding them to socket or backlog queue
2229 * @inputq: list of incoming buffers with potentially different destinations
2230 * @sk: socket where the buffers should be enqueued
2231 * @dport: port number for the socket
2232 *
2233 * Caller must hold socket lock
2234 */
2235static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2236 u32 dport, struct sk_buff_head *xmitq)
2237{
2238 unsigned long time_limit = jiffies + 2;
2239 struct sk_buff *skb;
2240 unsigned int lim;
2241 atomic_t *dcnt;
2242 u32 onode;
2243
2244 while (skb_queue_len(inputq)) {
2245 if (unlikely(time_after_eq(jiffies, time_limit)))
2246 return;
2247
2248 skb = tipc_skb_dequeue(inputq, dport);
2249 if (unlikely(!skb))
2250 return;
2251
2252 /* Add message directly to receive queue if possible */
2253 if (!sock_owned_by_user(sk)) {
2254 tipc_sk_filter_rcv(sk, skb, xmitq);
2255 continue;
2256 }
2257
2258 /* Try backlog, compensating for double-counted bytes */
2259 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2260 if (!sk->sk_backlog.len)
2261 atomic_set(dcnt, 0);
2262 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2263 if (likely(!sk_add_backlog(sk, skb, lim))) {
2264 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2265 "bklg & rcvq >90% allocated!");
2266 continue;
2267 }
2268
2269 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2270 /* Overload => reject message back to sender */
2271 onode = tipc_own_addr(sock_net(sk));
2272 atomic_inc(&sk->sk_drops);
2273 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2274 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2275 "@sk_enqueue!");
2276 __skb_queue_tail(xmitq, skb);
2277 }
2278 break;
2279 }
2280}
2281
2282/**
2283 * tipc_sk_rcv - handle a chain of incoming buffers
2284 * @inputq: buffer list containing the buffers
2285 * Consumes all buffers in list until inputq is empty
2286 * Note: may be called in multiple threads referring to the same queue
2287 */
2288void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2289{
2290 struct sk_buff_head xmitq;
2291 u32 dnode, dport = 0;
2292 int err;
2293 struct tipc_sock *tsk;
2294 struct sock *sk;
2295 struct sk_buff *skb;
2296
2297 __skb_queue_head_init(&xmitq);
2298 while (skb_queue_len(inputq)) {
2299 dport = tipc_skb_peek_port(inputq, dport);
2300 tsk = tipc_sk_lookup(net, dport);
2301
2302 if (likely(tsk)) {
2303 sk = &tsk->sk;
2304 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2305 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2306 spin_unlock_bh(&sk->sk_lock.slock);
2307 }
2308 /* Send pending response/rejected messages, if any */
2309 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2310 sock_put(sk);
2311 continue;
2312 }
2313 /* No destination socket => dequeue skb if still there */
2314 skb = tipc_skb_dequeue(inputq, dport);
2315 if (!skb)
2316 return;
2317
2318 /* Try secondary lookup if unresolved named message */
2319 err = TIPC_ERR_NO_PORT;
2320 if (tipc_msg_lookup_dest(net, skb, &err))
2321 goto xmit;
2322
2323 /* Prepare for message rejection */
2324 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2325 continue;
2326
2327 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2328xmit:
2329 dnode = msg_destnode(buf_msg(skb));
2330 tipc_node_xmit_skb(net, skb, dnode, dport);
2331 }
2332}
2333
2334static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2335{
2336 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2337 struct sock *sk = sock->sk;
2338 int done;
2339
2340 do {
2341 int err = sock_error(sk);
2342 if (err)
2343 return err;
2344 if (!*timeo_p)
2345 return -ETIMEDOUT;
2346 if (signal_pending(current))
2347 return sock_intr_errno(*timeo_p);
2348
2349 add_wait_queue(sk_sleep(sk), &wait);
2350 done = sk_wait_event(sk, timeo_p,
2351 sk->sk_state != TIPC_CONNECTING, &wait);
2352 remove_wait_queue(sk_sleep(sk), &wait);
2353 } while (!done);
2354 return 0;
2355}
2356
2357static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2358{
2359 if (addr->family != AF_TIPC)
2360 return false;
2361 if (addr->addrtype == TIPC_SERVICE_RANGE)
2362 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2363 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2364 addr->addrtype == TIPC_SOCKET_ADDR);
2365}
2366
2367/**
2368 * tipc_connect - establish a connection to another TIPC port
2369 * @sock: socket structure
2370 * @dest: socket address for destination port
2371 * @destlen: size of socket address data structure
2372 * @flags: file-related flags associated with socket
2373 *
2374 * Returns 0 on success, errno otherwise
2375 */
2376static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2377 int destlen, int flags)
2378{
2379 struct sock *sk = sock->sk;
2380 struct tipc_sock *tsk = tipc_sk(sk);
2381 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2382 struct msghdr m = {NULL,};
2383 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2384 int previous;
2385 int res = 0;
2386
2387 if (destlen != sizeof(struct sockaddr_tipc))
2388 return -EINVAL;
2389
2390 lock_sock(sk);
2391
2392 if (tsk->group) {
2393 res = -EINVAL;
2394 goto exit;
2395 }
2396
2397 if (dst->family == AF_UNSPEC) {
2398 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2399 if (!tipc_sk_type_connectionless(sk))
2400 res = -EINVAL;
2401 goto exit;
2402 }
2403 if (!tipc_sockaddr_is_sane(dst)) {
2404 res = -EINVAL;
2405 goto exit;
2406 }
2407 /* DGRAM/RDM connect(), just save the destaddr */
2408 if (tipc_sk_type_connectionless(sk)) {
2409 memcpy(&tsk->peer, dest, destlen);
2410 goto exit;
2411 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2412 res = -EINVAL;
2413 goto exit;
2414 }
2415
2416 previous = sk->sk_state;
2417
2418 switch (sk->sk_state) {
2419 case TIPC_OPEN:
2420 /* Send a 'SYN-' to destination */
2421 m.msg_name = dest;
2422 m.msg_namelen = destlen;
2423
2424 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2425 * indicate send_msg() is never blocked.
2426 */
2427 if (!timeout)
2428 m.msg_flags = MSG_DONTWAIT;
2429
2430 res = __tipc_sendmsg(sock, &m, 0);
2431 if ((res < 0) && (res != -EWOULDBLOCK))
2432 goto exit;
2433
2434 /* Just entered TIPC_CONNECTING state; the only
2435 * difference is that return value in non-blocking
2436 * case is EINPROGRESS, rather than EALREADY.
2437 */
2438 res = -EINPROGRESS;
2439 /* fall through */
2440 case TIPC_CONNECTING:
2441 if (!timeout) {
2442 if (previous == TIPC_CONNECTING)
2443 res = -EALREADY;
2444 goto exit;
2445 }
2446 timeout = msecs_to_jiffies(timeout);
2447 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2448 res = tipc_wait_for_connect(sock, &timeout);
2449 break;
2450 case TIPC_ESTABLISHED:
2451 res = -EISCONN;
2452 break;
2453 default:
2454 res = -EINVAL;
2455 }
2456
2457exit:
2458 release_sock(sk);
2459 return res;
2460}
2461
2462/**
2463 * tipc_listen - allow socket to listen for incoming connections
2464 * @sock: socket structure
2465 * @len: (unused)
2466 *
2467 * Returns 0 on success, errno otherwise
2468 */
2469static int tipc_listen(struct socket *sock, int len)
2470{
2471 struct sock *sk = sock->sk;
2472 int res;
2473
2474 lock_sock(sk);
2475 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2476 release_sock(sk);
2477
2478 return res;
2479}
2480
2481static int tipc_wait_for_accept(struct socket *sock, long timeo)
2482{
2483 struct sock *sk = sock->sk;
2484 DEFINE_WAIT(wait);
2485 int err;
2486
2487 /* True wake-one mechanism for incoming connections: only
2488 * one process gets woken up, not the 'whole herd'.
2489 * Since we do not 'race & poll' for established sockets
2490 * anymore, the common case will execute the loop only once.
2491 */
2492 for (;;) {
2493 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2494 TASK_INTERRUPTIBLE);
2495 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2496 release_sock(sk);
2497 timeo = schedule_timeout(timeo);
2498 lock_sock(sk);
2499 }
2500 err = 0;
2501 if (!skb_queue_empty(&sk->sk_receive_queue))
2502 break;
2503 err = -EAGAIN;
2504 if (!timeo)
2505 break;
2506 err = sock_intr_errno(timeo);
2507 if (signal_pending(current))
2508 break;
2509 }
2510 finish_wait(sk_sleep(sk), &wait);
2511 return err;
2512}
2513
2514/**
2515 * tipc_accept - wait for connection request
2516 * @sock: listening socket
2517 * @newsock: new socket that is to be connected
2518 * @flags: file-related flags associated with socket
2519 *
2520 * Returns 0 on success, errno otherwise
2521 */
2522static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2523 bool kern)
2524{
2525 struct sock *new_sk, *sk = sock->sk;
2526 struct sk_buff *buf;
2527 struct tipc_sock *new_tsock;
2528 struct tipc_msg *msg;
2529 long timeo;
2530 int res;
2531
2532 lock_sock(sk);
2533
2534 if (sk->sk_state != TIPC_LISTEN) {
2535 res = -EINVAL;
2536 goto exit;
2537 }
2538 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2539 res = tipc_wait_for_accept(sock, timeo);
2540 if (res)
2541 goto exit;
2542
2543 buf = skb_peek(&sk->sk_receive_queue);
2544
2545 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2546 if (res)
2547 goto exit;
2548 security_sk_clone(sock->sk, new_sock->sk);
2549
2550 new_sk = new_sock->sk;
2551 new_tsock = tipc_sk(new_sk);
2552 msg = buf_msg(buf);
2553
2554 /* we lock on new_sk; but lockdep sees the lock on sk */
2555 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2556
2557 /*
2558 * Reject any stray messages received by new socket
2559 * before the socket lock was taken (very, very unlikely)
2560 */
2561 tsk_rej_rx_queue(new_sk);
2562
2563 /* Connect new socket to it's peer */
2564 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2565
2566 tsk_set_importance(new_tsock, msg_importance(msg));
2567 if (msg_named(msg)) {
2568 new_tsock->conn_type = msg_nametype(msg);
2569 new_tsock->conn_instance = msg_nameinst(msg);
2570 }
2571
2572 /*
2573 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2574 * Respond to 'SYN+' by queuing it on new socket.
2575 */
2576 if (!msg_data_sz(msg)) {
2577 struct msghdr m = {NULL,};
2578
2579 tsk_advance_rx_queue(sk);
2580 __tipc_sendstream(new_sock, &m, 0);
2581 } else {
2582 __skb_dequeue(&sk->sk_receive_queue);
2583 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2584 skb_set_owner_r(buf, new_sk);
2585 }
2586 release_sock(new_sk);
2587exit:
2588 release_sock(sk);
2589 return res;
2590}
2591
2592/**
2593 * tipc_shutdown - shutdown socket connection
2594 * @sock: socket structure
2595 * @how: direction to close (must be SHUT_RDWR)
2596 *
2597 * Terminates connection (if necessary), then purges socket's receive queue.
2598 *
2599 * Returns 0 on success, errno otherwise
2600 */
2601static int tipc_shutdown(struct socket *sock, int how)
2602{
2603 struct sock *sk = sock->sk;
2604 int res;
2605
2606 if (how != SHUT_RDWR)
2607 return -EINVAL;
2608
2609 lock_sock(sk);
2610
2611 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2612 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2613 sk->sk_shutdown = SEND_SHUTDOWN;
2614
2615 if (sk->sk_state == TIPC_DISCONNECTING) {
2616 /* Discard any unreceived messages */
2617 __skb_queue_purge(&sk->sk_receive_queue);
2618
2619 /* Wake up anyone sleeping in poll */
2620 sk->sk_state_change(sk);
2621 res = 0;
2622 } else {
2623 res = -ENOTCONN;
2624 }
2625
2626 release_sock(sk);
2627 return res;
2628}
2629
2630static void tipc_sk_check_probing_state(struct sock *sk,
2631 struct sk_buff_head *list)
2632{
2633 struct tipc_sock *tsk = tipc_sk(sk);
2634 u32 pnode = tsk_peer_node(tsk);
2635 u32 pport = tsk_peer_port(tsk);
2636 u32 self = tsk_own_node(tsk);
2637 u32 oport = tsk->portid;
2638 struct sk_buff *skb;
2639
2640 if (tsk->probe_unacked) {
2641 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2642 sk->sk_err = ECONNABORTED;
2643 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2644 sk->sk_state_change(sk);
2645 return;
2646 }
2647 /* Prepare new probe */
2648 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2649 pnode, self, pport, oport, TIPC_OK);
2650 if (skb)
2651 __skb_queue_tail(list, skb);
2652 tsk->probe_unacked = true;
2653 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2654}
2655
2656static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2657{
2658 struct tipc_sock *tsk = tipc_sk(sk);
2659
2660 /* Try again later if dest link is congested */
2661 if (tsk->cong_link_cnt) {
2662 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2663 return;
2664 }
2665 /* Prepare SYN for retransmit */
2666 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2667}
2668
2669static void tipc_sk_timeout(struct timer_list *t)
2670{
2671 struct sock *sk = from_timer(sk, t, sk_timer);
2672 struct tipc_sock *tsk = tipc_sk(sk);
2673 u32 pnode = tsk_peer_node(tsk);
2674 struct sk_buff_head list;
2675 int rc = 0;
2676
2677 __skb_queue_head_init(&list);
2678 bh_lock_sock(sk);
2679
2680 /* Try again later if socket is busy */
2681 if (sock_owned_by_user(sk)) {
2682 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2683 bh_unlock_sock(sk);
2684 return;
2685 }
2686
2687 if (sk->sk_state == TIPC_ESTABLISHED)
2688 tipc_sk_check_probing_state(sk, &list);
2689 else if (sk->sk_state == TIPC_CONNECTING)
2690 tipc_sk_retry_connect(sk, &list);
2691
2692 bh_unlock_sock(sk);
2693
2694 if (!skb_queue_empty(&list))
2695 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2696
2697 /* SYN messages may cause link congestion */
2698 if (rc == -ELINKCONG) {
2699 tipc_dest_push(&tsk->cong_links, pnode, 0);
2700 tsk->cong_link_cnt = 1;
2701 }
2702 sock_put(sk);
2703}
2704
2705static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2706 struct tipc_name_seq const *seq)
2707{
2708 struct sock *sk = &tsk->sk;
2709 struct net *net = sock_net(sk);
2710 struct publication *publ;
2711 u32 key;
2712
2713 if (scope != TIPC_NODE_SCOPE)
2714 scope = TIPC_CLUSTER_SCOPE;
2715
2716 if (tipc_sk_connected(sk))
2717 return -EINVAL;
2718 key = tsk->portid + tsk->pub_count + 1;
2719 if (key == tsk->portid)
2720 return -EADDRINUSE;
2721
2722 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2723 scope, tsk->portid, key);
2724 if (unlikely(!publ))
2725 return -EINVAL;
2726
2727 list_add(&publ->binding_sock, &tsk->publications);
2728 tsk->pub_count++;
2729 tsk->published = 1;
2730 return 0;
2731}
2732
2733static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2734 struct tipc_name_seq const *seq)
2735{
2736 struct net *net = sock_net(&tsk->sk);
2737 struct publication *publ;
2738 struct publication *safe;
2739 int rc = -EINVAL;
2740
2741 if (scope != TIPC_NODE_SCOPE)
2742 scope = TIPC_CLUSTER_SCOPE;
2743
2744 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2745 if (seq) {
2746 if (publ->scope != scope)
2747 continue;
2748 if (publ->type != seq->type)
2749 continue;
2750 if (publ->lower != seq->lower)
2751 continue;
2752 if (publ->upper != seq->upper)
2753 break;
2754 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2755 publ->upper, publ->key);
2756 rc = 0;
2757 break;
2758 }
2759 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2760 publ->upper, publ->key);
2761 rc = 0;
2762 }
2763 if (list_empty(&tsk->publications))
2764 tsk->published = 0;
2765 return rc;
2766}
2767
2768/* tipc_sk_reinit: set non-zero address in all existing sockets
2769 * when we go from standalone to network mode.
2770 */
2771void tipc_sk_reinit(struct net *net)
2772{
2773 struct tipc_net *tn = net_generic(net, tipc_net_id);
2774 struct rhashtable_iter iter;
2775 struct tipc_sock *tsk;
2776 struct tipc_msg *msg;
2777
2778 rhashtable_walk_enter(&tn->sk_rht, &iter);
2779
2780 do {
2781 rhashtable_walk_start(&iter);
2782
2783 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2784 sock_hold(&tsk->sk);
2785 rhashtable_walk_stop(&iter);
2786 lock_sock(&tsk->sk);
2787 msg = &tsk->phdr;
2788 msg_set_prevnode(msg, tipc_own_addr(net));
2789 msg_set_orignode(msg, tipc_own_addr(net));
2790 release_sock(&tsk->sk);
2791 rhashtable_walk_start(&iter);
2792 sock_put(&tsk->sk);
2793 }
2794
2795 rhashtable_walk_stop(&iter);
2796 } while (tsk == ERR_PTR(-EAGAIN));
2797
2798 rhashtable_walk_exit(&iter);
2799}
2800
2801static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2802{
2803 struct tipc_net *tn = net_generic(net, tipc_net_id);
2804 struct tipc_sock *tsk;
2805
2806 rcu_read_lock();
2807 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2808 if (tsk)
2809 sock_hold(&tsk->sk);
2810 rcu_read_unlock();
2811
2812 return tsk;
2813}
2814
2815static int tipc_sk_insert(struct tipc_sock *tsk)
2816{
2817 struct sock *sk = &tsk->sk;
2818 struct net *net = sock_net(sk);
2819 struct tipc_net *tn = net_generic(net, tipc_net_id);
2820 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2821 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2822
2823 while (remaining--) {
2824 portid++;
2825 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2826 portid = TIPC_MIN_PORT;
2827 tsk->portid = portid;
2828 sock_hold(&tsk->sk);
2829 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2830 tsk_rht_params))
2831 return 0;
2832 sock_put(&tsk->sk);
2833 }
2834
2835 return -1;
2836}
2837
2838static void tipc_sk_remove(struct tipc_sock *tsk)
2839{
2840 struct sock *sk = &tsk->sk;
2841 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2842
2843 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2844 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2845 __sock_put(sk);
2846 }
2847}
2848
2849static const struct rhashtable_params tsk_rht_params = {
2850 .nelem_hint = 192,
2851 .head_offset = offsetof(struct tipc_sock, node),
2852 .key_offset = offsetof(struct tipc_sock, portid),
2853 .key_len = sizeof(u32), /* portid */
2854 .max_size = 1048576,
2855 .min_size = 256,
2856 .automatic_shrinking = true,
2857};
2858
2859int tipc_sk_rht_init(struct net *net)
2860{
2861 struct tipc_net *tn = net_generic(net, tipc_net_id);
2862
2863 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2864}
2865
2866void tipc_sk_rht_destroy(struct net *net)
2867{
2868 struct tipc_net *tn = net_generic(net, tipc_net_id);
2869
2870 /* Wait for socket readers to complete */
2871 synchronize_net();
2872
2873 rhashtable_destroy(&tn->sk_rht);
2874}
2875
2876static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2877{
2878 struct net *net = sock_net(&tsk->sk);
2879 struct tipc_group *grp = tsk->group;
2880 struct tipc_msg *hdr = &tsk->phdr;
2881 struct tipc_name_seq seq;
2882 int rc;
2883
2884 if (mreq->type < TIPC_RESERVED_TYPES)
2885 return -EACCES;
2886 if (mreq->scope > TIPC_NODE_SCOPE)
2887 return -EINVAL;
2888 if (grp)
2889 return -EACCES;
2890 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2891 if (!grp)
2892 return -ENOMEM;
2893 tsk->group = grp;
2894 msg_set_lookup_scope(hdr, mreq->scope);
2895 msg_set_nametype(hdr, mreq->type);
2896 msg_set_dest_droppable(hdr, true);
2897 seq.type = mreq->type;
2898 seq.lower = mreq->instance;
2899 seq.upper = seq.lower;
2900 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2901 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2902 if (rc) {
2903 tipc_group_delete(net, grp);
2904 tsk->group = NULL;
2905 return rc;
2906 }
2907 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2908 tsk->mc_method.rcast = true;
2909 tsk->mc_method.mandatory = true;
2910 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2911 return rc;
2912}
2913
2914static int tipc_sk_leave(struct tipc_sock *tsk)
2915{
2916 struct net *net = sock_net(&tsk->sk);
2917 struct tipc_group *grp = tsk->group;
2918 struct tipc_name_seq seq;
2919 int scope;
2920
2921 if (!grp)
2922 return -EINVAL;
2923 tipc_group_self(grp, &seq, &scope);
2924 tipc_group_delete(net, grp);
2925 tsk->group = NULL;
2926 tipc_sk_withdraw(tsk, scope, &seq);
2927 return 0;
2928}
2929
2930/**
2931 * tipc_setsockopt - set socket option
2932 * @sock: socket structure
2933 * @lvl: option level
2934 * @opt: option identifier
2935 * @ov: pointer to new option value
2936 * @ol: length of option value
2937 *
2938 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2939 * (to ease compatibility).
2940 *
2941 * Returns 0 on success, errno otherwise
2942 */
2943static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2944 char __user *ov, unsigned int ol)
2945{
2946 struct sock *sk = sock->sk;
2947 struct tipc_sock *tsk = tipc_sk(sk);
2948 struct tipc_group_req mreq;
2949 u32 value = 0;
2950 int res = 0;
2951
2952 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2953 return 0;
2954 if (lvl != SOL_TIPC)
2955 return -ENOPROTOOPT;
2956
2957 switch (opt) {
2958 case TIPC_IMPORTANCE:
2959 case TIPC_SRC_DROPPABLE:
2960 case TIPC_DEST_DROPPABLE:
2961 case TIPC_CONN_TIMEOUT:
2962 if (ol < sizeof(value))
2963 return -EINVAL;
2964 if (get_user(value, (u32 __user *)ov))
2965 return -EFAULT;
2966 break;
2967 case TIPC_GROUP_JOIN:
2968 if (ol < sizeof(mreq))
2969 return -EINVAL;
2970 if (copy_from_user(&mreq, ov, sizeof(mreq)))
2971 return -EFAULT;
2972 break;
2973 default:
2974 if (ov || ol)
2975 return -EINVAL;
2976 }
2977
2978 lock_sock(sk);
2979
2980 switch (opt) {
2981 case TIPC_IMPORTANCE:
2982 res = tsk_set_importance(tsk, value);
2983 break;
2984 case TIPC_SRC_DROPPABLE:
2985 if (sock->type != SOCK_STREAM)
2986 tsk_set_unreliable(tsk, value);
2987 else
2988 res = -ENOPROTOOPT;
2989 break;
2990 case TIPC_DEST_DROPPABLE:
2991 tsk_set_unreturnable(tsk, value);
2992 break;
2993 case TIPC_CONN_TIMEOUT:
2994 tipc_sk(sk)->conn_timeout = value;
2995 break;
2996 case TIPC_MCAST_BROADCAST:
2997 tsk->mc_method.rcast = false;
2998 tsk->mc_method.mandatory = true;
2999 break;
3000 case TIPC_MCAST_REPLICAST:
3001 tsk->mc_method.rcast = true;
3002 tsk->mc_method.mandatory = true;
3003 break;
3004 case TIPC_GROUP_JOIN:
3005 res = tipc_sk_join(tsk, &mreq);
3006 break;
3007 case TIPC_GROUP_LEAVE:
3008 res = tipc_sk_leave(tsk);
3009 break;
3010 default:
3011 res = -EINVAL;
3012 }
3013
3014 release_sock(sk);
3015
3016 return res;
3017}
3018
3019/**
3020 * tipc_getsockopt - get socket option
3021 * @sock: socket structure
3022 * @lvl: option level
3023 * @opt: option identifier
3024 * @ov: receptacle for option value
3025 * @ol: receptacle for length of option value
3026 *
3027 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3028 * (to ease compatibility).
3029 *
3030 * Returns 0 on success, errno otherwise
3031 */
3032static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3033 char __user *ov, int __user *ol)
3034{
3035 struct sock *sk = sock->sk;
3036 struct tipc_sock *tsk = tipc_sk(sk);
3037 struct tipc_name_seq seq;
3038 int len, scope;
3039 u32 value;
3040 int res;
3041
3042 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3043 return put_user(0, ol);
3044 if (lvl != SOL_TIPC)
3045 return -ENOPROTOOPT;
3046 res = get_user(len, ol);
3047 if (res)
3048 return res;
3049
3050 lock_sock(sk);
3051
3052 switch (opt) {
3053 case TIPC_IMPORTANCE:
3054 value = tsk_importance(tsk);
3055 break;
3056 case TIPC_SRC_DROPPABLE:
3057 value = tsk_unreliable(tsk);
3058 break;
3059 case TIPC_DEST_DROPPABLE:
3060 value = tsk_unreturnable(tsk);
3061 break;
3062 case TIPC_CONN_TIMEOUT:
3063 value = tsk->conn_timeout;
3064 /* no need to set "res", since already 0 at this point */
3065 break;
3066 case TIPC_NODE_RECVQ_DEPTH:
3067 value = 0; /* was tipc_queue_size, now obsolete */
3068 break;
3069 case TIPC_SOCK_RECVQ_DEPTH:
3070 value = skb_queue_len(&sk->sk_receive_queue);
3071 break;
3072 case TIPC_SOCK_RECVQ_USED:
3073 value = sk_rmem_alloc_get(sk);
3074 break;
3075 case TIPC_GROUP_JOIN:
3076 seq.type = 0;
3077 if (tsk->group)
3078 tipc_group_self(tsk->group, &seq, &scope);
3079 value = seq.type;
3080 break;
3081 default:
3082 res = -EINVAL;
3083 }
3084
3085 release_sock(sk);
3086
3087 if (res)
3088 return res; /* "get" failed */
3089
3090 if (len < sizeof(value))
3091 return -EINVAL;
3092
3093 if (copy_to_user(ov, &value, sizeof(value)))
3094 return -EFAULT;
3095
3096 return put_user(sizeof(value), ol);
3097}
3098
3099static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3100{
3101 struct net *net = sock_net(sock->sk);
3102 struct tipc_sioc_nodeid_req nr = {0};
3103 struct tipc_sioc_ln_req lnr;
3104 void __user *argp = (void __user *)arg;
3105
3106 switch (cmd) {
3107 case SIOCGETLINKNAME:
3108 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3109 return -EFAULT;
3110 if (!tipc_node_get_linkname(net,
3111 lnr.bearer_id & 0xffff, lnr.peer,
3112 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3113 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3114 return -EFAULT;
3115 return 0;
3116 }
3117 return -EADDRNOTAVAIL;
3118 case SIOCGETNODEID:
3119 if (copy_from_user(&nr, argp, sizeof(nr)))
3120 return -EFAULT;
3121 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3122 return -EADDRNOTAVAIL;
3123 if (copy_to_user(argp, &nr, sizeof(nr)))
3124 return -EFAULT;
3125 return 0;
3126 default:
3127 return -ENOIOCTLCMD;
3128 }
3129}
3130
3131static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3132{
3133 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3134 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3135 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3136
3137 tsk1->peer.family = AF_TIPC;
3138 tsk1->peer.addrtype = TIPC_ADDR_ID;
3139 tsk1->peer.scope = TIPC_NODE_SCOPE;
3140 tsk1->peer.addr.id.ref = tsk2->portid;
3141 tsk1->peer.addr.id.node = onode;
3142 tsk2->peer.family = AF_TIPC;
3143 tsk2->peer.addrtype = TIPC_ADDR_ID;
3144 tsk2->peer.scope = TIPC_NODE_SCOPE;
3145 tsk2->peer.addr.id.ref = tsk1->portid;
3146 tsk2->peer.addr.id.node = onode;
3147
3148 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3149 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3150 return 0;
3151}
3152
3153/* Protocol switches for the various types of TIPC sockets */
3154
3155static const struct proto_ops msg_ops = {
3156 .owner = THIS_MODULE,
3157 .family = AF_TIPC,
3158 .release = tipc_release,
3159 .bind = tipc_bind,
3160 .connect = tipc_connect,
3161 .socketpair = tipc_socketpair,
3162 .accept = sock_no_accept,
3163 .getname = tipc_getname,
3164 .poll = tipc_poll,
3165 .ioctl = tipc_ioctl,
3166 .listen = sock_no_listen,
3167 .shutdown = tipc_shutdown,
3168 .setsockopt = tipc_setsockopt,
3169 .getsockopt = tipc_getsockopt,
3170 .sendmsg = tipc_sendmsg,
3171 .recvmsg = tipc_recvmsg,
3172 .mmap = sock_no_mmap,
3173 .sendpage = sock_no_sendpage
3174};
3175
3176static const struct proto_ops packet_ops = {
3177 .owner = THIS_MODULE,
3178 .family = AF_TIPC,
3179 .release = tipc_release,
3180 .bind = tipc_bind,
3181 .connect = tipc_connect,
3182 .socketpair = tipc_socketpair,
3183 .accept = tipc_accept,
3184 .getname = tipc_getname,
3185 .poll = tipc_poll,
3186 .ioctl = tipc_ioctl,
3187 .listen = tipc_listen,
3188 .shutdown = tipc_shutdown,
3189 .setsockopt = tipc_setsockopt,
3190 .getsockopt = tipc_getsockopt,
3191 .sendmsg = tipc_send_packet,
3192 .recvmsg = tipc_recvmsg,
3193 .mmap = sock_no_mmap,
3194 .sendpage = sock_no_sendpage
3195};
3196
3197static const struct proto_ops stream_ops = {
3198 .owner = THIS_MODULE,
3199 .family = AF_TIPC,
3200 .release = tipc_release,
3201 .bind = tipc_bind,
3202 .connect = tipc_connect,
3203 .socketpair = tipc_socketpair,
3204 .accept = tipc_accept,
3205 .getname = tipc_getname,
3206 .poll = tipc_poll,
3207 .ioctl = tipc_ioctl,
3208 .listen = tipc_listen,
3209 .shutdown = tipc_shutdown,
3210 .setsockopt = tipc_setsockopt,
3211 .getsockopt = tipc_getsockopt,
3212 .sendmsg = tipc_sendstream,
3213 .recvmsg = tipc_recvstream,
3214 .mmap = sock_no_mmap,
3215 .sendpage = sock_no_sendpage
3216};
3217
3218static const struct net_proto_family tipc_family_ops = {
3219 .owner = THIS_MODULE,
3220 .family = AF_TIPC,
3221 .create = tipc_sk_create
3222};
3223
3224static struct proto tipc_proto = {
3225 .name = "TIPC",
3226 .owner = THIS_MODULE,
3227 .obj_size = sizeof(struct tipc_sock),
3228 .sysctl_rmem = sysctl_tipc_rmem
3229};
3230
3231/**
3232 * tipc_socket_init - initialize TIPC socket interface
3233 *
3234 * Returns 0 on success, errno otherwise
3235 */
3236int tipc_socket_init(void)
3237{
3238 int res;
3239
3240 res = proto_register(&tipc_proto, 1);
3241 if (res) {
3242 pr_err("Failed to register TIPC protocol type\n");
3243 goto out;
3244 }
3245
3246 res = sock_register(&tipc_family_ops);
3247 if (res) {
3248 pr_err("Failed to register TIPC socket type\n");
3249 proto_unregister(&tipc_proto);
3250 goto out;
3251 }
3252 out:
3253 return res;
3254}
3255
3256/**
3257 * tipc_socket_stop - stop TIPC socket interface
3258 */
3259void tipc_socket_stop(void)
3260{
3261 sock_unregister(tipc_family_ops.family);
3262 proto_unregister(&tipc_proto);
3263}
3264
3265/* Caller should hold socket lock for the passed tipc socket. */
3266static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3267{
3268 u32 peer_node;
3269 u32 peer_port;
3270 struct nlattr *nest;
3271
3272 peer_node = tsk_peer_node(tsk);
3273 peer_port = tsk_peer_port(tsk);
3274
3275 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3276 if (!nest)
3277 return -EMSGSIZE;
3278
3279 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3280 goto msg_full;
3281 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3282 goto msg_full;
3283
3284 if (tsk->conn_type != 0) {
3285 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3286 goto msg_full;
3287 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3288 goto msg_full;
3289 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3290 goto msg_full;
3291 }
3292 nla_nest_end(skb, nest);
3293
3294 return 0;
3295
3296msg_full:
3297 nla_nest_cancel(skb, nest);
3298
3299 return -EMSGSIZE;
3300}
3301
3302static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3303 *tsk)
3304{
3305 struct net *net = sock_net(skb->sk);
3306 struct sock *sk = &tsk->sk;
3307
3308 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3309 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3310 return -EMSGSIZE;
3311
3312 if (tipc_sk_connected(sk)) {
3313 if (__tipc_nl_add_sk_con(skb, tsk))
3314 return -EMSGSIZE;
3315 } else if (!list_empty(&tsk->publications)) {
3316 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3317 return -EMSGSIZE;
3318 }
3319 return 0;
3320}
3321
3322/* Caller should hold socket lock for the passed tipc socket. */
3323static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3324 struct tipc_sock *tsk)
3325{
3326 struct nlattr *attrs;
3327 void *hdr;
3328
3329 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3330 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3331 if (!hdr)
3332 goto msg_cancel;
3333
3334 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3335 if (!attrs)
3336 goto genlmsg_cancel;
3337
3338 if (__tipc_nl_add_sk_info(skb, tsk))
3339 goto attr_msg_cancel;
3340
3341 nla_nest_end(skb, attrs);
3342 genlmsg_end(skb, hdr);
3343
3344 return 0;
3345
3346attr_msg_cancel:
3347 nla_nest_cancel(skb, attrs);
3348genlmsg_cancel:
3349 genlmsg_cancel(skb, hdr);
3350msg_cancel:
3351 return -EMSGSIZE;
3352}
3353
3354int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3355 int (*skb_handler)(struct sk_buff *skb,
3356 struct netlink_callback *cb,
3357 struct tipc_sock *tsk))
3358{
3359 struct rhashtable_iter *iter = (void *)cb->args[4];
3360 struct tipc_sock *tsk;
3361 int err;
3362
3363 rhashtable_walk_start(iter);
3364 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3365 if (IS_ERR(tsk)) {
3366 err = PTR_ERR(tsk);
3367 if (err == -EAGAIN) {
3368 err = 0;
3369 continue;
3370 }
3371 break;
3372 }
3373
3374 sock_hold(&tsk->sk);
3375 rhashtable_walk_stop(iter);
3376 lock_sock(&tsk->sk);
3377 err = skb_handler(skb, cb, tsk);
3378 if (err) {
3379 release_sock(&tsk->sk);
3380 sock_put(&tsk->sk);
3381 goto out;
3382 }
3383 release_sock(&tsk->sk);
3384 rhashtable_walk_start(iter);
3385 sock_put(&tsk->sk);
3386 }
3387 rhashtable_walk_stop(iter);
3388out:
3389 return skb->len;
3390}
3391EXPORT_SYMBOL(tipc_nl_sk_walk);
3392
3393int tipc_dump_start(struct netlink_callback *cb)
3394{
3395 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3396}
3397EXPORT_SYMBOL(tipc_dump_start);
3398
3399int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3400{
3401 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3402 struct rhashtable_iter *iter = (void *)cb->args[4];
3403 struct tipc_net *tn = tipc_net(net);
3404
3405 if (!iter) {
3406 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3407 if (!iter)
3408 return -ENOMEM;
3409
3410 cb->args[4] = (long)iter;
3411 }
3412
3413 rhashtable_walk_enter(&tn->sk_rht, iter);
3414 return 0;
3415}
3416
3417int tipc_dump_done(struct netlink_callback *cb)
3418{
3419 struct rhashtable_iter *hti = (void *)cb->args[4];
3420
3421 rhashtable_walk_exit(hti);
3422 kfree(hti);
3423 return 0;
3424}
3425EXPORT_SYMBOL(tipc_dump_done);
3426
3427int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3428 struct tipc_sock *tsk, u32 sk_filter_state,
3429 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3430{
3431 struct sock *sk = &tsk->sk;
3432 struct nlattr *attrs;
3433 struct nlattr *stat;
3434
3435 /*filter response w.r.t sk_state*/
3436 if (!(sk_filter_state & (1 << sk->sk_state)))
3437 return 0;
3438
3439 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3440 if (!attrs)
3441 goto msg_cancel;
3442
3443 if (__tipc_nl_add_sk_info(skb, tsk))
3444 goto attr_msg_cancel;
3445
3446 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3447 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3448 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3449 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3450 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3451 sock_i_uid(sk))) ||
3452 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3453 tipc_diag_gen_cookie(sk),
3454 TIPC_NLA_SOCK_PAD))
3455 goto attr_msg_cancel;
3456
3457 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3458 if (!stat)
3459 goto attr_msg_cancel;
3460
3461 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3462 skb_queue_len(&sk->sk_receive_queue)) ||
3463 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3464 skb_queue_len(&sk->sk_write_queue)) ||
3465 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3466 atomic_read(&sk->sk_drops)))
3467 goto stat_msg_cancel;
3468
3469 if (tsk->cong_link_cnt &&
3470 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3471 goto stat_msg_cancel;
3472
3473 if (tsk_conn_cong(tsk) &&
3474 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3475 goto stat_msg_cancel;
3476
3477 nla_nest_end(skb, stat);
3478
3479 if (tsk->group)
3480 if (tipc_group_fill_sock_diag(tsk->group, skb))
3481 goto stat_msg_cancel;
3482
3483 nla_nest_end(skb, attrs);
3484
3485 return 0;
3486
3487stat_msg_cancel:
3488 nla_nest_cancel(skb, stat);
3489attr_msg_cancel:
3490 nla_nest_cancel(skb, attrs);
3491msg_cancel:
3492 return -EMSGSIZE;
3493}
3494EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3495
3496int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3497{
3498 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3499}
3500
3501/* Caller should hold socket lock for the passed tipc socket. */
3502static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3503 struct netlink_callback *cb,
3504 struct publication *publ)
3505{
3506 void *hdr;
3507 struct nlattr *attrs;
3508
3509 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3510 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3511 if (!hdr)
3512 goto msg_cancel;
3513
3514 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3515 if (!attrs)
3516 goto genlmsg_cancel;
3517
3518 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3519 goto attr_msg_cancel;
3520 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3521 goto attr_msg_cancel;
3522 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3523 goto attr_msg_cancel;
3524 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3525 goto attr_msg_cancel;
3526
3527 nla_nest_end(skb, attrs);
3528 genlmsg_end(skb, hdr);
3529
3530 return 0;
3531
3532attr_msg_cancel:
3533 nla_nest_cancel(skb, attrs);
3534genlmsg_cancel:
3535 genlmsg_cancel(skb, hdr);
3536msg_cancel:
3537 return -EMSGSIZE;
3538}
3539
3540/* Caller should hold socket lock for the passed tipc socket. */
3541static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3542 struct netlink_callback *cb,
3543 struct tipc_sock *tsk, u32 *last_publ)
3544{
3545 int err;
3546 struct publication *p;
3547
3548 if (*last_publ) {
3549 list_for_each_entry(p, &tsk->publications, binding_sock) {
3550 if (p->key == *last_publ)
3551 break;
3552 }
3553 if (p->key != *last_publ) {
3554 /* We never set seq or call nl_dump_check_consistent()
3555 * this means that setting prev_seq here will cause the
3556 * consistence check to fail in the netlink callback
3557 * handler. Resulting in the last NLMSG_DONE message
3558 * having the NLM_F_DUMP_INTR flag set.
3559 */
3560 cb->prev_seq = 1;
3561 *last_publ = 0;
3562 return -EPIPE;
3563 }
3564 } else {
3565 p = list_first_entry(&tsk->publications, struct publication,
3566 binding_sock);
3567 }
3568
3569 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3570 err = __tipc_nl_add_sk_publ(skb, cb, p);
3571 if (err) {
3572 *last_publ = p->key;
3573 return err;
3574 }
3575 }
3576 *last_publ = 0;
3577
3578 return 0;
3579}
3580
3581int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3582{
3583 int err;
3584 u32 tsk_portid = cb->args[0];
3585 u32 last_publ = cb->args[1];
3586 u32 done = cb->args[2];
3587 struct net *net = sock_net(skb->sk);
3588 struct tipc_sock *tsk;
3589
3590 if (!tsk_portid) {
3591 struct nlattr **attrs;
3592 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3593
3594 err = tipc_nlmsg_parse(cb->nlh, &attrs);
3595 if (err)
3596 return err;
3597
3598 if (!attrs[TIPC_NLA_SOCK])
3599 return -EINVAL;
3600
3601 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3602 attrs[TIPC_NLA_SOCK],
3603 tipc_nl_sock_policy, NULL);
3604 if (err)
3605 return err;
3606
3607 if (!sock[TIPC_NLA_SOCK_REF])
3608 return -EINVAL;
3609
3610 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3611 }
3612
3613 if (done)
3614 return 0;
3615
3616 tsk = tipc_sk_lookup(net, tsk_portid);
3617 if (!tsk)
3618 return -EINVAL;
3619
3620 lock_sock(&tsk->sk);
3621 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3622 if (!err)
3623 done = 1;
3624 release_sock(&tsk->sk);
3625 sock_put(&tsk->sk);
3626
3627 cb->args[0] = tsk_portid;
3628 cb->args[1] = last_publ;
3629 cb->args[2] = done;
3630
3631 return skb->len;
3632}
3633
3634/**
3635 * tipc_sk_filtering - check if a socket should be traced
3636 * @sk: the socket to be examined
3637 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3638 * (portid, sock type, name type, name lower, name upper)
3639 *
3640 * Returns true if the socket meets the socket tuple data
3641 * (value 0 = 'any') or when there is no tuple set (all = 0),
3642 * otherwise false
3643 */
3644bool tipc_sk_filtering(struct sock *sk)
3645{
3646 struct tipc_sock *tsk;
3647 struct publication *p;
3648 u32 _port, _sktype, _type, _lower, _upper;
3649 u32 type = 0, lower = 0, upper = 0;
3650
3651 if (!sk)
3652 return true;
3653
3654 tsk = tipc_sk(sk);
3655
3656 _port = sysctl_tipc_sk_filter[0];
3657 _sktype = sysctl_tipc_sk_filter[1];
3658 _type = sysctl_tipc_sk_filter[2];
3659 _lower = sysctl_tipc_sk_filter[3];
3660 _upper = sysctl_tipc_sk_filter[4];
3661
3662 if (!_port && !_sktype && !_type && !_lower && !_upper)
3663 return true;
3664
3665 if (_port)
3666 return (_port == tsk->portid);
3667
3668 if (_sktype && _sktype != sk->sk_type)
3669 return false;
3670
3671 if (tsk->published) {
3672 p = list_first_entry_or_null(&tsk->publications,
3673 struct publication, binding_sock);
3674 if (p) {
3675 type = p->type;
3676 lower = p->lower;
3677 upper = p->upper;
3678 }
3679 }
3680
3681 if (!tipc_sk_type_connectionless(sk)) {
3682 type = tsk->conn_type;
3683 lower = tsk->conn_instance;
3684 upper = tsk->conn_instance;
3685 }
3686
3687 if ((_type && _type != type) || (_lower && _lower != lower) ||
3688 (_upper && _upper != upper))
3689 return false;
3690
3691 return true;
3692}
3693
3694u32 tipc_sock_get_portid(struct sock *sk)
3695{
3696 return (sk) ? (tipc_sk(sk))->portid : 0;
3697}
3698
3699/**
3700 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3701 * both the rcv and backlog queues are considered
3702 * @sk: tipc sk to be checked
3703 * @skb: tipc msg to be checked
3704 *
3705 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3706 */
3707
3708bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3709{
3710 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3711 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3712 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3713
3714 return (qsize > lim * 90 / 100);
3715}
3716
3717/**
3718 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3719 * only the rcv queue is considered
3720 * @sk: tipc sk to be checked
3721 * @skb: tipc msg to be checked
3722 *
3723 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3724 */
3725
3726bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3727{
3728 unsigned int lim = rcvbuf_limit(sk, skb);
3729 unsigned int qsize = sk_rmem_alloc_get(sk);
3730
3731 return (qsize > lim * 90 / 100);
3732}
3733
3734/**
3735 * tipc_sk_dump - dump TIPC socket
3736 * @sk: tipc sk to be dumped
3737 * @dqueues: bitmask to decide if any socket queue to be dumped?
3738 * - TIPC_DUMP_NONE: don't dump socket queues
3739 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3740 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3741 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3742 * - TIPC_DUMP_ALL: dump all the socket queues above
3743 * @buf: returned buffer of dump data in format
3744 */
3745int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3746{
3747 int i = 0;
3748 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3749 struct tipc_sock *tsk;
3750 struct publication *p;
3751 bool tsk_connected;
3752
3753 if (!sk) {
3754 i += scnprintf(buf, sz, "sk data: (null)\n");
3755 return i;
3756 }
3757
3758 tsk = tipc_sk(sk);
3759 tsk_connected = !tipc_sk_type_connectionless(sk);
3760
3761 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3762 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3763 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3764 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3765 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3766 if (tsk_connected) {
3767 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3768 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3769 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3770 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3771 }
3772 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3773 if (tsk->published) {
3774 p = list_first_entry_or_null(&tsk->publications,
3775 struct publication, binding_sock);
3776 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3777 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3778 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3779 }
3780 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3781 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3782 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3783 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3784 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3785 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3786 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3787 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3788 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3789 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3790 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3791 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3792 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3793 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3794
3795 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3796 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3797 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3798 }
3799
3800 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3801 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3802 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3803 }
3804
3805 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3806 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3807 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3808 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3809 i += scnprintf(buf + i, sz - i, " tail ");
3810 i += tipc_skb_dump(sk->sk_backlog.tail, false,
3811 buf + i);
3812 }
3813 }
3814
3815 return i;
3816}