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 NAGLE_START_INIT 4
52#define NAGLE_START_MAX 1024
53#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
54#define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
55#define TIPC_FWD_MSG 1
56#define TIPC_MAX_PORT 0xffffffff
57#define TIPC_MIN_PORT 1
58#define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
59
60enum {
61 TIPC_LISTEN = TCP_LISTEN,
62 TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 TIPC_OPEN = TCP_CLOSE,
64 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 TIPC_CONNECTING = TCP_SYN_SENT,
66};
67
68struct sockaddr_pair {
69 struct sockaddr_tipc sock;
70 struct sockaddr_tipc member;
71};
72
73/**
74 * struct tipc_sock - TIPC socket structure
75 * @sk: socket - interacts with 'port' and with user via the socket API
76 * @conn_type: TIPC type used when connection was established
77 * @conn_instance: TIPC instance used when connection was established
78 * @published: non-zero if port has one or more associated names
79 * @max_pkt: maximum packet size "hint" used when building messages sent by port
80 * @maxnagle: maximum size of msg which can be subject to nagle
81 * @portid: unique port identity in TIPC socket hash table
82 * @phdr: preformatted message header used when sending messages
83 * #cong_links: list of congested links
84 * @publications: list of publications for port
85 * @blocking_link: address of the congested link we are currently sleeping on
86 * @pub_count: total # of publications port has made during its lifetime
87 * @conn_timeout: the time we can wait for an unresponded setup request
88 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
89 * @cong_link_cnt: number of congested links
90 * @snt_unacked: # messages sent by socket, and not yet acked by peer
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @peer: 'connected' peer for dgram/rdm
93 * @node: hash table node
94 * @mc_method: cookie for use between socket and broadcast layer
95 * @rcu: rcu struct for tipc_sock
96 */
97struct tipc_sock {
98 struct sock sk;
99 u32 conn_type;
100 u32 conn_instance;
101 int published;
102 u32 max_pkt;
103 u32 maxnagle;
104 u32 portid;
105 struct tipc_msg phdr;
106 struct list_head cong_links;
107 struct list_head publications;
108 u32 pub_count;
109 atomic_t dupl_rcvcnt;
110 u16 conn_timeout;
111 bool probe_unacked;
112 u16 cong_link_cnt;
113 u16 snt_unacked;
114 u16 snd_win;
115 u16 peer_caps;
116 u16 rcv_unacked;
117 u16 rcv_win;
118 struct sockaddr_tipc peer;
119 struct rhash_head node;
120 struct tipc_mc_method mc_method;
121 struct rcu_head rcu;
122 struct tipc_group *group;
123 u32 oneway;
124 u32 nagle_start;
125 u16 snd_backlog;
126 u16 msg_acc;
127 u16 pkt_cnt;
128 bool expect_ack;
129 bool nodelay;
130 bool group_is_open;
131};
132
133static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
134static void tipc_data_ready(struct sock *sk);
135static void tipc_write_space(struct sock *sk);
136static void tipc_sock_destruct(struct sock *sk);
137static int tipc_release(struct socket *sock);
138static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
139 bool kern);
140static void tipc_sk_timeout(struct timer_list *t);
141static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
142 struct tipc_name_seq const *seq);
143static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
144 struct tipc_name_seq const *seq);
145static int tipc_sk_leave(struct tipc_sock *tsk);
146static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
147static int tipc_sk_insert(struct tipc_sock *tsk);
148static void tipc_sk_remove(struct tipc_sock *tsk);
149static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
150static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
151static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
152
153static const struct proto_ops packet_ops;
154static const struct proto_ops stream_ops;
155static const struct proto_ops msg_ops;
156static struct proto tipc_proto;
157static const struct rhashtable_params tsk_rht_params;
158
159static u32 tsk_own_node(struct tipc_sock *tsk)
160{
161 return msg_prevnode(&tsk->phdr);
162}
163
164static u32 tsk_peer_node(struct tipc_sock *tsk)
165{
166 return msg_destnode(&tsk->phdr);
167}
168
169static u32 tsk_peer_port(struct tipc_sock *tsk)
170{
171 return msg_destport(&tsk->phdr);
172}
173
174static bool tsk_unreliable(struct tipc_sock *tsk)
175{
176 return msg_src_droppable(&tsk->phdr) != 0;
177}
178
179static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
180{
181 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
182}
183
184static bool tsk_unreturnable(struct tipc_sock *tsk)
185{
186 return msg_dest_droppable(&tsk->phdr) != 0;
187}
188
189static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
190{
191 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
192}
193
194static int tsk_importance(struct tipc_sock *tsk)
195{
196 return msg_importance(&tsk->phdr);
197}
198
199static struct tipc_sock *tipc_sk(const struct sock *sk)
200{
201 return container_of(sk, struct tipc_sock, sk);
202}
203
204int tsk_set_importance(struct sock *sk, int imp)
205{
206 if (imp > TIPC_CRITICAL_IMPORTANCE)
207 return -EINVAL;
208 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
209 return 0;
210}
211
212static bool tsk_conn_cong(struct tipc_sock *tsk)
213{
214 return tsk->snt_unacked > tsk->snd_win;
215}
216
217static u16 tsk_blocks(int len)
218{
219 return ((len / FLOWCTL_BLK_SZ) + 1);
220}
221
222/* tsk_blocks(): translate a buffer size in bytes to number of
223 * advertisable blocks, taking into account the ratio truesize(len)/len
224 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
225 */
226static u16 tsk_adv_blocks(int len)
227{
228 return len / FLOWCTL_BLK_SZ / 4;
229}
230
231/* tsk_inc(): increment counter for sent or received data
232 * - If block based flow control is not supported by peer we
233 * fall back to message based ditto, incrementing the counter
234 */
235static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
236{
237 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
238 return ((msglen / FLOWCTL_BLK_SZ) + 1);
239 return 1;
240}
241
242/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
243 */
244static void tsk_set_nagle(struct tipc_sock *tsk)
245{
246 struct sock *sk = &tsk->sk;
247
248 tsk->maxnagle = 0;
249 if (sk->sk_type != SOCK_STREAM)
250 return;
251 if (tsk->nodelay)
252 return;
253 if (!(tsk->peer_caps & TIPC_NAGLE))
254 return;
255 /* Limit node local buffer size to avoid receive queue overflow */
256 if (tsk->max_pkt == MAX_MSG_SIZE)
257 tsk->maxnagle = 1500;
258 else
259 tsk->maxnagle = tsk->max_pkt;
260}
261
262/**
263 * tsk_advance_rx_queue - discard first buffer in socket receive queue
264 *
265 * Caller must hold socket lock
266 */
267static void tsk_advance_rx_queue(struct sock *sk)
268{
269 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
270 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
271}
272
273/* tipc_sk_respond() : send response message back to sender
274 */
275static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
276{
277 u32 selector;
278 u32 dnode;
279 u32 onode = tipc_own_addr(sock_net(sk));
280
281 if (!tipc_msg_reverse(onode, &skb, err))
282 return;
283
284 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
285 dnode = msg_destnode(buf_msg(skb));
286 selector = msg_origport(buf_msg(skb));
287 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
288}
289
290/**
291 * tsk_rej_rx_queue - reject all buffers in socket receive queue
292 *
293 * Caller must hold socket lock
294 */
295static void tsk_rej_rx_queue(struct sock *sk, int error)
296{
297 struct sk_buff *skb;
298
299 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
300 tipc_sk_respond(sk, skb, error);
301}
302
303static bool tipc_sk_connected(struct sock *sk)
304{
305 return sk->sk_state == TIPC_ESTABLISHED;
306}
307
308/* tipc_sk_type_connectionless - check if the socket is datagram socket
309 * @sk: socket
310 *
311 * Returns true if connection less, false otherwise
312 */
313static bool tipc_sk_type_connectionless(struct sock *sk)
314{
315 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
316}
317
318/* tsk_peer_msg - verify if message was sent by connected port's peer
319 *
320 * Handles cases where the node's network address has changed from
321 * the default of <0.0.0> to its configured setting.
322 */
323static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
324{
325 struct sock *sk = &tsk->sk;
326 u32 self = tipc_own_addr(sock_net(sk));
327 u32 peer_port = tsk_peer_port(tsk);
328 u32 orig_node, peer_node;
329
330 if (unlikely(!tipc_sk_connected(sk)))
331 return false;
332
333 if (unlikely(msg_origport(msg) != peer_port))
334 return false;
335
336 orig_node = msg_orignode(msg);
337 peer_node = tsk_peer_node(tsk);
338
339 if (likely(orig_node == peer_node))
340 return true;
341
342 if (!orig_node && peer_node == self)
343 return true;
344
345 if (!peer_node && orig_node == self)
346 return true;
347
348 return false;
349}
350
351/* tipc_set_sk_state - set the sk_state of the socket
352 * @sk: socket
353 *
354 * Caller must hold socket lock
355 *
356 * Returns 0 on success, errno otherwise
357 */
358static int tipc_set_sk_state(struct sock *sk, int state)
359{
360 int oldsk_state = sk->sk_state;
361 int res = -EINVAL;
362
363 switch (state) {
364 case TIPC_OPEN:
365 res = 0;
366 break;
367 case TIPC_LISTEN:
368 case TIPC_CONNECTING:
369 if (oldsk_state == TIPC_OPEN)
370 res = 0;
371 break;
372 case TIPC_ESTABLISHED:
373 if (oldsk_state == TIPC_CONNECTING ||
374 oldsk_state == TIPC_OPEN)
375 res = 0;
376 break;
377 case TIPC_DISCONNECTING:
378 if (oldsk_state == TIPC_CONNECTING ||
379 oldsk_state == TIPC_ESTABLISHED)
380 res = 0;
381 break;
382 }
383
384 if (!res)
385 sk->sk_state = state;
386
387 return res;
388}
389
390static int tipc_sk_sock_err(struct socket *sock, long *timeout)
391{
392 struct sock *sk = sock->sk;
393 int err = sock_error(sk);
394 int typ = sock->type;
395
396 if (err)
397 return err;
398 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
399 if (sk->sk_state == TIPC_DISCONNECTING)
400 return -EPIPE;
401 else if (!tipc_sk_connected(sk))
402 return -ENOTCONN;
403 }
404 if (!*timeout)
405 return -EAGAIN;
406 if (signal_pending(current))
407 return sock_intr_errno(*timeout);
408
409 return 0;
410}
411
412#define tipc_wait_for_cond(sock_, timeo_, condition_) \
413({ \
414 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
415 struct sock *sk_; \
416 int rc_; \
417 \
418 while ((rc_ = !(condition_))) { \
419 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
420 smp_rmb(); \
421 sk_ = (sock_)->sk; \
422 rc_ = tipc_sk_sock_err((sock_), timeo_); \
423 if (rc_) \
424 break; \
425 add_wait_queue(sk_sleep(sk_), &wait_); \
426 release_sock(sk_); \
427 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
428 sched_annotate_sleep(); \
429 lock_sock(sk_); \
430 remove_wait_queue(sk_sleep(sk_), &wait_); \
431 } \
432 rc_; \
433})
434
435/**
436 * tipc_sk_create - create a TIPC socket
437 * @net: network namespace (must be default network)
438 * @sock: pre-allocated socket structure
439 * @protocol: protocol indicator (must be 0)
440 * @kern: caused by kernel or by userspace?
441 *
442 * This routine creates additional data structures used by the TIPC socket,
443 * initializes them, and links them together.
444 *
445 * Returns 0 on success, errno otherwise
446 */
447static int tipc_sk_create(struct net *net, struct socket *sock,
448 int protocol, int kern)
449{
450 const struct proto_ops *ops;
451 struct sock *sk;
452 struct tipc_sock *tsk;
453 struct tipc_msg *msg;
454
455 /* Validate arguments */
456 if (unlikely(protocol != 0))
457 return -EPROTONOSUPPORT;
458
459 switch (sock->type) {
460 case SOCK_STREAM:
461 ops = &stream_ops;
462 break;
463 case SOCK_SEQPACKET:
464 ops = &packet_ops;
465 break;
466 case SOCK_DGRAM:
467 case SOCK_RDM:
468 ops = &msg_ops;
469 break;
470 default:
471 return -EPROTOTYPE;
472 }
473
474 /* Allocate socket's protocol area */
475 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
476 if (sk == NULL)
477 return -ENOMEM;
478
479 tsk = tipc_sk(sk);
480 tsk->max_pkt = MAX_PKT_DEFAULT;
481 tsk->maxnagle = 0;
482 tsk->nagle_start = NAGLE_START_INIT;
483 INIT_LIST_HEAD(&tsk->publications);
484 INIT_LIST_HEAD(&tsk->cong_links);
485 msg = &tsk->phdr;
486
487 /* Finish initializing socket data structures */
488 sock->ops = ops;
489 sock_init_data(sock, sk);
490 tipc_set_sk_state(sk, TIPC_OPEN);
491 if (tipc_sk_insert(tsk)) {
492 pr_warn("Socket create failed; port number exhausted\n");
493 return -EINVAL;
494 }
495
496 /* Ensure tsk is visible before we read own_addr. */
497 smp_mb();
498
499 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
500 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
501
502 msg_set_origport(msg, tsk->portid);
503 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
504 sk->sk_shutdown = 0;
505 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
506 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
507 sk->sk_data_ready = tipc_data_ready;
508 sk->sk_write_space = tipc_write_space;
509 sk->sk_destruct = tipc_sock_destruct;
510 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
511 tsk->group_is_open = true;
512 atomic_set(&tsk->dupl_rcvcnt, 0);
513
514 /* Start out with safe limits until we receive an advertised window */
515 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
516 tsk->rcv_win = tsk->snd_win;
517
518 if (tipc_sk_type_connectionless(sk)) {
519 tsk_set_unreturnable(tsk, true);
520 if (sock->type == SOCK_DGRAM)
521 tsk_set_unreliable(tsk, true);
522 }
523 __skb_queue_head_init(&tsk->mc_method.deferredq);
524 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
525 return 0;
526}
527
528static void tipc_sk_callback(struct rcu_head *head)
529{
530 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
531
532 sock_put(&tsk->sk);
533}
534
535/* Caller should hold socket lock for the socket. */
536static void __tipc_shutdown(struct socket *sock, int error)
537{
538 struct sock *sk = sock->sk;
539 struct tipc_sock *tsk = tipc_sk(sk);
540 struct net *net = sock_net(sk);
541 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
542 u32 dnode = tsk_peer_node(tsk);
543 struct sk_buff *skb;
544
545 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
546 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
547 !tsk_conn_cong(tsk)));
548
549 /* Push out delayed messages if in Nagle mode */
550 tipc_sk_push_backlog(tsk, false);
551 /* Remove pending SYN */
552 __skb_queue_purge(&sk->sk_write_queue);
553
554 /* Remove partially received buffer if any */
555 skb = skb_peek(&sk->sk_receive_queue);
556 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
557 __skb_unlink(skb, &sk->sk_receive_queue);
558 kfree_skb(skb);
559 }
560
561 /* Reject all unreceived messages if connectionless */
562 if (tipc_sk_type_connectionless(sk)) {
563 tsk_rej_rx_queue(sk, error);
564 return;
565 }
566
567 switch (sk->sk_state) {
568 case TIPC_CONNECTING:
569 case TIPC_ESTABLISHED:
570 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
571 tipc_node_remove_conn(net, dnode, tsk->portid);
572 /* Send a FIN+/- to its peer */
573 skb = __skb_dequeue(&sk->sk_receive_queue);
574 if (skb) {
575 __skb_queue_purge(&sk->sk_receive_queue);
576 tipc_sk_respond(sk, skb, error);
577 break;
578 }
579 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
580 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
581 tsk_own_node(tsk), tsk_peer_port(tsk),
582 tsk->portid, error);
583 if (skb)
584 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
585 break;
586 case TIPC_LISTEN:
587 /* Reject all SYN messages */
588 tsk_rej_rx_queue(sk, error);
589 break;
590 default:
591 __skb_queue_purge(&sk->sk_receive_queue);
592 break;
593 }
594}
595
596/**
597 * tipc_release - destroy a TIPC socket
598 * @sock: socket to destroy
599 *
600 * This routine cleans up any messages that are still queued on the socket.
601 * For DGRAM and RDM socket types, all queued messages are rejected.
602 * For SEQPACKET and STREAM socket types, the first message is rejected
603 * and any others are discarded. (If the first message on a STREAM socket
604 * is partially-read, it is discarded and the next one is rejected instead.)
605 *
606 * NOTE: Rejected messages are not necessarily returned to the sender! They
607 * are returned or discarded according to the "destination droppable" setting
608 * specified for the message by the sender.
609 *
610 * Returns 0 on success, errno otherwise
611 */
612static int tipc_release(struct socket *sock)
613{
614 struct sock *sk = sock->sk;
615 struct tipc_sock *tsk;
616
617 /*
618 * Exit if socket isn't fully initialized (occurs when a failed accept()
619 * releases a pre-allocated child socket that was never used)
620 */
621 if (sk == NULL)
622 return 0;
623
624 tsk = tipc_sk(sk);
625 lock_sock(sk);
626
627 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
628 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
629 sk->sk_shutdown = SHUTDOWN_MASK;
630 tipc_sk_leave(tsk);
631 tipc_sk_withdraw(tsk, 0, NULL);
632 __skb_queue_purge(&tsk->mc_method.deferredq);
633 sk_stop_timer(sk, &sk->sk_timer);
634 tipc_sk_remove(tsk);
635
636 sock_orphan(sk);
637 /* Reject any messages that accumulated in backlog queue */
638 release_sock(sk);
639 tipc_dest_list_purge(&tsk->cong_links);
640 tsk->cong_link_cnt = 0;
641 call_rcu(&tsk->rcu, tipc_sk_callback);
642 sock->sk = NULL;
643
644 return 0;
645}
646
647/**
648 * tipc_bind - associate or disassocate TIPC name(s) with a socket
649 * @sock: socket structure
650 * @uaddr: socket address describing name(s) and desired operation
651 * @uaddr_len: size of socket address data structure
652 *
653 * Name and name sequence binding is indicated using a positive scope value;
654 * a negative scope value unbinds the specified name. Specifying no name
655 * (i.e. a socket address length of 0) unbinds all names from the socket.
656 *
657 * Returns 0 on success, errno otherwise
658 *
659 * NOTE: This routine doesn't need to take the socket lock since it doesn't
660 * access any non-constant socket information.
661 */
662static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
663 int uaddr_len)
664{
665 struct sock *sk = sock->sk;
666 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
667 struct tipc_sock *tsk = tipc_sk(sk);
668 int res = -EINVAL;
669
670 lock_sock(sk);
671 if (unlikely(!uaddr_len)) {
672 res = tipc_sk_withdraw(tsk, 0, NULL);
673 goto exit;
674 }
675 if (tsk->group) {
676 res = -EACCES;
677 goto exit;
678 }
679 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
680 res = -EINVAL;
681 goto exit;
682 }
683 if (addr->family != AF_TIPC) {
684 res = -EAFNOSUPPORT;
685 goto exit;
686 }
687
688 if (addr->addrtype == TIPC_ADDR_NAME)
689 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
690 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
691 res = -EAFNOSUPPORT;
692 goto exit;
693 }
694
695 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
696 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
697 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
698 res = -EACCES;
699 goto exit;
700 }
701
702 res = (addr->scope >= 0) ?
703 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
704 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
705exit:
706 release_sock(sk);
707 return res;
708}
709
710/**
711 * tipc_getname - get port ID of socket or peer socket
712 * @sock: socket structure
713 * @uaddr: area for returned socket address
714 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
715 *
716 * Returns 0 on success, errno otherwise
717 *
718 * NOTE: This routine doesn't need to take the socket lock since it only
719 * accesses socket information that is unchanging (or which changes in
720 * a completely predictable manner).
721 */
722static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
723 int peer)
724{
725 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
726 struct sock *sk = sock->sk;
727 struct tipc_sock *tsk = tipc_sk(sk);
728
729 memset(addr, 0, sizeof(*addr));
730 if (peer) {
731 if ((!tipc_sk_connected(sk)) &&
732 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
733 return -ENOTCONN;
734 addr->addr.id.ref = tsk_peer_port(tsk);
735 addr->addr.id.node = tsk_peer_node(tsk);
736 } else {
737 addr->addr.id.ref = tsk->portid;
738 addr->addr.id.node = tipc_own_addr(sock_net(sk));
739 }
740
741 addr->addrtype = TIPC_ADDR_ID;
742 addr->family = AF_TIPC;
743 addr->scope = 0;
744 addr->addr.name.domain = 0;
745
746 return sizeof(*addr);
747}
748
749/**
750 * tipc_poll - read and possibly block on pollmask
751 * @file: file structure associated with the socket
752 * @sock: socket for which to calculate the poll bits
753 * @wait: ???
754 *
755 * Returns pollmask value
756 *
757 * COMMENTARY:
758 * It appears that the usual socket locking mechanisms are not useful here
759 * since the pollmask info is potentially out-of-date the moment this routine
760 * exits. TCP and other protocols seem to rely on higher level poll routines
761 * to handle any preventable race conditions, so TIPC will do the same ...
762 *
763 * IMPORTANT: The fact that a read or write operation is indicated does NOT
764 * imply that the operation will succeed, merely that it should be performed
765 * and will not block.
766 */
767static __poll_t tipc_poll(struct file *file, struct socket *sock,
768 poll_table *wait)
769{
770 struct sock *sk = sock->sk;
771 struct tipc_sock *tsk = tipc_sk(sk);
772 __poll_t revents = 0;
773
774 sock_poll_wait(file, sock, wait);
775 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
776
777 if (sk->sk_shutdown & RCV_SHUTDOWN)
778 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
779 if (sk->sk_shutdown == SHUTDOWN_MASK)
780 revents |= EPOLLHUP;
781
782 switch (sk->sk_state) {
783 case TIPC_ESTABLISHED:
784 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
785 revents |= EPOLLOUT;
786 fallthrough;
787 case TIPC_LISTEN:
788 case TIPC_CONNECTING:
789 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
790 revents |= EPOLLIN | EPOLLRDNORM;
791 break;
792 case TIPC_OPEN:
793 if (tsk->group_is_open && !tsk->cong_link_cnt)
794 revents |= EPOLLOUT;
795 if (!tipc_sk_type_connectionless(sk))
796 break;
797 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
798 break;
799 revents |= EPOLLIN | EPOLLRDNORM;
800 break;
801 case TIPC_DISCONNECTING:
802 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
803 break;
804 }
805 return revents;
806}
807
808/**
809 * tipc_sendmcast - send multicast message
810 * @sock: socket structure
811 * @seq: destination address
812 * @msg: message to send
813 * @dlen: length of data to send
814 * @timeout: timeout to wait for wakeup
815 *
816 * Called from function tipc_sendmsg(), which has done all sanity checks
817 * Returns the number of bytes sent on success, or errno
818 */
819static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
820 struct msghdr *msg, size_t dlen, long timeout)
821{
822 struct sock *sk = sock->sk;
823 struct tipc_sock *tsk = tipc_sk(sk);
824 struct tipc_msg *hdr = &tsk->phdr;
825 struct net *net = sock_net(sk);
826 int mtu = tipc_bcast_get_mtu(net);
827 struct tipc_mc_method *method = &tsk->mc_method;
828 struct sk_buff_head pkts;
829 struct tipc_nlist dsts;
830 int rc;
831
832 if (tsk->group)
833 return -EACCES;
834
835 /* Block or return if any destination link is congested */
836 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
837 if (unlikely(rc))
838 return rc;
839
840 /* Lookup destination nodes */
841 tipc_nlist_init(&dsts, tipc_own_addr(net));
842 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
843 seq->upper, &dsts);
844 if (!dsts.local && !dsts.remote)
845 return -EHOSTUNREACH;
846
847 /* Build message header */
848 msg_set_type(hdr, TIPC_MCAST_MSG);
849 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
850 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
851 msg_set_destport(hdr, 0);
852 msg_set_destnode(hdr, 0);
853 msg_set_nametype(hdr, seq->type);
854 msg_set_namelower(hdr, seq->lower);
855 msg_set_nameupper(hdr, seq->upper);
856
857 /* Build message as chain of buffers */
858 __skb_queue_head_init(&pkts);
859 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
860
861 /* Send message if build was successful */
862 if (unlikely(rc == dlen)) {
863 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
864 TIPC_DUMP_SK_SNDQ, " ");
865 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
866 &tsk->cong_link_cnt);
867 }
868
869 tipc_nlist_purge(&dsts);
870
871 return rc ? rc : dlen;
872}
873
874/**
875 * tipc_send_group_msg - send a message to a member in the group
876 * @net: network namespace
877 * @m: message to send
878 * @mb: group member
879 * @dnode: destination node
880 * @dport: destination port
881 * @dlen: total length of message data
882 */
883static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
884 struct msghdr *m, struct tipc_member *mb,
885 u32 dnode, u32 dport, int dlen)
886{
887 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
888 struct tipc_mc_method *method = &tsk->mc_method;
889 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
890 struct tipc_msg *hdr = &tsk->phdr;
891 struct sk_buff_head pkts;
892 int mtu, rc;
893
894 /* Complete message header */
895 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
896 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
897 msg_set_destport(hdr, dport);
898 msg_set_destnode(hdr, dnode);
899 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
900
901 /* Build message as chain of buffers */
902 __skb_queue_head_init(&pkts);
903 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
904 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
905 if (unlikely(rc != dlen))
906 return rc;
907
908 /* Send message */
909 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
910 if (unlikely(rc == -ELINKCONG)) {
911 tipc_dest_push(&tsk->cong_links, dnode, 0);
912 tsk->cong_link_cnt++;
913 }
914
915 /* Update send window */
916 tipc_group_update_member(mb, blks);
917
918 /* A broadcast sent within next EXPIRE period must follow same path */
919 method->rcast = true;
920 method->mandatory = true;
921 return dlen;
922}
923
924/**
925 * tipc_send_group_unicast - send message to a member in the group
926 * @sock: socket structure
927 * @m: message to send
928 * @dlen: total length of message data
929 * @timeout: timeout to wait for wakeup
930 *
931 * Called from function tipc_sendmsg(), which has done all sanity checks
932 * Returns the number of bytes sent on success, or errno
933 */
934static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
935 int dlen, long timeout)
936{
937 struct sock *sk = sock->sk;
938 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
939 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
940 struct tipc_sock *tsk = tipc_sk(sk);
941 struct net *net = sock_net(sk);
942 struct tipc_member *mb = NULL;
943 u32 node, port;
944 int rc;
945
946 node = dest->addr.id.node;
947 port = dest->addr.id.ref;
948 if (!port && !node)
949 return -EHOSTUNREACH;
950
951 /* Block or return if destination link or member is congested */
952 rc = tipc_wait_for_cond(sock, &timeout,
953 !tipc_dest_find(&tsk->cong_links, node, 0) &&
954 tsk->group &&
955 !tipc_group_cong(tsk->group, node, port, blks,
956 &mb));
957 if (unlikely(rc))
958 return rc;
959
960 if (unlikely(!mb))
961 return -EHOSTUNREACH;
962
963 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
964
965 return rc ? rc : dlen;
966}
967
968/**
969 * tipc_send_group_anycast - send message to any member with given identity
970 * @sock: socket structure
971 * @m: message to send
972 * @dlen: total length of message data
973 * @timeout: timeout to wait for wakeup
974 *
975 * Called from function tipc_sendmsg(), which has done all sanity checks
976 * Returns the number of bytes sent on success, or errno
977 */
978static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
979 int dlen, long timeout)
980{
981 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
982 struct sock *sk = sock->sk;
983 struct tipc_sock *tsk = tipc_sk(sk);
984 struct list_head *cong_links = &tsk->cong_links;
985 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
986 struct tipc_msg *hdr = &tsk->phdr;
987 struct tipc_member *first = NULL;
988 struct tipc_member *mbr = NULL;
989 struct net *net = sock_net(sk);
990 u32 node, port, exclude;
991 struct list_head dsts;
992 u32 type, inst, scope;
993 int lookups = 0;
994 int dstcnt, rc;
995 bool cong;
996
997 INIT_LIST_HEAD(&dsts);
998
999 type = msg_nametype(hdr);
1000 inst = dest->addr.name.name.instance;
1001 scope = msg_lookup_scope(hdr);
1002
1003 while (++lookups < 4) {
1004 exclude = tipc_group_exclude(tsk->group);
1005
1006 first = NULL;
1007
1008 /* Look for a non-congested destination member, if any */
1009 while (1) {
1010 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1011 &dstcnt, exclude, false))
1012 return -EHOSTUNREACH;
1013 tipc_dest_pop(&dsts, &node, &port);
1014 cong = tipc_group_cong(tsk->group, node, port, blks,
1015 &mbr);
1016 if (!cong)
1017 break;
1018 if (mbr == first)
1019 break;
1020 if (!first)
1021 first = mbr;
1022 }
1023
1024 /* Start over if destination was not in member list */
1025 if (unlikely(!mbr))
1026 continue;
1027
1028 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1029 break;
1030
1031 /* Block or return if destination link or member is congested */
1032 rc = tipc_wait_for_cond(sock, &timeout,
1033 !tipc_dest_find(cong_links, node, 0) &&
1034 tsk->group &&
1035 !tipc_group_cong(tsk->group, node, port,
1036 blks, &mbr));
1037 if (unlikely(rc))
1038 return rc;
1039
1040 /* Send, unless destination disappeared while waiting */
1041 if (likely(mbr))
1042 break;
1043 }
1044
1045 if (unlikely(lookups >= 4))
1046 return -EHOSTUNREACH;
1047
1048 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1049
1050 return rc ? rc : dlen;
1051}
1052
1053/**
1054 * tipc_send_group_bcast - send message to all members in communication group
1055 * @sock: socket structure
1056 * @m: message to send
1057 * @dlen: total length of message data
1058 * @timeout: timeout to wait for wakeup
1059 *
1060 * Called from function tipc_sendmsg(), which has done all sanity checks
1061 * Returns the number of bytes sent on success, or errno
1062 */
1063static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1064 int dlen, long timeout)
1065{
1066 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1067 struct sock *sk = sock->sk;
1068 struct net *net = sock_net(sk);
1069 struct tipc_sock *tsk = tipc_sk(sk);
1070 struct tipc_nlist *dsts;
1071 struct tipc_mc_method *method = &tsk->mc_method;
1072 bool ack = method->mandatory && method->rcast;
1073 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1074 struct tipc_msg *hdr = &tsk->phdr;
1075 int mtu = tipc_bcast_get_mtu(net);
1076 struct sk_buff_head pkts;
1077 int rc = -EHOSTUNREACH;
1078
1079 /* Block or return if any destination link or member is congested */
1080 rc = tipc_wait_for_cond(sock, &timeout,
1081 !tsk->cong_link_cnt && tsk->group &&
1082 !tipc_group_bc_cong(tsk->group, blks));
1083 if (unlikely(rc))
1084 return rc;
1085
1086 dsts = tipc_group_dests(tsk->group);
1087 if (!dsts->local && !dsts->remote)
1088 return -EHOSTUNREACH;
1089
1090 /* Complete message header */
1091 if (dest) {
1092 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1093 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1094 } else {
1095 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1096 msg_set_nameinst(hdr, 0);
1097 }
1098 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1099 msg_set_destport(hdr, 0);
1100 msg_set_destnode(hdr, 0);
1101 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1102
1103 /* Avoid getting stuck with repeated forced replicasts */
1104 msg_set_grp_bc_ack_req(hdr, ack);
1105
1106 /* Build message as chain of buffers */
1107 __skb_queue_head_init(&pkts);
1108 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1109 if (unlikely(rc != dlen))
1110 return rc;
1111
1112 /* Send message */
1113 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1114 if (unlikely(rc))
1115 return rc;
1116
1117 /* Update broadcast sequence number and send windows */
1118 tipc_group_update_bc_members(tsk->group, blks, ack);
1119
1120 /* Broadcast link is now free to choose method for next broadcast */
1121 method->mandatory = false;
1122 method->expires = jiffies;
1123
1124 return dlen;
1125}
1126
1127/**
1128 * tipc_send_group_mcast - send message to all members with given identity
1129 * @sock: socket structure
1130 * @m: message to send
1131 * @dlen: total length of message data
1132 * @timeout: timeout to wait for wakeup
1133 *
1134 * Called from function tipc_sendmsg(), which has done all sanity checks
1135 * Returns the number of bytes sent on success, or errno
1136 */
1137static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1138 int dlen, long timeout)
1139{
1140 struct sock *sk = sock->sk;
1141 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1142 struct tipc_sock *tsk = tipc_sk(sk);
1143 struct tipc_group *grp = tsk->group;
1144 struct tipc_msg *hdr = &tsk->phdr;
1145 struct net *net = sock_net(sk);
1146 u32 type, inst, scope, exclude;
1147 struct list_head dsts;
1148 u32 dstcnt;
1149
1150 INIT_LIST_HEAD(&dsts);
1151
1152 type = msg_nametype(hdr);
1153 inst = dest->addr.name.name.instance;
1154 scope = msg_lookup_scope(hdr);
1155 exclude = tipc_group_exclude(grp);
1156
1157 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1158 &dstcnt, exclude, true))
1159 return -EHOSTUNREACH;
1160
1161 if (dstcnt == 1) {
1162 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1163 return tipc_send_group_unicast(sock, m, dlen, timeout);
1164 }
1165
1166 tipc_dest_list_purge(&dsts);
1167 return tipc_send_group_bcast(sock, m, dlen, timeout);
1168}
1169
1170/**
1171 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1172 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1173 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1174 *
1175 * Multi-threaded: parallel calls with reference to same queues may occur
1176 */
1177void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1178 struct sk_buff_head *inputq)
1179{
1180 u32 self = tipc_own_addr(net);
1181 u32 type, lower, upper, scope;
1182 struct sk_buff *skb, *_skb;
1183 u32 portid, onode;
1184 struct sk_buff_head tmpq;
1185 struct list_head dports;
1186 struct tipc_msg *hdr;
1187 int user, mtyp, hlen;
1188 bool exact;
1189
1190 __skb_queue_head_init(&tmpq);
1191 INIT_LIST_HEAD(&dports);
1192
1193 skb = tipc_skb_peek(arrvq, &inputq->lock);
1194 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1195 hdr = buf_msg(skb);
1196 user = msg_user(hdr);
1197 mtyp = msg_type(hdr);
1198 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1199 onode = msg_orignode(hdr);
1200 type = msg_nametype(hdr);
1201
1202 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1203 spin_lock_bh(&inputq->lock);
1204 if (skb_peek(arrvq) == skb) {
1205 __skb_dequeue(arrvq);
1206 __skb_queue_tail(inputq, skb);
1207 }
1208 kfree_skb(skb);
1209 spin_unlock_bh(&inputq->lock);
1210 continue;
1211 }
1212
1213 /* Group messages require exact scope match */
1214 if (msg_in_group(hdr)) {
1215 lower = 0;
1216 upper = ~0;
1217 scope = msg_lookup_scope(hdr);
1218 exact = true;
1219 } else {
1220 /* TIPC_NODE_SCOPE means "any scope" in this context */
1221 if (onode == self)
1222 scope = TIPC_NODE_SCOPE;
1223 else
1224 scope = TIPC_CLUSTER_SCOPE;
1225 exact = false;
1226 lower = msg_namelower(hdr);
1227 upper = msg_nameupper(hdr);
1228 }
1229
1230 /* Create destination port list: */
1231 tipc_nametbl_mc_lookup(net, type, lower, upper,
1232 scope, exact, &dports);
1233
1234 /* Clone message per destination */
1235 while (tipc_dest_pop(&dports, NULL, &portid)) {
1236 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1237 if (_skb) {
1238 msg_set_destport(buf_msg(_skb), portid);
1239 __skb_queue_tail(&tmpq, _skb);
1240 continue;
1241 }
1242 pr_warn("Failed to clone mcast rcv buffer\n");
1243 }
1244 /* Append to inputq if not already done by other thread */
1245 spin_lock_bh(&inputq->lock);
1246 if (skb_peek(arrvq) == skb) {
1247 skb_queue_splice_tail_init(&tmpq, inputq);
1248 kfree_skb(__skb_dequeue(arrvq));
1249 }
1250 spin_unlock_bh(&inputq->lock);
1251 __skb_queue_purge(&tmpq);
1252 kfree_skb(skb);
1253 }
1254 tipc_sk_rcv(net, inputq);
1255}
1256
1257/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1258 * when socket is in Nagle mode
1259 */
1260static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1261{
1262 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1263 struct sk_buff *skb = skb_peek_tail(txq);
1264 struct net *net = sock_net(&tsk->sk);
1265 u32 dnode = tsk_peer_node(tsk);
1266 int rc;
1267
1268 if (nagle_ack) {
1269 tsk->pkt_cnt += skb_queue_len(txq);
1270 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1271 tsk->oneway = 0;
1272 if (tsk->nagle_start < NAGLE_START_MAX)
1273 tsk->nagle_start *= 2;
1274 tsk->expect_ack = false;
1275 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1276 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1277 tsk->nagle_start);
1278 } else {
1279 tsk->nagle_start = NAGLE_START_INIT;
1280 if (skb) {
1281 msg_set_ack_required(buf_msg(skb));
1282 tsk->expect_ack = true;
1283 } else {
1284 tsk->expect_ack = false;
1285 }
1286 }
1287 tsk->msg_acc = 0;
1288 tsk->pkt_cnt = 0;
1289 }
1290
1291 if (!skb || tsk->cong_link_cnt)
1292 return;
1293
1294 /* Do not send SYN again after congestion */
1295 if (msg_is_syn(buf_msg(skb)))
1296 return;
1297
1298 if (tsk->msg_acc)
1299 tsk->pkt_cnt += skb_queue_len(txq);
1300 tsk->snt_unacked += tsk->snd_backlog;
1301 tsk->snd_backlog = 0;
1302 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1303 if (rc == -ELINKCONG)
1304 tsk->cong_link_cnt = 1;
1305}
1306
1307/**
1308 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1309 * @tsk: receiving socket
1310 * @skb: pointer to message buffer.
1311 */
1312static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1313 struct sk_buff_head *inputq,
1314 struct sk_buff_head *xmitq)
1315{
1316 struct tipc_msg *hdr = buf_msg(skb);
1317 u32 onode = tsk_own_node(tsk);
1318 struct sock *sk = &tsk->sk;
1319 int mtyp = msg_type(hdr);
1320 bool was_cong;
1321
1322 /* Ignore if connection cannot be validated: */
1323 if (!tsk_peer_msg(tsk, hdr)) {
1324 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1325 goto exit;
1326 }
1327
1328 if (unlikely(msg_errcode(hdr))) {
1329 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1330 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1331 tsk_peer_port(tsk));
1332 sk->sk_state_change(sk);
1333
1334 /* State change is ignored if socket already awake,
1335 * - convert msg to abort msg and add to inqueue
1336 */
1337 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1338 msg_set_type(hdr, TIPC_CONN_MSG);
1339 msg_set_size(hdr, BASIC_H_SIZE);
1340 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1341 __skb_queue_tail(inputq, skb);
1342 return;
1343 }
1344
1345 tsk->probe_unacked = false;
1346
1347 if (mtyp == CONN_PROBE) {
1348 msg_set_type(hdr, CONN_PROBE_REPLY);
1349 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1350 __skb_queue_tail(xmitq, skb);
1351 return;
1352 } else if (mtyp == CONN_ACK) {
1353 was_cong = tsk_conn_cong(tsk);
1354 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1355 tsk->snt_unacked -= msg_conn_ack(hdr);
1356 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1357 tsk->snd_win = msg_adv_win(hdr);
1358 if (was_cong && !tsk_conn_cong(tsk))
1359 sk->sk_write_space(sk);
1360 } else if (mtyp != CONN_PROBE_REPLY) {
1361 pr_warn("Received unknown CONN_PROTO msg\n");
1362 }
1363exit:
1364 kfree_skb(skb);
1365}
1366
1367/**
1368 * tipc_sendmsg - send message in connectionless manner
1369 * @sock: socket structure
1370 * @m: message to send
1371 * @dsz: amount of user data to be sent
1372 *
1373 * Message must have an destination specified explicitly.
1374 * Used for SOCK_RDM and SOCK_DGRAM messages,
1375 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1376 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1377 *
1378 * Returns the number of bytes sent on success, or errno otherwise
1379 */
1380static int tipc_sendmsg(struct socket *sock,
1381 struct msghdr *m, size_t dsz)
1382{
1383 struct sock *sk = sock->sk;
1384 int ret;
1385
1386 lock_sock(sk);
1387 ret = __tipc_sendmsg(sock, m, dsz);
1388 release_sock(sk);
1389
1390 return ret;
1391}
1392
1393static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1394{
1395 struct sock *sk = sock->sk;
1396 struct net *net = sock_net(sk);
1397 struct tipc_sock *tsk = tipc_sk(sk);
1398 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1399 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1400 struct list_head *clinks = &tsk->cong_links;
1401 bool syn = !tipc_sk_type_connectionless(sk);
1402 struct tipc_group *grp = tsk->group;
1403 struct tipc_msg *hdr = &tsk->phdr;
1404 struct tipc_name_seq *seq;
1405 struct sk_buff_head pkts;
1406 u32 dport = 0, dnode = 0;
1407 u32 type = 0, inst = 0;
1408 int mtu, rc;
1409
1410 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1411 return -EMSGSIZE;
1412
1413 if (likely(dest)) {
1414 if (unlikely(m->msg_namelen < sizeof(*dest)))
1415 return -EINVAL;
1416 if (unlikely(dest->family != AF_TIPC))
1417 return -EINVAL;
1418 }
1419
1420 if (grp) {
1421 if (!dest)
1422 return tipc_send_group_bcast(sock, m, dlen, timeout);
1423 if (dest->addrtype == TIPC_ADDR_NAME)
1424 return tipc_send_group_anycast(sock, m, dlen, timeout);
1425 if (dest->addrtype == TIPC_ADDR_ID)
1426 return tipc_send_group_unicast(sock, m, dlen, timeout);
1427 if (dest->addrtype == TIPC_ADDR_MCAST)
1428 return tipc_send_group_mcast(sock, m, dlen, timeout);
1429 return -EINVAL;
1430 }
1431
1432 if (unlikely(!dest)) {
1433 dest = &tsk->peer;
1434 if (!syn && dest->family != AF_TIPC)
1435 return -EDESTADDRREQ;
1436 }
1437
1438 if (unlikely(syn)) {
1439 if (sk->sk_state == TIPC_LISTEN)
1440 return -EPIPE;
1441 if (sk->sk_state != TIPC_OPEN)
1442 return -EISCONN;
1443 if (tsk->published)
1444 return -EOPNOTSUPP;
1445 if (dest->addrtype == TIPC_ADDR_NAME) {
1446 tsk->conn_type = dest->addr.name.name.type;
1447 tsk->conn_instance = dest->addr.name.name.instance;
1448 }
1449 msg_set_syn(hdr, 1);
1450 }
1451
1452 seq = &dest->addr.nameseq;
1453 if (dest->addrtype == TIPC_ADDR_MCAST)
1454 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1455
1456 if (dest->addrtype == TIPC_ADDR_NAME) {
1457 type = dest->addr.name.name.type;
1458 inst = dest->addr.name.name.instance;
1459 dnode = dest->addr.name.domain;
1460 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1461 if (unlikely(!dport && !dnode))
1462 return -EHOSTUNREACH;
1463 } else if (dest->addrtype == TIPC_ADDR_ID) {
1464 dnode = dest->addr.id.node;
1465 } else {
1466 return -EINVAL;
1467 }
1468
1469 /* Block or return if destination link is congested */
1470 rc = tipc_wait_for_cond(sock, &timeout,
1471 !tipc_dest_find(clinks, dnode, 0));
1472 if (unlikely(rc))
1473 return rc;
1474
1475 if (dest->addrtype == TIPC_ADDR_NAME) {
1476 msg_set_type(hdr, TIPC_NAMED_MSG);
1477 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1478 msg_set_nametype(hdr, type);
1479 msg_set_nameinst(hdr, inst);
1480 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1481 msg_set_destnode(hdr, dnode);
1482 msg_set_destport(hdr, dport);
1483 } else { /* TIPC_ADDR_ID */
1484 msg_set_type(hdr, TIPC_DIRECT_MSG);
1485 msg_set_lookup_scope(hdr, 0);
1486 msg_set_destnode(hdr, dnode);
1487 msg_set_destport(hdr, dest->addr.id.ref);
1488 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1489 }
1490
1491 __skb_queue_head_init(&pkts);
1492 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1493 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1494 if (unlikely(rc != dlen))
1495 return rc;
1496 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1497 __skb_queue_purge(&pkts);
1498 return -ENOMEM;
1499 }
1500
1501 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1502 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1503 if (unlikely(rc == -ELINKCONG)) {
1504 tipc_dest_push(clinks, dnode, 0);
1505 tsk->cong_link_cnt++;
1506 rc = 0;
1507 }
1508
1509 if (unlikely(syn && !rc))
1510 tipc_set_sk_state(sk, TIPC_CONNECTING);
1511
1512 return rc ? rc : dlen;
1513}
1514
1515/**
1516 * tipc_sendstream - send stream-oriented data
1517 * @sock: socket structure
1518 * @m: data to send
1519 * @dsz: total length of data to be transmitted
1520 *
1521 * Used for SOCK_STREAM data.
1522 *
1523 * Returns the number of bytes sent on success (or partial success),
1524 * or errno if no data sent
1525 */
1526static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1527{
1528 struct sock *sk = sock->sk;
1529 int ret;
1530
1531 lock_sock(sk);
1532 ret = __tipc_sendstream(sock, m, dsz);
1533 release_sock(sk);
1534
1535 return ret;
1536}
1537
1538static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1539{
1540 struct sock *sk = sock->sk;
1541 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1542 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1543 struct sk_buff_head *txq = &sk->sk_write_queue;
1544 struct tipc_sock *tsk = tipc_sk(sk);
1545 struct tipc_msg *hdr = &tsk->phdr;
1546 struct net *net = sock_net(sk);
1547 struct sk_buff *skb;
1548 u32 dnode = tsk_peer_node(tsk);
1549 int maxnagle = tsk->maxnagle;
1550 int maxpkt = tsk->max_pkt;
1551 int send, sent = 0;
1552 int blocks, rc = 0;
1553
1554 if (unlikely(dlen > INT_MAX))
1555 return -EMSGSIZE;
1556
1557 /* Handle implicit connection setup */
1558 if (unlikely(dest)) {
1559 rc = __tipc_sendmsg(sock, m, dlen);
1560 if (dlen && dlen == rc) {
1561 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1562 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1563 }
1564 return rc;
1565 }
1566
1567 do {
1568 rc = tipc_wait_for_cond(sock, &timeout,
1569 (!tsk->cong_link_cnt &&
1570 !tsk_conn_cong(tsk) &&
1571 tipc_sk_connected(sk)));
1572 if (unlikely(rc))
1573 break;
1574 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1575 blocks = tsk->snd_backlog;
1576 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1577 send <= maxnagle) {
1578 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1579 if (unlikely(rc < 0))
1580 break;
1581 blocks += rc;
1582 tsk->msg_acc++;
1583 if (blocks <= 64 && tsk->expect_ack) {
1584 tsk->snd_backlog = blocks;
1585 sent += send;
1586 break;
1587 } else if (blocks > 64) {
1588 tsk->pkt_cnt += skb_queue_len(txq);
1589 } else {
1590 skb = skb_peek_tail(txq);
1591 if (skb) {
1592 msg_set_ack_required(buf_msg(skb));
1593 tsk->expect_ack = true;
1594 } else {
1595 tsk->expect_ack = false;
1596 }
1597 tsk->msg_acc = 0;
1598 tsk->pkt_cnt = 0;
1599 }
1600 } else {
1601 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1602 if (unlikely(rc != send))
1603 break;
1604 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1605 }
1606 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1607 TIPC_DUMP_SK_SNDQ, " ");
1608 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1609 if (unlikely(rc == -ELINKCONG)) {
1610 tsk->cong_link_cnt = 1;
1611 rc = 0;
1612 }
1613 if (likely(!rc)) {
1614 tsk->snt_unacked += blocks;
1615 tsk->snd_backlog = 0;
1616 sent += send;
1617 }
1618 } while (sent < dlen && !rc);
1619
1620 return sent ? sent : rc;
1621}
1622
1623/**
1624 * tipc_send_packet - send a connection-oriented message
1625 * @sock: socket structure
1626 * @m: message to send
1627 * @dsz: length of data to be transmitted
1628 *
1629 * Used for SOCK_SEQPACKET messages.
1630 *
1631 * Returns the number of bytes sent on success, or errno otherwise
1632 */
1633static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1634{
1635 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1636 return -EMSGSIZE;
1637
1638 return tipc_sendstream(sock, m, dsz);
1639}
1640
1641/* tipc_sk_finish_conn - complete the setup of a connection
1642 */
1643static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1644 u32 peer_node)
1645{
1646 struct sock *sk = &tsk->sk;
1647 struct net *net = sock_net(sk);
1648 struct tipc_msg *msg = &tsk->phdr;
1649
1650 msg_set_syn(msg, 0);
1651 msg_set_destnode(msg, peer_node);
1652 msg_set_destport(msg, peer_port);
1653 msg_set_type(msg, TIPC_CONN_MSG);
1654 msg_set_lookup_scope(msg, 0);
1655 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1656
1657 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1658 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1659 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1660 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1661 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1662 tsk_set_nagle(tsk);
1663 __skb_queue_purge(&sk->sk_write_queue);
1664 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1665 return;
1666
1667 /* Fall back to message based flow control */
1668 tsk->rcv_win = FLOWCTL_MSG_WIN;
1669 tsk->snd_win = FLOWCTL_MSG_WIN;
1670}
1671
1672/**
1673 * tipc_sk_set_orig_addr - capture sender's address for received message
1674 * @m: descriptor for message info
1675 * @skb: received message
1676 *
1677 * Note: Address is not captured if not requested by receiver.
1678 */
1679static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1680{
1681 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1682 struct tipc_msg *hdr = buf_msg(skb);
1683
1684 if (!srcaddr)
1685 return;
1686
1687 srcaddr->sock.family = AF_TIPC;
1688 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1689 srcaddr->sock.scope = 0;
1690 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1691 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1692 srcaddr->sock.addr.name.domain = 0;
1693 m->msg_namelen = sizeof(struct sockaddr_tipc);
1694
1695 if (!msg_in_group(hdr))
1696 return;
1697
1698 /* Group message users may also want to know sending member's id */
1699 srcaddr->member.family = AF_TIPC;
1700 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1701 srcaddr->member.scope = 0;
1702 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1703 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1704 srcaddr->member.addr.name.domain = 0;
1705 m->msg_namelen = sizeof(*srcaddr);
1706}
1707
1708/**
1709 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1710 * @m: descriptor for message info
1711 * @skb: received message buffer
1712 * @tsk: TIPC port associated with message
1713 *
1714 * Note: Ancillary data is not captured if not requested by receiver.
1715 *
1716 * Returns 0 if successful, otherwise errno
1717 */
1718static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1719 struct tipc_sock *tsk)
1720{
1721 struct tipc_msg *msg;
1722 u32 anc_data[3];
1723 u32 err;
1724 u32 dest_type;
1725 int has_name;
1726 int res;
1727
1728 if (likely(m->msg_controllen == 0))
1729 return 0;
1730 msg = buf_msg(skb);
1731
1732 /* Optionally capture errored message object(s) */
1733 err = msg ? msg_errcode(msg) : 0;
1734 if (unlikely(err)) {
1735 anc_data[0] = err;
1736 anc_data[1] = msg_data_sz(msg);
1737 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1738 if (res)
1739 return res;
1740 if (anc_data[1]) {
1741 if (skb_linearize(skb))
1742 return -ENOMEM;
1743 msg = buf_msg(skb);
1744 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1745 msg_data(msg));
1746 if (res)
1747 return res;
1748 }
1749 }
1750
1751 /* Optionally capture message destination object */
1752 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1753 switch (dest_type) {
1754 case TIPC_NAMED_MSG:
1755 has_name = 1;
1756 anc_data[0] = msg_nametype(msg);
1757 anc_data[1] = msg_namelower(msg);
1758 anc_data[2] = msg_namelower(msg);
1759 break;
1760 case TIPC_MCAST_MSG:
1761 has_name = 1;
1762 anc_data[0] = msg_nametype(msg);
1763 anc_data[1] = msg_namelower(msg);
1764 anc_data[2] = msg_nameupper(msg);
1765 break;
1766 case TIPC_CONN_MSG:
1767 has_name = (tsk->conn_type != 0);
1768 anc_data[0] = tsk->conn_type;
1769 anc_data[1] = tsk->conn_instance;
1770 anc_data[2] = tsk->conn_instance;
1771 break;
1772 default:
1773 has_name = 0;
1774 }
1775 if (has_name) {
1776 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1777 if (res)
1778 return res;
1779 }
1780
1781 return 0;
1782}
1783
1784static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1785{
1786 struct sock *sk = &tsk->sk;
1787 struct sk_buff *skb = NULL;
1788 struct tipc_msg *msg;
1789 u32 peer_port = tsk_peer_port(tsk);
1790 u32 dnode = tsk_peer_node(tsk);
1791
1792 if (!tipc_sk_connected(sk))
1793 return NULL;
1794 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1795 dnode, tsk_own_node(tsk), peer_port,
1796 tsk->portid, TIPC_OK);
1797 if (!skb)
1798 return NULL;
1799 msg = buf_msg(skb);
1800 msg_set_conn_ack(msg, tsk->rcv_unacked);
1801 tsk->rcv_unacked = 0;
1802
1803 /* Adjust to and advertize the correct window limit */
1804 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1805 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1806 msg_set_adv_win(msg, tsk->rcv_win);
1807 }
1808 return skb;
1809}
1810
1811static void tipc_sk_send_ack(struct tipc_sock *tsk)
1812{
1813 struct sk_buff *skb;
1814
1815 skb = tipc_sk_build_ack(tsk);
1816 if (!skb)
1817 return;
1818
1819 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1820 msg_link_selector(buf_msg(skb)));
1821}
1822
1823static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1824{
1825 struct sock *sk = sock->sk;
1826 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1827 long timeo = *timeop;
1828 int err = sock_error(sk);
1829
1830 if (err)
1831 return err;
1832
1833 for (;;) {
1834 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1835 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1836 err = -ENOTCONN;
1837 break;
1838 }
1839 add_wait_queue(sk_sleep(sk), &wait);
1840 release_sock(sk);
1841 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1842 sched_annotate_sleep();
1843 lock_sock(sk);
1844 remove_wait_queue(sk_sleep(sk), &wait);
1845 }
1846 err = 0;
1847 if (!skb_queue_empty(&sk->sk_receive_queue))
1848 break;
1849 err = -EAGAIN;
1850 if (!timeo)
1851 break;
1852 err = sock_intr_errno(timeo);
1853 if (signal_pending(current))
1854 break;
1855
1856 err = sock_error(sk);
1857 if (err)
1858 break;
1859 }
1860 *timeop = timeo;
1861 return err;
1862}
1863
1864/**
1865 * tipc_recvmsg - receive packet-oriented message
1866 * @m: descriptor for message info
1867 * @buflen: length of user buffer area
1868 * @flags: receive flags
1869 *
1870 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1871 * If the complete message doesn't fit in user area, truncate it.
1872 *
1873 * Returns size of returned message data, errno otherwise
1874 */
1875static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1876 size_t buflen, int flags)
1877{
1878 struct sock *sk = sock->sk;
1879 bool connected = !tipc_sk_type_connectionless(sk);
1880 struct tipc_sock *tsk = tipc_sk(sk);
1881 int rc, err, hlen, dlen, copy;
1882 struct sk_buff_head xmitq;
1883 struct tipc_msg *hdr;
1884 struct sk_buff *skb;
1885 bool grp_evt;
1886 long timeout;
1887
1888 /* Catch invalid receive requests */
1889 if (unlikely(!buflen))
1890 return -EINVAL;
1891
1892 lock_sock(sk);
1893 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1894 rc = -ENOTCONN;
1895 goto exit;
1896 }
1897 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1898
1899 /* Step rcv queue to first msg with data or error; wait if necessary */
1900 do {
1901 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1902 if (unlikely(rc))
1903 goto exit;
1904 skb = skb_peek(&sk->sk_receive_queue);
1905 hdr = buf_msg(skb);
1906 dlen = msg_data_sz(hdr);
1907 hlen = msg_hdr_sz(hdr);
1908 err = msg_errcode(hdr);
1909 grp_evt = msg_is_grp_evt(hdr);
1910 if (likely(dlen || err))
1911 break;
1912 tsk_advance_rx_queue(sk);
1913 } while (1);
1914
1915 /* Collect msg meta data, including error code and rejected data */
1916 tipc_sk_set_orig_addr(m, skb);
1917 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1918 if (unlikely(rc))
1919 goto exit;
1920 hdr = buf_msg(skb);
1921
1922 /* Capture data if non-error msg, otherwise just set return value */
1923 if (likely(!err)) {
1924 copy = min_t(int, dlen, buflen);
1925 if (unlikely(copy != dlen))
1926 m->msg_flags |= MSG_TRUNC;
1927 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1928 } else {
1929 copy = 0;
1930 rc = 0;
1931 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1932 rc = -ECONNRESET;
1933 }
1934 if (unlikely(rc))
1935 goto exit;
1936
1937 /* Mark message as group event if applicable */
1938 if (unlikely(grp_evt)) {
1939 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1940 m->msg_flags |= MSG_EOR;
1941 m->msg_flags |= MSG_OOB;
1942 copy = 0;
1943 }
1944
1945 /* Caption of data or error code/rejected data was successful */
1946 if (unlikely(flags & MSG_PEEK))
1947 goto exit;
1948
1949 /* Send group flow control advertisement when applicable */
1950 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1951 __skb_queue_head_init(&xmitq);
1952 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1953 msg_orignode(hdr), msg_origport(hdr),
1954 &xmitq);
1955 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1956 }
1957
1958 tsk_advance_rx_queue(sk);
1959
1960 if (likely(!connected))
1961 goto exit;
1962
1963 /* Send connection flow control advertisement when applicable */
1964 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1965 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1966 tipc_sk_send_ack(tsk);
1967exit:
1968 release_sock(sk);
1969 return rc ? rc : copy;
1970}
1971
1972/**
1973 * tipc_recvstream - receive stream-oriented data
1974 * @m: descriptor for message info
1975 * @buflen: total size of user buffer area
1976 * @flags: receive flags
1977 *
1978 * Used for SOCK_STREAM messages only. If not enough data is available
1979 * will optionally wait for more; never truncates data.
1980 *
1981 * Returns size of returned message data, errno otherwise
1982 */
1983static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1984 size_t buflen, int flags)
1985{
1986 struct sock *sk = sock->sk;
1987 struct tipc_sock *tsk = tipc_sk(sk);
1988 struct sk_buff *skb;
1989 struct tipc_msg *hdr;
1990 struct tipc_skb_cb *skb_cb;
1991 bool peek = flags & MSG_PEEK;
1992 int offset, required, copy, copied = 0;
1993 int hlen, dlen, err, rc;
1994 long timeout;
1995
1996 /* Catch invalid receive attempts */
1997 if (unlikely(!buflen))
1998 return -EINVAL;
1999
2000 lock_sock(sk);
2001
2002 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2003 rc = -ENOTCONN;
2004 goto exit;
2005 }
2006 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2007 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2008
2009 do {
2010 /* Look at first msg in receive queue; wait if necessary */
2011 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2012 if (unlikely(rc))
2013 break;
2014 skb = skb_peek(&sk->sk_receive_queue);
2015 skb_cb = TIPC_SKB_CB(skb);
2016 hdr = buf_msg(skb);
2017 dlen = msg_data_sz(hdr);
2018 hlen = msg_hdr_sz(hdr);
2019 err = msg_errcode(hdr);
2020
2021 /* Discard any empty non-errored (SYN-) message */
2022 if (unlikely(!dlen && !err)) {
2023 tsk_advance_rx_queue(sk);
2024 continue;
2025 }
2026
2027 /* Collect msg meta data, incl. error code and rejected data */
2028 if (!copied) {
2029 tipc_sk_set_orig_addr(m, skb);
2030 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2031 if (rc)
2032 break;
2033 hdr = buf_msg(skb);
2034 }
2035
2036 /* Copy data if msg ok, otherwise return error/partial data */
2037 if (likely(!err)) {
2038 offset = skb_cb->bytes_read;
2039 copy = min_t(int, dlen - offset, buflen - copied);
2040 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2041 if (unlikely(rc))
2042 break;
2043 copied += copy;
2044 offset += copy;
2045 if (unlikely(offset < dlen)) {
2046 if (!peek)
2047 skb_cb->bytes_read = offset;
2048 break;
2049 }
2050 } else {
2051 rc = 0;
2052 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2053 rc = -ECONNRESET;
2054 if (copied || rc)
2055 break;
2056 }
2057
2058 if (unlikely(peek))
2059 break;
2060
2061 tsk_advance_rx_queue(sk);
2062
2063 /* Send connection flow control advertisement when applicable */
2064 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2065 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2066 tipc_sk_send_ack(tsk);
2067
2068 /* Exit if all requested data or FIN/error received */
2069 if (copied == buflen || err)
2070 break;
2071
2072 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2073exit:
2074 release_sock(sk);
2075 return copied ? copied : rc;
2076}
2077
2078/**
2079 * tipc_write_space - wake up thread if port congestion is released
2080 * @sk: socket
2081 */
2082static void tipc_write_space(struct sock *sk)
2083{
2084 struct socket_wq *wq;
2085
2086 rcu_read_lock();
2087 wq = rcu_dereference(sk->sk_wq);
2088 if (skwq_has_sleeper(wq))
2089 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2090 EPOLLWRNORM | EPOLLWRBAND);
2091 rcu_read_unlock();
2092}
2093
2094/**
2095 * tipc_data_ready - wake up threads to indicate messages have been received
2096 * @sk: socket
2097 */
2098static void tipc_data_ready(struct sock *sk)
2099{
2100 struct socket_wq *wq;
2101
2102 rcu_read_lock();
2103 wq = rcu_dereference(sk->sk_wq);
2104 if (skwq_has_sleeper(wq))
2105 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2106 EPOLLRDNORM | EPOLLRDBAND);
2107 rcu_read_unlock();
2108}
2109
2110static void tipc_sock_destruct(struct sock *sk)
2111{
2112 __skb_queue_purge(&sk->sk_receive_queue);
2113}
2114
2115static void tipc_sk_proto_rcv(struct sock *sk,
2116 struct sk_buff_head *inputq,
2117 struct sk_buff_head *xmitq)
2118{
2119 struct sk_buff *skb = __skb_dequeue(inputq);
2120 struct tipc_sock *tsk = tipc_sk(sk);
2121 struct tipc_msg *hdr = buf_msg(skb);
2122 struct tipc_group *grp = tsk->group;
2123 bool wakeup = false;
2124
2125 switch (msg_user(hdr)) {
2126 case CONN_MANAGER:
2127 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2128 return;
2129 case SOCK_WAKEUP:
2130 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2131 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2132 smp_wmb();
2133 tsk->cong_link_cnt--;
2134 wakeup = true;
2135 tipc_sk_push_backlog(tsk, false);
2136 break;
2137 case GROUP_PROTOCOL:
2138 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2139 break;
2140 case TOP_SRV:
2141 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2142 hdr, inputq, xmitq);
2143 break;
2144 default:
2145 break;
2146 }
2147
2148 if (wakeup)
2149 sk->sk_write_space(sk);
2150
2151 kfree_skb(skb);
2152}
2153
2154/**
2155 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2156 * @tsk: TIPC socket
2157 * @skb: pointer to message buffer.
2158 * @xmitq: for Nagle ACK if any
2159 * Returns true if message should be added to receive queue, false otherwise
2160 */
2161static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2162 struct sk_buff_head *xmitq)
2163{
2164 struct sock *sk = &tsk->sk;
2165 struct net *net = sock_net(sk);
2166 struct tipc_msg *hdr = buf_msg(skb);
2167 bool con_msg = msg_connected(hdr);
2168 u32 pport = tsk_peer_port(tsk);
2169 u32 pnode = tsk_peer_node(tsk);
2170 u32 oport = msg_origport(hdr);
2171 u32 onode = msg_orignode(hdr);
2172 int err = msg_errcode(hdr);
2173 unsigned long delay;
2174
2175 if (unlikely(msg_mcast(hdr)))
2176 return false;
2177 tsk->oneway = 0;
2178
2179 switch (sk->sk_state) {
2180 case TIPC_CONNECTING:
2181 /* Setup ACK */
2182 if (likely(con_msg)) {
2183 if (err)
2184 break;
2185 tipc_sk_finish_conn(tsk, oport, onode);
2186 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2187 /* ACK+ message with data is added to receive queue */
2188 if (msg_data_sz(hdr))
2189 return true;
2190 /* Empty ACK-, - wake up sleeping connect() and drop */
2191 sk->sk_state_change(sk);
2192 msg_set_dest_droppable(hdr, 1);
2193 return false;
2194 }
2195 /* Ignore connectionless message if not from listening socket */
2196 if (oport != pport || onode != pnode)
2197 return false;
2198
2199 /* Rejected SYN */
2200 if (err != TIPC_ERR_OVERLOAD)
2201 break;
2202
2203 /* Prepare for new setup attempt if we have a SYN clone */
2204 if (skb_queue_empty(&sk->sk_write_queue))
2205 break;
2206 get_random_bytes(&delay, 2);
2207 delay %= (tsk->conn_timeout / 4);
2208 delay = msecs_to_jiffies(delay + 100);
2209 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2210 return false;
2211 case TIPC_OPEN:
2212 case TIPC_DISCONNECTING:
2213 return false;
2214 case TIPC_LISTEN:
2215 /* Accept only SYN message */
2216 if (!msg_is_syn(hdr) &&
2217 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2218 return false;
2219 if (!con_msg && !err)
2220 return true;
2221 return false;
2222 case TIPC_ESTABLISHED:
2223 if (!skb_queue_empty(&sk->sk_write_queue))
2224 tipc_sk_push_backlog(tsk, false);
2225 /* Accept only connection-based messages sent by peer */
2226 if (likely(con_msg && !err && pport == oport &&
2227 pnode == onode)) {
2228 if (msg_ack_required(hdr)) {
2229 struct sk_buff *skb;
2230
2231 skb = tipc_sk_build_ack(tsk);
2232 if (skb) {
2233 msg_set_nagle_ack(buf_msg(skb));
2234 __skb_queue_tail(xmitq, skb);
2235 }
2236 }
2237 return true;
2238 }
2239 if (!tsk_peer_msg(tsk, hdr))
2240 return false;
2241 if (!err)
2242 return true;
2243 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2244 tipc_node_remove_conn(net, pnode, tsk->portid);
2245 sk->sk_state_change(sk);
2246 return true;
2247 default:
2248 pr_err("Unknown sk_state %u\n", sk->sk_state);
2249 }
2250 /* Abort connection setup attempt */
2251 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2252 sk->sk_err = ECONNREFUSED;
2253 sk->sk_state_change(sk);
2254 return true;
2255}
2256
2257/**
2258 * rcvbuf_limit - get proper overload limit of socket receive queue
2259 * @sk: socket
2260 * @skb: message
2261 *
2262 * For connection oriented messages, irrespective of importance,
2263 * default queue limit is 2 MB.
2264 *
2265 * For connectionless messages, queue limits are based on message
2266 * importance as follows:
2267 *
2268 * TIPC_LOW_IMPORTANCE (2 MB)
2269 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2270 * TIPC_HIGH_IMPORTANCE (8 MB)
2271 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2272 *
2273 * Returns overload limit according to corresponding message importance
2274 */
2275static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2276{
2277 struct tipc_sock *tsk = tipc_sk(sk);
2278 struct tipc_msg *hdr = buf_msg(skb);
2279
2280 if (unlikely(msg_in_group(hdr)))
2281 return READ_ONCE(sk->sk_rcvbuf);
2282
2283 if (unlikely(!msg_connected(hdr)))
2284 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2285
2286 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2287 return READ_ONCE(sk->sk_rcvbuf);
2288
2289 return FLOWCTL_MSG_LIM;
2290}
2291
2292/**
2293 * tipc_sk_filter_rcv - validate incoming message
2294 * @sk: socket
2295 * @skb: pointer to message.
2296 *
2297 * Enqueues message on receive queue if acceptable; optionally handles
2298 * disconnect indication for a connected socket.
2299 *
2300 * Called with socket lock already taken
2301 *
2302 */
2303static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2304 struct sk_buff_head *xmitq)
2305{
2306 bool sk_conn = !tipc_sk_type_connectionless(sk);
2307 struct tipc_sock *tsk = tipc_sk(sk);
2308 struct tipc_group *grp = tsk->group;
2309 struct tipc_msg *hdr = buf_msg(skb);
2310 struct net *net = sock_net(sk);
2311 struct sk_buff_head inputq;
2312 int mtyp = msg_type(hdr);
2313 int limit, err = TIPC_OK;
2314
2315 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2316 TIPC_SKB_CB(skb)->bytes_read = 0;
2317 __skb_queue_head_init(&inputq);
2318 __skb_queue_tail(&inputq, skb);
2319
2320 if (unlikely(!msg_isdata(hdr)))
2321 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2322
2323 if (unlikely(grp))
2324 tipc_group_filter_msg(grp, &inputq, xmitq);
2325
2326 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2327 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2328
2329 /* Validate and add to receive buffer if there is space */
2330 while ((skb = __skb_dequeue(&inputq))) {
2331 hdr = buf_msg(skb);
2332 limit = rcvbuf_limit(sk, skb);
2333 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2334 (!sk_conn && msg_connected(hdr)) ||
2335 (!grp && msg_in_group(hdr)))
2336 err = TIPC_ERR_NO_PORT;
2337 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2338 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2339 "err_overload2!");
2340 atomic_inc(&sk->sk_drops);
2341 err = TIPC_ERR_OVERLOAD;
2342 }
2343
2344 if (unlikely(err)) {
2345 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2346 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2347 "@filter_rcv!");
2348 __skb_queue_tail(xmitq, skb);
2349 }
2350 err = TIPC_OK;
2351 continue;
2352 }
2353 __skb_queue_tail(&sk->sk_receive_queue, skb);
2354 skb_set_owner_r(skb, sk);
2355 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2356 "rcvq >90% allocated!");
2357 sk->sk_data_ready(sk);
2358 }
2359}
2360
2361/**
2362 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2363 * @sk: socket
2364 * @skb: message
2365 *
2366 * Caller must hold socket lock
2367 */
2368static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2369{
2370 unsigned int before = sk_rmem_alloc_get(sk);
2371 struct sk_buff_head xmitq;
2372 unsigned int added;
2373
2374 __skb_queue_head_init(&xmitq);
2375
2376 tipc_sk_filter_rcv(sk, skb, &xmitq);
2377 added = sk_rmem_alloc_get(sk) - before;
2378 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2379
2380 /* Send pending response/rejected messages, if any */
2381 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2382 return 0;
2383}
2384
2385/**
2386 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2387 * inputq and try adding them to socket or backlog queue
2388 * @inputq: list of incoming buffers with potentially different destinations
2389 * @sk: socket where the buffers should be enqueued
2390 * @dport: port number for the socket
2391 *
2392 * Caller must hold socket lock
2393 */
2394static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2395 u32 dport, struct sk_buff_head *xmitq)
2396{
2397 unsigned long time_limit = jiffies + 2;
2398 struct sk_buff *skb;
2399 unsigned int lim;
2400 atomic_t *dcnt;
2401 u32 onode;
2402
2403 while (skb_queue_len(inputq)) {
2404 if (unlikely(time_after_eq(jiffies, time_limit)))
2405 return;
2406
2407 skb = tipc_skb_dequeue(inputq, dport);
2408 if (unlikely(!skb))
2409 return;
2410
2411 /* Add message directly to receive queue if possible */
2412 if (!sock_owned_by_user(sk)) {
2413 tipc_sk_filter_rcv(sk, skb, xmitq);
2414 continue;
2415 }
2416
2417 /* Try backlog, compensating for double-counted bytes */
2418 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2419 if (!sk->sk_backlog.len)
2420 atomic_set(dcnt, 0);
2421 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2422 if (likely(!sk_add_backlog(sk, skb, lim))) {
2423 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2424 "bklg & rcvq >90% allocated!");
2425 continue;
2426 }
2427
2428 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2429 /* Overload => reject message back to sender */
2430 onode = tipc_own_addr(sock_net(sk));
2431 atomic_inc(&sk->sk_drops);
2432 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2433 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2434 "@sk_enqueue!");
2435 __skb_queue_tail(xmitq, skb);
2436 }
2437 break;
2438 }
2439}
2440
2441/**
2442 * tipc_sk_rcv - handle a chain of incoming buffers
2443 * @inputq: buffer list containing the buffers
2444 * Consumes all buffers in list until inputq is empty
2445 * Note: may be called in multiple threads referring to the same queue
2446 */
2447void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2448{
2449 struct sk_buff_head xmitq;
2450 u32 dnode, dport = 0;
2451 int err;
2452 struct tipc_sock *tsk;
2453 struct sock *sk;
2454 struct sk_buff *skb;
2455
2456 __skb_queue_head_init(&xmitq);
2457 while (skb_queue_len(inputq)) {
2458 dport = tipc_skb_peek_port(inputq, dport);
2459 tsk = tipc_sk_lookup(net, dport);
2460
2461 if (likely(tsk)) {
2462 sk = &tsk->sk;
2463 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2464 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2465 spin_unlock_bh(&sk->sk_lock.slock);
2466 }
2467 /* Send pending response/rejected messages, if any */
2468 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2469 sock_put(sk);
2470 continue;
2471 }
2472 /* No destination socket => dequeue skb if still there */
2473 skb = tipc_skb_dequeue(inputq, dport);
2474 if (!skb)
2475 return;
2476
2477 /* Try secondary lookup if unresolved named message */
2478 err = TIPC_ERR_NO_PORT;
2479 if (tipc_msg_lookup_dest(net, skb, &err))
2480 goto xmit;
2481
2482 /* Prepare for message rejection */
2483 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2484 continue;
2485
2486 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2487xmit:
2488 dnode = msg_destnode(buf_msg(skb));
2489 tipc_node_xmit_skb(net, skb, dnode, dport);
2490 }
2491}
2492
2493static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2494{
2495 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2496 struct sock *sk = sock->sk;
2497 int done;
2498
2499 do {
2500 int err = sock_error(sk);
2501 if (err)
2502 return err;
2503 if (!*timeo_p)
2504 return -ETIMEDOUT;
2505 if (signal_pending(current))
2506 return sock_intr_errno(*timeo_p);
2507 if (sk->sk_state == TIPC_DISCONNECTING)
2508 break;
2509
2510 add_wait_queue(sk_sleep(sk), &wait);
2511 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2512 &wait);
2513 remove_wait_queue(sk_sleep(sk), &wait);
2514 } while (!done);
2515 return 0;
2516}
2517
2518static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2519{
2520 if (addr->family != AF_TIPC)
2521 return false;
2522 if (addr->addrtype == TIPC_SERVICE_RANGE)
2523 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2524 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2525 addr->addrtype == TIPC_SOCKET_ADDR);
2526}
2527
2528/**
2529 * tipc_connect - establish a connection to another TIPC port
2530 * @sock: socket structure
2531 * @dest: socket address for destination port
2532 * @destlen: size of socket address data structure
2533 * @flags: file-related flags associated with socket
2534 *
2535 * Returns 0 on success, errno otherwise
2536 */
2537static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2538 int destlen, int flags)
2539{
2540 struct sock *sk = sock->sk;
2541 struct tipc_sock *tsk = tipc_sk(sk);
2542 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2543 struct msghdr m = {NULL,};
2544 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2545 int previous;
2546 int res = 0;
2547
2548 if (destlen != sizeof(struct sockaddr_tipc))
2549 return -EINVAL;
2550
2551 lock_sock(sk);
2552
2553 if (tsk->group) {
2554 res = -EINVAL;
2555 goto exit;
2556 }
2557
2558 if (dst->family == AF_UNSPEC) {
2559 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2560 if (!tipc_sk_type_connectionless(sk))
2561 res = -EINVAL;
2562 goto exit;
2563 }
2564 if (!tipc_sockaddr_is_sane(dst)) {
2565 res = -EINVAL;
2566 goto exit;
2567 }
2568 /* DGRAM/RDM connect(), just save the destaddr */
2569 if (tipc_sk_type_connectionless(sk)) {
2570 memcpy(&tsk->peer, dest, destlen);
2571 goto exit;
2572 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2573 res = -EINVAL;
2574 goto exit;
2575 }
2576
2577 previous = sk->sk_state;
2578
2579 switch (sk->sk_state) {
2580 case TIPC_OPEN:
2581 /* Send a 'SYN-' to destination */
2582 m.msg_name = dest;
2583 m.msg_namelen = destlen;
2584
2585 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2586 * indicate send_msg() is never blocked.
2587 */
2588 if (!timeout)
2589 m.msg_flags = MSG_DONTWAIT;
2590
2591 res = __tipc_sendmsg(sock, &m, 0);
2592 if ((res < 0) && (res != -EWOULDBLOCK))
2593 goto exit;
2594
2595 /* Just entered TIPC_CONNECTING state; the only
2596 * difference is that return value in non-blocking
2597 * case is EINPROGRESS, rather than EALREADY.
2598 */
2599 res = -EINPROGRESS;
2600 fallthrough;
2601 case TIPC_CONNECTING:
2602 if (!timeout) {
2603 if (previous == TIPC_CONNECTING)
2604 res = -EALREADY;
2605 goto exit;
2606 }
2607 timeout = msecs_to_jiffies(timeout);
2608 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2609 res = tipc_wait_for_connect(sock, &timeout);
2610 break;
2611 case TIPC_ESTABLISHED:
2612 res = -EISCONN;
2613 break;
2614 default:
2615 res = -EINVAL;
2616 }
2617
2618exit:
2619 release_sock(sk);
2620 return res;
2621}
2622
2623/**
2624 * tipc_listen - allow socket to listen for incoming connections
2625 * @sock: socket structure
2626 * @len: (unused)
2627 *
2628 * Returns 0 on success, errno otherwise
2629 */
2630static int tipc_listen(struct socket *sock, int len)
2631{
2632 struct sock *sk = sock->sk;
2633 int res;
2634
2635 lock_sock(sk);
2636 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2637 release_sock(sk);
2638
2639 return res;
2640}
2641
2642static int tipc_wait_for_accept(struct socket *sock, long timeo)
2643{
2644 struct sock *sk = sock->sk;
2645 DEFINE_WAIT(wait);
2646 int err;
2647
2648 /* True wake-one mechanism for incoming connections: only
2649 * one process gets woken up, not the 'whole herd'.
2650 * Since we do not 'race & poll' for established sockets
2651 * anymore, the common case will execute the loop only once.
2652 */
2653 for (;;) {
2654 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2655 TASK_INTERRUPTIBLE);
2656 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2657 release_sock(sk);
2658 timeo = schedule_timeout(timeo);
2659 lock_sock(sk);
2660 }
2661 err = 0;
2662 if (!skb_queue_empty(&sk->sk_receive_queue))
2663 break;
2664 err = -EAGAIN;
2665 if (!timeo)
2666 break;
2667 err = sock_intr_errno(timeo);
2668 if (signal_pending(current))
2669 break;
2670 }
2671 finish_wait(sk_sleep(sk), &wait);
2672 return err;
2673}
2674
2675/**
2676 * tipc_accept - wait for connection request
2677 * @sock: listening socket
2678 * @new_sock: new socket that is to be connected
2679 * @flags: file-related flags associated with socket
2680 *
2681 * Returns 0 on success, errno otherwise
2682 */
2683static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2684 bool kern)
2685{
2686 struct sock *new_sk, *sk = sock->sk;
2687 struct sk_buff *buf;
2688 struct tipc_sock *new_tsock;
2689 struct tipc_msg *msg;
2690 long timeo;
2691 int res;
2692
2693 lock_sock(sk);
2694
2695 if (sk->sk_state != TIPC_LISTEN) {
2696 res = -EINVAL;
2697 goto exit;
2698 }
2699 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2700 res = tipc_wait_for_accept(sock, timeo);
2701 if (res)
2702 goto exit;
2703
2704 buf = skb_peek(&sk->sk_receive_queue);
2705
2706 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2707 if (res)
2708 goto exit;
2709 security_sk_clone(sock->sk, new_sock->sk);
2710
2711 new_sk = new_sock->sk;
2712 new_tsock = tipc_sk(new_sk);
2713 msg = buf_msg(buf);
2714
2715 /* we lock on new_sk; but lockdep sees the lock on sk */
2716 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2717
2718 /*
2719 * Reject any stray messages received by new socket
2720 * before the socket lock was taken (very, very unlikely)
2721 */
2722 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2723
2724 /* Connect new socket to it's peer */
2725 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2726
2727 tsk_set_importance(new_sk, msg_importance(msg));
2728 if (msg_named(msg)) {
2729 new_tsock->conn_type = msg_nametype(msg);
2730 new_tsock->conn_instance = msg_nameinst(msg);
2731 }
2732
2733 /*
2734 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2735 * Respond to 'SYN+' by queuing it on new socket.
2736 */
2737 if (!msg_data_sz(msg)) {
2738 struct msghdr m = {NULL,};
2739
2740 tsk_advance_rx_queue(sk);
2741 __tipc_sendstream(new_sock, &m, 0);
2742 } else {
2743 __skb_dequeue(&sk->sk_receive_queue);
2744 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2745 skb_set_owner_r(buf, new_sk);
2746 }
2747 release_sock(new_sk);
2748exit:
2749 release_sock(sk);
2750 return res;
2751}
2752
2753/**
2754 * tipc_shutdown - shutdown socket connection
2755 * @sock: socket structure
2756 * @how: direction to close (must be SHUT_RDWR)
2757 *
2758 * Terminates connection (if necessary), then purges socket's receive queue.
2759 *
2760 * Returns 0 on success, errno otherwise
2761 */
2762static int tipc_shutdown(struct socket *sock, int how)
2763{
2764 struct sock *sk = sock->sk;
2765 int res;
2766
2767 if (how != SHUT_RDWR)
2768 return -EINVAL;
2769
2770 lock_sock(sk);
2771
2772 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2773 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2774 sk->sk_shutdown = SHUTDOWN_MASK;
2775
2776 if (sk->sk_state == TIPC_DISCONNECTING) {
2777 /* Discard any unreceived messages */
2778 __skb_queue_purge(&sk->sk_receive_queue);
2779
2780 res = 0;
2781 } else {
2782 res = -ENOTCONN;
2783 }
2784 /* Wake up anyone sleeping in poll. */
2785 sk->sk_state_change(sk);
2786
2787 release_sock(sk);
2788 return res;
2789}
2790
2791static void tipc_sk_check_probing_state(struct sock *sk,
2792 struct sk_buff_head *list)
2793{
2794 struct tipc_sock *tsk = tipc_sk(sk);
2795 u32 pnode = tsk_peer_node(tsk);
2796 u32 pport = tsk_peer_port(tsk);
2797 u32 self = tsk_own_node(tsk);
2798 u32 oport = tsk->portid;
2799 struct sk_buff *skb;
2800
2801 if (tsk->probe_unacked) {
2802 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2803 sk->sk_err = ECONNABORTED;
2804 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2805 sk->sk_state_change(sk);
2806 return;
2807 }
2808 /* Prepare new probe */
2809 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2810 pnode, self, pport, oport, TIPC_OK);
2811 if (skb)
2812 __skb_queue_tail(list, skb);
2813 tsk->probe_unacked = true;
2814 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2815}
2816
2817static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2818{
2819 struct tipc_sock *tsk = tipc_sk(sk);
2820
2821 /* Try again later if dest link is congested */
2822 if (tsk->cong_link_cnt) {
2823 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2824 return;
2825 }
2826 /* Prepare SYN for retransmit */
2827 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2828}
2829
2830static void tipc_sk_timeout(struct timer_list *t)
2831{
2832 struct sock *sk = from_timer(sk, t, sk_timer);
2833 struct tipc_sock *tsk = tipc_sk(sk);
2834 u32 pnode = tsk_peer_node(tsk);
2835 struct sk_buff_head list;
2836 int rc = 0;
2837
2838 __skb_queue_head_init(&list);
2839 bh_lock_sock(sk);
2840
2841 /* Try again later if socket is busy */
2842 if (sock_owned_by_user(sk)) {
2843 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2844 bh_unlock_sock(sk);
2845 sock_put(sk);
2846 return;
2847 }
2848
2849 if (sk->sk_state == TIPC_ESTABLISHED)
2850 tipc_sk_check_probing_state(sk, &list);
2851 else if (sk->sk_state == TIPC_CONNECTING)
2852 tipc_sk_retry_connect(sk, &list);
2853
2854 bh_unlock_sock(sk);
2855
2856 if (!skb_queue_empty(&list))
2857 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2858
2859 /* SYN messages may cause link congestion */
2860 if (rc == -ELINKCONG) {
2861 tipc_dest_push(&tsk->cong_links, pnode, 0);
2862 tsk->cong_link_cnt = 1;
2863 }
2864 sock_put(sk);
2865}
2866
2867static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2868 struct tipc_name_seq const *seq)
2869{
2870 struct sock *sk = &tsk->sk;
2871 struct net *net = sock_net(sk);
2872 struct publication *publ;
2873 u32 key;
2874
2875 if (scope != TIPC_NODE_SCOPE)
2876 scope = TIPC_CLUSTER_SCOPE;
2877
2878 if (tipc_sk_connected(sk))
2879 return -EINVAL;
2880 key = tsk->portid + tsk->pub_count + 1;
2881 if (key == tsk->portid)
2882 return -EADDRINUSE;
2883
2884 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2885 scope, tsk->portid, key);
2886 if (unlikely(!publ))
2887 return -EINVAL;
2888
2889 list_add(&publ->binding_sock, &tsk->publications);
2890 tsk->pub_count++;
2891 tsk->published = 1;
2892 return 0;
2893}
2894
2895static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2896 struct tipc_name_seq const *seq)
2897{
2898 struct net *net = sock_net(&tsk->sk);
2899 struct publication *publ;
2900 struct publication *safe;
2901 int rc = -EINVAL;
2902
2903 if (scope != TIPC_NODE_SCOPE)
2904 scope = TIPC_CLUSTER_SCOPE;
2905
2906 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2907 if (seq) {
2908 if (publ->scope != scope)
2909 continue;
2910 if (publ->type != seq->type)
2911 continue;
2912 if (publ->lower != seq->lower)
2913 continue;
2914 if (publ->upper != seq->upper)
2915 break;
2916 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2917 publ->upper, publ->key);
2918 rc = 0;
2919 break;
2920 }
2921 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2922 publ->upper, publ->key);
2923 rc = 0;
2924 }
2925 if (list_empty(&tsk->publications))
2926 tsk->published = 0;
2927 return rc;
2928}
2929
2930/* tipc_sk_reinit: set non-zero address in all existing sockets
2931 * when we go from standalone to network mode.
2932 */
2933void tipc_sk_reinit(struct net *net)
2934{
2935 struct tipc_net *tn = net_generic(net, tipc_net_id);
2936 struct rhashtable_iter iter;
2937 struct tipc_sock *tsk;
2938 struct tipc_msg *msg;
2939
2940 rhashtable_walk_enter(&tn->sk_rht, &iter);
2941
2942 do {
2943 rhashtable_walk_start(&iter);
2944
2945 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2946 sock_hold(&tsk->sk);
2947 rhashtable_walk_stop(&iter);
2948 lock_sock(&tsk->sk);
2949 msg = &tsk->phdr;
2950 msg_set_prevnode(msg, tipc_own_addr(net));
2951 msg_set_orignode(msg, tipc_own_addr(net));
2952 release_sock(&tsk->sk);
2953 rhashtable_walk_start(&iter);
2954 sock_put(&tsk->sk);
2955 }
2956
2957 rhashtable_walk_stop(&iter);
2958 } while (tsk == ERR_PTR(-EAGAIN));
2959
2960 rhashtable_walk_exit(&iter);
2961}
2962
2963static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2964{
2965 struct tipc_net *tn = net_generic(net, tipc_net_id);
2966 struct tipc_sock *tsk;
2967
2968 rcu_read_lock();
2969 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2970 if (tsk)
2971 sock_hold(&tsk->sk);
2972 rcu_read_unlock();
2973
2974 return tsk;
2975}
2976
2977static int tipc_sk_insert(struct tipc_sock *tsk)
2978{
2979 struct sock *sk = &tsk->sk;
2980 struct net *net = sock_net(sk);
2981 struct tipc_net *tn = net_generic(net, tipc_net_id);
2982 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2983 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2984
2985 while (remaining--) {
2986 portid++;
2987 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2988 portid = TIPC_MIN_PORT;
2989 tsk->portid = portid;
2990 sock_hold(&tsk->sk);
2991 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2992 tsk_rht_params))
2993 return 0;
2994 sock_put(&tsk->sk);
2995 }
2996
2997 return -1;
2998}
2999
3000static void tipc_sk_remove(struct tipc_sock *tsk)
3001{
3002 struct sock *sk = &tsk->sk;
3003 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3004
3005 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3006 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3007 __sock_put(sk);
3008 }
3009}
3010
3011static const struct rhashtable_params tsk_rht_params = {
3012 .nelem_hint = 192,
3013 .head_offset = offsetof(struct tipc_sock, node),
3014 .key_offset = offsetof(struct tipc_sock, portid),
3015 .key_len = sizeof(u32), /* portid */
3016 .max_size = 1048576,
3017 .min_size = 256,
3018 .automatic_shrinking = true,
3019};
3020
3021int tipc_sk_rht_init(struct net *net)
3022{
3023 struct tipc_net *tn = net_generic(net, tipc_net_id);
3024
3025 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3026}
3027
3028void tipc_sk_rht_destroy(struct net *net)
3029{
3030 struct tipc_net *tn = net_generic(net, tipc_net_id);
3031
3032 /* Wait for socket readers to complete */
3033 synchronize_net();
3034
3035 rhashtable_destroy(&tn->sk_rht);
3036}
3037
3038static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3039{
3040 struct net *net = sock_net(&tsk->sk);
3041 struct tipc_group *grp = tsk->group;
3042 struct tipc_msg *hdr = &tsk->phdr;
3043 struct tipc_name_seq seq;
3044 int rc;
3045
3046 if (mreq->type < TIPC_RESERVED_TYPES)
3047 return -EACCES;
3048 if (mreq->scope > TIPC_NODE_SCOPE)
3049 return -EINVAL;
3050 if (grp)
3051 return -EACCES;
3052 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3053 if (!grp)
3054 return -ENOMEM;
3055 tsk->group = grp;
3056 msg_set_lookup_scope(hdr, mreq->scope);
3057 msg_set_nametype(hdr, mreq->type);
3058 msg_set_dest_droppable(hdr, true);
3059 seq.type = mreq->type;
3060 seq.lower = mreq->instance;
3061 seq.upper = seq.lower;
3062 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3063 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3064 if (rc) {
3065 tipc_group_delete(net, grp);
3066 tsk->group = NULL;
3067 return rc;
3068 }
3069 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3070 tsk->mc_method.rcast = true;
3071 tsk->mc_method.mandatory = true;
3072 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3073 return rc;
3074}
3075
3076static int tipc_sk_leave(struct tipc_sock *tsk)
3077{
3078 struct net *net = sock_net(&tsk->sk);
3079 struct tipc_group *grp = tsk->group;
3080 struct tipc_name_seq seq;
3081 int scope;
3082
3083 if (!grp)
3084 return -EINVAL;
3085 tipc_group_self(grp, &seq, &scope);
3086 tipc_group_delete(net, grp);
3087 tsk->group = NULL;
3088 tipc_sk_withdraw(tsk, scope, &seq);
3089 return 0;
3090}
3091
3092/**
3093 * tipc_setsockopt - set socket option
3094 * @sock: socket structure
3095 * @lvl: option level
3096 * @opt: option identifier
3097 * @ov: pointer to new option value
3098 * @ol: length of option value
3099 *
3100 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3101 * (to ease compatibility).
3102 *
3103 * Returns 0 on success, errno otherwise
3104 */
3105static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3106 sockptr_t ov, unsigned int ol)
3107{
3108 struct sock *sk = sock->sk;
3109 struct tipc_sock *tsk = tipc_sk(sk);
3110 struct tipc_group_req mreq;
3111 u32 value = 0;
3112 int res = 0;
3113
3114 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3115 return 0;
3116 if (lvl != SOL_TIPC)
3117 return -ENOPROTOOPT;
3118
3119 switch (opt) {
3120 case TIPC_IMPORTANCE:
3121 case TIPC_SRC_DROPPABLE:
3122 case TIPC_DEST_DROPPABLE:
3123 case TIPC_CONN_TIMEOUT:
3124 case TIPC_NODELAY:
3125 if (ol < sizeof(value))
3126 return -EINVAL;
3127 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3128 return -EFAULT;
3129 break;
3130 case TIPC_GROUP_JOIN:
3131 if (ol < sizeof(mreq))
3132 return -EINVAL;
3133 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3134 return -EFAULT;
3135 break;
3136 default:
3137 if (!sockptr_is_null(ov) || ol)
3138 return -EINVAL;
3139 }
3140
3141 lock_sock(sk);
3142
3143 switch (opt) {
3144 case TIPC_IMPORTANCE:
3145 res = tsk_set_importance(sk, value);
3146 break;
3147 case TIPC_SRC_DROPPABLE:
3148 if (sock->type != SOCK_STREAM)
3149 tsk_set_unreliable(tsk, value);
3150 else
3151 res = -ENOPROTOOPT;
3152 break;
3153 case TIPC_DEST_DROPPABLE:
3154 tsk_set_unreturnable(tsk, value);
3155 break;
3156 case TIPC_CONN_TIMEOUT:
3157 tipc_sk(sk)->conn_timeout = value;
3158 break;
3159 case TIPC_MCAST_BROADCAST:
3160 tsk->mc_method.rcast = false;
3161 tsk->mc_method.mandatory = true;
3162 break;
3163 case TIPC_MCAST_REPLICAST:
3164 tsk->mc_method.rcast = true;
3165 tsk->mc_method.mandatory = true;
3166 break;
3167 case TIPC_GROUP_JOIN:
3168 res = tipc_sk_join(tsk, &mreq);
3169 break;
3170 case TIPC_GROUP_LEAVE:
3171 res = tipc_sk_leave(tsk);
3172 break;
3173 case TIPC_NODELAY:
3174 tsk->nodelay = !!value;
3175 tsk_set_nagle(tsk);
3176 break;
3177 default:
3178 res = -EINVAL;
3179 }
3180
3181 release_sock(sk);
3182
3183 return res;
3184}
3185
3186/**
3187 * tipc_getsockopt - get socket option
3188 * @sock: socket structure
3189 * @lvl: option level
3190 * @opt: option identifier
3191 * @ov: receptacle for option value
3192 * @ol: receptacle for length of option value
3193 *
3194 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3195 * (to ease compatibility).
3196 *
3197 * Returns 0 on success, errno otherwise
3198 */
3199static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3200 char __user *ov, int __user *ol)
3201{
3202 struct sock *sk = sock->sk;
3203 struct tipc_sock *tsk = tipc_sk(sk);
3204 struct tipc_name_seq seq;
3205 int len, scope;
3206 u32 value;
3207 int res;
3208
3209 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3210 return put_user(0, ol);
3211 if (lvl != SOL_TIPC)
3212 return -ENOPROTOOPT;
3213 res = get_user(len, ol);
3214 if (res)
3215 return res;
3216
3217 lock_sock(sk);
3218
3219 switch (opt) {
3220 case TIPC_IMPORTANCE:
3221 value = tsk_importance(tsk);
3222 break;
3223 case TIPC_SRC_DROPPABLE:
3224 value = tsk_unreliable(tsk);
3225 break;
3226 case TIPC_DEST_DROPPABLE:
3227 value = tsk_unreturnable(tsk);
3228 break;
3229 case TIPC_CONN_TIMEOUT:
3230 value = tsk->conn_timeout;
3231 /* no need to set "res", since already 0 at this point */
3232 break;
3233 case TIPC_NODE_RECVQ_DEPTH:
3234 value = 0; /* was tipc_queue_size, now obsolete */
3235 break;
3236 case TIPC_SOCK_RECVQ_DEPTH:
3237 value = skb_queue_len(&sk->sk_receive_queue);
3238 break;
3239 case TIPC_SOCK_RECVQ_USED:
3240 value = sk_rmem_alloc_get(sk);
3241 break;
3242 case TIPC_GROUP_JOIN:
3243 seq.type = 0;
3244 if (tsk->group)
3245 tipc_group_self(tsk->group, &seq, &scope);
3246 value = seq.type;
3247 break;
3248 default:
3249 res = -EINVAL;
3250 }
3251
3252 release_sock(sk);
3253
3254 if (res)
3255 return res; /* "get" failed */
3256
3257 if (len < sizeof(value))
3258 return -EINVAL;
3259
3260 if (copy_to_user(ov, &value, sizeof(value)))
3261 return -EFAULT;
3262
3263 return put_user(sizeof(value), ol);
3264}
3265
3266static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3267{
3268 struct net *net = sock_net(sock->sk);
3269 struct tipc_sioc_nodeid_req nr = {0};
3270 struct tipc_sioc_ln_req lnr;
3271 void __user *argp = (void __user *)arg;
3272
3273 switch (cmd) {
3274 case SIOCGETLINKNAME:
3275 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3276 return -EFAULT;
3277 if (!tipc_node_get_linkname(net,
3278 lnr.bearer_id & 0xffff, lnr.peer,
3279 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3280 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3281 return -EFAULT;
3282 return 0;
3283 }
3284 return -EADDRNOTAVAIL;
3285 case SIOCGETNODEID:
3286 if (copy_from_user(&nr, argp, sizeof(nr)))
3287 return -EFAULT;
3288 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3289 return -EADDRNOTAVAIL;
3290 if (copy_to_user(argp, &nr, sizeof(nr)))
3291 return -EFAULT;
3292 return 0;
3293 default:
3294 return -ENOIOCTLCMD;
3295 }
3296}
3297
3298static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3299{
3300 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3301 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3302 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3303
3304 tsk1->peer.family = AF_TIPC;
3305 tsk1->peer.addrtype = TIPC_ADDR_ID;
3306 tsk1->peer.scope = TIPC_NODE_SCOPE;
3307 tsk1->peer.addr.id.ref = tsk2->portid;
3308 tsk1->peer.addr.id.node = onode;
3309 tsk2->peer.family = AF_TIPC;
3310 tsk2->peer.addrtype = TIPC_ADDR_ID;
3311 tsk2->peer.scope = TIPC_NODE_SCOPE;
3312 tsk2->peer.addr.id.ref = tsk1->portid;
3313 tsk2->peer.addr.id.node = onode;
3314
3315 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3316 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3317 return 0;
3318}
3319
3320/* Protocol switches for the various types of TIPC sockets */
3321
3322static const struct proto_ops msg_ops = {
3323 .owner = THIS_MODULE,
3324 .family = AF_TIPC,
3325 .release = tipc_release,
3326 .bind = tipc_bind,
3327 .connect = tipc_connect,
3328 .socketpair = tipc_socketpair,
3329 .accept = sock_no_accept,
3330 .getname = tipc_getname,
3331 .poll = tipc_poll,
3332 .ioctl = tipc_ioctl,
3333 .listen = sock_no_listen,
3334 .shutdown = tipc_shutdown,
3335 .setsockopt = tipc_setsockopt,
3336 .getsockopt = tipc_getsockopt,
3337 .sendmsg = tipc_sendmsg,
3338 .recvmsg = tipc_recvmsg,
3339 .mmap = sock_no_mmap,
3340 .sendpage = sock_no_sendpage
3341};
3342
3343static const struct proto_ops packet_ops = {
3344 .owner = THIS_MODULE,
3345 .family = AF_TIPC,
3346 .release = tipc_release,
3347 .bind = tipc_bind,
3348 .connect = tipc_connect,
3349 .socketpair = tipc_socketpair,
3350 .accept = tipc_accept,
3351 .getname = tipc_getname,
3352 .poll = tipc_poll,
3353 .ioctl = tipc_ioctl,
3354 .listen = tipc_listen,
3355 .shutdown = tipc_shutdown,
3356 .setsockopt = tipc_setsockopt,
3357 .getsockopt = tipc_getsockopt,
3358 .sendmsg = tipc_send_packet,
3359 .recvmsg = tipc_recvmsg,
3360 .mmap = sock_no_mmap,
3361 .sendpage = sock_no_sendpage
3362};
3363
3364static const struct proto_ops stream_ops = {
3365 .owner = THIS_MODULE,
3366 .family = AF_TIPC,
3367 .release = tipc_release,
3368 .bind = tipc_bind,
3369 .connect = tipc_connect,
3370 .socketpair = tipc_socketpair,
3371 .accept = tipc_accept,
3372 .getname = tipc_getname,
3373 .poll = tipc_poll,
3374 .ioctl = tipc_ioctl,
3375 .listen = tipc_listen,
3376 .shutdown = tipc_shutdown,
3377 .setsockopt = tipc_setsockopt,
3378 .getsockopt = tipc_getsockopt,
3379 .sendmsg = tipc_sendstream,
3380 .recvmsg = tipc_recvstream,
3381 .mmap = sock_no_mmap,
3382 .sendpage = sock_no_sendpage
3383};
3384
3385static const struct net_proto_family tipc_family_ops = {
3386 .owner = THIS_MODULE,
3387 .family = AF_TIPC,
3388 .create = tipc_sk_create
3389};
3390
3391static struct proto tipc_proto = {
3392 .name = "TIPC",
3393 .owner = THIS_MODULE,
3394 .obj_size = sizeof(struct tipc_sock),
3395 .sysctl_rmem = sysctl_tipc_rmem
3396};
3397
3398/**
3399 * tipc_socket_init - initialize TIPC socket interface
3400 *
3401 * Returns 0 on success, errno otherwise
3402 */
3403int tipc_socket_init(void)
3404{
3405 int res;
3406
3407 res = proto_register(&tipc_proto, 1);
3408 if (res) {
3409 pr_err("Failed to register TIPC protocol type\n");
3410 goto out;
3411 }
3412
3413 res = sock_register(&tipc_family_ops);
3414 if (res) {
3415 pr_err("Failed to register TIPC socket type\n");
3416 proto_unregister(&tipc_proto);
3417 goto out;
3418 }
3419 out:
3420 return res;
3421}
3422
3423/**
3424 * tipc_socket_stop - stop TIPC socket interface
3425 */
3426void tipc_socket_stop(void)
3427{
3428 sock_unregister(tipc_family_ops.family);
3429 proto_unregister(&tipc_proto);
3430}
3431
3432/* Caller should hold socket lock for the passed tipc socket. */
3433static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3434{
3435 u32 peer_node;
3436 u32 peer_port;
3437 struct nlattr *nest;
3438
3439 peer_node = tsk_peer_node(tsk);
3440 peer_port = tsk_peer_port(tsk);
3441
3442 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3443 if (!nest)
3444 return -EMSGSIZE;
3445
3446 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3447 goto msg_full;
3448 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3449 goto msg_full;
3450
3451 if (tsk->conn_type != 0) {
3452 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3453 goto msg_full;
3454 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3455 goto msg_full;
3456 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3457 goto msg_full;
3458 }
3459 nla_nest_end(skb, nest);
3460
3461 return 0;
3462
3463msg_full:
3464 nla_nest_cancel(skb, nest);
3465
3466 return -EMSGSIZE;
3467}
3468
3469static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3470 *tsk)
3471{
3472 struct net *net = sock_net(skb->sk);
3473 struct sock *sk = &tsk->sk;
3474
3475 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3476 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3477 return -EMSGSIZE;
3478
3479 if (tipc_sk_connected(sk)) {
3480 if (__tipc_nl_add_sk_con(skb, tsk))
3481 return -EMSGSIZE;
3482 } else if (!list_empty(&tsk->publications)) {
3483 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3484 return -EMSGSIZE;
3485 }
3486 return 0;
3487}
3488
3489/* Caller should hold socket lock for the passed tipc socket. */
3490static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3491 struct tipc_sock *tsk)
3492{
3493 struct nlattr *attrs;
3494 void *hdr;
3495
3496 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3497 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3498 if (!hdr)
3499 goto msg_cancel;
3500
3501 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3502 if (!attrs)
3503 goto genlmsg_cancel;
3504
3505 if (__tipc_nl_add_sk_info(skb, tsk))
3506 goto attr_msg_cancel;
3507
3508 nla_nest_end(skb, attrs);
3509 genlmsg_end(skb, hdr);
3510
3511 return 0;
3512
3513attr_msg_cancel:
3514 nla_nest_cancel(skb, attrs);
3515genlmsg_cancel:
3516 genlmsg_cancel(skb, hdr);
3517msg_cancel:
3518 return -EMSGSIZE;
3519}
3520
3521int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3522 int (*skb_handler)(struct sk_buff *skb,
3523 struct netlink_callback *cb,
3524 struct tipc_sock *tsk))
3525{
3526 struct rhashtable_iter *iter = (void *)cb->args[4];
3527 struct tipc_sock *tsk;
3528 int err;
3529
3530 rhashtable_walk_start(iter);
3531 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3532 if (IS_ERR(tsk)) {
3533 err = PTR_ERR(tsk);
3534 if (err == -EAGAIN) {
3535 err = 0;
3536 continue;
3537 }
3538 break;
3539 }
3540
3541 sock_hold(&tsk->sk);
3542 rhashtable_walk_stop(iter);
3543 lock_sock(&tsk->sk);
3544 err = skb_handler(skb, cb, tsk);
3545 if (err) {
3546 release_sock(&tsk->sk);
3547 sock_put(&tsk->sk);
3548 goto out;
3549 }
3550 release_sock(&tsk->sk);
3551 rhashtable_walk_start(iter);
3552 sock_put(&tsk->sk);
3553 }
3554 rhashtable_walk_stop(iter);
3555out:
3556 return skb->len;
3557}
3558EXPORT_SYMBOL(tipc_nl_sk_walk);
3559
3560int tipc_dump_start(struct netlink_callback *cb)
3561{
3562 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3563}
3564EXPORT_SYMBOL(tipc_dump_start);
3565
3566int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3567{
3568 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3569 struct rhashtable_iter *iter = (void *)cb->args[4];
3570 struct tipc_net *tn = tipc_net(net);
3571
3572 if (!iter) {
3573 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3574 if (!iter)
3575 return -ENOMEM;
3576
3577 cb->args[4] = (long)iter;
3578 }
3579
3580 rhashtable_walk_enter(&tn->sk_rht, iter);
3581 return 0;
3582}
3583
3584int tipc_dump_done(struct netlink_callback *cb)
3585{
3586 struct rhashtable_iter *hti = (void *)cb->args[4];
3587
3588 rhashtable_walk_exit(hti);
3589 kfree(hti);
3590 return 0;
3591}
3592EXPORT_SYMBOL(tipc_dump_done);
3593
3594int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3595 struct tipc_sock *tsk, u32 sk_filter_state,
3596 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3597{
3598 struct sock *sk = &tsk->sk;
3599 struct nlattr *attrs;
3600 struct nlattr *stat;
3601
3602 /*filter response w.r.t sk_state*/
3603 if (!(sk_filter_state & (1 << sk->sk_state)))
3604 return 0;
3605
3606 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3607 if (!attrs)
3608 goto msg_cancel;
3609
3610 if (__tipc_nl_add_sk_info(skb, tsk))
3611 goto attr_msg_cancel;
3612
3613 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3614 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3615 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3616 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3617 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3618 sock_i_uid(sk))) ||
3619 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3620 tipc_diag_gen_cookie(sk),
3621 TIPC_NLA_SOCK_PAD))
3622 goto attr_msg_cancel;
3623
3624 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3625 if (!stat)
3626 goto attr_msg_cancel;
3627
3628 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3629 skb_queue_len(&sk->sk_receive_queue)) ||
3630 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3631 skb_queue_len(&sk->sk_write_queue)) ||
3632 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3633 atomic_read(&sk->sk_drops)))
3634 goto stat_msg_cancel;
3635
3636 if (tsk->cong_link_cnt &&
3637 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3638 goto stat_msg_cancel;
3639
3640 if (tsk_conn_cong(tsk) &&
3641 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3642 goto stat_msg_cancel;
3643
3644 nla_nest_end(skb, stat);
3645
3646 if (tsk->group)
3647 if (tipc_group_fill_sock_diag(tsk->group, skb))
3648 goto stat_msg_cancel;
3649
3650 nla_nest_end(skb, attrs);
3651
3652 return 0;
3653
3654stat_msg_cancel:
3655 nla_nest_cancel(skb, stat);
3656attr_msg_cancel:
3657 nla_nest_cancel(skb, attrs);
3658msg_cancel:
3659 return -EMSGSIZE;
3660}
3661EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3662
3663int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3664{
3665 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3666}
3667
3668/* Caller should hold socket lock for the passed tipc socket. */
3669static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3670 struct netlink_callback *cb,
3671 struct publication *publ)
3672{
3673 void *hdr;
3674 struct nlattr *attrs;
3675
3676 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3677 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3678 if (!hdr)
3679 goto msg_cancel;
3680
3681 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3682 if (!attrs)
3683 goto genlmsg_cancel;
3684
3685 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3686 goto attr_msg_cancel;
3687 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3688 goto attr_msg_cancel;
3689 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3690 goto attr_msg_cancel;
3691 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3692 goto attr_msg_cancel;
3693
3694 nla_nest_end(skb, attrs);
3695 genlmsg_end(skb, hdr);
3696
3697 return 0;
3698
3699attr_msg_cancel:
3700 nla_nest_cancel(skb, attrs);
3701genlmsg_cancel:
3702 genlmsg_cancel(skb, hdr);
3703msg_cancel:
3704 return -EMSGSIZE;
3705}
3706
3707/* Caller should hold socket lock for the passed tipc socket. */
3708static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3709 struct netlink_callback *cb,
3710 struct tipc_sock *tsk, u32 *last_publ)
3711{
3712 int err;
3713 struct publication *p;
3714
3715 if (*last_publ) {
3716 list_for_each_entry(p, &tsk->publications, binding_sock) {
3717 if (p->key == *last_publ)
3718 break;
3719 }
3720 if (p->key != *last_publ) {
3721 /* We never set seq or call nl_dump_check_consistent()
3722 * this means that setting prev_seq here will cause the
3723 * consistence check to fail in the netlink callback
3724 * handler. Resulting in the last NLMSG_DONE message
3725 * having the NLM_F_DUMP_INTR flag set.
3726 */
3727 cb->prev_seq = 1;
3728 *last_publ = 0;
3729 return -EPIPE;
3730 }
3731 } else {
3732 p = list_first_entry(&tsk->publications, struct publication,
3733 binding_sock);
3734 }
3735
3736 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3737 err = __tipc_nl_add_sk_publ(skb, cb, p);
3738 if (err) {
3739 *last_publ = p->key;
3740 return err;
3741 }
3742 }
3743 *last_publ = 0;
3744
3745 return 0;
3746}
3747
3748int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3749{
3750 int err;
3751 u32 tsk_portid = cb->args[0];
3752 u32 last_publ = cb->args[1];
3753 u32 done = cb->args[2];
3754 struct net *net = sock_net(skb->sk);
3755 struct tipc_sock *tsk;
3756
3757 if (!tsk_portid) {
3758 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3759 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3760
3761 if (!attrs[TIPC_NLA_SOCK])
3762 return -EINVAL;
3763
3764 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3765 attrs[TIPC_NLA_SOCK],
3766 tipc_nl_sock_policy, NULL);
3767 if (err)
3768 return err;
3769
3770 if (!sock[TIPC_NLA_SOCK_REF])
3771 return -EINVAL;
3772
3773 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3774 }
3775
3776 if (done)
3777 return 0;
3778
3779 tsk = tipc_sk_lookup(net, tsk_portid);
3780 if (!tsk)
3781 return -EINVAL;
3782
3783 lock_sock(&tsk->sk);
3784 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3785 if (!err)
3786 done = 1;
3787 release_sock(&tsk->sk);
3788 sock_put(&tsk->sk);
3789
3790 cb->args[0] = tsk_portid;
3791 cb->args[1] = last_publ;
3792 cb->args[2] = done;
3793
3794 return skb->len;
3795}
3796
3797/**
3798 * tipc_sk_filtering - check if a socket should be traced
3799 * @sk: the socket to be examined
3800 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3801 * (portid, sock type, name type, name lower, name upper)
3802 *
3803 * Returns true if the socket meets the socket tuple data
3804 * (value 0 = 'any') or when there is no tuple set (all = 0),
3805 * otherwise false
3806 */
3807bool tipc_sk_filtering(struct sock *sk)
3808{
3809 struct tipc_sock *tsk;
3810 struct publication *p;
3811 u32 _port, _sktype, _type, _lower, _upper;
3812 u32 type = 0, lower = 0, upper = 0;
3813
3814 if (!sk)
3815 return true;
3816
3817 tsk = tipc_sk(sk);
3818
3819 _port = sysctl_tipc_sk_filter[0];
3820 _sktype = sysctl_tipc_sk_filter[1];
3821 _type = sysctl_tipc_sk_filter[2];
3822 _lower = sysctl_tipc_sk_filter[3];
3823 _upper = sysctl_tipc_sk_filter[4];
3824
3825 if (!_port && !_sktype && !_type && !_lower && !_upper)
3826 return true;
3827
3828 if (_port)
3829 return (_port == tsk->portid);
3830
3831 if (_sktype && _sktype != sk->sk_type)
3832 return false;
3833
3834 if (tsk->published) {
3835 p = list_first_entry_or_null(&tsk->publications,
3836 struct publication, binding_sock);
3837 if (p) {
3838 type = p->type;
3839 lower = p->lower;
3840 upper = p->upper;
3841 }
3842 }
3843
3844 if (!tipc_sk_type_connectionless(sk)) {
3845 type = tsk->conn_type;
3846 lower = tsk->conn_instance;
3847 upper = tsk->conn_instance;
3848 }
3849
3850 if ((_type && _type != type) || (_lower && _lower != lower) ||
3851 (_upper && _upper != upper))
3852 return false;
3853
3854 return true;
3855}
3856
3857u32 tipc_sock_get_portid(struct sock *sk)
3858{
3859 return (sk) ? (tipc_sk(sk))->portid : 0;
3860}
3861
3862/**
3863 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3864 * both the rcv and backlog queues are considered
3865 * @sk: tipc sk to be checked
3866 * @skb: tipc msg to be checked
3867 *
3868 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3869 */
3870
3871bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3872{
3873 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3874 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3875 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3876
3877 return (qsize > lim * 90 / 100);
3878}
3879
3880/**
3881 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3882 * only the rcv queue is considered
3883 * @sk: tipc sk to be checked
3884 * @skb: tipc msg to be checked
3885 *
3886 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3887 */
3888
3889bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3890{
3891 unsigned int lim = rcvbuf_limit(sk, skb);
3892 unsigned int qsize = sk_rmem_alloc_get(sk);
3893
3894 return (qsize > lim * 90 / 100);
3895}
3896
3897/**
3898 * tipc_sk_dump - dump TIPC socket
3899 * @sk: tipc sk to be dumped
3900 * @dqueues: bitmask to decide if any socket queue to be dumped?
3901 * - TIPC_DUMP_NONE: don't dump socket queues
3902 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3903 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3904 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3905 * - TIPC_DUMP_ALL: dump all the socket queues above
3906 * @buf: returned buffer of dump data in format
3907 */
3908int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3909{
3910 int i = 0;
3911 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3912 struct tipc_sock *tsk;
3913 struct publication *p;
3914 bool tsk_connected;
3915
3916 if (!sk) {
3917 i += scnprintf(buf, sz, "sk data: (null)\n");
3918 return i;
3919 }
3920
3921 tsk = tipc_sk(sk);
3922 tsk_connected = !tipc_sk_type_connectionless(sk);
3923
3924 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3925 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3926 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3927 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3928 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3929 if (tsk_connected) {
3930 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3931 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3932 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3933 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3934 }
3935 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3936 if (tsk->published) {
3937 p = list_first_entry_or_null(&tsk->publications,
3938 struct publication, binding_sock);
3939 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3940 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3941 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3942 }
3943 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3944 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3945 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3946 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3947 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3948 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3949 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3950 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3951 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3952 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3953 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3954 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3955 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3956 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3957
3958 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3959 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3960 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3961 }
3962
3963 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3964 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3965 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3966 }
3967
3968 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3969 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3970 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3971 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3972 i += scnprintf(buf + i, sz - i, " tail ");
3973 i += tipc_skb_dump(sk->sk_backlog.tail, false,
3974 buf + i);
3975 }
3976 }
3977
3978 return i;
3979}