Loading...
1/*
2 * net/tipc/socket.c: TIPC socket API
3 *
4 * Copyright (c) 2001-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
19 *
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
35 */
36
37#include <linux/rhashtable.h>
38#include "core.h"
39#include "name_table.h"
40#include "node.h"
41#include "link.h"
42#include "name_distr.h"
43#include "socket.h"
44#include "bcast.h"
45#include "netlink.h"
46
47#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
48#define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
49#define TIPC_FWD_MSG 1
50#define TIPC_MAX_PORT 0xffffffff
51#define TIPC_MIN_PORT 1
52
53enum {
54 TIPC_LISTEN = TCP_LISTEN,
55 TIPC_ESTABLISHED = TCP_ESTABLISHED,
56 TIPC_OPEN = TCP_CLOSE,
57 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
58 TIPC_CONNECTING = TCP_SYN_SENT,
59};
60
61/**
62 * struct tipc_sock - TIPC socket structure
63 * @sk: socket - interacts with 'port' and with user via the socket API
64 * @conn_type: TIPC type used when connection was established
65 * @conn_instance: TIPC instance used when connection was established
66 * @published: non-zero if port has one or more associated names
67 * @max_pkt: maximum packet size "hint" used when building messages sent by port
68 * @portid: unique port identity in TIPC socket hash table
69 * @phdr: preformatted message header used when sending messages
70 * @publications: list of publications for port
71 * @pub_count: total # of publications port has made during its lifetime
72 * @probing_state:
73 * @conn_timeout: the time we can wait for an unresponded setup request
74 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
75 * @link_cong: non-zero if owner must sleep because of link congestion
76 * @sent_unacked: # messages sent by socket, and not yet acked by peer
77 * @rcv_unacked: # messages read by user, but not yet acked back to peer
78 * @peer: 'connected' peer for dgram/rdm
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
81 */
82struct tipc_sock {
83 struct sock sk;
84 u32 conn_type;
85 u32 conn_instance;
86 int published;
87 u32 max_pkt;
88 u32 portid;
89 struct tipc_msg phdr;
90 struct list_head sock_list;
91 struct list_head publications;
92 u32 pub_count;
93 uint conn_timeout;
94 atomic_t dupl_rcvcnt;
95 bool probe_unacked;
96 bool link_cong;
97 u16 snt_unacked;
98 u16 snd_win;
99 u16 peer_caps;
100 u16 rcv_unacked;
101 u16 rcv_win;
102 struct sockaddr_tipc peer;
103 struct rhash_head node;
104 struct rcu_head rcu;
105};
106
107static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
108static void tipc_data_ready(struct sock *sk);
109static void tipc_write_space(struct sock *sk);
110static void tipc_sock_destruct(struct sock *sk);
111static int tipc_release(struct socket *sock);
112static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
113static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
114static void tipc_sk_timeout(unsigned long data);
115static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
116 struct tipc_name_seq const *seq);
117static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
118 struct tipc_name_seq const *seq);
119static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
120static int tipc_sk_insert(struct tipc_sock *tsk);
121static void tipc_sk_remove(struct tipc_sock *tsk);
122static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
123 size_t dsz);
124static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
125
126static const struct proto_ops packet_ops;
127static const struct proto_ops stream_ops;
128static const struct proto_ops msg_ops;
129static struct proto tipc_proto;
130static const struct rhashtable_params tsk_rht_params;
131
132static u32 tsk_own_node(struct tipc_sock *tsk)
133{
134 return msg_prevnode(&tsk->phdr);
135}
136
137static u32 tsk_peer_node(struct tipc_sock *tsk)
138{
139 return msg_destnode(&tsk->phdr);
140}
141
142static u32 tsk_peer_port(struct tipc_sock *tsk)
143{
144 return msg_destport(&tsk->phdr);
145}
146
147static bool tsk_unreliable(struct tipc_sock *tsk)
148{
149 return msg_src_droppable(&tsk->phdr) != 0;
150}
151
152static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
153{
154 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
155}
156
157static bool tsk_unreturnable(struct tipc_sock *tsk)
158{
159 return msg_dest_droppable(&tsk->phdr) != 0;
160}
161
162static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
163{
164 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
165}
166
167static int tsk_importance(struct tipc_sock *tsk)
168{
169 return msg_importance(&tsk->phdr);
170}
171
172static int tsk_set_importance(struct tipc_sock *tsk, int imp)
173{
174 if (imp > TIPC_CRITICAL_IMPORTANCE)
175 return -EINVAL;
176 msg_set_importance(&tsk->phdr, (u32)imp);
177 return 0;
178}
179
180static struct tipc_sock *tipc_sk(const struct sock *sk)
181{
182 return container_of(sk, struct tipc_sock, sk);
183}
184
185static bool tsk_conn_cong(struct tipc_sock *tsk)
186{
187 return tsk->snt_unacked > tsk->snd_win;
188}
189
190/* tsk_blocks(): translate a buffer size in bytes to number of
191 * advertisable blocks, taking into account the ratio truesize(len)/len
192 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
193 */
194static u16 tsk_adv_blocks(int len)
195{
196 return len / FLOWCTL_BLK_SZ / 4;
197}
198
199/* tsk_inc(): increment counter for sent or received data
200 * - If block based flow control is not supported by peer we
201 * fall back to message based ditto, incrementing the counter
202 */
203static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
204{
205 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
206 return ((msglen / FLOWCTL_BLK_SZ) + 1);
207 return 1;
208}
209
210/**
211 * tsk_advance_rx_queue - discard first buffer in socket receive queue
212 *
213 * Caller must hold socket lock
214 */
215static void tsk_advance_rx_queue(struct sock *sk)
216{
217 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
218}
219
220/* tipc_sk_respond() : send response message back to sender
221 */
222static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
223{
224 u32 selector;
225 u32 dnode;
226 u32 onode = tipc_own_addr(sock_net(sk));
227
228 if (!tipc_msg_reverse(onode, &skb, err))
229 return;
230
231 dnode = msg_destnode(buf_msg(skb));
232 selector = msg_origport(buf_msg(skb));
233 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
234}
235
236/**
237 * tsk_rej_rx_queue - reject all buffers in socket receive queue
238 *
239 * Caller must hold socket lock
240 */
241static void tsk_rej_rx_queue(struct sock *sk)
242{
243 struct sk_buff *skb;
244
245 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
246 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
247}
248
249static bool tipc_sk_connected(struct sock *sk)
250{
251 return sk->sk_state == TIPC_ESTABLISHED;
252}
253
254/* tipc_sk_type_connectionless - check if the socket is datagram socket
255 * @sk: socket
256 *
257 * Returns true if connection less, false otherwise
258 */
259static bool tipc_sk_type_connectionless(struct sock *sk)
260{
261 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
262}
263
264/* tsk_peer_msg - verify if message was sent by connected port's peer
265 *
266 * Handles cases where the node's network address has changed from
267 * the default of <0.0.0> to its configured setting.
268 */
269static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
270{
271 struct sock *sk = &tsk->sk;
272 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
273 u32 peer_port = tsk_peer_port(tsk);
274 u32 orig_node;
275 u32 peer_node;
276
277 if (unlikely(!tipc_sk_connected(sk)))
278 return false;
279
280 if (unlikely(msg_origport(msg) != peer_port))
281 return false;
282
283 orig_node = msg_orignode(msg);
284 peer_node = tsk_peer_node(tsk);
285
286 if (likely(orig_node == peer_node))
287 return true;
288
289 if (!orig_node && (peer_node == tn->own_addr))
290 return true;
291
292 if (!peer_node && (orig_node == tn->own_addr))
293 return true;
294
295 return false;
296}
297
298/* tipc_set_sk_state - set the sk_state of the socket
299 * @sk: socket
300 *
301 * Caller must hold socket lock
302 *
303 * Returns 0 on success, errno otherwise
304 */
305static int tipc_set_sk_state(struct sock *sk, int state)
306{
307 int oldsk_state = sk->sk_state;
308 int res = -EINVAL;
309
310 switch (state) {
311 case TIPC_OPEN:
312 res = 0;
313 break;
314 case TIPC_LISTEN:
315 case TIPC_CONNECTING:
316 if (oldsk_state == TIPC_OPEN)
317 res = 0;
318 break;
319 case TIPC_ESTABLISHED:
320 if (oldsk_state == TIPC_CONNECTING ||
321 oldsk_state == TIPC_OPEN)
322 res = 0;
323 break;
324 case TIPC_DISCONNECTING:
325 if (oldsk_state == TIPC_CONNECTING ||
326 oldsk_state == TIPC_ESTABLISHED)
327 res = 0;
328 break;
329 }
330
331 if (!res)
332 sk->sk_state = state;
333
334 return res;
335}
336
337/**
338 * tipc_sk_create - create a TIPC socket
339 * @net: network namespace (must be default network)
340 * @sock: pre-allocated socket structure
341 * @protocol: protocol indicator (must be 0)
342 * @kern: caused by kernel or by userspace?
343 *
344 * This routine creates additional data structures used by the TIPC socket,
345 * initializes them, and links them together.
346 *
347 * Returns 0 on success, errno otherwise
348 */
349static int tipc_sk_create(struct net *net, struct socket *sock,
350 int protocol, int kern)
351{
352 struct tipc_net *tn;
353 const struct proto_ops *ops;
354 struct sock *sk;
355 struct tipc_sock *tsk;
356 struct tipc_msg *msg;
357
358 /* Validate arguments */
359 if (unlikely(protocol != 0))
360 return -EPROTONOSUPPORT;
361
362 switch (sock->type) {
363 case SOCK_STREAM:
364 ops = &stream_ops;
365 break;
366 case SOCK_SEQPACKET:
367 ops = &packet_ops;
368 break;
369 case SOCK_DGRAM:
370 case SOCK_RDM:
371 ops = &msg_ops;
372 break;
373 default:
374 return -EPROTOTYPE;
375 }
376
377 /* Allocate socket's protocol area */
378 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
379 if (sk == NULL)
380 return -ENOMEM;
381
382 tsk = tipc_sk(sk);
383 tsk->max_pkt = MAX_PKT_DEFAULT;
384 INIT_LIST_HEAD(&tsk->publications);
385 msg = &tsk->phdr;
386 tn = net_generic(sock_net(sk), tipc_net_id);
387 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
388 NAMED_H_SIZE, 0);
389
390 /* Finish initializing socket data structures */
391 sock->ops = ops;
392 sock_init_data(sock, sk);
393 tipc_set_sk_state(sk, TIPC_OPEN);
394 if (tipc_sk_insert(tsk)) {
395 pr_warn("Socket create failed; port number exhausted\n");
396 return -EINVAL;
397 }
398 msg_set_origport(msg, tsk->portid);
399 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
400 sk->sk_shutdown = 0;
401 sk->sk_backlog_rcv = tipc_backlog_rcv;
402 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
403 sk->sk_data_ready = tipc_data_ready;
404 sk->sk_write_space = tipc_write_space;
405 sk->sk_destruct = tipc_sock_destruct;
406 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
407 atomic_set(&tsk->dupl_rcvcnt, 0);
408
409 /* Start out with safe limits until we receive an advertised window */
410 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
411 tsk->rcv_win = tsk->snd_win;
412
413 if (tipc_sk_type_connectionless(sk)) {
414 tsk_set_unreturnable(tsk, true);
415 if (sock->type == SOCK_DGRAM)
416 tsk_set_unreliable(tsk, true);
417 }
418
419 return 0;
420}
421
422static void tipc_sk_callback(struct rcu_head *head)
423{
424 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
425
426 sock_put(&tsk->sk);
427}
428
429/* Caller should hold socket lock for the socket. */
430static void __tipc_shutdown(struct socket *sock, int error)
431{
432 struct sock *sk = sock->sk;
433 struct tipc_sock *tsk = tipc_sk(sk);
434 struct net *net = sock_net(sk);
435 u32 dnode = tsk_peer_node(tsk);
436 struct sk_buff *skb;
437
438 /* Reject all unreceived messages, except on an active connection
439 * (which disconnects locally & sends a 'FIN+' to peer).
440 */
441 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
442 if (TIPC_SKB_CB(skb)->bytes_read) {
443 kfree_skb(skb);
444 continue;
445 }
446 if (!tipc_sk_type_connectionless(sk) &&
447 sk->sk_state != TIPC_DISCONNECTING) {
448 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
449 tipc_node_remove_conn(net, dnode, tsk->portid);
450 }
451 tipc_sk_respond(sk, skb, error);
452 }
453
454 if (tipc_sk_type_connectionless(sk))
455 return;
456
457 if (sk->sk_state != TIPC_DISCONNECTING) {
458 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
459 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
460 tsk_own_node(tsk), tsk_peer_port(tsk),
461 tsk->portid, error);
462 if (skb)
463 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
464 tipc_node_remove_conn(net, dnode, tsk->portid);
465 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
466 }
467}
468
469/**
470 * tipc_release - destroy a TIPC socket
471 * @sock: socket to destroy
472 *
473 * This routine cleans up any messages that are still queued on the socket.
474 * For DGRAM and RDM socket types, all queued messages are rejected.
475 * For SEQPACKET and STREAM socket types, the first message is rejected
476 * and any others are discarded. (If the first message on a STREAM socket
477 * is partially-read, it is discarded and the next one is rejected instead.)
478 *
479 * NOTE: Rejected messages are not necessarily returned to the sender! They
480 * are returned or discarded according to the "destination droppable" setting
481 * specified for the message by the sender.
482 *
483 * Returns 0 on success, errno otherwise
484 */
485static int tipc_release(struct socket *sock)
486{
487 struct sock *sk = sock->sk;
488 struct tipc_sock *tsk;
489
490 /*
491 * Exit if socket isn't fully initialized (occurs when a failed accept()
492 * releases a pre-allocated child socket that was never used)
493 */
494 if (sk == NULL)
495 return 0;
496
497 tsk = tipc_sk(sk);
498 lock_sock(sk);
499
500 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
501 sk->sk_shutdown = SHUTDOWN_MASK;
502 tipc_sk_withdraw(tsk, 0, NULL);
503 sk_stop_timer(sk, &sk->sk_timer);
504 tipc_sk_remove(tsk);
505
506 /* Reject any messages that accumulated in backlog queue */
507 release_sock(sk);
508
509 call_rcu(&tsk->rcu, tipc_sk_callback);
510 sock->sk = NULL;
511
512 return 0;
513}
514
515/**
516 * tipc_bind - associate or disassocate TIPC name(s) with a socket
517 * @sock: socket structure
518 * @uaddr: socket address describing name(s) and desired operation
519 * @uaddr_len: size of socket address data structure
520 *
521 * Name and name sequence binding is indicated using a positive scope value;
522 * a negative scope value unbinds the specified name. Specifying no name
523 * (i.e. a socket address length of 0) unbinds all names from the socket.
524 *
525 * Returns 0 on success, errno otherwise
526 *
527 * NOTE: This routine doesn't need to take the socket lock since it doesn't
528 * access any non-constant socket information.
529 */
530static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
531 int uaddr_len)
532{
533 struct sock *sk = sock->sk;
534 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
535 struct tipc_sock *tsk = tipc_sk(sk);
536 int res = -EINVAL;
537
538 lock_sock(sk);
539 if (unlikely(!uaddr_len)) {
540 res = tipc_sk_withdraw(tsk, 0, NULL);
541 goto exit;
542 }
543
544 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
545 res = -EINVAL;
546 goto exit;
547 }
548 if (addr->family != AF_TIPC) {
549 res = -EAFNOSUPPORT;
550 goto exit;
551 }
552
553 if (addr->addrtype == TIPC_ADDR_NAME)
554 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
555 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
556 res = -EAFNOSUPPORT;
557 goto exit;
558 }
559
560 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
561 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
562 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
563 res = -EACCES;
564 goto exit;
565 }
566
567 res = (addr->scope > 0) ?
568 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
569 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
570exit:
571 release_sock(sk);
572 return res;
573}
574
575/**
576 * tipc_getname - get port ID of socket or peer socket
577 * @sock: socket structure
578 * @uaddr: area for returned socket address
579 * @uaddr_len: area for returned length of socket address
580 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
581 *
582 * Returns 0 on success, errno otherwise
583 *
584 * NOTE: This routine doesn't need to take the socket lock since it only
585 * accesses socket information that is unchanging (or which changes in
586 * a completely predictable manner).
587 */
588static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
589 int *uaddr_len, int peer)
590{
591 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
592 struct sock *sk = sock->sk;
593 struct tipc_sock *tsk = tipc_sk(sk);
594 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
595
596 memset(addr, 0, sizeof(*addr));
597 if (peer) {
598 if ((!tipc_sk_connected(sk)) &&
599 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
600 return -ENOTCONN;
601 addr->addr.id.ref = tsk_peer_port(tsk);
602 addr->addr.id.node = tsk_peer_node(tsk);
603 } else {
604 addr->addr.id.ref = tsk->portid;
605 addr->addr.id.node = tn->own_addr;
606 }
607
608 *uaddr_len = sizeof(*addr);
609 addr->addrtype = TIPC_ADDR_ID;
610 addr->family = AF_TIPC;
611 addr->scope = 0;
612 addr->addr.name.domain = 0;
613
614 return 0;
615}
616
617/**
618 * tipc_poll - read and possibly block on pollmask
619 * @file: file structure associated with the socket
620 * @sock: socket for which to calculate the poll bits
621 * @wait: ???
622 *
623 * Returns pollmask value
624 *
625 * COMMENTARY:
626 * It appears that the usual socket locking mechanisms are not useful here
627 * since the pollmask info is potentially out-of-date the moment this routine
628 * exits. TCP and other protocols seem to rely on higher level poll routines
629 * to handle any preventable race conditions, so TIPC will do the same ...
630 *
631 * IMPORTANT: The fact that a read or write operation is indicated does NOT
632 * imply that the operation will succeed, merely that it should be performed
633 * and will not block.
634 */
635static unsigned int tipc_poll(struct file *file, struct socket *sock,
636 poll_table *wait)
637{
638 struct sock *sk = sock->sk;
639 struct tipc_sock *tsk = tipc_sk(sk);
640 u32 mask = 0;
641
642 sock_poll_wait(file, sk_sleep(sk), wait);
643
644 if (sk->sk_shutdown & RCV_SHUTDOWN)
645 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
646 if (sk->sk_shutdown == SHUTDOWN_MASK)
647 mask |= POLLHUP;
648
649 switch (sk->sk_state) {
650 case TIPC_ESTABLISHED:
651 if (!tsk->link_cong && !tsk_conn_cong(tsk))
652 mask |= POLLOUT;
653 /* fall thru' */
654 case TIPC_LISTEN:
655 case TIPC_CONNECTING:
656 if (!skb_queue_empty(&sk->sk_receive_queue))
657 mask |= (POLLIN | POLLRDNORM);
658 break;
659 case TIPC_OPEN:
660 if (!tsk->link_cong)
661 mask |= POLLOUT;
662 if (tipc_sk_type_connectionless(sk) &&
663 (!skb_queue_empty(&sk->sk_receive_queue)))
664 mask |= (POLLIN | POLLRDNORM);
665 break;
666 case TIPC_DISCONNECTING:
667 mask = (POLLIN | POLLRDNORM | POLLHUP);
668 break;
669 }
670
671 return mask;
672}
673
674/**
675 * tipc_sendmcast - send multicast message
676 * @sock: socket structure
677 * @seq: destination address
678 * @msg: message to send
679 * @dsz: total length of message data
680 * @timeo: timeout to wait for wakeup
681 *
682 * Called from function tipc_sendmsg(), which has done all sanity checks
683 * Returns the number of bytes sent on success, or errno
684 */
685static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
686 struct msghdr *msg, size_t dsz, long timeo)
687{
688 struct sock *sk = sock->sk;
689 struct tipc_sock *tsk = tipc_sk(sk);
690 struct net *net = sock_net(sk);
691 struct tipc_msg *mhdr = &tsk->phdr;
692 struct sk_buff_head pktchain;
693 struct iov_iter save = msg->msg_iter;
694 uint mtu;
695 int rc;
696
697 if (!timeo && tsk->link_cong)
698 return -ELINKCONG;
699
700 msg_set_type(mhdr, TIPC_MCAST_MSG);
701 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
702 msg_set_destport(mhdr, 0);
703 msg_set_destnode(mhdr, 0);
704 msg_set_nametype(mhdr, seq->type);
705 msg_set_namelower(mhdr, seq->lower);
706 msg_set_nameupper(mhdr, seq->upper);
707 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
708
709 skb_queue_head_init(&pktchain);
710
711new_mtu:
712 mtu = tipc_bcast_get_mtu(net);
713 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
714 if (unlikely(rc < 0))
715 return rc;
716
717 do {
718 rc = tipc_bcast_xmit(net, &pktchain);
719 if (likely(!rc))
720 return dsz;
721
722 if (rc == -ELINKCONG) {
723 tsk->link_cong = 1;
724 rc = tipc_wait_for_sndmsg(sock, &timeo);
725 if (!rc)
726 continue;
727 }
728 __skb_queue_purge(&pktchain);
729 if (rc == -EMSGSIZE) {
730 msg->msg_iter = save;
731 goto new_mtu;
732 }
733 break;
734 } while (1);
735 return rc;
736}
737
738/**
739 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
740 * @arrvq: queue with arriving messages, to be cloned after destination lookup
741 * @inputq: queue with cloned messages, delivered to socket after dest lookup
742 *
743 * Multi-threaded: parallel calls with reference to same queues may occur
744 */
745void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
746 struct sk_buff_head *inputq)
747{
748 struct tipc_msg *msg;
749 struct tipc_plist dports;
750 u32 portid;
751 u32 scope = TIPC_CLUSTER_SCOPE;
752 struct sk_buff_head tmpq;
753 uint hsz;
754 struct sk_buff *skb, *_skb;
755
756 __skb_queue_head_init(&tmpq);
757 tipc_plist_init(&dports);
758
759 skb = tipc_skb_peek(arrvq, &inputq->lock);
760 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
761 msg = buf_msg(skb);
762 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
763
764 if (in_own_node(net, msg_orignode(msg)))
765 scope = TIPC_NODE_SCOPE;
766
767 /* Create destination port list and message clones: */
768 tipc_nametbl_mc_translate(net,
769 msg_nametype(msg), msg_namelower(msg),
770 msg_nameupper(msg), scope, &dports);
771 portid = tipc_plist_pop(&dports);
772 for (; portid; portid = tipc_plist_pop(&dports)) {
773 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
774 if (_skb) {
775 msg_set_destport(buf_msg(_skb), portid);
776 __skb_queue_tail(&tmpq, _skb);
777 continue;
778 }
779 pr_warn("Failed to clone mcast rcv buffer\n");
780 }
781 /* Append to inputq if not already done by other thread */
782 spin_lock_bh(&inputq->lock);
783 if (skb_peek(arrvq) == skb) {
784 skb_queue_splice_tail_init(&tmpq, inputq);
785 kfree_skb(__skb_dequeue(arrvq));
786 }
787 spin_unlock_bh(&inputq->lock);
788 __skb_queue_purge(&tmpq);
789 kfree_skb(skb);
790 }
791 tipc_sk_rcv(net, inputq);
792}
793
794/**
795 * tipc_sk_proto_rcv - receive a connection mng protocol message
796 * @tsk: receiving socket
797 * @skb: pointer to message buffer.
798 */
799static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
800 struct sk_buff_head *xmitq)
801{
802 struct sock *sk = &tsk->sk;
803 u32 onode = tsk_own_node(tsk);
804 struct tipc_msg *hdr = buf_msg(skb);
805 int mtyp = msg_type(hdr);
806 bool conn_cong;
807
808 /* Ignore if connection cannot be validated: */
809 if (!tsk_peer_msg(tsk, hdr))
810 goto exit;
811
812 tsk->probe_unacked = false;
813
814 if (mtyp == CONN_PROBE) {
815 msg_set_type(hdr, CONN_PROBE_REPLY);
816 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
817 __skb_queue_tail(xmitq, skb);
818 return;
819 } else if (mtyp == CONN_ACK) {
820 conn_cong = tsk_conn_cong(tsk);
821 tsk->snt_unacked -= msg_conn_ack(hdr);
822 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
823 tsk->snd_win = msg_adv_win(hdr);
824 if (conn_cong)
825 sk->sk_write_space(sk);
826 } else if (mtyp != CONN_PROBE_REPLY) {
827 pr_warn("Received unknown CONN_PROTO msg\n");
828 }
829exit:
830 kfree_skb(skb);
831}
832
833static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
834{
835 DEFINE_WAIT_FUNC(wait, woken_wake_function);
836 struct sock *sk = sock->sk;
837 struct tipc_sock *tsk = tipc_sk(sk);
838 int done;
839
840 do {
841 int err = sock_error(sk);
842 if (err)
843 return err;
844 if (sk->sk_shutdown & SEND_SHUTDOWN)
845 return -EPIPE;
846 if (!*timeo_p)
847 return -EAGAIN;
848 if (signal_pending(current))
849 return sock_intr_errno(*timeo_p);
850
851 add_wait_queue(sk_sleep(sk), &wait);
852 done = sk_wait_event(sk, timeo_p, !tsk->link_cong, &wait);
853 remove_wait_queue(sk_sleep(sk), &wait);
854 } while (!done);
855 return 0;
856}
857
858/**
859 * tipc_sendmsg - send message in connectionless manner
860 * @sock: socket structure
861 * @m: message to send
862 * @dsz: amount of user data to be sent
863 *
864 * Message must have an destination specified explicitly.
865 * Used for SOCK_RDM and SOCK_DGRAM messages,
866 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
867 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
868 *
869 * Returns the number of bytes sent on success, or errno otherwise
870 */
871static int tipc_sendmsg(struct socket *sock,
872 struct msghdr *m, size_t dsz)
873{
874 struct sock *sk = sock->sk;
875 int ret;
876
877 lock_sock(sk);
878 ret = __tipc_sendmsg(sock, m, dsz);
879 release_sock(sk);
880
881 return ret;
882}
883
884static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
885{
886 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
887 struct sock *sk = sock->sk;
888 struct tipc_sock *tsk = tipc_sk(sk);
889 struct net *net = sock_net(sk);
890 struct tipc_msg *mhdr = &tsk->phdr;
891 u32 dnode, dport;
892 struct sk_buff_head pktchain;
893 bool is_connectionless = tipc_sk_type_connectionless(sk);
894 struct sk_buff *skb;
895 struct tipc_name_seq *seq;
896 struct iov_iter save;
897 u32 mtu;
898 long timeo;
899 int rc;
900
901 if (dsz > TIPC_MAX_USER_MSG_SIZE)
902 return -EMSGSIZE;
903 if (unlikely(!dest)) {
904 if (is_connectionless && tsk->peer.family == AF_TIPC)
905 dest = &tsk->peer;
906 else
907 return -EDESTADDRREQ;
908 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
909 dest->family != AF_TIPC) {
910 return -EINVAL;
911 }
912 if (!is_connectionless) {
913 if (sk->sk_state == TIPC_LISTEN)
914 return -EPIPE;
915 if (sk->sk_state != TIPC_OPEN)
916 return -EISCONN;
917 if (tsk->published)
918 return -EOPNOTSUPP;
919 if (dest->addrtype == TIPC_ADDR_NAME) {
920 tsk->conn_type = dest->addr.name.name.type;
921 tsk->conn_instance = dest->addr.name.name.instance;
922 }
923 }
924 seq = &dest->addr.nameseq;
925 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
926
927 if (dest->addrtype == TIPC_ADDR_MCAST) {
928 return tipc_sendmcast(sock, seq, m, dsz, timeo);
929 } else if (dest->addrtype == TIPC_ADDR_NAME) {
930 u32 type = dest->addr.name.name.type;
931 u32 inst = dest->addr.name.name.instance;
932 u32 domain = dest->addr.name.domain;
933
934 dnode = domain;
935 msg_set_type(mhdr, TIPC_NAMED_MSG);
936 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
937 msg_set_nametype(mhdr, type);
938 msg_set_nameinst(mhdr, inst);
939 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
940 dport = tipc_nametbl_translate(net, type, inst, &dnode);
941 msg_set_destnode(mhdr, dnode);
942 msg_set_destport(mhdr, dport);
943 if (unlikely(!dport && !dnode))
944 return -EHOSTUNREACH;
945 } else if (dest->addrtype == TIPC_ADDR_ID) {
946 dnode = dest->addr.id.node;
947 msg_set_type(mhdr, TIPC_DIRECT_MSG);
948 msg_set_lookup_scope(mhdr, 0);
949 msg_set_destnode(mhdr, dnode);
950 msg_set_destport(mhdr, dest->addr.id.ref);
951 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
952 }
953
954 skb_queue_head_init(&pktchain);
955 save = m->msg_iter;
956new_mtu:
957 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
958 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
959 if (rc < 0)
960 return rc;
961
962 do {
963 skb = skb_peek(&pktchain);
964 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
965 rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
966 if (likely(!rc)) {
967 if (!is_connectionless)
968 tipc_set_sk_state(sk, TIPC_CONNECTING);
969 return dsz;
970 }
971 if (rc == -ELINKCONG) {
972 tsk->link_cong = 1;
973 rc = tipc_wait_for_sndmsg(sock, &timeo);
974 if (!rc)
975 continue;
976 }
977 __skb_queue_purge(&pktchain);
978 if (rc == -EMSGSIZE) {
979 m->msg_iter = save;
980 goto new_mtu;
981 }
982 break;
983 } while (1);
984
985 return rc;
986}
987
988static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
989{
990 DEFINE_WAIT_FUNC(wait, woken_wake_function);
991 struct sock *sk = sock->sk;
992 struct tipc_sock *tsk = tipc_sk(sk);
993 int done;
994
995 do {
996 int err = sock_error(sk);
997 if (err)
998 return err;
999 if (sk->sk_state == TIPC_DISCONNECTING)
1000 return -EPIPE;
1001 else if (!tipc_sk_connected(sk))
1002 return -ENOTCONN;
1003 if (!*timeo_p)
1004 return -EAGAIN;
1005 if (signal_pending(current))
1006 return sock_intr_errno(*timeo_p);
1007
1008 add_wait_queue(sk_sleep(sk), &wait);
1009 done = sk_wait_event(sk, timeo_p,
1010 (!tsk->link_cong &&
1011 !tsk_conn_cong(tsk)) ||
1012 !tipc_sk_connected(sk), &wait);
1013 remove_wait_queue(sk_sleep(sk), &wait);
1014 } while (!done);
1015 return 0;
1016}
1017
1018/**
1019 * tipc_send_stream - send stream-oriented data
1020 * @sock: socket structure
1021 * @m: data to send
1022 * @dsz: total length of data to be transmitted
1023 *
1024 * Used for SOCK_STREAM data.
1025 *
1026 * Returns the number of bytes sent on success (or partial success),
1027 * or errno if no data sent
1028 */
1029static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1030{
1031 struct sock *sk = sock->sk;
1032 int ret;
1033
1034 lock_sock(sk);
1035 ret = __tipc_send_stream(sock, m, dsz);
1036 release_sock(sk);
1037
1038 return ret;
1039}
1040
1041static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1042{
1043 struct sock *sk = sock->sk;
1044 struct net *net = sock_net(sk);
1045 struct tipc_sock *tsk = tipc_sk(sk);
1046 struct tipc_msg *mhdr = &tsk->phdr;
1047 struct sk_buff_head pktchain;
1048 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1049 u32 portid = tsk->portid;
1050 int rc = -EINVAL;
1051 long timeo;
1052 u32 dnode;
1053 uint mtu, send, sent = 0;
1054 struct iov_iter save;
1055 int hlen = MIN_H_SIZE;
1056
1057 /* Handle implied connection establishment */
1058 if (unlikely(dest)) {
1059 rc = __tipc_sendmsg(sock, m, dsz);
1060 hlen = msg_hdr_sz(mhdr);
1061 if (dsz && (dsz == rc))
1062 tsk->snt_unacked = tsk_inc(tsk, dsz + hlen);
1063 return rc;
1064 }
1065 if (dsz > (uint)INT_MAX)
1066 return -EMSGSIZE;
1067
1068 if (unlikely(!tipc_sk_connected(sk))) {
1069 if (sk->sk_state == TIPC_DISCONNECTING)
1070 return -EPIPE;
1071 else
1072 return -ENOTCONN;
1073 }
1074
1075 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1076 if (!timeo && tsk->link_cong)
1077 return -ELINKCONG;
1078
1079 dnode = tsk_peer_node(tsk);
1080 skb_queue_head_init(&pktchain);
1081
1082next:
1083 save = m->msg_iter;
1084 mtu = tsk->max_pkt;
1085 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1086 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1087 if (unlikely(rc < 0))
1088 return rc;
1089
1090 do {
1091 if (likely(!tsk_conn_cong(tsk))) {
1092 rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1093 if (likely(!rc)) {
1094 tsk->snt_unacked += tsk_inc(tsk, send + hlen);
1095 sent += send;
1096 if (sent == dsz)
1097 return dsz;
1098 goto next;
1099 }
1100 if (rc == -EMSGSIZE) {
1101 __skb_queue_purge(&pktchain);
1102 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1103 portid);
1104 m->msg_iter = save;
1105 goto next;
1106 }
1107 if (rc != -ELINKCONG)
1108 break;
1109
1110 tsk->link_cong = 1;
1111 }
1112 rc = tipc_wait_for_sndpkt(sock, &timeo);
1113 } while (!rc);
1114
1115 __skb_queue_purge(&pktchain);
1116 return sent ? sent : rc;
1117}
1118
1119/**
1120 * tipc_send_packet - send a connection-oriented message
1121 * @sock: socket structure
1122 * @m: message to send
1123 * @dsz: length of data to be transmitted
1124 *
1125 * Used for SOCK_SEQPACKET messages.
1126 *
1127 * Returns the number of bytes sent on success, or errno otherwise
1128 */
1129static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1130{
1131 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1132 return -EMSGSIZE;
1133
1134 return tipc_send_stream(sock, m, dsz);
1135}
1136
1137/* tipc_sk_finish_conn - complete the setup of a connection
1138 */
1139static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1140 u32 peer_node)
1141{
1142 struct sock *sk = &tsk->sk;
1143 struct net *net = sock_net(sk);
1144 struct tipc_msg *msg = &tsk->phdr;
1145
1146 msg_set_destnode(msg, peer_node);
1147 msg_set_destport(msg, peer_port);
1148 msg_set_type(msg, TIPC_CONN_MSG);
1149 msg_set_lookup_scope(msg, 0);
1150 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1151
1152 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
1153 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1154 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1155 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1156 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1157 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1158 return;
1159
1160 /* Fall back to message based flow control */
1161 tsk->rcv_win = FLOWCTL_MSG_WIN;
1162 tsk->snd_win = FLOWCTL_MSG_WIN;
1163}
1164
1165/**
1166 * set_orig_addr - capture sender's address for received message
1167 * @m: descriptor for message info
1168 * @msg: received message header
1169 *
1170 * Note: Address is not captured if not requested by receiver.
1171 */
1172static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1173{
1174 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1175
1176 if (addr) {
1177 addr->family = AF_TIPC;
1178 addr->addrtype = TIPC_ADDR_ID;
1179 memset(&addr->addr, 0, sizeof(addr->addr));
1180 addr->addr.id.ref = msg_origport(msg);
1181 addr->addr.id.node = msg_orignode(msg);
1182 addr->addr.name.domain = 0; /* could leave uninitialized */
1183 addr->scope = 0; /* could leave uninitialized */
1184 m->msg_namelen = sizeof(struct sockaddr_tipc);
1185 }
1186}
1187
1188/**
1189 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1190 * @m: descriptor for message info
1191 * @msg: received message header
1192 * @tsk: TIPC port associated with message
1193 *
1194 * Note: Ancillary data is not captured if not requested by receiver.
1195 *
1196 * Returns 0 if successful, otherwise errno
1197 */
1198static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1199 struct tipc_sock *tsk)
1200{
1201 u32 anc_data[3];
1202 u32 err;
1203 u32 dest_type;
1204 int has_name;
1205 int res;
1206
1207 if (likely(m->msg_controllen == 0))
1208 return 0;
1209
1210 /* Optionally capture errored message object(s) */
1211 err = msg ? msg_errcode(msg) : 0;
1212 if (unlikely(err)) {
1213 anc_data[0] = err;
1214 anc_data[1] = msg_data_sz(msg);
1215 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1216 if (res)
1217 return res;
1218 if (anc_data[1]) {
1219 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1220 msg_data(msg));
1221 if (res)
1222 return res;
1223 }
1224 }
1225
1226 /* Optionally capture message destination object */
1227 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1228 switch (dest_type) {
1229 case TIPC_NAMED_MSG:
1230 has_name = 1;
1231 anc_data[0] = msg_nametype(msg);
1232 anc_data[1] = msg_namelower(msg);
1233 anc_data[2] = msg_namelower(msg);
1234 break;
1235 case TIPC_MCAST_MSG:
1236 has_name = 1;
1237 anc_data[0] = msg_nametype(msg);
1238 anc_data[1] = msg_namelower(msg);
1239 anc_data[2] = msg_nameupper(msg);
1240 break;
1241 case TIPC_CONN_MSG:
1242 has_name = (tsk->conn_type != 0);
1243 anc_data[0] = tsk->conn_type;
1244 anc_data[1] = tsk->conn_instance;
1245 anc_data[2] = tsk->conn_instance;
1246 break;
1247 default:
1248 has_name = 0;
1249 }
1250 if (has_name) {
1251 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1252 if (res)
1253 return res;
1254 }
1255
1256 return 0;
1257}
1258
1259static void tipc_sk_send_ack(struct tipc_sock *tsk)
1260{
1261 struct sock *sk = &tsk->sk;
1262 struct net *net = sock_net(sk);
1263 struct sk_buff *skb = NULL;
1264 struct tipc_msg *msg;
1265 u32 peer_port = tsk_peer_port(tsk);
1266 u32 dnode = tsk_peer_node(tsk);
1267
1268 if (!tipc_sk_connected(sk))
1269 return;
1270 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1271 dnode, tsk_own_node(tsk), peer_port,
1272 tsk->portid, TIPC_OK);
1273 if (!skb)
1274 return;
1275 msg = buf_msg(skb);
1276 msg_set_conn_ack(msg, tsk->rcv_unacked);
1277 tsk->rcv_unacked = 0;
1278
1279 /* Adjust to and advertize the correct window limit */
1280 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1281 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1282 msg_set_adv_win(msg, tsk->rcv_win);
1283 }
1284 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1285}
1286
1287static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1288{
1289 struct sock *sk = sock->sk;
1290 DEFINE_WAIT(wait);
1291 long timeo = *timeop;
1292 int err;
1293
1294 for (;;) {
1295 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1296 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1297 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1298 err = -ENOTCONN;
1299 break;
1300 }
1301 release_sock(sk);
1302 timeo = schedule_timeout(timeo);
1303 lock_sock(sk);
1304 }
1305 err = 0;
1306 if (!skb_queue_empty(&sk->sk_receive_queue))
1307 break;
1308 err = -EAGAIN;
1309 if (!timeo)
1310 break;
1311 err = sock_intr_errno(timeo);
1312 if (signal_pending(current))
1313 break;
1314 }
1315 finish_wait(sk_sleep(sk), &wait);
1316 *timeop = timeo;
1317 return err;
1318}
1319
1320/**
1321 * tipc_recvmsg - receive packet-oriented message
1322 * @m: descriptor for message info
1323 * @buf_len: total size of user buffer area
1324 * @flags: receive flags
1325 *
1326 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1327 * If the complete message doesn't fit in user area, truncate it.
1328 *
1329 * Returns size of returned message data, errno otherwise
1330 */
1331static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1332 int flags)
1333{
1334 struct sock *sk = sock->sk;
1335 struct tipc_sock *tsk = tipc_sk(sk);
1336 struct sk_buff *buf;
1337 struct tipc_msg *msg;
1338 bool is_connectionless = tipc_sk_type_connectionless(sk);
1339 long timeo;
1340 unsigned int sz;
1341 u32 err;
1342 int res, hlen;
1343
1344 /* Catch invalid receive requests */
1345 if (unlikely(!buf_len))
1346 return -EINVAL;
1347
1348 lock_sock(sk);
1349
1350 if (!is_connectionless && unlikely(sk->sk_state == TIPC_OPEN)) {
1351 res = -ENOTCONN;
1352 goto exit;
1353 }
1354
1355 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1356restart:
1357
1358 /* Look for a message in receive queue; wait if necessary */
1359 res = tipc_wait_for_rcvmsg(sock, &timeo);
1360 if (res)
1361 goto exit;
1362
1363 /* Look at first message in receive queue */
1364 buf = skb_peek(&sk->sk_receive_queue);
1365 msg = buf_msg(buf);
1366 sz = msg_data_sz(msg);
1367 hlen = msg_hdr_sz(msg);
1368 err = msg_errcode(msg);
1369
1370 /* Discard an empty non-errored message & try again */
1371 if ((!sz) && (!err)) {
1372 tsk_advance_rx_queue(sk);
1373 goto restart;
1374 }
1375
1376 /* Capture sender's address (optional) */
1377 set_orig_addr(m, msg);
1378
1379 /* Capture ancillary data (optional) */
1380 res = tipc_sk_anc_data_recv(m, msg, tsk);
1381 if (res)
1382 goto exit;
1383
1384 /* Capture message data (if valid) & compute return value (always) */
1385 if (!err) {
1386 if (unlikely(buf_len < sz)) {
1387 sz = buf_len;
1388 m->msg_flags |= MSG_TRUNC;
1389 }
1390 res = skb_copy_datagram_msg(buf, hlen, m, sz);
1391 if (res)
1392 goto exit;
1393 res = sz;
1394 } else {
1395 if (is_connectionless || err == TIPC_CONN_SHUTDOWN ||
1396 m->msg_control)
1397 res = 0;
1398 else
1399 res = -ECONNRESET;
1400 }
1401
1402 if (unlikely(flags & MSG_PEEK))
1403 goto exit;
1404
1405 if (likely(!is_connectionless)) {
1406 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1407 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1408 tipc_sk_send_ack(tsk);
1409 }
1410 tsk_advance_rx_queue(sk);
1411exit:
1412 release_sock(sk);
1413 return res;
1414}
1415
1416/**
1417 * tipc_recv_stream - receive stream-oriented data
1418 * @m: descriptor for message info
1419 * @buf_len: total size of user buffer area
1420 * @flags: receive flags
1421 *
1422 * Used for SOCK_STREAM messages only. If not enough data is available
1423 * will optionally wait for more; never truncates data.
1424 *
1425 * Returns size of returned message data, errno otherwise
1426 */
1427static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1428 size_t buf_len, int flags)
1429{
1430 struct sock *sk = sock->sk;
1431 struct tipc_sock *tsk = tipc_sk(sk);
1432 struct sk_buff *buf;
1433 struct tipc_msg *msg;
1434 long timeo;
1435 unsigned int sz;
1436 int target;
1437 int sz_copied = 0;
1438 u32 err;
1439 int res = 0, hlen;
1440
1441 /* Catch invalid receive attempts */
1442 if (unlikely(!buf_len))
1443 return -EINVAL;
1444
1445 lock_sock(sk);
1446
1447 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1448 res = -ENOTCONN;
1449 goto exit;
1450 }
1451
1452 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1453 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1454
1455restart:
1456 /* Look for a message in receive queue; wait if necessary */
1457 res = tipc_wait_for_rcvmsg(sock, &timeo);
1458 if (res)
1459 goto exit;
1460
1461 /* Look at first message in receive queue */
1462 buf = skb_peek(&sk->sk_receive_queue);
1463 msg = buf_msg(buf);
1464 sz = msg_data_sz(msg);
1465 hlen = msg_hdr_sz(msg);
1466 err = msg_errcode(msg);
1467
1468 /* Discard an empty non-errored message & try again */
1469 if ((!sz) && (!err)) {
1470 tsk_advance_rx_queue(sk);
1471 goto restart;
1472 }
1473
1474 /* Optionally capture sender's address & ancillary data of first msg */
1475 if (sz_copied == 0) {
1476 set_orig_addr(m, msg);
1477 res = tipc_sk_anc_data_recv(m, msg, tsk);
1478 if (res)
1479 goto exit;
1480 }
1481
1482 /* Capture message data (if valid) & compute return value (always) */
1483 if (!err) {
1484 u32 offset = TIPC_SKB_CB(buf)->bytes_read;
1485 u32 needed;
1486 int sz_to_copy;
1487
1488 sz -= offset;
1489 needed = (buf_len - sz_copied);
1490 sz_to_copy = min(sz, needed);
1491
1492 res = skb_copy_datagram_msg(buf, hlen + offset, m, sz_to_copy);
1493 if (res)
1494 goto exit;
1495
1496 sz_copied += sz_to_copy;
1497
1498 if (sz_to_copy < sz) {
1499 if (!(flags & MSG_PEEK))
1500 TIPC_SKB_CB(buf)->bytes_read =
1501 offset + sz_to_copy;
1502 goto exit;
1503 }
1504 } else {
1505 if (sz_copied != 0)
1506 goto exit; /* can't add error msg to valid data */
1507
1508 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1509 res = 0;
1510 else
1511 res = -ECONNRESET;
1512 }
1513
1514 if (unlikely(flags & MSG_PEEK))
1515 goto exit;
1516
1517 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1518 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1519 tipc_sk_send_ack(tsk);
1520 tsk_advance_rx_queue(sk);
1521
1522 /* Loop around if more data is required */
1523 if ((sz_copied < buf_len) && /* didn't get all requested data */
1524 (!skb_queue_empty(&sk->sk_receive_queue) ||
1525 (sz_copied < target)) && /* and more is ready or required */
1526 (!err)) /* and haven't reached a FIN */
1527 goto restart;
1528
1529exit:
1530 release_sock(sk);
1531 return sz_copied ? sz_copied : res;
1532}
1533
1534/**
1535 * tipc_write_space - wake up thread if port congestion is released
1536 * @sk: socket
1537 */
1538static void tipc_write_space(struct sock *sk)
1539{
1540 struct socket_wq *wq;
1541
1542 rcu_read_lock();
1543 wq = rcu_dereference(sk->sk_wq);
1544 if (skwq_has_sleeper(wq))
1545 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1546 POLLWRNORM | POLLWRBAND);
1547 rcu_read_unlock();
1548}
1549
1550/**
1551 * tipc_data_ready - wake up threads to indicate messages have been received
1552 * @sk: socket
1553 * @len: the length of messages
1554 */
1555static void tipc_data_ready(struct sock *sk)
1556{
1557 struct socket_wq *wq;
1558
1559 rcu_read_lock();
1560 wq = rcu_dereference(sk->sk_wq);
1561 if (skwq_has_sleeper(wq))
1562 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1563 POLLRDNORM | POLLRDBAND);
1564 rcu_read_unlock();
1565}
1566
1567static void tipc_sock_destruct(struct sock *sk)
1568{
1569 __skb_queue_purge(&sk->sk_receive_queue);
1570}
1571
1572/**
1573 * filter_connect - Handle all incoming messages for a connection-based socket
1574 * @tsk: TIPC socket
1575 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1576 *
1577 * Returns true if everything ok, false otherwise
1578 */
1579static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1580{
1581 struct sock *sk = &tsk->sk;
1582 struct net *net = sock_net(sk);
1583 struct tipc_msg *hdr = buf_msg(skb);
1584
1585 if (unlikely(msg_mcast(hdr)))
1586 return false;
1587
1588 switch (sk->sk_state) {
1589 case TIPC_CONNECTING:
1590 /* Accept only ACK or NACK message */
1591 if (unlikely(!msg_connected(hdr)))
1592 return false;
1593
1594 if (unlikely(msg_errcode(hdr))) {
1595 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1596 sk->sk_err = ECONNREFUSED;
1597 return true;
1598 }
1599
1600 if (unlikely(!msg_isdata(hdr))) {
1601 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1602 sk->sk_err = EINVAL;
1603 return true;
1604 }
1605
1606 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1607 msg_set_importance(&tsk->phdr, msg_importance(hdr));
1608
1609 /* If 'ACK+' message, add to socket receive queue */
1610 if (msg_data_sz(hdr))
1611 return true;
1612
1613 /* If empty 'ACK-' message, wake up sleeping connect() */
1614 if (waitqueue_active(sk_sleep(sk)))
1615 wake_up_interruptible(sk_sleep(sk));
1616
1617 /* 'ACK-' message is neither accepted nor rejected: */
1618 msg_set_dest_droppable(hdr, 1);
1619 return false;
1620
1621 case TIPC_OPEN:
1622 case TIPC_DISCONNECTING:
1623 break;
1624 case TIPC_LISTEN:
1625 /* Accept only SYN message */
1626 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1627 return true;
1628 break;
1629 case TIPC_ESTABLISHED:
1630 /* Accept only connection-based messages sent by peer */
1631 if (unlikely(!tsk_peer_msg(tsk, hdr)))
1632 return false;
1633
1634 if (unlikely(msg_errcode(hdr))) {
1635 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1636 /* Let timer expire on it's own */
1637 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1638 tsk->portid);
1639 sk->sk_state_change(sk);
1640 }
1641 return true;
1642 default:
1643 pr_err("Unknown sk_state %u\n", sk->sk_state);
1644 }
1645
1646 return false;
1647}
1648
1649/**
1650 * rcvbuf_limit - get proper overload limit of socket receive queue
1651 * @sk: socket
1652 * @skb: message
1653 *
1654 * For connection oriented messages, irrespective of importance,
1655 * default queue limit is 2 MB.
1656 *
1657 * For connectionless messages, queue limits are based on message
1658 * importance as follows:
1659 *
1660 * TIPC_LOW_IMPORTANCE (2 MB)
1661 * TIPC_MEDIUM_IMPORTANCE (4 MB)
1662 * TIPC_HIGH_IMPORTANCE (8 MB)
1663 * TIPC_CRITICAL_IMPORTANCE (16 MB)
1664 *
1665 * Returns overload limit according to corresponding message importance
1666 */
1667static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
1668{
1669 struct tipc_sock *tsk = tipc_sk(sk);
1670 struct tipc_msg *hdr = buf_msg(skb);
1671
1672 if (unlikely(!msg_connected(hdr)))
1673 return sk->sk_rcvbuf << msg_importance(hdr);
1674
1675 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
1676 return sk->sk_rcvbuf;
1677
1678 return FLOWCTL_MSG_LIM;
1679}
1680
1681/**
1682 * filter_rcv - validate incoming message
1683 * @sk: socket
1684 * @skb: pointer to message.
1685 *
1686 * Enqueues message on receive queue if acceptable; optionally handles
1687 * disconnect indication for a connected socket.
1688 *
1689 * Called with socket lock already taken
1690 *
1691 * Returns true if message was added to socket receive queue, otherwise false
1692 */
1693static bool filter_rcv(struct sock *sk, struct sk_buff *skb,
1694 struct sk_buff_head *xmitq)
1695{
1696 struct tipc_sock *tsk = tipc_sk(sk);
1697 struct tipc_msg *hdr = buf_msg(skb);
1698 unsigned int limit = rcvbuf_limit(sk, skb);
1699 int err = TIPC_OK;
1700 int usr = msg_user(hdr);
1701
1702 if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1703 tipc_sk_proto_rcv(tsk, skb, xmitq);
1704 return false;
1705 }
1706
1707 if (unlikely(usr == SOCK_WAKEUP)) {
1708 kfree_skb(skb);
1709 tsk->link_cong = 0;
1710 sk->sk_write_space(sk);
1711 return false;
1712 }
1713
1714 /* Drop if illegal message type */
1715 if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1716 kfree_skb(skb);
1717 return false;
1718 }
1719
1720 /* Reject if wrong message type for current socket state */
1721 if (tipc_sk_type_connectionless(sk)) {
1722 if (msg_connected(hdr)) {
1723 err = TIPC_ERR_NO_PORT;
1724 goto reject;
1725 }
1726 } else if (unlikely(!filter_connect(tsk, skb))) {
1727 err = TIPC_ERR_NO_PORT;
1728 goto reject;
1729 }
1730
1731 /* Reject message if there isn't room to queue it */
1732 if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1733 err = TIPC_ERR_OVERLOAD;
1734 goto reject;
1735 }
1736
1737 /* Enqueue message */
1738 TIPC_SKB_CB(skb)->bytes_read = 0;
1739 __skb_queue_tail(&sk->sk_receive_queue, skb);
1740 skb_set_owner_r(skb, sk);
1741
1742 sk->sk_data_ready(sk);
1743 return true;
1744
1745reject:
1746 if (tipc_msg_reverse(tsk_own_node(tsk), &skb, err))
1747 __skb_queue_tail(xmitq, skb);
1748 return false;
1749}
1750
1751/**
1752 * tipc_backlog_rcv - handle incoming message from backlog queue
1753 * @sk: socket
1754 * @skb: message
1755 *
1756 * Caller must hold socket lock
1757 *
1758 * Returns 0
1759 */
1760static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1761{
1762 unsigned int truesize = skb->truesize;
1763 struct sk_buff_head xmitq;
1764 u32 dnode, selector;
1765
1766 __skb_queue_head_init(&xmitq);
1767
1768 if (likely(filter_rcv(sk, skb, &xmitq))) {
1769 atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1770 return 0;
1771 }
1772
1773 if (skb_queue_empty(&xmitq))
1774 return 0;
1775
1776 /* Send response/rejected message */
1777 skb = __skb_dequeue(&xmitq);
1778 dnode = msg_destnode(buf_msg(skb));
1779 selector = msg_origport(buf_msg(skb));
1780 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
1781 return 0;
1782}
1783
1784/**
1785 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1786 * inputq and try adding them to socket or backlog queue
1787 * @inputq: list of incoming buffers with potentially different destinations
1788 * @sk: socket where the buffers should be enqueued
1789 * @dport: port number for the socket
1790 *
1791 * Caller must hold socket lock
1792 */
1793static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1794 u32 dport, struct sk_buff_head *xmitq)
1795{
1796 unsigned long time_limit = jiffies + 2;
1797 struct sk_buff *skb;
1798 unsigned int lim;
1799 atomic_t *dcnt;
1800 u32 onode;
1801
1802 while (skb_queue_len(inputq)) {
1803 if (unlikely(time_after_eq(jiffies, time_limit)))
1804 return;
1805
1806 skb = tipc_skb_dequeue(inputq, dport);
1807 if (unlikely(!skb))
1808 return;
1809
1810 /* Add message directly to receive queue if possible */
1811 if (!sock_owned_by_user(sk)) {
1812 filter_rcv(sk, skb, xmitq);
1813 continue;
1814 }
1815
1816 /* Try backlog, compensating for double-counted bytes */
1817 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1818 if (!sk->sk_backlog.len)
1819 atomic_set(dcnt, 0);
1820 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1821 if (likely(!sk_add_backlog(sk, skb, lim)))
1822 continue;
1823
1824 /* Overload => reject message back to sender */
1825 onode = tipc_own_addr(sock_net(sk));
1826 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
1827 __skb_queue_tail(xmitq, skb);
1828 break;
1829 }
1830}
1831
1832/**
1833 * tipc_sk_rcv - handle a chain of incoming buffers
1834 * @inputq: buffer list containing the buffers
1835 * Consumes all buffers in list until inputq is empty
1836 * Note: may be called in multiple threads referring to the same queue
1837 */
1838void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1839{
1840 struct sk_buff_head xmitq;
1841 u32 dnode, dport = 0;
1842 int err;
1843 struct tipc_sock *tsk;
1844 struct sock *sk;
1845 struct sk_buff *skb;
1846
1847 __skb_queue_head_init(&xmitq);
1848 while (skb_queue_len(inputq)) {
1849 dport = tipc_skb_peek_port(inputq, dport);
1850 tsk = tipc_sk_lookup(net, dport);
1851
1852 if (likely(tsk)) {
1853 sk = &tsk->sk;
1854 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1855 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
1856 spin_unlock_bh(&sk->sk_lock.slock);
1857 }
1858 /* Send pending response/rejected messages, if any */
1859 while ((skb = __skb_dequeue(&xmitq))) {
1860 dnode = msg_destnode(buf_msg(skb));
1861 tipc_node_xmit_skb(net, skb, dnode, dport);
1862 }
1863 sock_put(sk);
1864 continue;
1865 }
1866
1867 /* No destination socket => dequeue skb if still there */
1868 skb = tipc_skb_dequeue(inputq, dport);
1869 if (!skb)
1870 return;
1871
1872 /* Try secondary lookup if unresolved named message */
1873 err = TIPC_ERR_NO_PORT;
1874 if (tipc_msg_lookup_dest(net, skb, &err))
1875 goto xmit;
1876
1877 /* Prepare for message rejection */
1878 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1879 continue;
1880xmit:
1881 dnode = msg_destnode(buf_msg(skb));
1882 tipc_node_xmit_skb(net, skb, dnode, dport);
1883 }
1884}
1885
1886static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1887{
1888 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1889 struct sock *sk = sock->sk;
1890 int done;
1891
1892 do {
1893 int err = sock_error(sk);
1894 if (err)
1895 return err;
1896 if (!*timeo_p)
1897 return -ETIMEDOUT;
1898 if (signal_pending(current))
1899 return sock_intr_errno(*timeo_p);
1900
1901 add_wait_queue(sk_sleep(sk), &wait);
1902 done = sk_wait_event(sk, timeo_p,
1903 sk->sk_state != TIPC_CONNECTING, &wait);
1904 remove_wait_queue(sk_sleep(sk), &wait);
1905 } while (!done);
1906 return 0;
1907}
1908
1909/**
1910 * tipc_connect - establish a connection to another TIPC port
1911 * @sock: socket structure
1912 * @dest: socket address for destination port
1913 * @destlen: size of socket address data structure
1914 * @flags: file-related flags associated with socket
1915 *
1916 * Returns 0 on success, errno otherwise
1917 */
1918static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1919 int destlen, int flags)
1920{
1921 struct sock *sk = sock->sk;
1922 struct tipc_sock *tsk = tipc_sk(sk);
1923 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1924 struct msghdr m = {NULL,};
1925 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1926 int previous;
1927 int res = 0;
1928
1929 lock_sock(sk);
1930
1931 /* DGRAM/RDM connect(), just save the destaddr */
1932 if (tipc_sk_type_connectionless(sk)) {
1933 if (dst->family == AF_UNSPEC) {
1934 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
1935 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1936 res = -EINVAL;
1937 } else {
1938 memcpy(&tsk->peer, dest, destlen);
1939 }
1940 goto exit;
1941 }
1942
1943 /*
1944 * Reject connection attempt using multicast address
1945 *
1946 * Note: send_msg() validates the rest of the address fields,
1947 * so there's no need to do it here
1948 */
1949 if (dst->addrtype == TIPC_ADDR_MCAST) {
1950 res = -EINVAL;
1951 goto exit;
1952 }
1953
1954 previous = sk->sk_state;
1955
1956 switch (sk->sk_state) {
1957 case TIPC_OPEN:
1958 /* Send a 'SYN-' to destination */
1959 m.msg_name = dest;
1960 m.msg_namelen = destlen;
1961
1962 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1963 * indicate send_msg() is never blocked.
1964 */
1965 if (!timeout)
1966 m.msg_flags = MSG_DONTWAIT;
1967
1968 res = __tipc_sendmsg(sock, &m, 0);
1969 if ((res < 0) && (res != -EWOULDBLOCK))
1970 goto exit;
1971
1972 /* Just entered TIPC_CONNECTING state; the only
1973 * difference is that return value in non-blocking
1974 * case is EINPROGRESS, rather than EALREADY.
1975 */
1976 res = -EINPROGRESS;
1977 /* fall thru' */
1978 case TIPC_CONNECTING:
1979 if (!timeout) {
1980 if (previous == TIPC_CONNECTING)
1981 res = -EALREADY;
1982 goto exit;
1983 }
1984 timeout = msecs_to_jiffies(timeout);
1985 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1986 res = tipc_wait_for_connect(sock, &timeout);
1987 break;
1988 case TIPC_ESTABLISHED:
1989 res = -EISCONN;
1990 break;
1991 default:
1992 res = -EINVAL;
1993 }
1994
1995exit:
1996 release_sock(sk);
1997 return res;
1998}
1999
2000/**
2001 * tipc_listen - allow socket to listen for incoming connections
2002 * @sock: socket structure
2003 * @len: (unused)
2004 *
2005 * Returns 0 on success, errno otherwise
2006 */
2007static int tipc_listen(struct socket *sock, int len)
2008{
2009 struct sock *sk = sock->sk;
2010 int res;
2011
2012 lock_sock(sk);
2013 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2014 release_sock(sk);
2015
2016 return res;
2017}
2018
2019static int tipc_wait_for_accept(struct socket *sock, long timeo)
2020{
2021 struct sock *sk = sock->sk;
2022 DEFINE_WAIT(wait);
2023 int err;
2024
2025 /* True wake-one mechanism for incoming connections: only
2026 * one process gets woken up, not the 'whole herd'.
2027 * Since we do not 'race & poll' for established sockets
2028 * anymore, the common case will execute the loop only once.
2029 */
2030 for (;;) {
2031 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2032 TASK_INTERRUPTIBLE);
2033 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2034 release_sock(sk);
2035 timeo = schedule_timeout(timeo);
2036 lock_sock(sk);
2037 }
2038 err = 0;
2039 if (!skb_queue_empty(&sk->sk_receive_queue))
2040 break;
2041 err = -EAGAIN;
2042 if (!timeo)
2043 break;
2044 err = sock_intr_errno(timeo);
2045 if (signal_pending(current))
2046 break;
2047 }
2048 finish_wait(sk_sleep(sk), &wait);
2049 return err;
2050}
2051
2052/**
2053 * tipc_accept - wait for connection request
2054 * @sock: listening socket
2055 * @newsock: new socket that is to be connected
2056 * @flags: file-related flags associated with socket
2057 *
2058 * Returns 0 on success, errno otherwise
2059 */
2060static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2061{
2062 struct sock *new_sk, *sk = sock->sk;
2063 struct sk_buff *buf;
2064 struct tipc_sock *new_tsock;
2065 struct tipc_msg *msg;
2066 long timeo;
2067 int res;
2068
2069 lock_sock(sk);
2070
2071 if (sk->sk_state != TIPC_LISTEN) {
2072 res = -EINVAL;
2073 goto exit;
2074 }
2075 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2076 res = tipc_wait_for_accept(sock, timeo);
2077 if (res)
2078 goto exit;
2079
2080 buf = skb_peek(&sk->sk_receive_queue);
2081
2082 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 0);
2083 if (res)
2084 goto exit;
2085 security_sk_clone(sock->sk, new_sock->sk);
2086
2087 new_sk = new_sock->sk;
2088 new_tsock = tipc_sk(new_sk);
2089 msg = buf_msg(buf);
2090
2091 /* we lock on new_sk; but lockdep sees the lock on sk */
2092 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2093
2094 /*
2095 * Reject any stray messages received by new socket
2096 * before the socket lock was taken (very, very unlikely)
2097 */
2098 tsk_rej_rx_queue(new_sk);
2099
2100 /* Connect new socket to it's peer */
2101 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2102
2103 tsk_set_importance(new_tsock, msg_importance(msg));
2104 if (msg_named(msg)) {
2105 new_tsock->conn_type = msg_nametype(msg);
2106 new_tsock->conn_instance = msg_nameinst(msg);
2107 }
2108
2109 /*
2110 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2111 * Respond to 'SYN+' by queuing it on new socket.
2112 */
2113 if (!msg_data_sz(msg)) {
2114 struct msghdr m = {NULL,};
2115
2116 tsk_advance_rx_queue(sk);
2117 __tipc_send_stream(new_sock, &m, 0);
2118 } else {
2119 __skb_dequeue(&sk->sk_receive_queue);
2120 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2121 skb_set_owner_r(buf, new_sk);
2122 }
2123 release_sock(new_sk);
2124exit:
2125 release_sock(sk);
2126 return res;
2127}
2128
2129/**
2130 * tipc_shutdown - shutdown socket connection
2131 * @sock: socket structure
2132 * @how: direction to close (must be SHUT_RDWR)
2133 *
2134 * Terminates connection (if necessary), then purges socket's receive queue.
2135 *
2136 * Returns 0 on success, errno otherwise
2137 */
2138static int tipc_shutdown(struct socket *sock, int how)
2139{
2140 struct sock *sk = sock->sk;
2141 int res;
2142
2143 if (how != SHUT_RDWR)
2144 return -EINVAL;
2145
2146 lock_sock(sk);
2147
2148 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2149 sk->sk_shutdown = SEND_SHUTDOWN;
2150
2151 if (sk->sk_state == TIPC_DISCONNECTING) {
2152 /* Discard any unreceived messages */
2153 __skb_queue_purge(&sk->sk_receive_queue);
2154
2155 /* Wake up anyone sleeping in poll */
2156 sk->sk_state_change(sk);
2157 res = 0;
2158 } else {
2159 res = -ENOTCONN;
2160 }
2161
2162 release_sock(sk);
2163 return res;
2164}
2165
2166static void tipc_sk_timeout(unsigned long data)
2167{
2168 struct tipc_sock *tsk = (struct tipc_sock *)data;
2169 struct sock *sk = &tsk->sk;
2170 struct sk_buff *skb = NULL;
2171 u32 peer_port, peer_node;
2172 u32 own_node = tsk_own_node(tsk);
2173
2174 bh_lock_sock(sk);
2175 if (!tipc_sk_connected(sk)) {
2176 bh_unlock_sock(sk);
2177 goto exit;
2178 }
2179 peer_port = tsk_peer_port(tsk);
2180 peer_node = tsk_peer_node(tsk);
2181
2182 if (tsk->probe_unacked) {
2183 if (!sock_owned_by_user(sk)) {
2184 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2185 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2186 tsk_peer_port(tsk));
2187 sk->sk_state_change(sk);
2188 } else {
2189 /* Try again later */
2190 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2191 }
2192
2193 bh_unlock_sock(sk);
2194 goto exit;
2195 }
2196
2197 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2198 INT_H_SIZE, 0, peer_node, own_node,
2199 peer_port, tsk->portid, TIPC_OK);
2200 tsk->probe_unacked = true;
2201 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
2202 bh_unlock_sock(sk);
2203 if (skb)
2204 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2205exit:
2206 sock_put(sk);
2207}
2208
2209static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2210 struct tipc_name_seq const *seq)
2211{
2212 struct sock *sk = &tsk->sk;
2213 struct net *net = sock_net(sk);
2214 struct publication *publ;
2215 u32 key;
2216
2217 if (tipc_sk_connected(sk))
2218 return -EINVAL;
2219 key = tsk->portid + tsk->pub_count + 1;
2220 if (key == tsk->portid)
2221 return -EADDRINUSE;
2222
2223 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2224 scope, tsk->portid, key);
2225 if (unlikely(!publ))
2226 return -EINVAL;
2227
2228 list_add(&publ->pport_list, &tsk->publications);
2229 tsk->pub_count++;
2230 tsk->published = 1;
2231 return 0;
2232}
2233
2234static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2235 struct tipc_name_seq const *seq)
2236{
2237 struct net *net = sock_net(&tsk->sk);
2238 struct publication *publ;
2239 struct publication *safe;
2240 int rc = -EINVAL;
2241
2242 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2243 if (seq) {
2244 if (publ->scope != scope)
2245 continue;
2246 if (publ->type != seq->type)
2247 continue;
2248 if (publ->lower != seq->lower)
2249 continue;
2250 if (publ->upper != seq->upper)
2251 break;
2252 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2253 publ->ref, publ->key);
2254 rc = 0;
2255 break;
2256 }
2257 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2258 publ->ref, publ->key);
2259 rc = 0;
2260 }
2261 if (list_empty(&tsk->publications))
2262 tsk->published = 0;
2263 return rc;
2264}
2265
2266/* tipc_sk_reinit: set non-zero address in all existing sockets
2267 * when we go from standalone to network mode.
2268 */
2269void tipc_sk_reinit(struct net *net)
2270{
2271 struct tipc_net *tn = net_generic(net, tipc_net_id);
2272 const struct bucket_table *tbl;
2273 struct rhash_head *pos;
2274 struct tipc_sock *tsk;
2275 struct tipc_msg *msg;
2276 int i;
2277
2278 rcu_read_lock();
2279 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2280 for (i = 0; i < tbl->size; i++) {
2281 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2282 spin_lock_bh(&tsk->sk.sk_lock.slock);
2283 msg = &tsk->phdr;
2284 msg_set_prevnode(msg, tn->own_addr);
2285 msg_set_orignode(msg, tn->own_addr);
2286 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2287 }
2288 }
2289 rcu_read_unlock();
2290}
2291
2292static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2293{
2294 struct tipc_net *tn = net_generic(net, tipc_net_id);
2295 struct tipc_sock *tsk;
2296
2297 rcu_read_lock();
2298 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2299 if (tsk)
2300 sock_hold(&tsk->sk);
2301 rcu_read_unlock();
2302
2303 return tsk;
2304}
2305
2306static int tipc_sk_insert(struct tipc_sock *tsk)
2307{
2308 struct sock *sk = &tsk->sk;
2309 struct net *net = sock_net(sk);
2310 struct tipc_net *tn = net_generic(net, tipc_net_id);
2311 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2312 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2313
2314 while (remaining--) {
2315 portid++;
2316 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2317 portid = TIPC_MIN_PORT;
2318 tsk->portid = portid;
2319 sock_hold(&tsk->sk);
2320 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2321 tsk_rht_params))
2322 return 0;
2323 sock_put(&tsk->sk);
2324 }
2325
2326 return -1;
2327}
2328
2329static void tipc_sk_remove(struct tipc_sock *tsk)
2330{
2331 struct sock *sk = &tsk->sk;
2332 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2333
2334 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2335 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2336 __sock_put(sk);
2337 }
2338}
2339
2340static const struct rhashtable_params tsk_rht_params = {
2341 .nelem_hint = 192,
2342 .head_offset = offsetof(struct tipc_sock, node),
2343 .key_offset = offsetof(struct tipc_sock, portid),
2344 .key_len = sizeof(u32), /* portid */
2345 .max_size = 1048576,
2346 .min_size = 256,
2347 .automatic_shrinking = true,
2348};
2349
2350int tipc_sk_rht_init(struct net *net)
2351{
2352 struct tipc_net *tn = net_generic(net, tipc_net_id);
2353
2354 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2355}
2356
2357void tipc_sk_rht_destroy(struct net *net)
2358{
2359 struct tipc_net *tn = net_generic(net, tipc_net_id);
2360
2361 /* Wait for socket readers to complete */
2362 synchronize_net();
2363
2364 rhashtable_destroy(&tn->sk_rht);
2365}
2366
2367/**
2368 * tipc_setsockopt - set socket option
2369 * @sock: socket structure
2370 * @lvl: option level
2371 * @opt: option identifier
2372 * @ov: pointer to new option value
2373 * @ol: length of option value
2374 *
2375 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2376 * (to ease compatibility).
2377 *
2378 * Returns 0 on success, errno otherwise
2379 */
2380static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2381 char __user *ov, unsigned int ol)
2382{
2383 struct sock *sk = sock->sk;
2384 struct tipc_sock *tsk = tipc_sk(sk);
2385 u32 value;
2386 int res;
2387
2388 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2389 return 0;
2390 if (lvl != SOL_TIPC)
2391 return -ENOPROTOOPT;
2392 if (ol < sizeof(value))
2393 return -EINVAL;
2394 res = get_user(value, (u32 __user *)ov);
2395 if (res)
2396 return res;
2397
2398 lock_sock(sk);
2399
2400 switch (opt) {
2401 case TIPC_IMPORTANCE:
2402 res = tsk_set_importance(tsk, value);
2403 break;
2404 case TIPC_SRC_DROPPABLE:
2405 if (sock->type != SOCK_STREAM)
2406 tsk_set_unreliable(tsk, value);
2407 else
2408 res = -ENOPROTOOPT;
2409 break;
2410 case TIPC_DEST_DROPPABLE:
2411 tsk_set_unreturnable(tsk, value);
2412 break;
2413 case TIPC_CONN_TIMEOUT:
2414 tipc_sk(sk)->conn_timeout = value;
2415 /* no need to set "res", since already 0 at this point */
2416 break;
2417 default:
2418 res = -EINVAL;
2419 }
2420
2421 release_sock(sk);
2422
2423 return res;
2424}
2425
2426/**
2427 * tipc_getsockopt - get socket option
2428 * @sock: socket structure
2429 * @lvl: option level
2430 * @opt: option identifier
2431 * @ov: receptacle for option value
2432 * @ol: receptacle for length of option value
2433 *
2434 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2435 * (to ease compatibility).
2436 *
2437 * Returns 0 on success, errno otherwise
2438 */
2439static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2440 char __user *ov, int __user *ol)
2441{
2442 struct sock *sk = sock->sk;
2443 struct tipc_sock *tsk = tipc_sk(sk);
2444 int len;
2445 u32 value;
2446 int res;
2447
2448 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2449 return put_user(0, ol);
2450 if (lvl != SOL_TIPC)
2451 return -ENOPROTOOPT;
2452 res = get_user(len, ol);
2453 if (res)
2454 return res;
2455
2456 lock_sock(sk);
2457
2458 switch (opt) {
2459 case TIPC_IMPORTANCE:
2460 value = tsk_importance(tsk);
2461 break;
2462 case TIPC_SRC_DROPPABLE:
2463 value = tsk_unreliable(tsk);
2464 break;
2465 case TIPC_DEST_DROPPABLE:
2466 value = tsk_unreturnable(tsk);
2467 break;
2468 case TIPC_CONN_TIMEOUT:
2469 value = tsk->conn_timeout;
2470 /* no need to set "res", since already 0 at this point */
2471 break;
2472 case TIPC_NODE_RECVQ_DEPTH:
2473 value = 0; /* was tipc_queue_size, now obsolete */
2474 break;
2475 case TIPC_SOCK_RECVQ_DEPTH:
2476 value = skb_queue_len(&sk->sk_receive_queue);
2477 break;
2478 default:
2479 res = -EINVAL;
2480 }
2481
2482 release_sock(sk);
2483
2484 if (res)
2485 return res; /* "get" failed */
2486
2487 if (len < sizeof(value))
2488 return -EINVAL;
2489
2490 if (copy_to_user(ov, &value, sizeof(value)))
2491 return -EFAULT;
2492
2493 return put_user(sizeof(value), ol);
2494}
2495
2496static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2497{
2498 struct sock *sk = sock->sk;
2499 struct tipc_sioc_ln_req lnr;
2500 void __user *argp = (void __user *)arg;
2501
2502 switch (cmd) {
2503 case SIOCGETLINKNAME:
2504 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2505 return -EFAULT;
2506 if (!tipc_node_get_linkname(sock_net(sk),
2507 lnr.bearer_id & 0xffff, lnr.peer,
2508 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2509 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2510 return -EFAULT;
2511 return 0;
2512 }
2513 return -EADDRNOTAVAIL;
2514 default:
2515 return -ENOIOCTLCMD;
2516 }
2517}
2518
2519/* Protocol switches for the various types of TIPC sockets */
2520
2521static const struct proto_ops msg_ops = {
2522 .owner = THIS_MODULE,
2523 .family = AF_TIPC,
2524 .release = tipc_release,
2525 .bind = tipc_bind,
2526 .connect = tipc_connect,
2527 .socketpair = sock_no_socketpair,
2528 .accept = sock_no_accept,
2529 .getname = tipc_getname,
2530 .poll = tipc_poll,
2531 .ioctl = tipc_ioctl,
2532 .listen = sock_no_listen,
2533 .shutdown = tipc_shutdown,
2534 .setsockopt = tipc_setsockopt,
2535 .getsockopt = tipc_getsockopt,
2536 .sendmsg = tipc_sendmsg,
2537 .recvmsg = tipc_recvmsg,
2538 .mmap = sock_no_mmap,
2539 .sendpage = sock_no_sendpage
2540};
2541
2542static const struct proto_ops packet_ops = {
2543 .owner = THIS_MODULE,
2544 .family = AF_TIPC,
2545 .release = tipc_release,
2546 .bind = tipc_bind,
2547 .connect = tipc_connect,
2548 .socketpair = sock_no_socketpair,
2549 .accept = tipc_accept,
2550 .getname = tipc_getname,
2551 .poll = tipc_poll,
2552 .ioctl = tipc_ioctl,
2553 .listen = tipc_listen,
2554 .shutdown = tipc_shutdown,
2555 .setsockopt = tipc_setsockopt,
2556 .getsockopt = tipc_getsockopt,
2557 .sendmsg = tipc_send_packet,
2558 .recvmsg = tipc_recvmsg,
2559 .mmap = sock_no_mmap,
2560 .sendpage = sock_no_sendpage
2561};
2562
2563static const struct proto_ops stream_ops = {
2564 .owner = THIS_MODULE,
2565 .family = AF_TIPC,
2566 .release = tipc_release,
2567 .bind = tipc_bind,
2568 .connect = tipc_connect,
2569 .socketpair = sock_no_socketpair,
2570 .accept = tipc_accept,
2571 .getname = tipc_getname,
2572 .poll = tipc_poll,
2573 .ioctl = tipc_ioctl,
2574 .listen = tipc_listen,
2575 .shutdown = tipc_shutdown,
2576 .setsockopt = tipc_setsockopt,
2577 .getsockopt = tipc_getsockopt,
2578 .sendmsg = tipc_send_stream,
2579 .recvmsg = tipc_recv_stream,
2580 .mmap = sock_no_mmap,
2581 .sendpage = sock_no_sendpage
2582};
2583
2584static const struct net_proto_family tipc_family_ops = {
2585 .owner = THIS_MODULE,
2586 .family = AF_TIPC,
2587 .create = tipc_sk_create
2588};
2589
2590static struct proto tipc_proto = {
2591 .name = "TIPC",
2592 .owner = THIS_MODULE,
2593 .obj_size = sizeof(struct tipc_sock),
2594 .sysctl_rmem = sysctl_tipc_rmem
2595};
2596
2597/**
2598 * tipc_socket_init - initialize TIPC socket interface
2599 *
2600 * Returns 0 on success, errno otherwise
2601 */
2602int tipc_socket_init(void)
2603{
2604 int res;
2605
2606 res = proto_register(&tipc_proto, 1);
2607 if (res) {
2608 pr_err("Failed to register TIPC protocol type\n");
2609 goto out;
2610 }
2611
2612 res = sock_register(&tipc_family_ops);
2613 if (res) {
2614 pr_err("Failed to register TIPC socket type\n");
2615 proto_unregister(&tipc_proto);
2616 goto out;
2617 }
2618 out:
2619 return res;
2620}
2621
2622/**
2623 * tipc_socket_stop - stop TIPC socket interface
2624 */
2625void tipc_socket_stop(void)
2626{
2627 sock_unregister(tipc_family_ops.family);
2628 proto_unregister(&tipc_proto);
2629}
2630
2631/* Caller should hold socket lock for the passed tipc socket. */
2632static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2633{
2634 u32 peer_node;
2635 u32 peer_port;
2636 struct nlattr *nest;
2637
2638 peer_node = tsk_peer_node(tsk);
2639 peer_port = tsk_peer_port(tsk);
2640
2641 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2642
2643 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2644 goto msg_full;
2645 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2646 goto msg_full;
2647
2648 if (tsk->conn_type != 0) {
2649 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2650 goto msg_full;
2651 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2652 goto msg_full;
2653 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2654 goto msg_full;
2655 }
2656 nla_nest_end(skb, nest);
2657
2658 return 0;
2659
2660msg_full:
2661 nla_nest_cancel(skb, nest);
2662
2663 return -EMSGSIZE;
2664}
2665
2666/* Caller should hold socket lock for the passed tipc socket. */
2667static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2668 struct tipc_sock *tsk)
2669{
2670 int err;
2671 void *hdr;
2672 struct nlattr *attrs;
2673 struct net *net = sock_net(skb->sk);
2674 struct tipc_net *tn = net_generic(net, tipc_net_id);
2675 struct sock *sk = &tsk->sk;
2676
2677 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2678 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2679 if (!hdr)
2680 goto msg_cancel;
2681
2682 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2683 if (!attrs)
2684 goto genlmsg_cancel;
2685 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2686 goto attr_msg_cancel;
2687 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2688 goto attr_msg_cancel;
2689
2690 if (tipc_sk_connected(sk)) {
2691 err = __tipc_nl_add_sk_con(skb, tsk);
2692 if (err)
2693 goto attr_msg_cancel;
2694 } else if (!list_empty(&tsk->publications)) {
2695 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2696 goto attr_msg_cancel;
2697 }
2698 nla_nest_end(skb, attrs);
2699 genlmsg_end(skb, hdr);
2700
2701 return 0;
2702
2703attr_msg_cancel:
2704 nla_nest_cancel(skb, attrs);
2705genlmsg_cancel:
2706 genlmsg_cancel(skb, hdr);
2707msg_cancel:
2708 return -EMSGSIZE;
2709}
2710
2711int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2712{
2713 int err;
2714 struct tipc_sock *tsk;
2715 const struct bucket_table *tbl;
2716 struct rhash_head *pos;
2717 struct net *net = sock_net(skb->sk);
2718 struct tipc_net *tn = net_generic(net, tipc_net_id);
2719 u32 tbl_id = cb->args[0];
2720 u32 prev_portid = cb->args[1];
2721
2722 rcu_read_lock();
2723 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2724 for (; tbl_id < tbl->size; tbl_id++) {
2725 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2726 spin_lock_bh(&tsk->sk.sk_lock.slock);
2727 if (prev_portid && prev_portid != tsk->portid) {
2728 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2729 continue;
2730 }
2731
2732 err = __tipc_nl_add_sk(skb, cb, tsk);
2733 if (err) {
2734 prev_portid = tsk->portid;
2735 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2736 goto out;
2737 }
2738 prev_portid = 0;
2739 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2740 }
2741 }
2742out:
2743 rcu_read_unlock();
2744 cb->args[0] = tbl_id;
2745 cb->args[1] = prev_portid;
2746
2747 return skb->len;
2748}
2749
2750/* Caller should hold socket lock for the passed tipc socket. */
2751static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2752 struct netlink_callback *cb,
2753 struct publication *publ)
2754{
2755 void *hdr;
2756 struct nlattr *attrs;
2757
2758 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2759 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2760 if (!hdr)
2761 goto msg_cancel;
2762
2763 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2764 if (!attrs)
2765 goto genlmsg_cancel;
2766
2767 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2768 goto attr_msg_cancel;
2769 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2770 goto attr_msg_cancel;
2771 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2772 goto attr_msg_cancel;
2773 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2774 goto attr_msg_cancel;
2775
2776 nla_nest_end(skb, attrs);
2777 genlmsg_end(skb, hdr);
2778
2779 return 0;
2780
2781attr_msg_cancel:
2782 nla_nest_cancel(skb, attrs);
2783genlmsg_cancel:
2784 genlmsg_cancel(skb, hdr);
2785msg_cancel:
2786 return -EMSGSIZE;
2787}
2788
2789/* Caller should hold socket lock for the passed tipc socket. */
2790static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2791 struct netlink_callback *cb,
2792 struct tipc_sock *tsk, u32 *last_publ)
2793{
2794 int err;
2795 struct publication *p;
2796
2797 if (*last_publ) {
2798 list_for_each_entry(p, &tsk->publications, pport_list) {
2799 if (p->key == *last_publ)
2800 break;
2801 }
2802 if (p->key != *last_publ) {
2803 /* We never set seq or call nl_dump_check_consistent()
2804 * this means that setting prev_seq here will cause the
2805 * consistence check to fail in the netlink callback
2806 * handler. Resulting in the last NLMSG_DONE message
2807 * having the NLM_F_DUMP_INTR flag set.
2808 */
2809 cb->prev_seq = 1;
2810 *last_publ = 0;
2811 return -EPIPE;
2812 }
2813 } else {
2814 p = list_first_entry(&tsk->publications, struct publication,
2815 pport_list);
2816 }
2817
2818 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2819 err = __tipc_nl_add_sk_publ(skb, cb, p);
2820 if (err) {
2821 *last_publ = p->key;
2822 return err;
2823 }
2824 }
2825 *last_publ = 0;
2826
2827 return 0;
2828}
2829
2830int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2831{
2832 int err;
2833 u32 tsk_portid = cb->args[0];
2834 u32 last_publ = cb->args[1];
2835 u32 done = cb->args[2];
2836 struct net *net = sock_net(skb->sk);
2837 struct tipc_sock *tsk;
2838
2839 if (!tsk_portid) {
2840 struct nlattr **attrs;
2841 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2842
2843 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2844 if (err)
2845 return err;
2846
2847 if (!attrs[TIPC_NLA_SOCK])
2848 return -EINVAL;
2849
2850 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2851 attrs[TIPC_NLA_SOCK],
2852 tipc_nl_sock_policy);
2853 if (err)
2854 return err;
2855
2856 if (!sock[TIPC_NLA_SOCK_REF])
2857 return -EINVAL;
2858
2859 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2860 }
2861
2862 if (done)
2863 return 0;
2864
2865 tsk = tipc_sk_lookup(net, tsk_portid);
2866 if (!tsk)
2867 return -EINVAL;
2868
2869 lock_sock(&tsk->sk);
2870 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2871 if (!err)
2872 done = 1;
2873 release_sock(&tsk->sk);
2874 sock_put(&tsk->sk);
2875
2876 cb->args[0] = tsk_portid;
2877 cb->args[1] = last_publ;
2878 cb->args[2] = done;
2879
2880 return skb->len;
2881}
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 <linux/export.h>
38#include <net/sock.h>
39
40#include "core.h"
41#include "port.h"
42
43#define SS_LISTENING -1 /* socket is listening */
44#define SS_READY -2 /* socket is connectionless */
45
46#define OVERLOAD_LIMIT_BASE 5000
47#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
48
49struct tipc_sock {
50 struct sock sk;
51 struct tipc_port *p;
52 struct tipc_portid peer_name;
53 unsigned int conn_timeout;
54};
55
56#define tipc_sk(sk) ((struct tipc_sock *)(sk))
57#define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
58
59#define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
60 (sock->state == SS_DISCONNECTING))
61
62static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
63static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
64static void wakeupdispatch(struct tipc_port *tport);
65
66static const struct proto_ops packet_ops;
67static const struct proto_ops stream_ops;
68static const struct proto_ops msg_ops;
69
70static struct proto tipc_proto;
71
72static int sockets_enabled;
73
74static atomic_t tipc_queue_size = ATOMIC_INIT(0);
75
76/*
77 * Revised TIPC socket locking policy:
78 *
79 * Most socket operations take the standard socket lock when they start
80 * and hold it until they finish (or until they need to sleep). Acquiring
81 * this lock grants the owner exclusive access to the fields of the socket
82 * data structures, with the exception of the backlog queue. A few socket
83 * operations can be done without taking the socket lock because they only
84 * read socket information that never changes during the life of the socket.
85 *
86 * Socket operations may acquire the lock for the associated TIPC port if they
87 * need to perform an operation on the port. If any routine needs to acquire
88 * both the socket lock and the port lock it must take the socket lock first
89 * to avoid the risk of deadlock.
90 *
91 * The dispatcher handling incoming messages cannot grab the socket lock in
92 * the standard fashion, since invoked it runs at the BH level and cannot block.
93 * Instead, it checks to see if the socket lock is currently owned by someone,
94 * and either handles the message itself or adds it to the socket's backlog
95 * queue; in the latter case the queued message is processed once the process
96 * owning the socket lock releases it.
97 *
98 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
99 * the problem of a blocked socket operation preventing any other operations
100 * from occurring. However, applications must be careful if they have
101 * multiple threads trying to send (or receive) on the same socket, as these
102 * operations might interfere with each other. For example, doing a connect
103 * and a receive at the same time might allow the receive to consume the
104 * ACK message meant for the connect. While additional work could be done
105 * to try and overcome this, it doesn't seem to be worthwhile at the present.
106 *
107 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
108 * that another operation that must be performed in a non-blocking manner is
109 * not delayed for very long because the lock has already been taken.
110 *
111 * NOTE: This code assumes that certain fields of a port/socket pair are
112 * constant over its lifetime; such fields can be examined without taking
113 * the socket lock and/or port lock, and do not need to be re-read even
114 * after resuming processing after waiting. These fields include:
115 * - socket type
116 * - pointer to socket sk structure (aka tipc_sock structure)
117 * - pointer to port structure
118 * - port reference
119 */
120
121/**
122 * advance_rx_queue - discard first buffer in socket receive queue
123 *
124 * Caller must hold socket lock
125 */
126static void advance_rx_queue(struct sock *sk)
127{
128 kfree_skb(__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 */
137static void discard_rx_queue(struct sock *sk)
138{
139 struct sk_buff *buf;
140
141 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
142 atomic_dec(&tipc_queue_size);
143 kfree_skb(buf);
144 }
145}
146
147/**
148 * reject_rx_queue - reject all buffers in socket receive queue
149 *
150 * Caller must hold socket lock
151 */
152static void reject_rx_queue(struct sock *sk)
153{
154 struct sk_buff *buf;
155
156 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
157 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
158 atomic_dec(&tipc_queue_size);
159 }
160}
161
162/**
163 * tipc_create - create a TIPC socket
164 * @net: network namespace (must be default network)
165 * @sock: pre-allocated socket structure
166 * @protocol: protocol indicator (must be 0)
167 * @kern: caused by kernel or by userspace?
168 *
169 * This routine creates additional data structures used by the TIPC socket,
170 * initializes them, and links them together.
171 *
172 * Returns 0 on success, errno otherwise
173 */
174static int tipc_create(struct net *net, struct socket *sock, int protocol,
175 int kern)
176{
177 const struct proto_ops *ops;
178 socket_state state;
179 struct sock *sk;
180 struct tipc_port *tp_ptr;
181
182 /* Validate arguments */
183 if (unlikely(protocol != 0))
184 return -EPROTONOSUPPORT;
185
186 switch (sock->type) {
187 case SOCK_STREAM:
188 ops = &stream_ops;
189 state = SS_UNCONNECTED;
190 break;
191 case SOCK_SEQPACKET:
192 ops = &packet_ops;
193 state = SS_UNCONNECTED;
194 break;
195 case SOCK_DGRAM:
196 case SOCK_RDM:
197 ops = &msg_ops;
198 state = SS_READY;
199 break;
200 default:
201 return -EPROTOTYPE;
202 }
203
204 /* Allocate socket's protocol area */
205 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
206 if (sk == NULL)
207 return -ENOMEM;
208
209 /* Allocate TIPC port for socket to use */
210 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
211 TIPC_LOW_IMPORTANCE);
212 if (unlikely(!tp_ptr)) {
213 sk_free(sk);
214 return -ENOMEM;
215 }
216
217 /* Finish initializing socket data structures */
218 sock->ops = ops;
219 sock->state = state;
220
221 sock_init_data(sock, sk);
222 sk->sk_backlog_rcv = backlog_rcv;
223 tipc_sk(sk)->p = tp_ptr;
224 tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT;
225
226 spin_unlock_bh(tp_ptr->lock);
227
228 if (sock->state == SS_READY) {
229 tipc_set_portunreturnable(tp_ptr->ref, 1);
230 if (sock->type == SOCK_DGRAM)
231 tipc_set_portunreliable(tp_ptr->ref, 1);
232 }
233
234 return 0;
235}
236
237/**
238 * release - destroy a TIPC socket
239 * @sock: socket to destroy
240 *
241 * This routine cleans up any messages that are still queued on the socket.
242 * For DGRAM and RDM socket types, all queued messages are rejected.
243 * For SEQPACKET and STREAM socket types, the first message is rejected
244 * and any others are discarded. (If the first message on a STREAM socket
245 * is partially-read, it is discarded and the next one is rejected instead.)
246 *
247 * NOTE: Rejected messages are not necessarily returned to the sender! They
248 * are returned or discarded according to the "destination droppable" setting
249 * specified for the message by the sender.
250 *
251 * Returns 0 on success, errno otherwise
252 */
253static int release(struct socket *sock)
254{
255 struct sock *sk = sock->sk;
256 struct tipc_port *tport;
257 struct sk_buff *buf;
258 int res;
259
260 /*
261 * Exit if socket isn't fully initialized (occurs when a failed accept()
262 * releases a pre-allocated child socket that was never used)
263 */
264 if (sk == NULL)
265 return 0;
266
267 tport = tipc_sk_port(sk);
268 lock_sock(sk);
269
270 /*
271 * Reject all unreceived messages, except on an active connection
272 * (which disconnects locally & sends a 'FIN+' to peer)
273 */
274 while (sock->state != SS_DISCONNECTING) {
275 buf = __skb_dequeue(&sk->sk_receive_queue);
276 if (buf == NULL)
277 break;
278 atomic_dec(&tipc_queue_size);
279 if (TIPC_SKB_CB(buf)->handle != 0)
280 kfree_skb(buf);
281 else {
282 if ((sock->state == SS_CONNECTING) ||
283 (sock->state == SS_CONNECTED)) {
284 sock->state = SS_DISCONNECTING;
285 tipc_disconnect(tport->ref);
286 }
287 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
288 }
289 }
290
291 /*
292 * Delete TIPC port; this ensures no more messages are queued
293 * (also disconnects an active connection & sends a 'FIN-' to peer)
294 */
295 res = tipc_deleteport(tport->ref);
296
297 /* Discard any remaining (connection-based) messages in receive queue */
298 discard_rx_queue(sk);
299
300 /* Reject any messages that accumulated in backlog queue */
301 sock->state = SS_DISCONNECTING;
302 release_sock(sk);
303
304 sock_put(sk);
305 sock->sk = NULL;
306
307 return res;
308}
309
310/**
311 * bind - associate or disassocate TIPC name(s) with a socket
312 * @sock: socket structure
313 * @uaddr: socket address describing name(s) and desired operation
314 * @uaddr_len: size of socket address data structure
315 *
316 * Name and name sequence binding is indicated using a positive scope value;
317 * a negative scope value unbinds the specified name. Specifying no name
318 * (i.e. a socket address length of 0) unbinds all names from the socket.
319 *
320 * Returns 0 on success, errno otherwise
321 *
322 * NOTE: This routine doesn't need to take the socket lock since it doesn't
323 * access any non-constant socket information.
324 */
325static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
326{
327 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
328 u32 portref = tipc_sk_port(sock->sk)->ref;
329
330 if (unlikely(!uaddr_len))
331 return tipc_withdraw(portref, 0, NULL);
332
333 if (uaddr_len < sizeof(struct sockaddr_tipc))
334 return -EINVAL;
335 if (addr->family != AF_TIPC)
336 return -EAFNOSUPPORT;
337
338 if (addr->addrtype == TIPC_ADDR_NAME)
339 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
340 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
341 return -EAFNOSUPPORT;
342
343 if (addr->addr.nameseq.type < TIPC_RESERVED_TYPES)
344 return -EACCES;
345
346 return (addr->scope > 0) ?
347 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
348 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
349}
350
351/**
352 * get_name - get port ID of socket or peer socket
353 * @sock: socket structure
354 * @uaddr: area for returned socket address
355 * @uaddr_len: area for returned length of socket address
356 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
357 *
358 * Returns 0 on success, errno otherwise
359 *
360 * NOTE: This routine doesn't need to take the socket lock since it only
361 * accesses socket information that is unchanging (or which changes in
362 * a completely predictable manner).
363 */
364static int get_name(struct socket *sock, struct sockaddr *uaddr,
365 int *uaddr_len, int peer)
366{
367 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
368 struct tipc_sock *tsock = tipc_sk(sock->sk);
369
370 memset(addr, 0, sizeof(*addr));
371 if (peer) {
372 if ((sock->state != SS_CONNECTED) &&
373 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
374 return -ENOTCONN;
375 addr->addr.id.ref = tsock->peer_name.ref;
376 addr->addr.id.node = tsock->peer_name.node;
377 } else {
378 addr->addr.id.ref = tsock->p->ref;
379 addr->addr.id.node = tipc_own_addr;
380 }
381
382 *uaddr_len = sizeof(*addr);
383 addr->addrtype = TIPC_ADDR_ID;
384 addr->family = AF_TIPC;
385 addr->scope = 0;
386 addr->addr.name.domain = 0;
387
388 return 0;
389}
390
391/**
392 * poll - read and possibly block on pollmask
393 * @file: file structure associated with the socket
394 * @sock: socket for which to calculate the poll bits
395 * @wait: ???
396 *
397 * Returns pollmask value
398 *
399 * COMMENTARY:
400 * It appears that the usual socket locking mechanisms are not useful here
401 * since the pollmask info is potentially out-of-date the moment this routine
402 * exits. TCP and other protocols seem to rely on higher level poll routines
403 * to handle any preventable race conditions, so TIPC will do the same ...
404 *
405 * TIPC sets the returned events as follows:
406 *
407 * socket state flags set
408 * ------------ ---------
409 * unconnected no read flags
410 * no write flags
411 *
412 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
413 * no write flags
414 *
415 * connected POLLIN/POLLRDNORM if data in rx queue
416 * POLLOUT if port is not congested
417 *
418 * disconnecting POLLIN/POLLRDNORM/POLLHUP
419 * no write flags
420 *
421 * listening POLLIN if SYN in rx queue
422 * no write flags
423 *
424 * ready POLLIN/POLLRDNORM if data in rx queue
425 * [connectionless] POLLOUT (since port cannot be congested)
426 *
427 * IMPORTANT: The fact that a read or write operation is indicated does NOT
428 * imply that the operation will succeed, merely that it should be performed
429 * and will not block.
430 */
431static unsigned int poll(struct file *file, struct socket *sock,
432 poll_table *wait)
433{
434 struct sock *sk = sock->sk;
435 u32 mask = 0;
436
437 poll_wait(file, sk_sleep(sk), wait);
438
439 switch ((int)sock->state) {
440 case SS_READY:
441 case SS_CONNECTED:
442 if (!tipc_sk_port(sk)->congested)
443 mask |= POLLOUT;
444 /* fall thru' */
445 case SS_CONNECTING:
446 case SS_LISTENING:
447 if (!skb_queue_empty(&sk->sk_receive_queue))
448 mask |= (POLLIN | POLLRDNORM);
449 break;
450 case SS_DISCONNECTING:
451 mask = (POLLIN | POLLRDNORM | POLLHUP);
452 break;
453 }
454
455 return mask;
456}
457
458/**
459 * dest_name_check - verify user is permitted to send to specified port name
460 * @dest: destination address
461 * @m: descriptor for message to be sent
462 *
463 * Prevents restricted configuration commands from being issued by
464 * unauthorized users.
465 *
466 * Returns 0 if permission is granted, otherwise errno
467 */
468static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
469{
470 struct tipc_cfg_msg_hdr hdr;
471
472 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
473 return 0;
474 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
475 return 0;
476 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
477 return -EACCES;
478
479 if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
480 return -EMSGSIZE;
481 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
482 return -EFAULT;
483 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
484 return -EACCES;
485
486 return 0;
487}
488
489/**
490 * send_msg - send message in connectionless manner
491 * @iocb: if NULL, indicates that socket lock is already held
492 * @sock: socket structure
493 * @m: message to send
494 * @total_len: length of message
495 *
496 * Message must have an destination specified explicitly.
497 * Used for SOCK_RDM and SOCK_DGRAM messages,
498 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
499 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
500 *
501 * Returns the number of bytes sent on success, or errno otherwise
502 */
503static int send_msg(struct kiocb *iocb, struct socket *sock,
504 struct msghdr *m, size_t total_len)
505{
506 struct sock *sk = sock->sk;
507 struct tipc_port *tport = tipc_sk_port(sk);
508 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
509 int needs_conn;
510 long timeout_val;
511 int res = -EINVAL;
512
513 if (unlikely(!dest))
514 return -EDESTADDRREQ;
515 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
516 (dest->family != AF_TIPC)))
517 return -EINVAL;
518 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
519 (m->msg_iovlen > (unsigned int)INT_MAX))
520 return -EMSGSIZE;
521
522 if (iocb)
523 lock_sock(sk);
524
525 needs_conn = (sock->state != SS_READY);
526 if (unlikely(needs_conn)) {
527 if (sock->state == SS_LISTENING) {
528 res = -EPIPE;
529 goto exit;
530 }
531 if (sock->state != SS_UNCONNECTED) {
532 res = -EISCONN;
533 goto exit;
534 }
535 if ((tport->published) ||
536 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
537 res = -EOPNOTSUPP;
538 goto exit;
539 }
540 if (dest->addrtype == TIPC_ADDR_NAME) {
541 tport->conn_type = dest->addr.name.name.type;
542 tport->conn_instance = dest->addr.name.name.instance;
543 }
544
545 /* Abort any pending connection attempts (very unlikely) */
546 reject_rx_queue(sk);
547 }
548
549 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
550
551 do {
552 if (dest->addrtype == TIPC_ADDR_NAME) {
553 res = dest_name_check(dest, m);
554 if (res)
555 break;
556 res = tipc_send2name(tport->ref,
557 &dest->addr.name.name,
558 dest->addr.name.domain,
559 m->msg_iovlen,
560 m->msg_iov,
561 total_len);
562 } else if (dest->addrtype == TIPC_ADDR_ID) {
563 res = tipc_send2port(tport->ref,
564 &dest->addr.id,
565 m->msg_iovlen,
566 m->msg_iov,
567 total_len);
568 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
569 if (needs_conn) {
570 res = -EOPNOTSUPP;
571 break;
572 }
573 res = dest_name_check(dest, m);
574 if (res)
575 break;
576 res = tipc_multicast(tport->ref,
577 &dest->addr.nameseq,
578 m->msg_iovlen,
579 m->msg_iov,
580 total_len);
581 }
582 if (likely(res != -ELINKCONG)) {
583 if (needs_conn && (res >= 0))
584 sock->state = SS_CONNECTING;
585 break;
586 }
587 if (timeout_val <= 0L) {
588 res = timeout_val ? timeout_val : -EWOULDBLOCK;
589 break;
590 }
591 release_sock(sk);
592 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
593 !tport->congested, timeout_val);
594 lock_sock(sk);
595 } while (1);
596
597exit:
598 if (iocb)
599 release_sock(sk);
600 return res;
601}
602
603/**
604 * send_packet - send a connection-oriented message
605 * @iocb: if NULL, indicates that socket lock is already held
606 * @sock: socket structure
607 * @m: message to send
608 * @total_len: length of message
609 *
610 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
611 *
612 * Returns the number of bytes sent on success, or errno otherwise
613 */
614static int send_packet(struct kiocb *iocb, struct socket *sock,
615 struct msghdr *m, size_t total_len)
616{
617 struct sock *sk = sock->sk;
618 struct tipc_port *tport = tipc_sk_port(sk);
619 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
620 long timeout_val;
621 int res;
622
623 /* Handle implied connection establishment */
624 if (unlikely(dest))
625 return send_msg(iocb, sock, m, total_len);
626
627 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
628 (m->msg_iovlen > (unsigned int)INT_MAX))
629 return -EMSGSIZE;
630
631 if (iocb)
632 lock_sock(sk);
633
634 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
635
636 do {
637 if (unlikely(sock->state != SS_CONNECTED)) {
638 if (sock->state == SS_DISCONNECTING)
639 res = -EPIPE;
640 else
641 res = -ENOTCONN;
642 break;
643 }
644
645 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
646 total_len);
647 if (likely(res != -ELINKCONG))
648 break;
649 if (timeout_val <= 0L) {
650 res = timeout_val ? timeout_val : -EWOULDBLOCK;
651 break;
652 }
653 release_sock(sk);
654 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
655 (!tport->congested || !tport->connected), timeout_val);
656 lock_sock(sk);
657 } while (1);
658
659 if (iocb)
660 release_sock(sk);
661 return res;
662}
663
664/**
665 * send_stream - send stream-oriented data
666 * @iocb: (unused)
667 * @sock: socket structure
668 * @m: data to send
669 * @total_len: total length of data to be sent
670 *
671 * Used for SOCK_STREAM data.
672 *
673 * Returns the number of bytes sent on success (or partial success),
674 * or errno if no data sent
675 */
676static int send_stream(struct kiocb *iocb, struct socket *sock,
677 struct msghdr *m, size_t total_len)
678{
679 struct sock *sk = sock->sk;
680 struct tipc_port *tport = tipc_sk_port(sk);
681 struct msghdr my_msg;
682 struct iovec my_iov;
683 struct iovec *curr_iov;
684 int curr_iovlen;
685 char __user *curr_start;
686 u32 hdr_size;
687 int curr_left;
688 int bytes_to_send;
689 int bytes_sent;
690 int res;
691
692 lock_sock(sk);
693
694 /* Handle special cases where there is no connection */
695 if (unlikely(sock->state != SS_CONNECTED)) {
696 if (sock->state == SS_UNCONNECTED) {
697 res = send_packet(NULL, sock, m, total_len);
698 goto exit;
699 } else if (sock->state == SS_DISCONNECTING) {
700 res = -EPIPE;
701 goto exit;
702 } else {
703 res = -ENOTCONN;
704 goto exit;
705 }
706 }
707
708 if (unlikely(m->msg_name)) {
709 res = -EISCONN;
710 goto exit;
711 }
712
713 if ((total_len > (unsigned int)INT_MAX) ||
714 (m->msg_iovlen > (unsigned int)INT_MAX)) {
715 res = -EMSGSIZE;
716 goto exit;
717 }
718
719 /*
720 * Send each iovec entry using one or more messages
721 *
722 * Note: This algorithm is good for the most likely case
723 * (i.e. one large iovec entry), but could be improved to pass sets
724 * of small iovec entries into send_packet().
725 */
726 curr_iov = m->msg_iov;
727 curr_iovlen = m->msg_iovlen;
728 my_msg.msg_iov = &my_iov;
729 my_msg.msg_iovlen = 1;
730 my_msg.msg_flags = m->msg_flags;
731 my_msg.msg_name = NULL;
732 bytes_sent = 0;
733
734 hdr_size = msg_hdr_sz(&tport->phdr);
735
736 while (curr_iovlen--) {
737 curr_start = curr_iov->iov_base;
738 curr_left = curr_iov->iov_len;
739
740 while (curr_left) {
741 bytes_to_send = tport->max_pkt - hdr_size;
742 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
743 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
744 if (curr_left < bytes_to_send)
745 bytes_to_send = curr_left;
746 my_iov.iov_base = curr_start;
747 my_iov.iov_len = bytes_to_send;
748 res = send_packet(NULL, sock, &my_msg, bytes_to_send);
749 if (res < 0) {
750 if (bytes_sent)
751 res = bytes_sent;
752 goto exit;
753 }
754 curr_left -= bytes_to_send;
755 curr_start += bytes_to_send;
756 bytes_sent += bytes_to_send;
757 }
758
759 curr_iov++;
760 }
761 res = bytes_sent;
762exit:
763 release_sock(sk);
764 return res;
765}
766
767/**
768 * auto_connect - complete connection setup to a remote port
769 * @sock: socket structure
770 * @msg: peer's response message
771 *
772 * Returns 0 on success, errno otherwise
773 */
774static int auto_connect(struct socket *sock, struct tipc_msg *msg)
775{
776 struct tipc_sock *tsock = tipc_sk(sock->sk);
777
778 if (msg_errcode(msg)) {
779 sock->state = SS_DISCONNECTING;
780 return -ECONNREFUSED;
781 }
782
783 tsock->peer_name.ref = msg_origport(msg);
784 tsock->peer_name.node = msg_orignode(msg);
785 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
786 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
787 sock->state = SS_CONNECTED;
788 return 0;
789}
790
791/**
792 * set_orig_addr - capture sender's address for received message
793 * @m: descriptor for message info
794 * @msg: received message header
795 *
796 * Note: Address is not captured if not requested by receiver.
797 */
798static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
799{
800 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
801
802 if (addr) {
803 addr->family = AF_TIPC;
804 addr->addrtype = TIPC_ADDR_ID;
805 addr->addr.id.ref = msg_origport(msg);
806 addr->addr.id.node = msg_orignode(msg);
807 addr->addr.name.domain = 0; /* could leave uninitialized */
808 addr->scope = 0; /* could leave uninitialized */
809 m->msg_namelen = sizeof(struct sockaddr_tipc);
810 }
811}
812
813/**
814 * anc_data_recv - optionally capture ancillary data for received message
815 * @m: descriptor for message info
816 * @msg: received message header
817 * @tport: TIPC port associated with message
818 *
819 * Note: Ancillary data is not captured if not requested by receiver.
820 *
821 * Returns 0 if successful, otherwise errno
822 */
823static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
824 struct tipc_port *tport)
825{
826 u32 anc_data[3];
827 u32 err;
828 u32 dest_type;
829 int has_name;
830 int res;
831
832 if (likely(m->msg_controllen == 0))
833 return 0;
834
835 /* Optionally capture errored message object(s) */
836 err = msg ? msg_errcode(msg) : 0;
837 if (unlikely(err)) {
838 anc_data[0] = err;
839 anc_data[1] = msg_data_sz(msg);
840 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
841 if (res)
842 return res;
843 if (anc_data[1]) {
844 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
845 msg_data(msg));
846 if (res)
847 return res;
848 }
849 }
850
851 /* Optionally capture message destination object */
852 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
853 switch (dest_type) {
854 case TIPC_NAMED_MSG:
855 has_name = 1;
856 anc_data[0] = msg_nametype(msg);
857 anc_data[1] = msg_namelower(msg);
858 anc_data[2] = msg_namelower(msg);
859 break;
860 case TIPC_MCAST_MSG:
861 has_name = 1;
862 anc_data[0] = msg_nametype(msg);
863 anc_data[1] = msg_namelower(msg);
864 anc_data[2] = msg_nameupper(msg);
865 break;
866 case TIPC_CONN_MSG:
867 has_name = (tport->conn_type != 0);
868 anc_data[0] = tport->conn_type;
869 anc_data[1] = tport->conn_instance;
870 anc_data[2] = tport->conn_instance;
871 break;
872 default:
873 has_name = 0;
874 }
875 if (has_name) {
876 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
877 if (res)
878 return res;
879 }
880
881 return 0;
882}
883
884/**
885 * recv_msg - receive packet-oriented message
886 * @iocb: (unused)
887 * @m: descriptor for message info
888 * @buf_len: total size of user buffer area
889 * @flags: receive flags
890 *
891 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
892 * If the complete message doesn't fit in user area, truncate it.
893 *
894 * Returns size of returned message data, errno otherwise
895 */
896static int recv_msg(struct kiocb *iocb, struct socket *sock,
897 struct msghdr *m, size_t buf_len, int flags)
898{
899 struct sock *sk = sock->sk;
900 struct tipc_port *tport = tipc_sk_port(sk);
901 struct sk_buff *buf;
902 struct tipc_msg *msg;
903 long timeout;
904 unsigned int sz;
905 u32 err;
906 int res;
907
908 /* Catch invalid receive requests */
909 if (unlikely(!buf_len))
910 return -EINVAL;
911
912 lock_sock(sk);
913
914 if (unlikely(sock->state == SS_UNCONNECTED)) {
915 res = -ENOTCONN;
916 goto exit;
917 }
918
919 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
920restart:
921
922 /* Look for a message in receive queue; wait if necessary */
923 while (skb_queue_empty(&sk->sk_receive_queue)) {
924 if (sock->state == SS_DISCONNECTING) {
925 res = -ENOTCONN;
926 goto exit;
927 }
928 if (timeout <= 0L) {
929 res = timeout ? timeout : -EWOULDBLOCK;
930 goto exit;
931 }
932 release_sock(sk);
933 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
934 tipc_rx_ready(sock),
935 timeout);
936 lock_sock(sk);
937 }
938
939 /* Look at first message in receive queue */
940 buf = skb_peek(&sk->sk_receive_queue);
941 msg = buf_msg(buf);
942 sz = msg_data_sz(msg);
943 err = msg_errcode(msg);
944
945 /* Complete connection setup for an implied connect */
946 if (unlikely(sock->state == SS_CONNECTING)) {
947 res = auto_connect(sock, msg);
948 if (res)
949 goto exit;
950 }
951
952 /* Discard an empty non-errored message & try again */
953 if ((!sz) && (!err)) {
954 advance_rx_queue(sk);
955 goto restart;
956 }
957
958 /* Capture sender's address (optional) */
959 set_orig_addr(m, msg);
960
961 /* Capture ancillary data (optional) */
962 res = anc_data_recv(m, msg, tport);
963 if (res)
964 goto exit;
965
966 /* Capture message data (if valid) & compute return value (always) */
967 if (!err) {
968 if (unlikely(buf_len < sz)) {
969 sz = buf_len;
970 m->msg_flags |= MSG_TRUNC;
971 }
972 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
973 m->msg_iov, sz);
974 if (res)
975 goto exit;
976 res = sz;
977 } else {
978 if ((sock->state == SS_READY) ||
979 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
980 res = 0;
981 else
982 res = -ECONNRESET;
983 }
984
985 /* Consume received message (optional) */
986 if (likely(!(flags & MSG_PEEK))) {
987 if ((sock->state != SS_READY) &&
988 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
989 tipc_acknowledge(tport->ref, tport->conn_unacked);
990 advance_rx_queue(sk);
991 }
992exit:
993 release_sock(sk);
994 return res;
995}
996
997/**
998 * recv_stream - receive stream-oriented data
999 * @iocb: (unused)
1000 * @m: descriptor for message info
1001 * @buf_len: total size of user buffer area
1002 * @flags: receive flags
1003 *
1004 * Used for SOCK_STREAM messages only. If not enough data is available
1005 * will optionally wait for more; never truncates data.
1006 *
1007 * Returns size of returned message data, errno otherwise
1008 */
1009static int recv_stream(struct kiocb *iocb, struct socket *sock,
1010 struct msghdr *m, size_t buf_len, int flags)
1011{
1012 struct sock *sk = sock->sk;
1013 struct tipc_port *tport = tipc_sk_port(sk);
1014 struct sk_buff *buf;
1015 struct tipc_msg *msg;
1016 long timeout;
1017 unsigned int sz;
1018 int sz_to_copy, target, needed;
1019 int sz_copied = 0;
1020 u32 err;
1021 int res = 0;
1022
1023 /* Catch invalid receive attempts */
1024 if (unlikely(!buf_len))
1025 return -EINVAL;
1026
1027 lock_sock(sk);
1028
1029 if (unlikely((sock->state == SS_UNCONNECTED) ||
1030 (sock->state == SS_CONNECTING))) {
1031 res = -ENOTCONN;
1032 goto exit;
1033 }
1034
1035 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1036 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1037
1038restart:
1039 /* Look for a message in receive queue; wait if necessary */
1040 while (skb_queue_empty(&sk->sk_receive_queue)) {
1041 if (sock->state == SS_DISCONNECTING) {
1042 res = -ENOTCONN;
1043 goto exit;
1044 }
1045 if (timeout <= 0L) {
1046 res = timeout ? timeout : -EWOULDBLOCK;
1047 goto exit;
1048 }
1049 release_sock(sk);
1050 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1051 tipc_rx_ready(sock),
1052 timeout);
1053 lock_sock(sk);
1054 }
1055
1056 /* Look at first message in receive queue */
1057 buf = skb_peek(&sk->sk_receive_queue);
1058 msg = buf_msg(buf);
1059 sz = msg_data_sz(msg);
1060 err = msg_errcode(msg);
1061
1062 /* Discard an empty non-errored message & try again */
1063 if ((!sz) && (!err)) {
1064 advance_rx_queue(sk);
1065 goto restart;
1066 }
1067
1068 /* Optionally capture sender's address & ancillary data of first msg */
1069 if (sz_copied == 0) {
1070 set_orig_addr(m, msg);
1071 res = anc_data_recv(m, msg, tport);
1072 if (res)
1073 goto exit;
1074 }
1075
1076 /* Capture message data (if valid) & compute return value (always) */
1077 if (!err) {
1078 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1079
1080 sz -= offset;
1081 needed = (buf_len - sz_copied);
1082 sz_to_copy = (sz <= needed) ? sz : needed;
1083
1084 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1085 m->msg_iov, sz_to_copy);
1086 if (res)
1087 goto exit;
1088
1089 sz_copied += sz_to_copy;
1090
1091 if (sz_to_copy < sz) {
1092 if (!(flags & MSG_PEEK))
1093 TIPC_SKB_CB(buf)->handle =
1094 (void *)(unsigned long)(offset + sz_to_copy);
1095 goto exit;
1096 }
1097 } else {
1098 if (sz_copied != 0)
1099 goto exit; /* can't add error msg to valid data */
1100
1101 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1102 res = 0;
1103 else
1104 res = -ECONNRESET;
1105 }
1106
1107 /* Consume received message (optional) */
1108 if (likely(!(flags & MSG_PEEK))) {
1109 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1110 tipc_acknowledge(tport->ref, tport->conn_unacked);
1111 advance_rx_queue(sk);
1112 }
1113
1114 /* Loop around if more data is required */
1115 if ((sz_copied < buf_len) && /* didn't get all requested data */
1116 (!skb_queue_empty(&sk->sk_receive_queue) ||
1117 (sz_copied < target)) && /* and more is ready or required */
1118 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1119 (!err)) /* and haven't reached a FIN */
1120 goto restart;
1121
1122exit:
1123 release_sock(sk);
1124 return sz_copied ? sz_copied : res;
1125}
1126
1127/**
1128 * rx_queue_full - determine if receive queue can accept another message
1129 * @msg: message to be added to queue
1130 * @queue_size: current size of queue
1131 * @base: nominal maximum size of queue
1132 *
1133 * Returns 1 if queue is unable to accept message, 0 otherwise
1134 */
1135static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1136{
1137 u32 threshold;
1138 u32 imp = msg_importance(msg);
1139
1140 if (imp == TIPC_LOW_IMPORTANCE)
1141 threshold = base;
1142 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1143 threshold = base * 2;
1144 else if (imp == TIPC_HIGH_IMPORTANCE)
1145 threshold = base * 100;
1146 else
1147 return 0;
1148
1149 if (msg_connected(msg))
1150 threshold *= 4;
1151
1152 return queue_size >= threshold;
1153}
1154
1155/**
1156 * filter_rcv - validate incoming message
1157 * @sk: socket
1158 * @buf: message
1159 *
1160 * Enqueues message on receive queue if acceptable; optionally handles
1161 * disconnect indication for a connected socket.
1162 *
1163 * Called with socket lock already taken; port lock may also be taken.
1164 *
1165 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1166 */
1167static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1168{
1169 struct socket *sock = sk->sk_socket;
1170 struct tipc_msg *msg = buf_msg(buf);
1171 u32 recv_q_len;
1172
1173 /* Reject message if it is wrong sort of message for socket */
1174 if (msg_type(msg) > TIPC_DIRECT_MSG)
1175 return TIPC_ERR_NO_PORT;
1176
1177 if (sock->state == SS_READY) {
1178 if (msg_connected(msg))
1179 return TIPC_ERR_NO_PORT;
1180 } else {
1181 if (msg_mcast(msg))
1182 return TIPC_ERR_NO_PORT;
1183 if (sock->state == SS_CONNECTED) {
1184 if (!msg_connected(msg) ||
1185 !tipc_port_peer_msg(tipc_sk_port(sk), msg))
1186 return TIPC_ERR_NO_PORT;
1187 } else if (sock->state == SS_CONNECTING) {
1188 if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1189 return TIPC_ERR_NO_PORT;
1190 } else if (sock->state == SS_LISTENING) {
1191 if (msg_connected(msg) || msg_errcode(msg))
1192 return TIPC_ERR_NO_PORT;
1193 } else if (sock->state == SS_DISCONNECTING) {
1194 return TIPC_ERR_NO_PORT;
1195 } else /* (sock->state == SS_UNCONNECTED) */ {
1196 if (msg_connected(msg) || msg_errcode(msg))
1197 return TIPC_ERR_NO_PORT;
1198 }
1199 }
1200
1201 /* Reject message if there isn't room to queue it */
1202 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1203 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1204 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1205 return TIPC_ERR_OVERLOAD;
1206 }
1207 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1208 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1209 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1210 return TIPC_ERR_OVERLOAD;
1211 }
1212
1213 /* Enqueue message (finally!) */
1214 TIPC_SKB_CB(buf)->handle = 0;
1215 atomic_inc(&tipc_queue_size);
1216 __skb_queue_tail(&sk->sk_receive_queue, buf);
1217
1218 /* Initiate connection termination for an incoming 'FIN' */
1219 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1220 sock->state = SS_DISCONNECTING;
1221 tipc_disconnect_port(tipc_sk_port(sk));
1222 }
1223
1224 if (waitqueue_active(sk_sleep(sk)))
1225 wake_up_interruptible(sk_sleep(sk));
1226 return TIPC_OK;
1227}
1228
1229/**
1230 * backlog_rcv - handle incoming message from backlog queue
1231 * @sk: socket
1232 * @buf: message
1233 *
1234 * Caller must hold socket lock, but not port lock.
1235 *
1236 * Returns 0
1237 */
1238static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1239{
1240 u32 res;
1241
1242 res = filter_rcv(sk, buf);
1243 if (res)
1244 tipc_reject_msg(buf, res);
1245 return 0;
1246}
1247
1248/**
1249 * dispatch - handle incoming message
1250 * @tport: TIPC port that received message
1251 * @buf: message
1252 *
1253 * Called with port lock already taken.
1254 *
1255 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1256 */
1257static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1258{
1259 struct sock *sk = (struct sock *)tport->usr_handle;
1260 u32 res;
1261
1262 /*
1263 * Process message if socket is unlocked; otherwise add to backlog queue
1264 *
1265 * This code is based on sk_receive_skb(), but must be distinct from it
1266 * since a TIPC-specific filter/reject mechanism is utilized
1267 */
1268 bh_lock_sock(sk);
1269 if (!sock_owned_by_user(sk)) {
1270 res = filter_rcv(sk, buf);
1271 } else {
1272 if (sk_add_backlog(sk, buf, sk->sk_rcvbuf))
1273 res = TIPC_ERR_OVERLOAD;
1274 else
1275 res = TIPC_OK;
1276 }
1277 bh_unlock_sock(sk);
1278
1279 return res;
1280}
1281
1282/**
1283 * wakeupdispatch - wake up port after congestion
1284 * @tport: port to wakeup
1285 *
1286 * Called with port lock already taken.
1287 */
1288static void wakeupdispatch(struct tipc_port *tport)
1289{
1290 struct sock *sk = (struct sock *)tport->usr_handle;
1291
1292 if (waitqueue_active(sk_sleep(sk)))
1293 wake_up_interruptible(sk_sleep(sk));
1294}
1295
1296/**
1297 * connect - establish a connection to another TIPC port
1298 * @sock: socket structure
1299 * @dest: socket address for destination port
1300 * @destlen: size of socket address data structure
1301 * @flags: file-related flags associated with socket
1302 *
1303 * Returns 0 on success, errno otherwise
1304 */
1305static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1306 int flags)
1307{
1308 struct sock *sk = sock->sk;
1309 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1310 struct msghdr m = {NULL,};
1311 struct sk_buff *buf;
1312 struct tipc_msg *msg;
1313 unsigned int timeout;
1314 int res;
1315
1316 lock_sock(sk);
1317
1318 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1319 if (sock->state == SS_READY) {
1320 res = -EOPNOTSUPP;
1321 goto exit;
1322 }
1323
1324 /* For now, TIPC does not support the non-blocking form of connect() */
1325 if (flags & O_NONBLOCK) {
1326 res = -EOPNOTSUPP;
1327 goto exit;
1328 }
1329
1330 /* Issue Posix-compliant error code if socket is in the wrong state */
1331 if (sock->state == SS_LISTENING) {
1332 res = -EOPNOTSUPP;
1333 goto exit;
1334 }
1335 if (sock->state == SS_CONNECTING) {
1336 res = -EALREADY;
1337 goto exit;
1338 }
1339 if (sock->state != SS_UNCONNECTED) {
1340 res = -EISCONN;
1341 goto exit;
1342 }
1343
1344 /*
1345 * Reject connection attempt using multicast address
1346 *
1347 * Note: send_msg() validates the rest of the address fields,
1348 * so there's no need to do it here
1349 */
1350 if (dst->addrtype == TIPC_ADDR_MCAST) {
1351 res = -EINVAL;
1352 goto exit;
1353 }
1354
1355 /* Reject any messages already in receive queue (very unlikely) */
1356 reject_rx_queue(sk);
1357
1358 /* Send a 'SYN-' to destination */
1359 m.msg_name = dest;
1360 m.msg_namelen = destlen;
1361 res = send_msg(NULL, sock, &m, 0);
1362 if (res < 0)
1363 goto exit;
1364
1365 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1366 timeout = tipc_sk(sk)->conn_timeout;
1367 release_sock(sk);
1368 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1369 (!skb_queue_empty(&sk->sk_receive_queue) ||
1370 (sock->state != SS_CONNECTING)),
1371 timeout ? (long)msecs_to_jiffies(timeout)
1372 : MAX_SCHEDULE_TIMEOUT);
1373 lock_sock(sk);
1374
1375 if (res > 0) {
1376 buf = skb_peek(&sk->sk_receive_queue);
1377 if (buf != NULL) {
1378 msg = buf_msg(buf);
1379 res = auto_connect(sock, msg);
1380 if (!res) {
1381 if (!msg_data_sz(msg))
1382 advance_rx_queue(sk);
1383 }
1384 } else {
1385 if (sock->state == SS_CONNECTED)
1386 res = -EISCONN;
1387 else
1388 res = -ECONNREFUSED;
1389 }
1390 } else {
1391 if (res == 0)
1392 res = -ETIMEDOUT;
1393 else
1394 ; /* leave "res" unchanged */
1395 sock->state = SS_DISCONNECTING;
1396 }
1397
1398exit:
1399 release_sock(sk);
1400 return res;
1401}
1402
1403/**
1404 * listen - allow socket to listen for incoming connections
1405 * @sock: socket structure
1406 * @len: (unused)
1407 *
1408 * Returns 0 on success, errno otherwise
1409 */
1410static int listen(struct socket *sock, int len)
1411{
1412 struct sock *sk = sock->sk;
1413 int res;
1414
1415 lock_sock(sk);
1416
1417 if (sock->state != SS_UNCONNECTED)
1418 res = -EINVAL;
1419 else {
1420 sock->state = SS_LISTENING;
1421 res = 0;
1422 }
1423
1424 release_sock(sk);
1425 return res;
1426}
1427
1428/**
1429 * accept - wait for connection request
1430 * @sock: listening socket
1431 * @newsock: new socket that is to be connected
1432 * @flags: file-related flags associated with socket
1433 *
1434 * Returns 0 on success, errno otherwise
1435 */
1436static int accept(struct socket *sock, struct socket *new_sock, int flags)
1437{
1438 struct sock *sk = sock->sk;
1439 struct sk_buff *buf;
1440 int res;
1441
1442 lock_sock(sk);
1443
1444 if (sock->state != SS_LISTENING) {
1445 res = -EINVAL;
1446 goto exit;
1447 }
1448
1449 while (skb_queue_empty(&sk->sk_receive_queue)) {
1450 if (flags & O_NONBLOCK) {
1451 res = -EWOULDBLOCK;
1452 goto exit;
1453 }
1454 release_sock(sk);
1455 res = wait_event_interruptible(*sk_sleep(sk),
1456 (!skb_queue_empty(&sk->sk_receive_queue)));
1457 lock_sock(sk);
1458 if (res)
1459 goto exit;
1460 }
1461
1462 buf = skb_peek(&sk->sk_receive_queue);
1463
1464 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1465 if (!res) {
1466 struct sock *new_sk = new_sock->sk;
1467 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1468 struct tipc_port *new_tport = new_tsock->p;
1469 u32 new_ref = new_tport->ref;
1470 struct tipc_msg *msg = buf_msg(buf);
1471
1472 lock_sock(new_sk);
1473
1474 /*
1475 * Reject any stray messages received by new socket
1476 * before the socket lock was taken (very, very unlikely)
1477 */
1478 reject_rx_queue(new_sk);
1479
1480 /* Connect new socket to it's peer */
1481 new_tsock->peer_name.ref = msg_origport(msg);
1482 new_tsock->peer_name.node = msg_orignode(msg);
1483 tipc_connect2port(new_ref, &new_tsock->peer_name);
1484 new_sock->state = SS_CONNECTED;
1485
1486 tipc_set_portimportance(new_ref, msg_importance(msg));
1487 if (msg_named(msg)) {
1488 new_tport->conn_type = msg_nametype(msg);
1489 new_tport->conn_instance = msg_nameinst(msg);
1490 }
1491
1492 /*
1493 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1494 * Respond to 'SYN+' by queuing it on new socket.
1495 */
1496 if (!msg_data_sz(msg)) {
1497 struct msghdr m = {NULL,};
1498
1499 advance_rx_queue(sk);
1500 send_packet(NULL, new_sock, &m, 0);
1501 } else {
1502 __skb_dequeue(&sk->sk_receive_queue);
1503 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1504 }
1505 release_sock(new_sk);
1506 }
1507exit:
1508 release_sock(sk);
1509 return res;
1510}
1511
1512/**
1513 * shutdown - shutdown socket connection
1514 * @sock: socket structure
1515 * @how: direction to close (must be SHUT_RDWR)
1516 *
1517 * Terminates connection (if necessary), then purges socket's receive queue.
1518 *
1519 * Returns 0 on success, errno otherwise
1520 */
1521static int shutdown(struct socket *sock, int how)
1522{
1523 struct sock *sk = sock->sk;
1524 struct tipc_port *tport = tipc_sk_port(sk);
1525 struct sk_buff *buf;
1526 int res;
1527
1528 if (how != SHUT_RDWR)
1529 return -EINVAL;
1530
1531 lock_sock(sk);
1532
1533 switch (sock->state) {
1534 case SS_CONNECTING:
1535 case SS_CONNECTED:
1536
1537restart:
1538 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1539 buf = __skb_dequeue(&sk->sk_receive_queue);
1540 if (buf) {
1541 atomic_dec(&tipc_queue_size);
1542 if (TIPC_SKB_CB(buf)->handle != 0) {
1543 kfree_skb(buf);
1544 goto restart;
1545 }
1546 tipc_disconnect(tport->ref);
1547 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1548 } else {
1549 tipc_shutdown(tport->ref);
1550 }
1551
1552 sock->state = SS_DISCONNECTING;
1553
1554 /* fall through */
1555
1556 case SS_DISCONNECTING:
1557
1558 /* Discard any unreceived messages; wake up sleeping tasks */
1559 discard_rx_queue(sk);
1560 if (waitqueue_active(sk_sleep(sk)))
1561 wake_up_interruptible(sk_sleep(sk));
1562 res = 0;
1563 break;
1564
1565 default:
1566 res = -ENOTCONN;
1567 }
1568
1569 release_sock(sk);
1570 return res;
1571}
1572
1573/**
1574 * setsockopt - set socket option
1575 * @sock: socket structure
1576 * @lvl: option level
1577 * @opt: option identifier
1578 * @ov: pointer to new option value
1579 * @ol: length of option value
1580 *
1581 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1582 * (to ease compatibility).
1583 *
1584 * Returns 0 on success, errno otherwise
1585 */
1586static int setsockopt(struct socket *sock,
1587 int lvl, int opt, char __user *ov, unsigned int ol)
1588{
1589 struct sock *sk = sock->sk;
1590 struct tipc_port *tport = tipc_sk_port(sk);
1591 u32 value;
1592 int res;
1593
1594 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1595 return 0;
1596 if (lvl != SOL_TIPC)
1597 return -ENOPROTOOPT;
1598 if (ol < sizeof(value))
1599 return -EINVAL;
1600 res = get_user(value, (u32 __user *)ov);
1601 if (res)
1602 return res;
1603
1604 lock_sock(sk);
1605
1606 switch (opt) {
1607 case TIPC_IMPORTANCE:
1608 res = tipc_set_portimportance(tport->ref, value);
1609 break;
1610 case TIPC_SRC_DROPPABLE:
1611 if (sock->type != SOCK_STREAM)
1612 res = tipc_set_portunreliable(tport->ref, value);
1613 else
1614 res = -ENOPROTOOPT;
1615 break;
1616 case TIPC_DEST_DROPPABLE:
1617 res = tipc_set_portunreturnable(tport->ref, value);
1618 break;
1619 case TIPC_CONN_TIMEOUT:
1620 tipc_sk(sk)->conn_timeout = value;
1621 /* no need to set "res", since already 0 at this point */
1622 break;
1623 default:
1624 res = -EINVAL;
1625 }
1626
1627 release_sock(sk);
1628
1629 return res;
1630}
1631
1632/**
1633 * getsockopt - get socket option
1634 * @sock: socket structure
1635 * @lvl: option level
1636 * @opt: option identifier
1637 * @ov: receptacle for option value
1638 * @ol: receptacle for length of option value
1639 *
1640 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1641 * (to ease compatibility).
1642 *
1643 * Returns 0 on success, errno otherwise
1644 */
1645static int getsockopt(struct socket *sock,
1646 int lvl, int opt, char __user *ov, int __user *ol)
1647{
1648 struct sock *sk = sock->sk;
1649 struct tipc_port *tport = tipc_sk_port(sk);
1650 int len;
1651 u32 value;
1652 int res;
1653
1654 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1655 return put_user(0, ol);
1656 if (lvl != SOL_TIPC)
1657 return -ENOPROTOOPT;
1658 res = get_user(len, ol);
1659 if (res)
1660 return res;
1661
1662 lock_sock(sk);
1663
1664 switch (opt) {
1665 case TIPC_IMPORTANCE:
1666 res = tipc_portimportance(tport->ref, &value);
1667 break;
1668 case TIPC_SRC_DROPPABLE:
1669 res = tipc_portunreliable(tport->ref, &value);
1670 break;
1671 case TIPC_DEST_DROPPABLE:
1672 res = tipc_portunreturnable(tport->ref, &value);
1673 break;
1674 case TIPC_CONN_TIMEOUT:
1675 value = tipc_sk(sk)->conn_timeout;
1676 /* no need to set "res", since already 0 at this point */
1677 break;
1678 case TIPC_NODE_RECVQ_DEPTH:
1679 value = (u32)atomic_read(&tipc_queue_size);
1680 break;
1681 case TIPC_SOCK_RECVQ_DEPTH:
1682 value = skb_queue_len(&sk->sk_receive_queue);
1683 break;
1684 default:
1685 res = -EINVAL;
1686 }
1687
1688 release_sock(sk);
1689
1690 if (res)
1691 return res; /* "get" failed */
1692
1693 if (len < sizeof(value))
1694 return -EINVAL;
1695
1696 if (copy_to_user(ov, &value, sizeof(value)))
1697 return -EFAULT;
1698
1699 return put_user(sizeof(value), ol);
1700}
1701
1702/**
1703 * Protocol switches for the various types of TIPC sockets
1704 */
1705static const struct proto_ops msg_ops = {
1706 .owner = THIS_MODULE,
1707 .family = AF_TIPC,
1708 .release = release,
1709 .bind = bind,
1710 .connect = connect,
1711 .socketpair = sock_no_socketpair,
1712 .accept = sock_no_accept,
1713 .getname = get_name,
1714 .poll = poll,
1715 .ioctl = sock_no_ioctl,
1716 .listen = sock_no_listen,
1717 .shutdown = shutdown,
1718 .setsockopt = setsockopt,
1719 .getsockopt = getsockopt,
1720 .sendmsg = send_msg,
1721 .recvmsg = recv_msg,
1722 .mmap = sock_no_mmap,
1723 .sendpage = sock_no_sendpage
1724};
1725
1726static const struct proto_ops packet_ops = {
1727 .owner = THIS_MODULE,
1728 .family = AF_TIPC,
1729 .release = release,
1730 .bind = bind,
1731 .connect = connect,
1732 .socketpair = sock_no_socketpair,
1733 .accept = accept,
1734 .getname = get_name,
1735 .poll = poll,
1736 .ioctl = sock_no_ioctl,
1737 .listen = listen,
1738 .shutdown = shutdown,
1739 .setsockopt = setsockopt,
1740 .getsockopt = getsockopt,
1741 .sendmsg = send_packet,
1742 .recvmsg = recv_msg,
1743 .mmap = sock_no_mmap,
1744 .sendpage = sock_no_sendpage
1745};
1746
1747static const struct proto_ops stream_ops = {
1748 .owner = THIS_MODULE,
1749 .family = AF_TIPC,
1750 .release = release,
1751 .bind = bind,
1752 .connect = connect,
1753 .socketpair = sock_no_socketpair,
1754 .accept = accept,
1755 .getname = get_name,
1756 .poll = poll,
1757 .ioctl = sock_no_ioctl,
1758 .listen = listen,
1759 .shutdown = shutdown,
1760 .setsockopt = setsockopt,
1761 .getsockopt = getsockopt,
1762 .sendmsg = send_stream,
1763 .recvmsg = recv_stream,
1764 .mmap = sock_no_mmap,
1765 .sendpage = sock_no_sendpage
1766};
1767
1768static const struct net_proto_family tipc_family_ops = {
1769 .owner = THIS_MODULE,
1770 .family = AF_TIPC,
1771 .create = tipc_create
1772};
1773
1774static struct proto tipc_proto = {
1775 .name = "TIPC",
1776 .owner = THIS_MODULE,
1777 .obj_size = sizeof(struct tipc_sock)
1778};
1779
1780/**
1781 * tipc_socket_init - initialize TIPC socket interface
1782 *
1783 * Returns 0 on success, errno otherwise
1784 */
1785int tipc_socket_init(void)
1786{
1787 int res;
1788
1789 res = proto_register(&tipc_proto, 1);
1790 if (res) {
1791 err("Failed to register TIPC protocol type\n");
1792 goto out;
1793 }
1794
1795 res = sock_register(&tipc_family_ops);
1796 if (res) {
1797 err("Failed to register TIPC socket type\n");
1798 proto_unregister(&tipc_proto);
1799 goto out;
1800 }
1801
1802 sockets_enabled = 1;
1803 out:
1804 return res;
1805}
1806
1807/**
1808 * tipc_socket_stop - stop TIPC socket interface
1809 */
1810void tipc_socket_stop(void)
1811{
1812 if (!sockets_enabled)
1813 return;
1814
1815 sockets_enabled = 0;
1816 sock_unregister(tipc_family_ops.family);
1817 proto_unregister(&tipc_proto);
1818}