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