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, 2012-2019, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * Copyright (c) 2020-2021, Red Hat Inc
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions are met:
11 *
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the names of the copyright holders nor the names of its
18 * contributors may be used to endorse or promote products derived from
19 * this software without specific prior written permission.
20 *
21 * Alternatively, this software may be distributed under the terms of the
22 * GNU General Public License ("GPL") version 2 as published by the Free
23 * Software Foundation.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38#include <linux/rhashtable.h>
39#include <linux/sched/signal.h>
40#include <trace/events/sock.h>
41
42#include "core.h"
43#include "name_table.h"
44#include "node.h"
45#include "link.h"
46#include "name_distr.h"
47#include "socket.h"
48#include "bcast.h"
49#include "netlink.h"
50#include "group.h"
51#include "trace.h"
52
53#define NAGLE_START_INIT 4
54#define NAGLE_START_MAX 1024
55#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
56#define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
57#define TIPC_MAX_PORT 0xffffffff
58#define TIPC_MIN_PORT 1
59#define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */
60
61enum {
62 TIPC_LISTEN = TCP_LISTEN,
63 TIPC_ESTABLISHED = TCP_ESTABLISHED,
64 TIPC_OPEN = TCP_CLOSE,
65 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
66 TIPC_CONNECTING = TCP_SYN_SENT,
67};
68
69struct sockaddr_pair {
70 struct sockaddr_tipc sock;
71 struct sockaddr_tipc member;
72};
73
74/**
75 * struct tipc_sock - TIPC socket structure
76 * @sk: socket - interacts with 'port' and with user via the socket API
77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
78 * @maxnagle: maximum size of msg which can be subject to nagle
79 * @portid: unique port identity in TIPC socket hash table
80 * @phdr: preformatted message header used when sending messages
81 * @cong_links: list of congested links
82 * @publications: list of publications for port
83 * @pub_count: total # of publications port has made during its lifetime
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @probe_unacked: probe has not received ack yet
86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87 * @cong_link_cnt: number of congested links
88 * @snt_unacked: # messages sent by socket, and not yet acked by peer
89 * @snd_win: send window size
90 * @peer_caps: peer capabilities mask
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @rcv_win: receive window size
93 * @peer: 'connected' peer for dgram/rdm
94 * @node: hash table node
95 * @mc_method: cookie for use between socket and broadcast layer
96 * @rcu: rcu struct for tipc_sock
97 * @group: TIPC communications group
98 * @oneway: message count in one direction (FIXME)
99 * @nagle_start: current nagle value
100 * @snd_backlog: send backlog count
101 * @msg_acc: messages accepted; used in managing backlog and nagle
102 * @pkt_cnt: TIPC socket packet count
103 * @expect_ack: whether this TIPC socket is expecting an ack
104 * @nodelay: setsockopt() TIPC_NODELAY setting
105 * @group_is_open: TIPC socket group is fully open (FIXME)
106 * @published: true if port has one or more associated names
107 * @conn_addrtype: address type used when establishing connection
108 */
109struct tipc_sock {
110 struct sock sk;
111 u32 max_pkt;
112 u32 maxnagle;
113 u32 portid;
114 struct tipc_msg phdr;
115 struct list_head cong_links;
116 struct list_head publications;
117 u32 pub_count;
118 atomic_t dupl_rcvcnt;
119 u16 conn_timeout;
120 bool probe_unacked;
121 u16 cong_link_cnt;
122 u16 snt_unacked;
123 u16 snd_win;
124 u16 peer_caps;
125 u16 rcv_unacked;
126 u16 rcv_win;
127 struct sockaddr_tipc peer;
128 struct rhash_head node;
129 struct tipc_mc_method mc_method;
130 struct rcu_head rcu;
131 struct tipc_group *group;
132 u32 oneway;
133 u32 nagle_start;
134 u16 snd_backlog;
135 u16 msg_acc;
136 u16 pkt_cnt;
137 bool expect_ack;
138 bool nodelay;
139 bool group_is_open;
140 bool published;
141 u8 conn_addrtype;
142};
143
144static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
145static void tipc_data_ready(struct sock *sk);
146static void tipc_write_space(struct sock *sk);
147static void tipc_sock_destruct(struct sock *sk);
148static int tipc_release(struct socket *sock);
149static void tipc_sk_timeout(struct timer_list *t);
150static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
151static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
152static int tipc_sk_leave(struct tipc_sock *tsk);
153static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
154static int tipc_sk_insert(struct tipc_sock *tsk);
155static void tipc_sk_remove(struct tipc_sock *tsk);
156static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
157static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
158static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
159static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
160
161static const struct proto_ops packet_ops;
162static const struct proto_ops stream_ops;
163static const struct proto_ops msg_ops;
164static struct proto tipc_proto;
165static const struct rhashtable_params tsk_rht_params;
166
167static u32 tsk_own_node(struct tipc_sock *tsk)
168{
169 return msg_prevnode(&tsk->phdr);
170}
171
172static u32 tsk_peer_node(struct tipc_sock *tsk)
173{
174 return msg_destnode(&tsk->phdr);
175}
176
177static u32 tsk_peer_port(struct tipc_sock *tsk)
178{
179 return msg_destport(&tsk->phdr);
180}
181
182static bool tsk_unreliable(struct tipc_sock *tsk)
183{
184 return msg_src_droppable(&tsk->phdr) != 0;
185}
186
187static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
188{
189 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
190}
191
192static bool tsk_unreturnable(struct tipc_sock *tsk)
193{
194 return msg_dest_droppable(&tsk->phdr) != 0;
195}
196
197static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
198{
199 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
200}
201
202static int tsk_importance(struct tipc_sock *tsk)
203{
204 return msg_importance(&tsk->phdr);
205}
206
207static struct tipc_sock *tipc_sk(const struct sock *sk)
208{
209 return container_of(sk, struct tipc_sock, sk);
210}
211
212int tsk_set_importance(struct sock *sk, int imp)
213{
214 if (imp > TIPC_CRITICAL_IMPORTANCE)
215 return -EINVAL;
216 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
217 return 0;
218}
219
220static bool tsk_conn_cong(struct tipc_sock *tsk)
221{
222 return tsk->snt_unacked > tsk->snd_win;
223}
224
225static u16 tsk_blocks(int len)
226{
227 return ((len / FLOWCTL_BLK_SZ) + 1);
228}
229
230/* tsk_blocks(): translate a buffer size in bytes to number of
231 * advertisable blocks, taking into account the ratio truesize(len)/len
232 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
233 */
234static u16 tsk_adv_blocks(int len)
235{
236 return len / FLOWCTL_BLK_SZ / 4;
237}
238
239/* tsk_inc(): increment counter for sent or received data
240 * - If block based flow control is not supported by peer we
241 * fall back to message based ditto, incrementing the counter
242 */
243static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
244{
245 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
246 return ((msglen / FLOWCTL_BLK_SZ) + 1);
247 return 1;
248}
249
250/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
251 */
252static void tsk_set_nagle(struct tipc_sock *tsk)
253{
254 struct sock *sk = &tsk->sk;
255
256 tsk->maxnagle = 0;
257 if (sk->sk_type != SOCK_STREAM)
258 return;
259 if (tsk->nodelay)
260 return;
261 if (!(tsk->peer_caps & TIPC_NAGLE))
262 return;
263 /* Limit node local buffer size to avoid receive queue overflow */
264 if (tsk->max_pkt == MAX_MSG_SIZE)
265 tsk->maxnagle = 1500;
266 else
267 tsk->maxnagle = tsk->max_pkt;
268}
269
270/**
271 * tsk_advance_rx_queue - discard first buffer in socket receive queue
272 * @sk: network socket
273 *
274 * Caller must hold socket lock
275 */
276static void tsk_advance_rx_queue(struct sock *sk)
277{
278 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
279 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
280}
281
282/* tipc_sk_respond() : send response message back to sender
283 */
284static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
285{
286 u32 selector;
287 u32 dnode;
288 u32 onode = tipc_own_addr(sock_net(sk));
289
290 if (!tipc_msg_reverse(onode, &skb, err))
291 return;
292
293 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
294 dnode = msg_destnode(buf_msg(skb));
295 selector = msg_origport(buf_msg(skb));
296 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
297}
298
299/**
300 * tsk_rej_rx_queue - reject all buffers in socket receive queue
301 * @sk: network socket
302 * @error: response error code
303 *
304 * Caller must hold socket lock
305 */
306static void tsk_rej_rx_queue(struct sock *sk, int error)
307{
308 struct sk_buff *skb;
309
310 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
311 tipc_sk_respond(sk, skb, error);
312}
313
314static bool tipc_sk_connected(const struct sock *sk)
315{
316 return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED;
317}
318
319/* tipc_sk_type_connectionless - check if the socket is datagram socket
320 * @sk: socket
321 *
322 * Returns true if connection less, false otherwise
323 */
324static bool tipc_sk_type_connectionless(struct sock *sk)
325{
326 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
327}
328
329/* tsk_peer_msg - verify if message was sent by connected port's peer
330 *
331 * Handles cases where the node's network address has changed from
332 * the default of <0.0.0> to its configured setting.
333 */
334static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
335{
336 struct sock *sk = &tsk->sk;
337 u32 self = tipc_own_addr(sock_net(sk));
338 u32 peer_port = tsk_peer_port(tsk);
339 u32 orig_node, peer_node;
340
341 if (unlikely(!tipc_sk_connected(sk)))
342 return false;
343
344 if (unlikely(msg_origport(msg) != peer_port))
345 return false;
346
347 orig_node = msg_orignode(msg);
348 peer_node = tsk_peer_node(tsk);
349
350 if (likely(orig_node == peer_node))
351 return true;
352
353 if (!orig_node && peer_node == self)
354 return true;
355
356 if (!peer_node && orig_node == self)
357 return true;
358
359 return false;
360}
361
362/* tipc_set_sk_state - set the sk_state of the socket
363 * @sk: socket
364 *
365 * Caller must hold socket lock
366 *
367 * Returns 0 on success, errno otherwise
368 */
369static int tipc_set_sk_state(struct sock *sk, int state)
370{
371 int oldsk_state = sk->sk_state;
372 int res = -EINVAL;
373
374 switch (state) {
375 case TIPC_OPEN:
376 res = 0;
377 break;
378 case TIPC_LISTEN:
379 case TIPC_CONNECTING:
380 if (oldsk_state == TIPC_OPEN)
381 res = 0;
382 break;
383 case TIPC_ESTABLISHED:
384 if (oldsk_state == TIPC_CONNECTING ||
385 oldsk_state == TIPC_OPEN)
386 res = 0;
387 break;
388 case TIPC_DISCONNECTING:
389 if (oldsk_state == TIPC_CONNECTING ||
390 oldsk_state == TIPC_ESTABLISHED)
391 res = 0;
392 break;
393 }
394
395 if (!res)
396 sk->sk_state = state;
397
398 return res;
399}
400
401static int tipc_sk_sock_err(struct socket *sock, long *timeout)
402{
403 struct sock *sk = sock->sk;
404 int err = sock_error(sk);
405 int typ = sock->type;
406
407 if (err)
408 return err;
409 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
410 if (sk->sk_state == TIPC_DISCONNECTING)
411 return -EPIPE;
412 else if (!tipc_sk_connected(sk))
413 return -ENOTCONN;
414 }
415 if (!*timeout)
416 return -EAGAIN;
417 if (signal_pending(current))
418 return sock_intr_errno(*timeout);
419
420 return 0;
421}
422
423#define tipc_wait_for_cond(sock_, timeo_, condition_) \
424({ \
425 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
426 struct sock *sk_; \
427 int rc_; \
428 \
429 while ((rc_ = !(condition_))) { \
430 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
431 smp_rmb(); \
432 sk_ = (sock_)->sk; \
433 rc_ = tipc_sk_sock_err((sock_), timeo_); \
434 if (rc_) \
435 break; \
436 add_wait_queue(sk_sleep(sk_), &wait_); \
437 release_sock(sk_); \
438 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
439 sched_annotate_sleep(); \
440 lock_sock(sk_); \
441 remove_wait_queue(sk_sleep(sk_), &wait_); \
442 } \
443 rc_; \
444})
445
446/**
447 * tipc_sk_create - create a TIPC socket
448 * @net: network namespace (must be default network)
449 * @sock: pre-allocated socket structure
450 * @protocol: protocol indicator (must be 0)
451 * @kern: caused by kernel or by userspace?
452 *
453 * This routine creates additional data structures used by the TIPC socket,
454 * initializes them, and links them together.
455 *
456 * Return: 0 on success, errno otherwise
457 */
458static int tipc_sk_create(struct net *net, struct socket *sock,
459 int protocol, int kern)
460{
461 const struct proto_ops *ops;
462 struct sock *sk;
463 struct tipc_sock *tsk;
464 struct tipc_msg *msg;
465
466 /* Validate arguments */
467 if (unlikely(protocol != 0))
468 return -EPROTONOSUPPORT;
469
470 switch (sock->type) {
471 case SOCK_STREAM:
472 ops = &stream_ops;
473 break;
474 case SOCK_SEQPACKET:
475 ops = &packet_ops;
476 break;
477 case SOCK_DGRAM:
478 case SOCK_RDM:
479 ops = &msg_ops;
480 break;
481 default:
482 return -EPROTOTYPE;
483 }
484
485 /* Allocate socket's protocol area */
486 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
487 if (sk == NULL)
488 return -ENOMEM;
489
490 tsk = tipc_sk(sk);
491 tsk->max_pkt = MAX_PKT_DEFAULT;
492 tsk->maxnagle = 0;
493 tsk->nagle_start = NAGLE_START_INIT;
494 INIT_LIST_HEAD(&tsk->publications);
495 INIT_LIST_HEAD(&tsk->cong_links);
496 msg = &tsk->phdr;
497
498 /* Finish initializing socket data structures */
499 sock->ops = ops;
500 sock_init_data(sock, sk);
501 tipc_set_sk_state(sk, TIPC_OPEN);
502 if (tipc_sk_insert(tsk)) {
503 sk_free(sk);
504 pr_warn("Socket create failed; port number exhausted\n");
505 return -EINVAL;
506 }
507
508 /* Ensure tsk is visible before we read own_addr. */
509 smp_mb();
510
511 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
512 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
513
514 msg_set_origport(msg, tsk->portid);
515 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
516 sk->sk_shutdown = 0;
517 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
518 sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
519 sk->sk_data_ready = tipc_data_ready;
520 sk->sk_write_space = tipc_write_space;
521 sk->sk_destruct = tipc_sock_destruct;
522 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
523 tsk->group_is_open = true;
524 atomic_set(&tsk->dupl_rcvcnt, 0);
525
526 /* Start out with safe limits until we receive an advertised window */
527 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
528 tsk->rcv_win = tsk->snd_win;
529
530 if (tipc_sk_type_connectionless(sk)) {
531 tsk_set_unreturnable(tsk, true);
532 if (sock->type == SOCK_DGRAM)
533 tsk_set_unreliable(tsk, true);
534 }
535 __skb_queue_head_init(&tsk->mc_method.deferredq);
536 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
537 return 0;
538}
539
540static void tipc_sk_callback(struct rcu_head *head)
541{
542 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
543
544 sock_put(&tsk->sk);
545}
546
547/* Caller should hold socket lock for the socket. */
548static void __tipc_shutdown(struct socket *sock, int error)
549{
550 struct sock *sk = sock->sk;
551 struct tipc_sock *tsk = tipc_sk(sk);
552 struct net *net = sock_net(sk);
553 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
554 u32 dnode = tsk_peer_node(tsk);
555 struct sk_buff *skb;
556
557 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
558 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
559 !tsk_conn_cong(tsk)));
560
561 /* Push out delayed messages if in Nagle mode */
562 tipc_sk_push_backlog(tsk, false);
563 /* Remove pending SYN */
564 __skb_queue_purge(&sk->sk_write_queue);
565
566 /* Remove partially received buffer if any */
567 skb = skb_peek(&sk->sk_receive_queue);
568 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
569 __skb_unlink(skb, &sk->sk_receive_queue);
570 kfree_skb(skb);
571 }
572
573 /* Reject all unreceived messages if connectionless */
574 if (tipc_sk_type_connectionless(sk)) {
575 tsk_rej_rx_queue(sk, error);
576 return;
577 }
578
579 switch (sk->sk_state) {
580 case TIPC_CONNECTING:
581 case TIPC_ESTABLISHED:
582 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
583 tipc_node_remove_conn(net, dnode, tsk->portid);
584 /* Send a FIN+/- to its peer */
585 skb = __skb_dequeue(&sk->sk_receive_queue);
586 if (skb) {
587 __skb_queue_purge(&sk->sk_receive_queue);
588 tipc_sk_respond(sk, skb, error);
589 break;
590 }
591 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
592 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
593 tsk_own_node(tsk), tsk_peer_port(tsk),
594 tsk->portid, error);
595 if (skb)
596 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
597 break;
598 case TIPC_LISTEN:
599 /* Reject all SYN messages */
600 tsk_rej_rx_queue(sk, error);
601 break;
602 default:
603 __skb_queue_purge(&sk->sk_receive_queue);
604 break;
605 }
606}
607
608/**
609 * tipc_release - destroy a TIPC socket
610 * @sock: socket to destroy
611 *
612 * This routine cleans up any messages that are still queued on the socket.
613 * For DGRAM and RDM socket types, all queued messages are rejected.
614 * For SEQPACKET and STREAM socket types, the first message is rejected
615 * and any others are discarded. (If the first message on a STREAM socket
616 * is partially-read, it is discarded and the next one is rejected instead.)
617 *
618 * NOTE: Rejected messages are not necessarily returned to the sender! They
619 * are returned or discarded according to the "destination droppable" setting
620 * specified for the message by the sender.
621 *
622 * Return: 0 on success, errno otherwise
623 */
624static int tipc_release(struct socket *sock)
625{
626 struct sock *sk = sock->sk;
627 struct tipc_sock *tsk;
628
629 /*
630 * Exit if socket isn't fully initialized (occurs when a failed accept()
631 * releases a pre-allocated child socket that was never used)
632 */
633 if (sk == NULL)
634 return 0;
635
636 tsk = tipc_sk(sk);
637 lock_sock(sk);
638
639 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
640 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
641 sk->sk_shutdown = SHUTDOWN_MASK;
642 tipc_sk_leave(tsk);
643 tipc_sk_withdraw(tsk, NULL);
644 __skb_queue_purge(&tsk->mc_method.deferredq);
645 sk_stop_timer(sk, &sk->sk_timer);
646 tipc_sk_remove(tsk);
647
648 sock_orphan(sk);
649 /* Reject any messages that accumulated in backlog queue */
650 release_sock(sk);
651 tipc_dest_list_purge(&tsk->cong_links);
652 tsk->cong_link_cnt = 0;
653 call_rcu(&tsk->rcu, tipc_sk_callback);
654 sock->sk = NULL;
655
656 return 0;
657}
658
659/**
660 * __tipc_bind - associate or disassociate TIPC name(s) with a socket
661 * @sock: socket structure
662 * @skaddr: socket address describing name(s) and desired operation
663 * @alen: size of socket address data structure
664 *
665 * Name and name sequence binding are indicated using a positive scope value;
666 * a negative scope value unbinds the specified name. Specifying no name
667 * (i.e. a socket address length of 0) unbinds all names from the socket.
668 *
669 * Return: 0 on success, errno otherwise
670 *
671 * NOTE: This routine doesn't need to take the socket lock since it doesn't
672 * access any non-constant socket information.
673 */
674static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
675{
676 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
677 struct tipc_sock *tsk = tipc_sk(sock->sk);
678 bool unbind = false;
679
680 if (unlikely(!alen))
681 return tipc_sk_withdraw(tsk, NULL);
682
683 if (ua->addrtype == TIPC_SERVICE_ADDR) {
684 ua->addrtype = TIPC_SERVICE_RANGE;
685 ua->sr.upper = ua->sr.lower;
686 }
687 if (ua->scope < 0) {
688 unbind = true;
689 ua->scope = -ua->scope;
690 }
691 /* Users may still use deprecated TIPC_ZONE_SCOPE */
692 if (ua->scope != TIPC_NODE_SCOPE)
693 ua->scope = TIPC_CLUSTER_SCOPE;
694
695 if (tsk->group)
696 return -EACCES;
697
698 if (unbind)
699 return tipc_sk_withdraw(tsk, ua);
700 return tipc_sk_publish(tsk, ua);
701}
702
703int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
704{
705 int res;
706
707 lock_sock(sock->sk);
708 res = __tipc_bind(sock, skaddr, alen);
709 release_sock(sock->sk);
710 return res;
711}
712
713static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
714{
715 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
716 u32 atype = ua->addrtype;
717
718 if (alen) {
719 if (!tipc_uaddr_valid(ua, alen))
720 return -EINVAL;
721 if (atype == TIPC_SOCKET_ADDR)
722 return -EAFNOSUPPORT;
723 if (ua->sr.type < TIPC_RESERVED_TYPES) {
724 pr_warn_once("Can't bind to reserved service type %u\n",
725 ua->sr.type);
726 return -EACCES;
727 }
728 }
729 return tipc_sk_bind(sock, skaddr, alen);
730}
731
732/**
733 * tipc_getname - get port ID of socket or peer socket
734 * @sock: socket structure
735 * @uaddr: area for returned socket address
736 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
737 *
738 * Return: 0 on success, errno otherwise
739 *
740 * NOTE: This routine doesn't need to take the socket lock since it only
741 * accesses socket information that is unchanging (or which changes in
742 * a completely predictable manner).
743 */
744static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
745 int peer)
746{
747 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
748 struct sock *sk = sock->sk;
749 struct tipc_sock *tsk = tipc_sk(sk);
750
751 memset(addr, 0, sizeof(*addr));
752 if (peer) {
753 if ((!tipc_sk_connected(sk)) &&
754 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
755 return -ENOTCONN;
756 addr->addr.id.ref = tsk_peer_port(tsk);
757 addr->addr.id.node = tsk_peer_node(tsk);
758 } else {
759 addr->addr.id.ref = tsk->portid;
760 addr->addr.id.node = tipc_own_addr(sock_net(sk));
761 }
762
763 addr->addrtype = TIPC_SOCKET_ADDR;
764 addr->family = AF_TIPC;
765 addr->scope = 0;
766 addr->addr.name.domain = 0;
767
768 return sizeof(*addr);
769}
770
771/**
772 * tipc_poll - read and possibly block on pollmask
773 * @file: file structure associated with the socket
774 * @sock: socket for which to calculate the poll bits
775 * @wait: ???
776 *
777 * Return: pollmask value
778 *
779 * COMMENTARY:
780 * It appears that the usual socket locking mechanisms are not useful here
781 * since the pollmask info is potentially out-of-date the moment this routine
782 * exits. TCP and other protocols seem to rely on higher level poll routines
783 * to handle any preventable race conditions, so TIPC will do the same ...
784 *
785 * IMPORTANT: The fact that a read or write operation is indicated does NOT
786 * imply that the operation will succeed, merely that it should be performed
787 * and will not block.
788 */
789static __poll_t tipc_poll(struct file *file, struct socket *sock,
790 poll_table *wait)
791{
792 struct sock *sk = sock->sk;
793 struct tipc_sock *tsk = tipc_sk(sk);
794 __poll_t revents = 0;
795
796 sock_poll_wait(file, sock, wait);
797 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
798
799 if (sk->sk_shutdown & RCV_SHUTDOWN)
800 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
801 if (sk->sk_shutdown == SHUTDOWN_MASK)
802 revents |= EPOLLHUP;
803
804 switch (sk->sk_state) {
805 case TIPC_ESTABLISHED:
806 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
807 revents |= EPOLLOUT;
808 fallthrough;
809 case TIPC_LISTEN:
810 case TIPC_CONNECTING:
811 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
812 revents |= EPOLLIN | EPOLLRDNORM;
813 break;
814 case TIPC_OPEN:
815 if (tsk->group_is_open && !tsk->cong_link_cnt)
816 revents |= EPOLLOUT;
817 if (!tipc_sk_type_connectionless(sk))
818 break;
819 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
820 break;
821 revents |= EPOLLIN | EPOLLRDNORM;
822 break;
823 case TIPC_DISCONNECTING:
824 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
825 break;
826 }
827 return revents;
828}
829
830/**
831 * tipc_sendmcast - send multicast message
832 * @sock: socket structure
833 * @ua: destination address struct
834 * @msg: message to send
835 * @dlen: length of data to send
836 * @timeout: timeout to wait for wakeup
837 *
838 * Called from function tipc_sendmsg(), which has done all sanity checks
839 * Return: the number of bytes sent on success, or errno
840 */
841static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
842 struct msghdr *msg, size_t dlen, long timeout)
843{
844 struct sock *sk = sock->sk;
845 struct tipc_sock *tsk = tipc_sk(sk);
846 struct tipc_msg *hdr = &tsk->phdr;
847 struct net *net = sock_net(sk);
848 int mtu = tipc_bcast_get_mtu(net);
849 struct sk_buff_head pkts;
850 struct tipc_nlist dsts;
851 int rc;
852
853 if (tsk->group)
854 return -EACCES;
855
856 /* Block or return if any destination link is congested */
857 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
858 if (unlikely(rc))
859 return rc;
860
861 /* Lookup destination nodes */
862 tipc_nlist_init(&dsts, tipc_own_addr(net));
863 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
864 if (!dsts.local && !dsts.remote)
865 return -EHOSTUNREACH;
866
867 /* Build message header */
868 msg_set_type(hdr, TIPC_MCAST_MSG);
869 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
870 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
871 msg_set_destport(hdr, 0);
872 msg_set_destnode(hdr, 0);
873 msg_set_nametype(hdr, ua->sr.type);
874 msg_set_namelower(hdr, ua->sr.lower);
875 msg_set_nameupper(hdr, ua->sr.upper);
876
877 /* Build message as chain of buffers */
878 __skb_queue_head_init(&pkts);
879 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
880
881 /* Send message if build was successful */
882 if (unlikely(rc == dlen)) {
883 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
884 TIPC_DUMP_SK_SNDQ, " ");
885 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
886 &tsk->cong_link_cnt);
887 }
888
889 tipc_nlist_purge(&dsts);
890
891 return rc ? rc : dlen;
892}
893
894/**
895 * tipc_send_group_msg - send a message to a member in the group
896 * @net: network namespace
897 * @tsk: tipc socket
898 * @m: message to send
899 * @mb: group member
900 * @dnode: destination node
901 * @dport: destination port
902 * @dlen: total length of message data
903 */
904static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
905 struct msghdr *m, struct tipc_member *mb,
906 u32 dnode, u32 dport, int dlen)
907{
908 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
909 struct tipc_mc_method *method = &tsk->mc_method;
910 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
911 struct tipc_msg *hdr = &tsk->phdr;
912 struct sk_buff_head pkts;
913 int mtu, rc;
914
915 /* Complete message header */
916 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
917 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
918 msg_set_destport(hdr, dport);
919 msg_set_destnode(hdr, dnode);
920 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
921
922 /* Build message as chain of buffers */
923 __skb_queue_head_init(&pkts);
924 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
925 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
926 if (unlikely(rc != dlen))
927 return rc;
928
929 /* Send message */
930 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
931 if (unlikely(rc == -ELINKCONG)) {
932 tipc_dest_push(&tsk->cong_links, dnode, 0);
933 tsk->cong_link_cnt++;
934 }
935
936 /* Update send window */
937 tipc_group_update_member(mb, blks);
938
939 /* A broadcast sent within next EXPIRE period must follow same path */
940 method->rcast = true;
941 method->mandatory = true;
942 return dlen;
943}
944
945/**
946 * tipc_send_group_unicast - send message to a member in the group
947 * @sock: socket structure
948 * @m: message to send
949 * @dlen: total length of message data
950 * @timeout: timeout to wait for wakeup
951 *
952 * Called from function tipc_sendmsg(), which has done all sanity checks
953 * Return: the number of bytes sent on success, or errno
954 */
955static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
956 int dlen, long timeout)
957{
958 struct sock *sk = sock->sk;
959 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
960 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
961 struct tipc_sock *tsk = tipc_sk(sk);
962 struct net *net = sock_net(sk);
963 struct tipc_member *mb = NULL;
964 u32 node, port;
965 int rc;
966
967 node = ua->sk.node;
968 port = ua->sk.ref;
969 if (!port && !node)
970 return -EHOSTUNREACH;
971
972 /* Block or return if destination link or member is congested */
973 rc = tipc_wait_for_cond(sock, &timeout,
974 !tipc_dest_find(&tsk->cong_links, node, 0) &&
975 tsk->group &&
976 !tipc_group_cong(tsk->group, node, port, blks,
977 &mb));
978 if (unlikely(rc))
979 return rc;
980
981 if (unlikely(!mb))
982 return -EHOSTUNREACH;
983
984 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
985
986 return rc ? rc : dlen;
987}
988
989/**
990 * tipc_send_group_anycast - send message to any member with given identity
991 * @sock: socket structure
992 * @m: message to send
993 * @dlen: total length of message data
994 * @timeout: timeout to wait for wakeup
995 *
996 * Called from function tipc_sendmsg(), which has done all sanity checks
997 * Return: the number of bytes sent on success, or errno
998 */
999static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1000 int dlen, long timeout)
1001{
1002 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1003 struct sock *sk = sock->sk;
1004 struct tipc_sock *tsk = tipc_sk(sk);
1005 struct list_head *cong_links = &tsk->cong_links;
1006 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1007 struct tipc_msg *hdr = &tsk->phdr;
1008 struct tipc_member *first = NULL;
1009 struct tipc_member *mbr = NULL;
1010 struct net *net = sock_net(sk);
1011 u32 node, port, exclude;
1012 LIST_HEAD(dsts);
1013 int lookups = 0;
1014 int dstcnt, rc;
1015 bool cong;
1016
1017 ua->sa.type = msg_nametype(hdr);
1018 ua->scope = msg_lookup_scope(hdr);
1019
1020 while (++lookups < 4) {
1021 exclude = tipc_group_exclude(tsk->group);
1022
1023 first = NULL;
1024
1025 /* Look for a non-congested destination member, if any */
1026 while (1) {
1027 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1028 exclude, false))
1029 return -EHOSTUNREACH;
1030 tipc_dest_pop(&dsts, &node, &port);
1031 cong = tipc_group_cong(tsk->group, node, port, blks,
1032 &mbr);
1033 if (!cong)
1034 break;
1035 if (mbr == first)
1036 break;
1037 if (!first)
1038 first = mbr;
1039 }
1040
1041 /* Start over if destination was not in member list */
1042 if (unlikely(!mbr))
1043 continue;
1044
1045 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1046 break;
1047
1048 /* Block or return if destination link or member is congested */
1049 rc = tipc_wait_for_cond(sock, &timeout,
1050 !tipc_dest_find(cong_links, node, 0) &&
1051 tsk->group &&
1052 !tipc_group_cong(tsk->group, node, port,
1053 blks, &mbr));
1054 if (unlikely(rc))
1055 return rc;
1056
1057 /* Send, unless destination disappeared while waiting */
1058 if (likely(mbr))
1059 break;
1060 }
1061
1062 if (unlikely(lookups >= 4))
1063 return -EHOSTUNREACH;
1064
1065 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1066
1067 return rc ? rc : dlen;
1068}
1069
1070/**
1071 * tipc_send_group_bcast - send message to all members in communication group
1072 * @sock: socket structure
1073 * @m: message to send
1074 * @dlen: total length of message data
1075 * @timeout: timeout to wait for wakeup
1076 *
1077 * Called from function tipc_sendmsg(), which has done all sanity checks
1078 * Return: the number of bytes sent on success, or errno
1079 */
1080static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1081 int dlen, long timeout)
1082{
1083 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1084 struct sock *sk = sock->sk;
1085 struct net *net = sock_net(sk);
1086 struct tipc_sock *tsk = tipc_sk(sk);
1087 struct tipc_nlist *dsts;
1088 struct tipc_mc_method *method = &tsk->mc_method;
1089 bool ack = method->mandatory && method->rcast;
1090 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1091 struct tipc_msg *hdr = &tsk->phdr;
1092 int mtu = tipc_bcast_get_mtu(net);
1093 struct sk_buff_head pkts;
1094 int rc = -EHOSTUNREACH;
1095
1096 /* Block or return if any destination link or member is congested */
1097 rc = tipc_wait_for_cond(sock, &timeout,
1098 !tsk->cong_link_cnt && tsk->group &&
1099 !tipc_group_bc_cong(tsk->group, blks));
1100 if (unlikely(rc))
1101 return rc;
1102
1103 dsts = tipc_group_dests(tsk->group);
1104 if (!dsts->local && !dsts->remote)
1105 return -EHOSTUNREACH;
1106
1107 /* Complete message header */
1108 if (ua) {
1109 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1110 msg_set_nameinst(hdr, ua->sa.instance);
1111 } else {
1112 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1113 msg_set_nameinst(hdr, 0);
1114 }
1115 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1116 msg_set_destport(hdr, 0);
1117 msg_set_destnode(hdr, 0);
1118 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1119
1120 /* Avoid getting stuck with repeated forced replicasts */
1121 msg_set_grp_bc_ack_req(hdr, ack);
1122
1123 /* Build message as chain of buffers */
1124 __skb_queue_head_init(&pkts);
1125 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1126 if (unlikely(rc != dlen))
1127 return rc;
1128
1129 /* Send message */
1130 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1131 if (unlikely(rc))
1132 return rc;
1133
1134 /* Update broadcast sequence number and send windows */
1135 tipc_group_update_bc_members(tsk->group, blks, ack);
1136
1137 /* Broadcast link is now free to choose method for next broadcast */
1138 method->mandatory = false;
1139 method->expires = jiffies;
1140
1141 return dlen;
1142}
1143
1144/**
1145 * tipc_send_group_mcast - send message to all members with given identity
1146 * @sock: socket structure
1147 * @m: message to send
1148 * @dlen: total length of message data
1149 * @timeout: timeout to wait for wakeup
1150 *
1151 * Called from function tipc_sendmsg(), which has done all sanity checks
1152 * Return: the number of bytes sent on success, or errno
1153 */
1154static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1155 int dlen, long timeout)
1156{
1157 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1158 struct sock *sk = sock->sk;
1159 struct tipc_sock *tsk = tipc_sk(sk);
1160 struct tipc_group *grp = tsk->group;
1161 struct tipc_msg *hdr = &tsk->phdr;
1162 struct net *net = sock_net(sk);
1163 u32 dstcnt, exclude;
1164 LIST_HEAD(dsts);
1165
1166 ua->sa.type = msg_nametype(hdr);
1167 ua->scope = msg_lookup_scope(hdr);
1168 exclude = tipc_group_exclude(grp);
1169
1170 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1171 return -EHOSTUNREACH;
1172
1173 if (dstcnt == 1) {
1174 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1175 return tipc_send_group_unicast(sock, m, dlen, timeout);
1176 }
1177
1178 tipc_dest_list_purge(&dsts);
1179 return tipc_send_group_bcast(sock, m, dlen, timeout);
1180}
1181
1182/**
1183 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1184 * @net: the associated network namespace
1185 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1186 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1187 *
1188 * Multi-threaded: parallel calls with reference to same queues may occur
1189 */
1190void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1191 struct sk_buff_head *inputq)
1192{
1193 u32 self = tipc_own_addr(net);
1194 struct sk_buff *skb, *_skb;
1195 u32 portid, onode;
1196 struct sk_buff_head tmpq;
1197 struct list_head dports;
1198 struct tipc_msg *hdr;
1199 struct tipc_uaddr ua;
1200 int user, mtyp, hlen;
1201
1202 __skb_queue_head_init(&tmpq);
1203 INIT_LIST_HEAD(&dports);
1204 ua.addrtype = TIPC_SERVICE_RANGE;
1205
1206 /* tipc_skb_peek() increments the head skb's reference counter */
1207 skb = tipc_skb_peek(arrvq, &inputq->lock);
1208 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1209 hdr = buf_msg(skb);
1210 user = msg_user(hdr);
1211 mtyp = msg_type(hdr);
1212 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1213 onode = msg_orignode(hdr);
1214 ua.sr.type = msg_nametype(hdr);
1215 ua.sr.lower = msg_namelower(hdr);
1216 ua.sr.upper = msg_nameupper(hdr);
1217 if (onode == self)
1218 ua.scope = TIPC_ANY_SCOPE;
1219 else
1220 ua.scope = TIPC_CLUSTER_SCOPE;
1221
1222 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1223 spin_lock_bh(&inputq->lock);
1224 if (skb_peek(arrvq) == skb) {
1225 __skb_dequeue(arrvq);
1226 __skb_queue_tail(inputq, skb);
1227 }
1228 kfree_skb(skb);
1229 spin_unlock_bh(&inputq->lock);
1230 continue;
1231 }
1232
1233 /* Group messages require exact scope match */
1234 if (msg_in_group(hdr)) {
1235 ua.sr.lower = 0;
1236 ua.sr.upper = ~0;
1237 ua.scope = msg_lookup_scope(hdr);
1238 }
1239
1240 /* Create destination port list: */
1241 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1242
1243 /* Clone message per destination */
1244 while (tipc_dest_pop(&dports, NULL, &portid)) {
1245 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1246 if (_skb) {
1247 msg_set_destport(buf_msg(_skb), portid);
1248 __skb_queue_tail(&tmpq, _skb);
1249 continue;
1250 }
1251 pr_warn("Failed to clone mcast rcv buffer\n");
1252 }
1253 /* Append clones to inputq only if skb is still head of arrvq */
1254 spin_lock_bh(&inputq->lock);
1255 if (skb_peek(arrvq) == skb) {
1256 skb_queue_splice_tail_init(&tmpq, inputq);
1257 /* Decrement the skb's refcnt */
1258 kfree_skb(__skb_dequeue(arrvq));
1259 }
1260 spin_unlock_bh(&inputq->lock);
1261 __skb_queue_purge(&tmpq);
1262 kfree_skb(skb);
1263 }
1264 tipc_sk_rcv(net, inputq);
1265}
1266
1267/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1268 * when socket is in Nagle mode
1269 */
1270static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1271{
1272 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1273 struct sk_buff *skb = skb_peek_tail(txq);
1274 struct net *net = sock_net(&tsk->sk);
1275 u32 dnode = tsk_peer_node(tsk);
1276 int rc;
1277
1278 if (nagle_ack) {
1279 tsk->pkt_cnt += skb_queue_len(txq);
1280 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1281 tsk->oneway = 0;
1282 if (tsk->nagle_start < NAGLE_START_MAX)
1283 tsk->nagle_start *= 2;
1284 tsk->expect_ack = false;
1285 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1286 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1287 tsk->nagle_start);
1288 } else {
1289 tsk->nagle_start = NAGLE_START_INIT;
1290 if (skb) {
1291 msg_set_ack_required(buf_msg(skb));
1292 tsk->expect_ack = true;
1293 } else {
1294 tsk->expect_ack = false;
1295 }
1296 }
1297 tsk->msg_acc = 0;
1298 tsk->pkt_cnt = 0;
1299 }
1300
1301 if (!skb || tsk->cong_link_cnt)
1302 return;
1303
1304 /* Do not send SYN again after congestion */
1305 if (msg_is_syn(buf_msg(skb)))
1306 return;
1307
1308 if (tsk->msg_acc)
1309 tsk->pkt_cnt += skb_queue_len(txq);
1310 tsk->snt_unacked += tsk->snd_backlog;
1311 tsk->snd_backlog = 0;
1312 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1313 if (rc == -ELINKCONG)
1314 tsk->cong_link_cnt = 1;
1315}
1316
1317/**
1318 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1319 * @tsk: receiving socket
1320 * @skb: pointer to message buffer.
1321 * @inputq: buffer list containing the buffers
1322 * @xmitq: output message area
1323 */
1324static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1325 struct sk_buff_head *inputq,
1326 struct sk_buff_head *xmitq)
1327{
1328 struct tipc_msg *hdr = buf_msg(skb);
1329 u32 onode = tsk_own_node(tsk);
1330 struct sock *sk = &tsk->sk;
1331 int mtyp = msg_type(hdr);
1332 bool was_cong;
1333
1334 /* Ignore if connection cannot be validated: */
1335 if (!tsk_peer_msg(tsk, hdr)) {
1336 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1337 goto exit;
1338 }
1339
1340 if (unlikely(msg_errcode(hdr))) {
1341 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1342 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1343 tsk_peer_port(tsk));
1344 sk->sk_state_change(sk);
1345
1346 /* State change is ignored if socket already awake,
1347 * - convert msg to abort msg and add to inqueue
1348 */
1349 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1350 msg_set_type(hdr, TIPC_CONN_MSG);
1351 msg_set_size(hdr, BASIC_H_SIZE);
1352 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1353 __skb_queue_tail(inputq, skb);
1354 return;
1355 }
1356
1357 tsk->probe_unacked = false;
1358
1359 if (mtyp == CONN_PROBE) {
1360 msg_set_type(hdr, CONN_PROBE_REPLY);
1361 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1362 __skb_queue_tail(xmitq, skb);
1363 return;
1364 } else if (mtyp == CONN_ACK) {
1365 was_cong = tsk_conn_cong(tsk);
1366 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1367 tsk->snt_unacked -= msg_conn_ack(hdr);
1368 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1369 tsk->snd_win = msg_adv_win(hdr);
1370 if (was_cong && !tsk_conn_cong(tsk))
1371 sk->sk_write_space(sk);
1372 } else if (mtyp != CONN_PROBE_REPLY) {
1373 pr_warn("Received unknown CONN_PROTO msg\n");
1374 }
1375exit:
1376 kfree_skb(skb);
1377}
1378
1379/**
1380 * tipc_sendmsg - send message in connectionless manner
1381 * @sock: socket structure
1382 * @m: message to send
1383 * @dsz: amount of user data to be sent
1384 *
1385 * Message must have an destination specified explicitly.
1386 * Used for SOCK_RDM and SOCK_DGRAM messages,
1387 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1388 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1389 *
1390 * Return: the number of bytes sent on success, or errno otherwise
1391 */
1392static int tipc_sendmsg(struct socket *sock,
1393 struct msghdr *m, size_t dsz)
1394{
1395 struct sock *sk = sock->sk;
1396 int ret;
1397
1398 lock_sock(sk);
1399 ret = __tipc_sendmsg(sock, m, dsz);
1400 release_sock(sk);
1401
1402 return ret;
1403}
1404
1405static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1406{
1407 struct sock *sk = sock->sk;
1408 struct net *net = sock_net(sk);
1409 struct tipc_sock *tsk = tipc_sk(sk);
1410 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1411 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1412 struct list_head *clinks = &tsk->cong_links;
1413 bool syn = !tipc_sk_type_connectionless(sk);
1414 struct tipc_group *grp = tsk->group;
1415 struct tipc_msg *hdr = &tsk->phdr;
1416 struct tipc_socket_addr skaddr;
1417 struct sk_buff_head pkts;
1418 int atype, mtu, rc;
1419
1420 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1421 return -EMSGSIZE;
1422
1423 if (ua) {
1424 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1425 return -EINVAL;
1426 atype = ua->addrtype;
1427 }
1428
1429 /* If socket belongs to a communication group follow other paths */
1430 if (grp) {
1431 if (!ua)
1432 return tipc_send_group_bcast(sock, m, dlen, timeout);
1433 if (atype == TIPC_SERVICE_ADDR)
1434 return tipc_send_group_anycast(sock, m, dlen, timeout);
1435 if (atype == TIPC_SOCKET_ADDR)
1436 return tipc_send_group_unicast(sock, m, dlen, timeout);
1437 if (atype == TIPC_SERVICE_RANGE)
1438 return tipc_send_group_mcast(sock, m, dlen, timeout);
1439 return -EINVAL;
1440 }
1441
1442 if (!ua) {
1443 ua = (struct tipc_uaddr *)&tsk->peer;
1444 if (!syn && ua->family != AF_TIPC)
1445 return -EDESTADDRREQ;
1446 atype = ua->addrtype;
1447 }
1448
1449 if (unlikely(syn)) {
1450 if (sk->sk_state == TIPC_LISTEN)
1451 return -EPIPE;
1452 if (sk->sk_state != TIPC_OPEN)
1453 return -EISCONN;
1454 if (tsk->published)
1455 return -EOPNOTSUPP;
1456 if (atype == TIPC_SERVICE_ADDR)
1457 tsk->conn_addrtype = atype;
1458 msg_set_syn(hdr, 1);
1459 }
1460
1461 memset(&skaddr, 0, sizeof(skaddr));
1462
1463 /* Determine destination */
1464 if (atype == TIPC_SERVICE_RANGE) {
1465 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1466 } else if (atype == TIPC_SERVICE_ADDR) {
1467 skaddr.node = ua->lookup_node;
1468 ua->scope = tipc_node2scope(skaddr.node);
1469 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1470 return -EHOSTUNREACH;
1471 } else if (atype == TIPC_SOCKET_ADDR) {
1472 skaddr = ua->sk;
1473 } else {
1474 return -EINVAL;
1475 }
1476
1477 /* Block or return if destination link is congested */
1478 rc = tipc_wait_for_cond(sock, &timeout,
1479 !tipc_dest_find(clinks, skaddr.node, 0));
1480 if (unlikely(rc))
1481 return rc;
1482
1483 /* Finally build message header */
1484 msg_set_destnode(hdr, skaddr.node);
1485 msg_set_destport(hdr, skaddr.ref);
1486 if (atype == TIPC_SERVICE_ADDR) {
1487 msg_set_type(hdr, TIPC_NAMED_MSG);
1488 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1489 msg_set_nametype(hdr, ua->sa.type);
1490 msg_set_nameinst(hdr, ua->sa.instance);
1491 msg_set_lookup_scope(hdr, ua->scope);
1492 } else { /* TIPC_SOCKET_ADDR */
1493 msg_set_type(hdr, TIPC_DIRECT_MSG);
1494 msg_set_lookup_scope(hdr, 0);
1495 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1496 }
1497
1498 /* Add message body */
1499 __skb_queue_head_init(&pkts);
1500 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1501 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1502 if (unlikely(rc != dlen))
1503 return rc;
1504 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1505 __skb_queue_purge(&pkts);
1506 return -ENOMEM;
1507 }
1508
1509 /* Send message */
1510 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1511 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1512 if (unlikely(rc == -ELINKCONG)) {
1513 tipc_dest_push(clinks, skaddr.node, 0);
1514 tsk->cong_link_cnt++;
1515 rc = 0;
1516 }
1517
1518 if (unlikely(syn && !rc)) {
1519 tipc_set_sk_state(sk, TIPC_CONNECTING);
1520 if (dlen && timeout) {
1521 timeout = msecs_to_jiffies(timeout);
1522 tipc_wait_for_connect(sock, &timeout);
1523 }
1524 }
1525
1526 return rc ? rc : dlen;
1527}
1528
1529/**
1530 * tipc_sendstream - send stream-oriented data
1531 * @sock: socket structure
1532 * @m: data to send
1533 * @dsz: total length of data to be transmitted
1534 *
1535 * Used for SOCK_STREAM data.
1536 *
1537 * Return: the number of bytes sent on success (or partial success),
1538 * or errno if no data sent
1539 */
1540static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1541{
1542 struct sock *sk = sock->sk;
1543 int ret;
1544
1545 lock_sock(sk);
1546 ret = __tipc_sendstream(sock, m, dsz);
1547 release_sock(sk);
1548
1549 return ret;
1550}
1551
1552static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1553{
1554 struct sock *sk = sock->sk;
1555 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1556 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1557 struct sk_buff_head *txq = &sk->sk_write_queue;
1558 struct tipc_sock *tsk = tipc_sk(sk);
1559 struct tipc_msg *hdr = &tsk->phdr;
1560 struct net *net = sock_net(sk);
1561 struct sk_buff *skb;
1562 u32 dnode = tsk_peer_node(tsk);
1563 int maxnagle = tsk->maxnagle;
1564 int maxpkt = tsk->max_pkt;
1565 int send, sent = 0;
1566 int blocks, rc = 0;
1567
1568 if (unlikely(dlen > INT_MAX))
1569 return -EMSGSIZE;
1570
1571 /* Handle implicit connection setup */
1572 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1573 rc = __tipc_sendmsg(sock, m, dlen);
1574 if (dlen && dlen == rc) {
1575 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1576 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1577 }
1578 return rc;
1579 }
1580
1581 do {
1582 rc = tipc_wait_for_cond(sock, &timeout,
1583 (!tsk->cong_link_cnt &&
1584 !tsk_conn_cong(tsk) &&
1585 tipc_sk_connected(sk)));
1586 if (unlikely(rc))
1587 break;
1588 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1589 blocks = tsk->snd_backlog;
1590 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1591 send <= maxnagle) {
1592 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1593 if (unlikely(rc < 0))
1594 break;
1595 blocks += rc;
1596 tsk->msg_acc++;
1597 if (blocks <= 64 && tsk->expect_ack) {
1598 tsk->snd_backlog = blocks;
1599 sent += send;
1600 break;
1601 } else if (blocks > 64) {
1602 tsk->pkt_cnt += skb_queue_len(txq);
1603 } else {
1604 skb = skb_peek_tail(txq);
1605 if (skb) {
1606 msg_set_ack_required(buf_msg(skb));
1607 tsk->expect_ack = true;
1608 } else {
1609 tsk->expect_ack = false;
1610 }
1611 tsk->msg_acc = 0;
1612 tsk->pkt_cnt = 0;
1613 }
1614 } else {
1615 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1616 if (unlikely(rc != send))
1617 break;
1618 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1619 }
1620 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1621 TIPC_DUMP_SK_SNDQ, " ");
1622 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1623 if (unlikely(rc == -ELINKCONG)) {
1624 tsk->cong_link_cnt = 1;
1625 rc = 0;
1626 }
1627 if (likely(!rc)) {
1628 tsk->snt_unacked += blocks;
1629 tsk->snd_backlog = 0;
1630 sent += send;
1631 }
1632 } while (sent < dlen && !rc);
1633
1634 return sent ? sent : rc;
1635}
1636
1637/**
1638 * tipc_send_packet - send a connection-oriented message
1639 * @sock: socket structure
1640 * @m: message to send
1641 * @dsz: length of data to be transmitted
1642 *
1643 * Used for SOCK_SEQPACKET messages.
1644 *
1645 * Return: the number of bytes sent on success, or errno otherwise
1646 */
1647static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1648{
1649 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1650 return -EMSGSIZE;
1651
1652 return tipc_sendstream(sock, m, dsz);
1653}
1654
1655/* tipc_sk_finish_conn - complete the setup of a connection
1656 */
1657static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1658 u32 peer_node)
1659{
1660 struct sock *sk = &tsk->sk;
1661 struct net *net = sock_net(sk);
1662 struct tipc_msg *msg = &tsk->phdr;
1663
1664 msg_set_syn(msg, 0);
1665 msg_set_destnode(msg, peer_node);
1666 msg_set_destport(msg, peer_port);
1667 msg_set_type(msg, TIPC_CONN_MSG);
1668 msg_set_lookup_scope(msg, 0);
1669 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1670
1671 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1672 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1673 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1674 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1675 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1676 tsk_set_nagle(tsk);
1677 __skb_queue_purge(&sk->sk_write_queue);
1678 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1679 return;
1680
1681 /* Fall back to message based flow control */
1682 tsk->rcv_win = FLOWCTL_MSG_WIN;
1683 tsk->snd_win = FLOWCTL_MSG_WIN;
1684}
1685
1686/**
1687 * tipc_sk_set_orig_addr - capture sender's address for received message
1688 * @m: descriptor for message info
1689 * @skb: received message
1690 *
1691 * Note: Address is not captured if not requested by receiver.
1692 */
1693static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1694{
1695 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1696 struct tipc_msg *hdr = buf_msg(skb);
1697
1698 if (!srcaddr)
1699 return;
1700
1701 srcaddr->sock.family = AF_TIPC;
1702 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1703 srcaddr->sock.scope = 0;
1704 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1705 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1706 srcaddr->sock.addr.name.domain = 0;
1707 m->msg_namelen = sizeof(struct sockaddr_tipc);
1708
1709 if (!msg_in_group(hdr))
1710 return;
1711
1712 /* Group message users may also want to know sending member's id */
1713 srcaddr->member.family = AF_TIPC;
1714 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1715 srcaddr->member.scope = 0;
1716 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1717 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1718 srcaddr->member.addr.name.domain = 0;
1719 m->msg_namelen = sizeof(*srcaddr);
1720}
1721
1722/**
1723 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1724 * @m: descriptor for message info
1725 * @skb: received message buffer
1726 * @tsk: TIPC port associated with message
1727 *
1728 * Note: Ancillary data is not captured if not requested by receiver.
1729 *
1730 * Return: 0 if successful, otherwise errno
1731 */
1732static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1733 struct tipc_sock *tsk)
1734{
1735 struct tipc_msg *hdr;
1736 u32 data[3] = {0,};
1737 bool has_addr;
1738 int dlen, rc;
1739
1740 if (likely(m->msg_controllen == 0))
1741 return 0;
1742
1743 hdr = buf_msg(skb);
1744 dlen = msg_data_sz(hdr);
1745
1746 /* Capture errored message object, if any */
1747 if (msg_errcode(hdr)) {
1748 if (skb_linearize(skb))
1749 return -ENOMEM;
1750 hdr = buf_msg(skb);
1751 data[0] = msg_errcode(hdr);
1752 data[1] = dlen;
1753 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1754 if (rc || !dlen)
1755 return rc;
1756 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1757 if (rc)
1758 return rc;
1759 }
1760
1761 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1762 switch (msg_type(hdr)) {
1763 case TIPC_NAMED_MSG:
1764 has_addr = true;
1765 data[0] = msg_nametype(hdr);
1766 data[1] = msg_namelower(hdr);
1767 data[2] = data[1];
1768 break;
1769 case TIPC_MCAST_MSG:
1770 has_addr = true;
1771 data[0] = msg_nametype(hdr);
1772 data[1] = msg_namelower(hdr);
1773 data[2] = msg_nameupper(hdr);
1774 break;
1775 case TIPC_CONN_MSG:
1776 has_addr = !!tsk->conn_addrtype;
1777 data[0] = msg_nametype(&tsk->phdr);
1778 data[1] = msg_nameinst(&tsk->phdr);
1779 data[2] = data[1];
1780 break;
1781 default:
1782 has_addr = false;
1783 }
1784 if (!has_addr)
1785 return 0;
1786 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1787}
1788
1789static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1790{
1791 struct sock *sk = &tsk->sk;
1792 struct sk_buff *skb = NULL;
1793 struct tipc_msg *msg;
1794 u32 peer_port = tsk_peer_port(tsk);
1795 u32 dnode = tsk_peer_node(tsk);
1796
1797 if (!tipc_sk_connected(sk))
1798 return NULL;
1799 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1800 dnode, tsk_own_node(tsk), peer_port,
1801 tsk->portid, TIPC_OK);
1802 if (!skb)
1803 return NULL;
1804 msg = buf_msg(skb);
1805 msg_set_conn_ack(msg, tsk->rcv_unacked);
1806 tsk->rcv_unacked = 0;
1807
1808 /* Adjust to and advertize the correct window limit */
1809 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1810 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1811 msg_set_adv_win(msg, tsk->rcv_win);
1812 }
1813 return skb;
1814}
1815
1816static void tipc_sk_send_ack(struct tipc_sock *tsk)
1817{
1818 struct sk_buff *skb;
1819
1820 skb = tipc_sk_build_ack(tsk);
1821 if (!skb)
1822 return;
1823
1824 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1825 msg_link_selector(buf_msg(skb)));
1826}
1827
1828static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1829{
1830 struct sock *sk = sock->sk;
1831 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1832 long timeo = *timeop;
1833 int err = sock_error(sk);
1834
1835 if (err)
1836 return err;
1837
1838 for (;;) {
1839 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1840 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1841 err = -ENOTCONN;
1842 break;
1843 }
1844 add_wait_queue(sk_sleep(sk), &wait);
1845 release_sock(sk);
1846 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1847 sched_annotate_sleep();
1848 lock_sock(sk);
1849 remove_wait_queue(sk_sleep(sk), &wait);
1850 }
1851 err = 0;
1852 if (!skb_queue_empty(&sk->sk_receive_queue))
1853 break;
1854 err = -EAGAIN;
1855 if (!timeo)
1856 break;
1857 err = sock_intr_errno(timeo);
1858 if (signal_pending(current))
1859 break;
1860
1861 err = sock_error(sk);
1862 if (err)
1863 break;
1864 }
1865 *timeop = timeo;
1866 return err;
1867}
1868
1869/**
1870 * tipc_recvmsg - receive packet-oriented message
1871 * @sock: network socket
1872 * @m: descriptor for message info
1873 * @buflen: length of user buffer area
1874 * @flags: receive flags
1875 *
1876 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1877 * If the complete message doesn't fit in user area, truncate it.
1878 *
1879 * Return: size of returned message data, errno otherwise
1880 */
1881static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1882 size_t buflen, int flags)
1883{
1884 struct sock *sk = sock->sk;
1885 bool connected = !tipc_sk_type_connectionless(sk);
1886 struct tipc_sock *tsk = tipc_sk(sk);
1887 int rc, err, hlen, dlen, copy;
1888 struct tipc_skb_cb *skb_cb;
1889 struct sk_buff_head xmitq;
1890 struct tipc_msg *hdr;
1891 struct sk_buff *skb;
1892 bool grp_evt;
1893 long timeout;
1894
1895 /* Catch invalid receive requests */
1896 if (unlikely(!buflen))
1897 return -EINVAL;
1898
1899 lock_sock(sk);
1900 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1901 rc = -ENOTCONN;
1902 goto exit;
1903 }
1904 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1905
1906 /* Step rcv queue to first msg with data or error; wait if necessary */
1907 do {
1908 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1909 if (unlikely(rc))
1910 goto exit;
1911 skb = skb_peek(&sk->sk_receive_queue);
1912 skb_cb = TIPC_SKB_CB(skb);
1913 hdr = buf_msg(skb);
1914 dlen = msg_data_sz(hdr);
1915 hlen = msg_hdr_sz(hdr);
1916 err = msg_errcode(hdr);
1917 grp_evt = msg_is_grp_evt(hdr);
1918 if (likely(dlen || err))
1919 break;
1920 tsk_advance_rx_queue(sk);
1921 } while (1);
1922
1923 /* Collect msg meta data, including error code and rejected data */
1924 tipc_sk_set_orig_addr(m, skb);
1925 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1926 if (unlikely(rc))
1927 goto exit;
1928 hdr = buf_msg(skb);
1929
1930 /* Capture data if non-error msg, otherwise just set return value */
1931 if (likely(!err)) {
1932 int offset = skb_cb->bytes_read;
1933
1934 copy = min_t(int, dlen - offset, buflen);
1935 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1936 if (unlikely(rc))
1937 goto exit;
1938 if (unlikely(offset + copy < dlen)) {
1939 if (flags & MSG_EOR) {
1940 if (!(flags & MSG_PEEK))
1941 skb_cb->bytes_read = offset + copy;
1942 } else {
1943 m->msg_flags |= MSG_TRUNC;
1944 skb_cb->bytes_read = 0;
1945 }
1946 } else {
1947 if (flags & MSG_EOR)
1948 m->msg_flags |= MSG_EOR;
1949 skb_cb->bytes_read = 0;
1950 }
1951 } else {
1952 copy = 0;
1953 rc = 0;
1954 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1955 rc = -ECONNRESET;
1956 goto exit;
1957 }
1958 }
1959
1960 /* Mark message as group event if applicable */
1961 if (unlikely(grp_evt)) {
1962 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1963 m->msg_flags |= MSG_EOR;
1964 m->msg_flags |= MSG_OOB;
1965 copy = 0;
1966 }
1967
1968 /* Caption of data or error code/rejected data was successful */
1969 if (unlikely(flags & MSG_PEEK))
1970 goto exit;
1971
1972 /* Send group flow control advertisement when applicable */
1973 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1974 __skb_queue_head_init(&xmitq);
1975 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1976 msg_orignode(hdr), msg_origport(hdr),
1977 &xmitq);
1978 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1979 }
1980
1981 if (skb_cb->bytes_read)
1982 goto exit;
1983
1984 tsk_advance_rx_queue(sk);
1985
1986 if (likely(!connected))
1987 goto exit;
1988
1989 /* Send connection flow control advertisement when applicable */
1990 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1991 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1992 tipc_sk_send_ack(tsk);
1993exit:
1994 release_sock(sk);
1995 return rc ? rc : copy;
1996}
1997
1998/**
1999 * tipc_recvstream - receive stream-oriented data
2000 * @sock: network socket
2001 * @m: descriptor for message info
2002 * @buflen: total size of user buffer area
2003 * @flags: receive flags
2004 *
2005 * Used for SOCK_STREAM messages only. If not enough data is available
2006 * will optionally wait for more; never truncates data.
2007 *
2008 * Return: size of returned message data, errno otherwise
2009 */
2010static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2011 size_t buflen, int flags)
2012{
2013 struct sock *sk = sock->sk;
2014 struct tipc_sock *tsk = tipc_sk(sk);
2015 struct sk_buff *skb;
2016 struct tipc_msg *hdr;
2017 struct tipc_skb_cb *skb_cb;
2018 bool peek = flags & MSG_PEEK;
2019 int offset, required, copy, copied = 0;
2020 int hlen, dlen, err, rc;
2021 long timeout;
2022
2023 /* Catch invalid receive attempts */
2024 if (unlikely(!buflen))
2025 return -EINVAL;
2026
2027 lock_sock(sk);
2028
2029 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2030 rc = -ENOTCONN;
2031 goto exit;
2032 }
2033 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2034 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2035
2036 do {
2037 /* Look at first msg in receive queue; wait if necessary */
2038 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2039 if (unlikely(rc))
2040 break;
2041 skb = skb_peek(&sk->sk_receive_queue);
2042 skb_cb = TIPC_SKB_CB(skb);
2043 hdr = buf_msg(skb);
2044 dlen = msg_data_sz(hdr);
2045 hlen = msg_hdr_sz(hdr);
2046 err = msg_errcode(hdr);
2047
2048 /* Discard any empty non-errored (SYN-) message */
2049 if (unlikely(!dlen && !err)) {
2050 tsk_advance_rx_queue(sk);
2051 continue;
2052 }
2053
2054 /* Collect msg meta data, incl. error code and rejected data */
2055 if (!copied) {
2056 tipc_sk_set_orig_addr(m, skb);
2057 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2058 if (rc)
2059 break;
2060 hdr = buf_msg(skb);
2061 }
2062
2063 /* Copy data if msg ok, otherwise return error/partial data */
2064 if (likely(!err)) {
2065 offset = skb_cb->bytes_read;
2066 copy = min_t(int, dlen - offset, buflen - copied);
2067 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2068 if (unlikely(rc))
2069 break;
2070 copied += copy;
2071 offset += copy;
2072 if (unlikely(offset < dlen)) {
2073 if (!peek)
2074 skb_cb->bytes_read = offset;
2075 break;
2076 }
2077 } else {
2078 rc = 0;
2079 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2080 rc = -ECONNRESET;
2081 if (copied || rc)
2082 break;
2083 }
2084
2085 if (unlikely(peek))
2086 break;
2087
2088 tsk_advance_rx_queue(sk);
2089
2090 /* Send connection flow control advertisement when applicable */
2091 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2092 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2093 tipc_sk_send_ack(tsk);
2094
2095 /* Exit if all requested data or FIN/error received */
2096 if (copied == buflen || err)
2097 break;
2098
2099 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2100exit:
2101 release_sock(sk);
2102 return copied ? copied : rc;
2103}
2104
2105/**
2106 * tipc_write_space - wake up thread if port congestion is released
2107 * @sk: socket
2108 */
2109static void tipc_write_space(struct sock *sk)
2110{
2111 struct socket_wq *wq;
2112
2113 rcu_read_lock();
2114 wq = rcu_dereference(sk->sk_wq);
2115 if (skwq_has_sleeper(wq))
2116 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2117 EPOLLWRNORM | EPOLLWRBAND);
2118 rcu_read_unlock();
2119}
2120
2121/**
2122 * tipc_data_ready - wake up threads to indicate messages have been received
2123 * @sk: socket
2124 */
2125static void tipc_data_ready(struct sock *sk)
2126{
2127 struct socket_wq *wq;
2128
2129 trace_sk_data_ready(sk);
2130
2131 rcu_read_lock();
2132 wq = rcu_dereference(sk->sk_wq);
2133 if (skwq_has_sleeper(wq))
2134 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2135 EPOLLRDNORM | EPOLLRDBAND);
2136 rcu_read_unlock();
2137}
2138
2139static void tipc_sock_destruct(struct sock *sk)
2140{
2141 __skb_queue_purge(&sk->sk_receive_queue);
2142}
2143
2144static void tipc_sk_proto_rcv(struct sock *sk,
2145 struct sk_buff_head *inputq,
2146 struct sk_buff_head *xmitq)
2147{
2148 struct sk_buff *skb = __skb_dequeue(inputq);
2149 struct tipc_sock *tsk = tipc_sk(sk);
2150 struct tipc_msg *hdr = buf_msg(skb);
2151 struct tipc_group *grp = tsk->group;
2152 bool wakeup = false;
2153
2154 switch (msg_user(hdr)) {
2155 case CONN_MANAGER:
2156 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2157 return;
2158 case SOCK_WAKEUP:
2159 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2160 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2161 smp_wmb();
2162 tsk->cong_link_cnt--;
2163 wakeup = true;
2164 tipc_sk_push_backlog(tsk, false);
2165 break;
2166 case GROUP_PROTOCOL:
2167 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2168 break;
2169 case TOP_SRV:
2170 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2171 hdr, inputq, xmitq);
2172 break;
2173 default:
2174 break;
2175 }
2176
2177 if (wakeup)
2178 sk->sk_write_space(sk);
2179
2180 kfree_skb(skb);
2181}
2182
2183/**
2184 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2185 * @tsk: TIPC socket
2186 * @skb: pointer to message buffer.
2187 * @xmitq: for Nagle ACK if any
2188 * Return: true if message should be added to receive queue, false otherwise
2189 */
2190static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2191 struct sk_buff_head *xmitq)
2192{
2193 struct sock *sk = &tsk->sk;
2194 struct net *net = sock_net(sk);
2195 struct tipc_msg *hdr = buf_msg(skb);
2196 bool con_msg = msg_connected(hdr);
2197 u32 pport = tsk_peer_port(tsk);
2198 u32 pnode = tsk_peer_node(tsk);
2199 u32 oport = msg_origport(hdr);
2200 u32 onode = msg_orignode(hdr);
2201 int err = msg_errcode(hdr);
2202 unsigned long delay;
2203
2204 if (unlikely(msg_mcast(hdr)))
2205 return false;
2206 tsk->oneway = 0;
2207
2208 switch (sk->sk_state) {
2209 case TIPC_CONNECTING:
2210 /* Setup ACK */
2211 if (likely(con_msg)) {
2212 if (err)
2213 break;
2214 tipc_sk_finish_conn(tsk, oport, onode);
2215 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2216 /* ACK+ message with data is added to receive queue */
2217 if (msg_data_sz(hdr))
2218 return true;
2219 /* Empty ACK-, - wake up sleeping connect() and drop */
2220 sk->sk_state_change(sk);
2221 msg_set_dest_droppable(hdr, 1);
2222 return false;
2223 }
2224 /* Ignore connectionless message if not from listening socket */
2225 if (oport != pport || onode != pnode)
2226 return false;
2227
2228 /* Rejected SYN */
2229 if (err != TIPC_ERR_OVERLOAD)
2230 break;
2231
2232 /* Prepare for new setup attempt if we have a SYN clone */
2233 if (skb_queue_empty(&sk->sk_write_queue))
2234 break;
2235 get_random_bytes(&delay, 2);
2236 delay %= (tsk->conn_timeout / 4);
2237 delay = msecs_to_jiffies(delay + 100);
2238 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2239 return false;
2240 case TIPC_OPEN:
2241 case TIPC_DISCONNECTING:
2242 return false;
2243 case TIPC_LISTEN:
2244 /* Accept only SYN message */
2245 if (!msg_is_syn(hdr) &&
2246 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2247 return false;
2248 if (!con_msg && !err)
2249 return true;
2250 return false;
2251 case TIPC_ESTABLISHED:
2252 if (!skb_queue_empty(&sk->sk_write_queue))
2253 tipc_sk_push_backlog(tsk, false);
2254 /* Accept only connection-based messages sent by peer */
2255 if (likely(con_msg && !err && pport == oport &&
2256 pnode == onode)) {
2257 if (msg_ack_required(hdr)) {
2258 struct sk_buff *skb;
2259
2260 skb = tipc_sk_build_ack(tsk);
2261 if (skb) {
2262 msg_set_nagle_ack(buf_msg(skb));
2263 __skb_queue_tail(xmitq, skb);
2264 }
2265 }
2266 return true;
2267 }
2268 if (!tsk_peer_msg(tsk, hdr))
2269 return false;
2270 if (!err)
2271 return true;
2272 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2273 tipc_node_remove_conn(net, pnode, tsk->portid);
2274 sk->sk_state_change(sk);
2275 return true;
2276 default:
2277 pr_err("Unknown sk_state %u\n", sk->sk_state);
2278 }
2279 /* Abort connection setup attempt */
2280 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2281 sk->sk_err = ECONNREFUSED;
2282 sk->sk_state_change(sk);
2283 return true;
2284}
2285
2286/**
2287 * rcvbuf_limit - get proper overload limit of socket receive queue
2288 * @sk: socket
2289 * @skb: message
2290 *
2291 * For connection oriented messages, irrespective of importance,
2292 * default queue limit is 2 MB.
2293 *
2294 * For connectionless messages, queue limits are based on message
2295 * importance as follows:
2296 *
2297 * TIPC_LOW_IMPORTANCE (2 MB)
2298 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2299 * TIPC_HIGH_IMPORTANCE (8 MB)
2300 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2301 *
2302 * Return: overload limit according to corresponding message importance
2303 */
2304static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2305{
2306 struct tipc_sock *tsk = tipc_sk(sk);
2307 struct tipc_msg *hdr = buf_msg(skb);
2308
2309 if (unlikely(msg_in_group(hdr)))
2310 return READ_ONCE(sk->sk_rcvbuf);
2311
2312 if (unlikely(!msg_connected(hdr)))
2313 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2314
2315 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2316 return READ_ONCE(sk->sk_rcvbuf);
2317
2318 return FLOWCTL_MSG_LIM;
2319}
2320
2321/**
2322 * tipc_sk_filter_rcv - validate incoming message
2323 * @sk: socket
2324 * @skb: pointer to message.
2325 * @xmitq: output message area (FIXME)
2326 *
2327 * Enqueues message on receive queue if acceptable; optionally handles
2328 * disconnect indication for a connected socket.
2329 *
2330 * Called with socket lock already taken
2331 */
2332static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2333 struct sk_buff_head *xmitq)
2334{
2335 bool sk_conn = !tipc_sk_type_connectionless(sk);
2336 struct tipc_sock *tsk = tipc_sk(sk);
2337 struct tipc_group *grp = tsk->group;
2338 struct tipc_msg *hdr = buf_msg(skb);
2339 struct net *net = sock_net(sk);
2340 struct sk_buff_head inputq;
2341 int mtyp = msg_type(hdr);
2342 int limit, err = TIPC_OK;
2343
2344 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2345 TIPC_SKB_CB(skb)->bytes_read = 0;
2346 __skb_queue_head_init(&inputq);
2347 __skb_queue_tail(&inputq, skb);
2348
2349 if (unlikely(!msg_isdata(hdr)))
2350 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2351
2352 if (unlikely(grp))
2353 tipc_group_filter_msg(grp, &inputq, xmitq);
2354
2355 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2356 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2357
2358 /* Validate and add to receive buffer if there is space */
2359 while ((skb = __skb_dequeue(&inputq))) {
2360 hdr = buf_msg(skb);
2361 limit = rcvbuf_limit(sk, skb);
2362 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2363 (!sk_conn && msg_connected(hdr)) ||
2364 (!grp && msg_in_group(hdr)))
2365 err = TIPC_ERR_NO_PORT;
2366 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2367 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2368 "err_overload2!");
2369 atomic_inc(&sk->sk_drops);
2370 err = TIPC_ERR_OVERLOAD;
2371 }
2372
2373 if (unlikely(err)) {
2374 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2375 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2376 "@filter_rcv!");
2377 __skb_queue_tail(xmitq, skb);
2378 }
2379 err = TIPC_OK;
2380 continue;
2381 }
2382 __skb_queue_tail(&sk->sk_receive_queue, skb);
2383 skb_set_owner_r(skb, sk);
2384 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2385 "rcvq >90% allocated!");
2386 sk->sk_data_ready(sk);
2387 }
2388}
2389
2390/**
2391 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2392 * @sk: socket
2393 * @skb: message
2394 *
2395 * Caller must hold socket lock
2396 */
2397static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2398{
2399 unsigned int before = sk_rmem_alloc_get(sk);
2400 struct sk_buff_head xmitq;
2401 unsigned int added;
2402
2403 __skb_queue_head_init(&xmitq);
2404
2405 tipc_sk_filter_rcv(sk, skb, &xmitq);
2406 added = sk_rmem_alloc_get(sk) - before;
2407 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2408
2409 /* Send pending response/rejected messages, if any */
2410 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2411 return 0;
2412}
2413
2414/**
2415 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2416 * inputq and try adding them to socket or backlog queue
2417 * @inputq: list of incoming buffers with potentially different destinations
2418 * @sk: socket where the buffers should be enqueued
2419 * @dport: port number for the socket
2420 * @xmitq: output queue
2421 *
2422 * Caller must hold socket lock
2423 */
2424static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2425 u32 dport, struct sk_buff_head *xmitq)
2426{
2427 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2428 struct sk_buff *skb;
2429 unsigned int lim;
2430 atomic_t *dcnt;
2431 u32 onode;
2432
2433 while (skb_queue_len(inputq)) {
2434 if (unlikely(time_after_eq(jiffies, time_limit)))
2435 return;
2436
2437 skb = tipc_skb_dequeue(inputq, dport);
2438 if (unlikely(!skb))
2439 return;
2440
2441 /* Add message directly to receive queue if possible */
2442 if (!sock_owned_by_user(sk)) {
2443 tipc_sk_filter_rcv(sk, skb, xmitq);
2444 continue;
2445 }
2446
2447 /* Try backlog, compensating for double-counted bytes */
2448 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2449 if (!sk->sk_backlog.len)
2450 atomic_set(dcnt, 0);
2451 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2452 if (likely(!sk_add_backlog(sk, skb, lim))) {
2453 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2454 "bklg & rcvq >90% allocated!");
2455 continue;
2456 }
2457
2458 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2459 /* Overload => reject message back to sender */
2460 onode = tipc_own_addr(sock_net(sk));
2461 atomic_inc(&sk->sk_drops);
2462 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2463 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2464 "@sk_enqueue!");
2465 __skb_queue_tail(xmitq, skb);
2466 }
2467 break;
2468 }
2469}
2470
2471/**
2472 * tipc_sk_rcv - handle a chain of incoming buffers
2473 * @net: the associated network namespace
2474 * @inputq: buffer list containing the buffers
2475 * Consumes all buffers in list until inputq is empty
2476 * Note: may be called in multiple threads referring to the same queue
2477 */
2478void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2479{
2480 struct sk_buff_head xmitq;
2481 u32 dnode, dport = 0;
2482 int err;
2483 struct tipc_sock *tsk;
2484 struct sock *sk;
2485 struct sk_buff *skb;
2486
2487 __skb_queue_head_init(&xmitq);
2488 while (skb_queue_len(inputq)) {
2489 dport = tipc_skb_peek_port(inputq, dport);
2490 tsk = tipc_sk_lookup(net, dport);
2491
2492 if (likely(tsk)) {
2493 sk = &tsk->sk;
2494 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2495 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2496 spin_unlock_bh(&sk->sk_lock.slock);
2497 }
2498 /* Send pending response/rejected messages, if any */
2499 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2500 sock_put(sk);
2501 continue;
2502 }
2503 /* No destination socket => dequeue skb if still there */
2504 skb = tipc_skb_dequeue(inputq, dport);
2505 if (!skb)
2506 return;
2507
2508 /* Try secondary lookup if unresolved named message */
2509 err = TIPC_ERR_NO_PORT;
2510 if (tipc_msg_lookup_dest(net, skb, &err))
2511 goto xmit;
2512
2513 /* Prepare for message rejection */
2514 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2515 continue;
2516
2517 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2518xmit:
2519 dnode = msg_destnode(buf_msg(skb));
2520 tipc_node_xmit_skb(net, skb, dnode, dport);
2521 }
2522}
2523
2524static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2525{
2526 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2527 struct sock *sk = sock->sk;
2528 int done;
2529
2530 do {
2531 int err = sock_error(sk);
2532 if (err)
2533 return err;
2534 if (!*timeo_p)
2535 return -ETIMEDOUT;
2536 if (signal_pending(current))
2537 return sock_intr_errno(*timeo_p);
2538 if (sk->sk_state == TIPC_DISCONNECTING)
2539 break;
2540
2541 add_wait_queue(sk_sleep(sk), &wait);
2542 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2543 &wait);
2544 remove_wait_queue(sk_sleep(sk), &wait);
2545 } while (!done);
2546 return 0;
2547}
2548
2549static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2550{
2551 if (addr->family != AF_TIPC)
2552 return false;
2553 if (addr->addrtype == TIPC_SERVICE_RANGE)
2554 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2555 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2556 addr->addrtype == TIPC_SOCKET_ADDR);
2557}
2558
2559/**
2560 * tipc_connect - establish a connection to another TIPC port
2561 * @sock: socket structure
2562 * @dest: socket address for destination port
2563 * @destlen: size of socket address data structure
2564 * @flags: file-related flags associated with socket
2565 *
2566 * Return: 0 on success, errno otherwise
2567 */
2568static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2569 int destlen, int flags)
2570{
2571 struct sock *sk = sock->sk;
2572 struct tipc_sock *tsk = tipc_sk(sk);
2573 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2574 struct msghdr m = {NULL,};
2575 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2576 int previous;
2577 int res = 0;
2578
2579 if (destlen != sizeof(struct sockaddr_tipc))
2580 return -EINVAL;
2581
2582 lock_sock(sk);
2583
2584 if (tsk->group) {
2585 res = -EINVAL;
2586 goto exit;
2587 }
2588
2589 if (dst->family == AF_UNSPEC) {
2590 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2591 if (!tipc_sk_type_connectionless(sk))
2592 res = -EINVAL;
2593 goto exit;
2594 }
2595 if (!tipc_sockaddr_is_sane(dst)) {
2596 res = -EINVAL;
2597 goto exit;
2598 }
2599 /* DGRAM/RDM connect(), just save the destaddr */
2600 if (tipc_sk_type_connectionless(sk)) {
2601 memcpy(&tsk->peer, dest, destlen);
2602 goto exit;
2603 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2604 res = -EINVAL;
2605 goto exit;
2606 }
2607
2608 previous = sk->sk_state;
2609
2610 switch (sk->sk_state) {
2611 case TIPC_OPEN:
2612 /* Send a 'SYN-' to destination */
2613 m.msg_name = dest;
2614 m.msg_namelen = destlen;
2615 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2616
2617 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2618 * indicate send_msg() is never blocked.
2619 */
2620 if (!timeout)
2621 m.msg_flags = MSG_DONTWAIT;
2622
2623 res = __tipc_sendmsg(sock, &m, 0);
2624 if ((res < 0) && (res != -EWOULDBLOCK))
2625 goto exit;
2626
2627 /* Just entered TIPC_CONNECTING state; the only
2628 * difference is that return value in non-blocking
2629 * case is EINPROGRESS, rather than EALREADY.
2630 */
2631 res = -EINPROGRESS;
2632 fallthrough;
2633 case TIPC_CONNECTING:
2634 if (!timeout) {
2635 if (previous == TIPC_CONNECTING)
2636 res = -EALREADY;
2637 goto exit;
2638 }
2639 timeout = msecs_to_jiffies(timeout);
2640 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2641 res = tipc_wait_for_connect(sock, &timeout);
2642 break;
2643 case TIPC_ESTABLISHED:
2644 res = -EISCONN;
2645 break;
2646 default:
2647 res = -EINVAL;
2648 }
2649
2650exit:
2651 release_sock(sk);
2652 return res;
2653}
2654
2655/**
2656 * tipc_listen - allow socket to listen for incoming connections
2657 * @sock: socket structure
2658 * @len: (unused)
2659 *
2660 * Return: 0 on success, errno otherwise
2661 */
2662static int tipc_listen(struct socket *sock, int len)
2663{
2664 struct sock *sk = sock->sk;
2665 int res;
2666
2667 lock_sock(sk);
2668 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2669 release_sock(sk);
2670
2671 return res;
2672}
2673
2674static int tipc_wait_for_accept(struct socket *sock, long timeo)
2675{
2676 struct sock *sk = sock->sk;
2677 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2678 int err;
2679
2680 /* True wake-one mechanism for incoming connections: only
2681 * one process gets woken up, not the 'whole herd'.
2682 * Since we do not 'race & poll' for established sockets
2683 * anymore, the common case will execute the loop only once.
2684 */
2685 for (;;) {
2686 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2687 add_wait_queue(sk_sleep(sk), &wait);
2688 release_sock(sk);
2689 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2690 lock_sock(sk);
2691 remove_wait_queue(sk_sleep(sk), &wait);
2692 }
2693 err = 0;
2694 if (!skb_queue_empty(&sk->sk_receive_queue))
2695 break;
2696 err = -EAGAIN;
2697 if (!timeo)
2698 break;
2699 err = sock_intr_errno(timeo);
2700 if (signal_pending(current))
2701 break;
2702 }
2703 return err;
2704}
2705
2706/**
2707 * tipc_accept - wait for connection request
2708 * @sock: listening socket
2709 * @new_sock: new socket that is to be connected
2710 * @arg: arguments for accept
2711 *
2712 * Return: 0 on success, errno otherwise
2713 */
2714static int tipc_accept(struct socket *sock, struct socket *new_sock,
2715 struct proto_accept_arg *arg)
2716{
2717 struct sock *new_sk, *sk = sock->sk;
2718 struct tipc_sock *new_tsock;
2719 struct msghdr m = {NULL,};
2720 struct tipc_msg *msg;
2721 struct sk_buff *buf;
2722 long timeo;
2723 int res;
2724
2725 lock_sock(sk);
2726
2727 if (sk->sk_state != TIPC_LISTEN) {
2728 res = -EINVAL;
2729 goto exit;
2730 }
2731 timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK);
2732 res = tipc_wait_for_accept(sock, timeo);
2733 if (res)
2734 goto exit;
2735
2736 buf = skb_peek(&sk->sk_receive_queue);
2737
2738 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, arg->kern);
2739 if (res)
2740 goto exit;
2741 security_sk_clone(sock->sk, new_sock->sk);
2742
2743 new_sk = new_sock->sk;
2744 new_tsock = tipc_sk(new_sk);
2745 msg = buf_msg(buf);
2746
2747 /* we lock on new_sk; but lockdep sees the lock on sk */
2748 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2749
2750 /*
2751 * Reject any stray messages received by new socket
2752 * before the socket lock was taken (very, very unlikely)
2753 */
2754 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2755
2756 /* Connect new socket to it's peer */
2757 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2758
2759 tsk_set_importance(new_sk, msg_importance(msg));
2760 if (msg_named(msg)) {
2761 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2762 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2763 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2764 }
2765
2766 /*
2767 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2768 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2769 */
2770 if (!msg_data_sz(msg)) {
2771 tsk_advance_rx_queue(sk);
2772 } else {
2773 __skb_dequeue(&sk->sk_receive_queue);
2774 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2775 skb_set_owner_r(buf, new_sk);
2776 }
2777 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2778 __tipc_sendstream(new_sock, &m, 0);
2779 release_sock(new_sk);
2780exit:
2781 release_sock(sk);
2782 return res;
2783}
2784
2785/**
2786 * tipc_shutdown - shutdown socket connection
2787 * @sock: socket structure
2788 * @how: direction to close (must be SHUT_RDWR)
2789 *
2790 * Terminates connection (if necessary), then purges socket's receive queue.
2791 *
2792 * Return: 0 on success, errno otherwise
2793 */
2794static int tipc_shutdown(struct socket *sock, int how)
2795{
2796 struct sock *sk = sock->sk;
2797 int res;
2798
2799 if (how != SHUT_RDWR)
2800 return -EINVAL;
2801
2802 lock_sock(sk);
2803
2804 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2805 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2806 sk->sk_shutdown = SHUTDOWN_MASK;
2807
2808 if (sk->sk_state == TIPC_DISCONNECTING) {
2809 /* Discard any unreceived messages */
2810 __skb_queue_purge(&sk->sk_receive_queue);
2811
2812 res = 0;
2813 } else {
2814 res = -ENOTCONN;
2815 }
2816 /* Wake up anyone sleeping in poll. */
2817 sk->sk_state_change(sk);
2818
2819 release_sock(sk);
2820 return res;
2821}
2822
2823static void tipc_sk_check_probing_state(struct sock *sk,
2824 struct sk_buff_head *list)
2825{
2826 struct tipc_sock *tsk = tipc_sk(sk);
2827 u32 pnode = tsk_peer_node(tsk);
2828 u32 pport = tsk_peer_port(tsk);
2829 u32 self = tsk_own_node(tsk);
2830 u32 oport = tsk->portid;
2831 struct sk_buff *skb;
2832
2833 if (tsk->probe_unacked) {
2834 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2835 sk->sk_err = ECONNABORTED;
2836 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2837 sk->sk_state_change(sk);
2838 return;
2839 }
2840 /* Prepare new probe */
2841 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2842 pnode, self, pport, oport, TIPC_OK);
2843 if (skb)
2844 __skb_queue_tail(list, skb);
2845 tsk->probe_unacked = true;
2846 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2847}
2848
2849static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2850{
2851 struct tipc_sock *tsk = tipc_sk(sk);
2852
2853 /* Try again later if dest link is congested */
2854 if (tsk->cong_link_cnt) {
2855 sk_reset_timer(sk, &sk->sk_timer,
2856 jiffies + msecs_to_jiffies(100));
2857 return;
2858 }
2859 /* Prepare SYN for retransmit */
2860 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2861}
2862
2863static void tipc_sk_timeout(struct timer_list *t)
2864{
2865 struct sock *sk = from_timer(sk, t, sk_timer);
2866 struct tipc_sock *tsk = tipc_sk(sk);
2867 u32 pnode = tsk_peer_node(tsk);
2868 struct sk_buff_head list;
2869 int rc = 0;
2870
2871 __skb_queue_head_init(&list);
2872 bh_lock_sock(sk);
2873
2874 /* Try again later if socket is busy */
2875 if (sock_owned_by_user(sk)) {
2876 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2877 bh_unlock_sock(sk);
2878 sock_put(sk);
2879 return;
2880 }
2881
2882 if (sk->sk_state == TIPC_ESTABLISHED)
2883 tipc_sk_check_probing_state(sk, &list);
2884 else if (sk->sk_state == TIPC_CONNECTING)
2885 tipc_sk_retry_connect(sk, &list);
2886
2887 bh_unlock_sock(sk);
2888
2889 if (!skb_queue_empty(&list))
2890 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2891
2892 /* SYN messages may cause link congestion */
2893 if (rc == -ELINKCONG) {
2894 tipc_dest_push(&tsk->cong_links, pnode, 0);
2895 tsk->cong_link_cnt = 1;
2896 }
2897 sock_put(sk);
2898}
2899
2900static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2901{
2902 struct sock *sk = &tsk->sk;
2903 struct net *net = sock_net(sk);
2904 struct tipc_socket_addr skaddr;
2905 struct publication *p;
2906 u32 key;
2907
2908 if (tipc_sk_connected(sk))
2909 return -EINVAL;
2910 key = tsk->portid + tsk->pub_count + 1;
2911 if (key == tsk->portid)
2912 return -EADDRINUSE;
2913 skaddr.ref = tsk->portid;
2914 skaddr.node = tipc_own_addr(net);
2915 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2916 if (unlikely(!p))
2917 return -EINVAL;
2918
2919 list_add(&p->binding_sock, &tsk->publications);
2920 tsk->pub_count++;
2921 tsk->published = true;
2922 return 0;
2923}
2924
2925static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2926{
2927 struct net *net = sock_net(&tsk->sk);
2928 struct publication *safe, *p;
2929 struct tipc_uaddr _ua;
2930 int rc = -EINVAL;
2931
2932 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2933 if (!ua) {
2934 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2935 p->sr.type, p->sr.lower, p->sr.upper);
2936 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2937 continue;
2938 }
2939 /* Unbind specific publication */
2940 if (p->scope != ua->scope)
2941 continue;
2942 if (p->sr.type != ua->sr.type)
2943 continue;
2944 if (p->sr.lower != ua->sr.lower)
2945 continue;
2946 if (p->sr.upper != ua->sr.upper)
2947 break;
2948 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2949 rc = 0;
2950 break;
2951 }
2952 if (list_empty(&tsk->publications)) {
2953 tsk->published = 0;
2954 rc = 0;
2955 }
2956 return rc;
2957}
2958
2959/* tipc_sk_reinit: set non-zero address in all existing sockets
2960 * when we go from standalone to network mode.
2961 */
2962void tipc_sk_reinit(struct net *net)
2963{
2964 struct tipc_net *tn = net_generic(net, tipc_net_id);
2965 struct rhashtable_iter iter;
2966 struct tipc_sock *tsk;
2967 struct tipc_msg *msg;
2968
2969 rhashtable_walk_enter(&tn->sk_rht, &iter);
2970
2971 do {
2972 rhashtable_walk_start(&iter);
2973
2974 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2975 sock_hold(&tsk->sk);
2976 rhashtable_walk_stop(&iter);
2977 lock_sock(&tsk->sk);
2978 msg = &tsk->phdr;
2979 msg_set_prevnode(msg, tipc_own_addr(net));
2980 msg_set_orignode(msg, tipc_own_addr(net));
2981 release_sock(&tsk->sk);
2982 rhashtable_walk_start(&iter);
2983 sock_put(&tsk->sk);
2984 }
2985
2986 rhashtable_walk_stop(&iter);
2987 } while (tsk == ERR_PTR(-EAGAIN));
2988
2989 rhashtable_walk_exit(&iter);
2990}
2991
2992static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2993{
2994 struct tipc_net *tn = net_generic(net, tipc_net_id);
2995 struct tipc_sock *tsk;
2996
2997 rcu_read_lock();
2998 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2999 if (tsk)
3000 sock_hold(&tsk->sk);
3001 rcu_read_unlock();
3002
3003 return tsk;
3004}
3005
3006static int tipc_sk_insert(struct tipc_sock *tsk)
3007{
3008 struct sock *sk = &tsk->sk;
3009 struct net *net = sock_net(sk);
3010 struct tipc_net *tn = net_generic(net, tipc_net_id);
3011 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3012 u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT;
3013
3014 while (remaining--) {
3015 portid++;
3016 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3017 portid = TIPC_MIN_PORT;
3018 tsk->portid = portid;
3019 sock_hold(&tsk->sk);
3020 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3021 tsk_rht_params))
3022 return 0;
3023 sock_put(&tsk->sk);
3024 }
3025
3026 return -1;
3027}
3028
3029static void tipc_sk_remove(struct tipc_sock *tsk)
3030{
3031 struct sock *sk = &tsk->sk;
3032 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3033
3034 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3035 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3036 __sock_put(sk);
3037 }
3038}
3039
3040static const struct rhashtable_params tsk_rht_params = {
3041 .nelem_hint = 192,
3042 .head_offset = offsetof(struct tipc_sock, node),
3043 .key_offset = offsetof(struct tipc_sock, portid),
3044 .key_len = sizeof(u32), /* portid */
3045 .max_size = 1048576,
3046 .min_size = 256,
3047 .automatic_shrinking = true,
3048};
3049
3050int tipc_sk_rht_init(struct net *net)
3051{
3052 struct tipc_net *tn = net_generic(net, tipc_net_id);
3053
3054 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3055}
3056
3057void tipc_sk_rht_destroy(struct net *net)
3058{
3059 struct tipc_net *tn = net_generic(net, tipc_net_id);
3060
3061 /* Wait for socket readers to complete */
3062 synchronize_net();
3063
3064 rhashtable_destroy(&tn->sk_rht);
3065}
3066
3067static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3068{
3069 struct net *net = sock_net(&tsk->sk);
3070 struct tipc_group *grp = tsk->group;
3071 struct tipc_msg *hdr = &tsk->phdr;
3072 struct tipc_uaddr ua;
3073 int rc;
3074
3075 if (mreq->type < TIPC_RESERVED_TYPES)
3076 return -EACCES;
3077 if (mreq->scope > TIPC_NODE_SCOPE)
3078 return -EINVAL;
3079 if (mreq->scope != TIPC_NODE_SCOPE)
3080 mreq->scope = TIPC_CLUSTER_SCOPE;
3081 if (grp)
3082 return -EACCES;
3083 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3084 if (!grp)
3085 return -ENOMEM;
3086 tsk->group = grp;
3087 msg_set_lookup_scope(hdr, mreq->scope);
3088 msg_set_nametype(hdr, mreq->type);
3089 msg_set_dest_droppable(hdr, true);
3090 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3091 mreq->type, mreq->instance, mreq->instance);
3092 tipc_nametbl_build_group(net, grp, &ua);
3093 rc = tipc_sk_publish(tsk, &ua);
3094 if (rc) {
3095 tipc_group_delete(net, grp);
3096 tsk->group = NULL;
3097 return rc;
3098 }
3099 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3100 tsk->mc_method.rcast = true;
3101 tsk->mc_method.mandatory = true;
3102 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3103 return rc;
3104}
3105
3106static int tipc_sk_leave(struct tipc_sock *tsk)
3107{
3108 struct net *net = sock_net(&tsk->sk);
3109 struct tipc_group *grp = tsk->group;
3110 struct tipc_uaddr ua;
3111 int scope;
3112
3113 if (!grp)
3114 return -EINVAL;
3115 ua.addrtype = TIPC_SERVICE_RANGE;
3116 tipc_group_self(grp, &ua.sr, &scope);
3117 ua.scope = scope;
3118 tipc_group_delete(net, grp);
3119 tsk->group = NULL;
3120 tipc_sk_withdraw(tsk, &ua);
3121 return 0;
3122}
3123
3124/**
3125 * tipc_setsockopt - set socket option
3126 * @sock: socket structure
3127 * @lvl: option level
3128 * @opt: option identifier
3129 * @ov: pointer to new option value
3130 * @ol: length of option value
3131 *
3132 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3133 * (to ease compatibility).
3134 *
3135 * Return: 0 on success, errno otherwise
3136 */
3137static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3138 sockptr_t ov, unsigned int ol)
3139{
3140 struct sock *sk = sock->sk;
3141 struct tipc_sock *tsk = tipc_sk(sk);
3142 struct tipc_group_req mreq;
3143 u32 value = 0;
3144 int res = 0;
3145
3146 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3147 return 0;
3148 if (lvl != SOL_TIPC)
3149 return -ENOPROTOOPT;
3150
3151 switch (opt) {
3152 case TIPC_IMPORTANCE:
3153 case TIPC_SRC_DROPPABLE:
3154 case TIPC_DEST_DROPPABLE:
3155 case TIPC_CONN_TIMEOUT:
3156 case TIPC_NODELAY:
3157 if (ol < sizeof(value))
3158 return -EINVAL;
3159 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3160 return -EFAULT;
3161 break;
3162 case TIPC_GROUP_JOIN:
3163 if (ol < sizeof(mreq))
3164 return -EINVAL;
3165 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3166 return -EFAULT;
3167 break;
3168 default:
3169 if (!sockptr_is_null(ov) || ol)
3170 return -EINVAL;
3171 }
3172
3173 lock_sock(sk);
3174
3175 switch (opt) {
3176 case TIPC_IMPORTANCE:
3177 res = tsk_set_importance(sk, value);
3178 break;
3179 case TIPC_SRC_DROPPABLE:
3180 if (sock->type != SOCK_STREAM)
3181 tsk_set_unreliable(tsk, value);
3182 else
3183 res = -ENOPROTOOPT;
3184 break;
3185 case TIPC_DEST_DROPPABLE:
3186 tsk_set_unreturnable(tsk, value);
3187 break;
3188 case TIPC_CONN_TIMEOUT:
3189 tipc_sk(sk)->conn_timeout = value;
3190 break;
3191 case TIPC_MCAST_BROADCAST:
3192 tsk->mc_method.rcast = false;
3193 tsk->mc_method.mandatory = true;
3194 break;
3195 case TIPC_MCAST_REPLICAST:
3196 tsk->mc_method.rcast = true;
3197 tsk->mc_method.mandatory = true;
3198 break;
3199 case TIPC_GROUP_JOIN:
3200 res = tipc_sk_join(tsk, &mreq);
3201 break;
3202 case TIPC_GROUP_LEAVE:
3203 res = tipc_sk_leave(tsk);
3204 break;
3205 case TIPC_NODELAY:
3206 tsk->nodelay = !!value;
3207 tsk_set_nagle(tsk);
3208 break;
3209 default:
3210 res = -EINVAL;
3211 }
3212
3213 release_sock(sk);
3214
3215 return res;
3216}
3217
3218/**
3219 * tipc_getsockopt - get socket option
3220 * @sock: socket structure
3221 * @lvl: option level
3222 * @opt: option identifier
3223 * @ov: receptacle for option value
3224 * @ol: receptacle for length of option value
3225 *
3226 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3227 * (to ease compatibility).
3228 *
3229 * Return: 0 on success, errno otherwise
3230 */
3231static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3232 char __user *ov, int __user *ol)
3233{
3234 struct sock *sk = sock->sk;
3235 struct tipc_sock *tsk = tipc_sk(sk);
3236 struct tipc_service_range seq;
3237 int len, scope;
3238 u32 value;
3239 int res;
3240
3241 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3242 return put_user(0, ol);
3243 if (lvl != SOL_TIPC)
3244 return -ENOPROTOOPT;
3245 res = get_user(len, ol);
3246 if (res)
3247 return res;
3248
3249 lock_sock(sk);
3250
3251 switch (opt) {
3252 case TIPC_IMPORTANCE:
3253 value = tsk_importance(tsk);
3254 break;
3255 case TIPC_SRC_DROPPABLE:
3256 value = tsk_unreliable(tsk);
3257 break;
3258 case TIPC_DEST_DROPPABLE:
3259 value = tsk_unreturnable(tsk);
3260 break;
3261 case TIPC_CONN_TIMEOUT:
3262 value = tsk->conn_timeout;
3263 /* no need to set "res", since already 0 at this point */
3264 break;
3265 case TIPC_NODE_RECVQ_DEPTH:
3266 value = 0; /* was tipc_queue_size, now obsolete */
3267 break;
3268 case TIPC_SOCK_RECVQ_DEPTH:
3269 value = skb_queue_len(&sk->sk_receive_queue);
3270 break;
3271 case TIPC_SOCK_RECVQ_USED:
3272 value = sk_rmem_alloc_get(sk);
3273 break;
3274 case TIPC_GROUP_JOIN:
3275 seq.type = 0;
3276 if (tsk->group)
3277 tipc_group_self(tsk->group, &seq, &scope);
3278 value = seq.type;
3279 break;
3280 default:
3281 res = -EINVAL;
3282 }
3283
3284 release_sock(sk);
3285
3286 if (res)
3287 return res; /* "get" failed */
3288
3289 if (len < sizeof(value))
3290 return -EINVAL;
3291
3292 if (copy_to_user(ov, &value, sizeof(value)))
3293 return -EFAULT;
3294
3295 return put_user(sizeof(value), ol);
3296}
3297
3298static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3299{
3300 struct net *net = sock_net(sock->sk);
3301 struct tipc_sioc_nodeid_req nr = {0};
3302 struct tipc_sioc_ln_req lnr;
3303 void __user *argp = (void __user *)arg;
3304
3305 switch (cmd) {
3306 case SIOCGETLINKNAME:
3307 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3308 return -EFAULT;
3309 if (!tipc_node_get_linkname(net,
3310 lnr.bearer_id & 0xffff, lnr.peer,
3311 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3312 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3313 return -EFAULT;
3314 return 0;
3315 }
3316 return -EADDRNOTAVAIL;
3317 case SIOCGETNODEID:
3318 if (copy_from_user(&nr, argp, sizeof(nr)))
3319 return -EFAULT;
3320 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3321 return -EADDRNOTAVAIL;
3322 if (copy_to_user(argp, &nr, sizeof(nr)))
3323 return -EFAULT;
3324 return 0;
3325 default:
3326 return -ENOIOCTLCMD;
3327 }
3328}
3329
3330static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3331{
3332 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3333 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3334 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3335
3336 tsk1->peer.family = AF_TIPC;
3337 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3338 tsk1->peer.scope = TIPC_NODE_SCOPE;
3339 tsk1->peer.addr.id.ref = tsk2->portid;
3340 tsk1->peer.addr.id.node = onode;
3341 tsk2->peer.family = AF_TIPC;
3342 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3343 tsk2->peer.scope = TIPC_NODE_SCOPE;
3344 tsk2->peer.addr.id.ref = tsk1->portid;
3345 tsk2->peer.addr.id.node = onode;
3346
3347 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3348 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3349 return 0;
3350}
3351
3352/* Protocol switches for the various types of TIPC sockets */
3353
3354static const struct proto_ops msg_ops = {
3355 .owner = THIS_MODULE,
3356 .family = AF_TIPC,
3357 .release = tipc_release,
3358 .bind = tipc_bind,
3359 .connect = tipc_connect,
3360 .socketpair = tipc_socketpair,
3361 .accept = sock_no_accept,
3362 .getname = tipc_getname,
3363 .poll = tipc_poll,
3364 .ioctl = tipc_ioctl,
3365 .listen = sock_no_listen,
3366 .shutdown = tipc_shutdown,
3367 .setsockopt = tipc_setsockopt,
3368 .getsockopt = tipc_getsockopt,
3369 .sendmsg = tipc_sendmsg,
3370 .recvmsg = tipc_recvmsg,
3371 .mmap = sock_no_mmap,
3372};
3373
3374static const struct proto_ops packet_ops = {
3375 .owner = THIS_MODULE,
3376 .family = AF_TIPC,
3377 .release = tipc_release,
3378 .bind = tipc_bind,
3379 .connect = tipc_connect,
3380 .socketpair = tipc_socketpair,
3381 .accept = tipc_accept,
3382 .getname = tipc_getname,
3383 .poll = tipc_poll,
3384 .ioctl = tipc_ioctl,
3385 .listen = tipc_listen,
3386 .shutdown = tipc_shutdown,
3387 .setsockopt = tipc_setsockopt,
3388 .getsockopt = tipc_getsockopt,
3389 .sendmsg = tipc_send_packet,
3390 .recvmsg = tipc_recvmsg,
3391 .mmap = sock_no_mmap,
3392};
3393
3394static const struct proto_ops stream_ops = {
3395 .owner = THIS_MODULE,
3396 .family = AF_TIPC,
3397 .release = tipc_release,
3398 .bind = tipc_bind,
3399 .connect = tipc_connect,
3400 .socketpair = tipc_socketpair,
3401 .accept = tipc_accept,
3402 .getname = tipc_getname,
3403 .poll = tipc_poll,
3404 .ioctl = tipc_ioctl,
3405 .listen = tipc_listen,
3406 .shutdown = tipc_shutdown,
3407 .setsockopt = tipc_setsockopt,
3408 .getsockopt = tipc_getsockopt,
3409 .sendmsg = tipc_sendstream,
3410 .recvmsg = tipc_recvstream,
3411 .mmap = sock_no_mmap,
3412};
3413
3414static const struct net_proto_family tipc_family_ops = {
3415 .owner = THIS_MODULE,
3416 .family = AF_TIPC,
3417 .create = tipc_sk_create
3418};
3419
3420static struct proto tipc_proto = {
3421 .name = "TIPC",
3422 .owner = THIS_MODULE,
3423 .obj_size = sizeof(struct tipc_sock),
3424 .sysctl_rmem = sysctl_tipc_rmem
3425};
3426
3427/**
3428 * tipc_socket_init - initialize TIPC socket interface
3429 *
3430 * Return: 0 on success, errno otherwise
3431 */
3432int tipc_socket_init(void)
3433{
3434 int res;
3435
3436 res = proto_register(&tipc_proto, 1);
3437 if (res) {
3438 pr_err("Failed to register TIPC protocol type\n");
3439 goto out;
3440 }
3441
3442 res = sock_register(&tipc_family_ops);
3443 if (res) {
3444 pr_err("Failed to register TIPC socket type\n");
3445 proto_unregister(&tipc_proto);
3446 goto out;
3447 }
3448 out:
3449 return res;
3450}
3451
3452/**
3453 * tipc_socket_stop - stop TIPC socket interface
3454 */
3455void tipc_socket_stop(void)
3456{
3457 sock_unregister(tipc_family_ops.family);
3458 proto_unregister(&tipc_proto);
3459}
3460
3461/* Caller should hold socket lock for the passed tipc socket. */
3462static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3463{
3464 u32 peer_node, peer_port;
3465 u32 conn_type, conn_instance;
3466 struct nlattr *nest;
3467
3468 peer_node = tsk_peer_node(tsk);
3469 peer_port = tsk_peer_port(tsk);
3470 conn_type = msg_nametype(&tsk->phdr);
3471 conn_instance = msg_nameinst(&tsk->phdr);
3472 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3473 if (!nest)
3474 return -EMSGSIZE;
3475
3476 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3477 goto msg_full;
3478 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3479 goto msg_full;
3480
3481 if (tsk->conn_addrtype != 0) {
3482 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3483 goto msg_full;
3484 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3485 goto msg_full;
3486 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3487 goto msg_full;
3488 }
3489 nla_nest_end(skb, nest);
3490
3491 return 0;
3492
3493msg_full:
3494 nla_nest_cancel(skb, nest);
3495
3496 return -EMSGSIZE;
3497}
3498
3499static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3500 *tsk)
3501{
3502 struct net *net = sock_net(skb->sk);
3503 struct sock *sk = &tsk->sk;
3504
3505 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3506 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3507 return -EMSGSIZE;
3508
3509 if (tipc_sk_connected(sk)) {
3510 if (__tipc_nl_add_sk_con(skb, tsk))
3511 return -EMSGSIZE;
3512 } else if (!list_empty(&tsk->publications)) {
3513 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3514 return -EMSGSIZE;
3515 }
3516 return 0;
3517}
3518
3519/* Caller should hold socket lock for the passed tipc socket. */
3520static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3521 struct tipc_sock *tsk)
3522{
3523 struct nlattr *attrs;
3524 void *hdr;
3525
3526 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3527 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3528 if (!hdr)
3529 goto msg_cancel;
3530
3531 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3532 if (!attrs)
3533 goto genlmsg_cancel;
3534
3535 if (__tipc_nl_add_sk_info(skb, tsk))
3536 goto attr_msg_cancel;
3537
3538 nla_nest_end(skb, attrs);
3539 genlmsg_end(skb, hdr);
3540
3541 return 0;
3542
3543attr_msg_cancel:
3544 nla_nest_cancel(skb, attrs);
3545genlmsg_cancel:
3546 genlmsg_cancel(skb, hdr);
3547msg_cancel:
3548 return -EMSGSIZE;
3549}
3550
3551int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3552 int (*skb_handler)(struct sk_buff *skb,
3553 struct netlink_callback *cb,
3554 struct tipc_sock *tsk))
3555{
3556 struct rhashtable_iter *iter = (void *)cb->args[4];
3557 struct tipc_sock *tsk;
3558 int err;
3559
3560 rhashtable_walk_start(iter);
3561 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3562 if (IS_ERR(tsk)) {
3563 if (PTR_ERR(tsk) == -EAGAIN)
3564 continue;
3565 break;
3566 }
3567
3568 sock_hold(&tsk->sk);
3569 rhashtable_walk_stop(iter);
3570 lock_sock(&tsk->sk);
3571 err = skb_handler(skb, cb, tsk);
3572 if (err) {
3573 release_sock(&tsk->sk);
3574 sock_put(&tsk->sk);
3575 goto out;
3576 }
3577 release_sock(&tsk->sk);
3578 rhashtable_walk_start(iter);
3579 sock_put(&tsk->sk);
3580 }
3581 rhashtable_walk_stop(iter);
3582out:
3583 return skb->len;
3584}
3585EXPORT_SYMBOL(tipc_nl_sk_walk);
3586
3587int tipc_dump_start(struct netlink_callback *cb)
3588{
3589 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3590}
3591EXPORT_SYMBOL(tipc_dump_start);
3592
3593int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3594{
3595 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3596 struct rhashtable_iter *iter = (void *)cb->args[4];
3597 struct tipc_net *tn = tipc_net(net);
3598
3599 if (!iter) {
3600 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3601 if (!iter)
3602 return -ENOMEM;
3603
3604 cb->args[4] = (long)iter;
3605 }
3606
3607 rhashtable_walk_enter(&tn->sk_rht, iter);
3608 return 0;
3609}
3610
3611int tipc_dump_done(struct netlink_callback *cb)
3612{
3613 struct rhashtable_iter *hti = (void *)cb->args[4];
3614
3615 rhashtable_walk_exit(hti);
3616 kfree(hti);
3617 return 0;
3618}
3619EXPORT_SYMBOL(tipc_dump_done);
3620
3621int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3622 struct tipc_sock *tsk, u32 sk_filter_state,
3623 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3624{
3625 struct sock *sk = &tsk->sk;
3626 struct nlattr *attrs;
3627 struct nlattr *stat;
3628
3629 /*filter response w.r.t sk_state*/
3630 if (!(sk_filter_state & (1 << sk->sk_state)))
3631 return 0;
3632
3633 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3634 if (!attrs)
3635 goto msg_cancel;
3636
3637 if (__tipc_nl_add_sk_info(skb, tsk))
3638 goto attr_msg_cancel;
3639
3640 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3641 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3642 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3643 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3644 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3645 sock_i_uid(sk))) ||
3646 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3647 tipc_diag_gen_cookie(sk),
3648 TIPC_NLA_SOCK_PAD))
3649 goto attr_msg_cancel;
3650
3651 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3652 if (!stat)
3653 goto attr_msg_cancel;
3654
3655 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3656 skb_queue_len(&sk->sk_receive_queue)) ||
3657 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3658 skb_queue_len(&sk->sk_write_queue)) ||
3659 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3660 atomic_read(&sk->sk_drops)))
3661 goto stat_msg_cancel;
3662
3663 if (tsk->cong_link_cnt &&
3664 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3665 goto stat_msg_cancel;
3666
3667 if (tsk_conn_cong(tsk) &&
3668 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3669 goto stat_msg_cancel;
3670
3671 nla_nest_end(skb, stat);
3672
3673 if (tsk->group)
3674 if (tipc_group_fill_sock_diag(tsk->group, skb))
3675 goto stat_msg_cancel;
3676
3677 nla_nest_end(skb, attrs);
3678
3679 return 0;
3680
3681stat_msg_cancel:
3682 nla_nest_cancel(skb, stat);
3683attr_msg_cancel:
3684 nla_nest_cancel(skb, attrs);
3685msg_cancel:
3686 return -EMSGSIZE;
3687}
3688EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3689
3690int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3691{
3692 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3693}
3694
3695/* Caller should hold socket lock for the passed tipc socket. */
3696static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3697 struct netlink_callback *cb,
3698 struct publication *publ)
3699{
3700 void *hdr;
3701 struct nlattr *attrs;
3702
3703 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3704 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3705 if (!hdr)
3706 goto msg_cancel;
3707
3708 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3709 if (!attrs)
3710 goto genlmsg_cancel;
3711
3712 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3713 goto attr_msg_cancel;
3714 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3715 goto attr_msg_cancel;
3716 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3717 goto attr_msg_cancel;
3718 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3719 goto attr_msg_cancel;
3720
3721 nla_nest_end(skb, attrs);
3722 genlmsg_end(skb, hdr);
3723
3724 return 0;
3725
3726attr_msg_cancel:
3727 nla_nest_cancel(skb, attrs);
3728genlmsg_cancel:
3729 genlmsg_cancel(skb, hdr);
3730msg_cancel:
3731 return -EMSGSIZE;
3732}
3733
3734/* Caller should hold socket lock for the passed tipc socket. */
3735static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3736 struct netlink_callback *cb,
3737 struct tipc_sock *tsk, u32 *last_publ)
3738{
3739 int err;
3740 struct publication *p;
3741
3742 if (*last_publ) {
3743 list_for_each_entry(p, &tsk->publications, binding_sock) {
3744 if (p->key == *last_publ)
3745 break;
3746 }
3747 if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3748 /* We never set seq or call nl_dump_check_consistent()
3749 * this means that setting prev_seq here will cause the
3750 * consistence check to fail in the netlink callback
3751 * handler. Resulting in the last NLMSG_DONE message
3752 * having the NLM_F_DUMP_INTR flag set.
3753 */
3754 cb->prev_seq = 1;
3755 *last_publ = 0;
3756 return -EPIPE;
3757 }
3758 } else {
3759 p = list_first_entry(&tsk->publications, struct publication,
3760 binding_sock);
3761 }
3762
3763 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3764 err = __tipc_nl_add_sk_publ(skb, cb, p);
3765 if (err) {
3766 *last_publ = p->key;
3767 return err;
3768 }
3769 }
3770 *last_publ = 0;
3771
3772 return 0;
3773}
3774
3775int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3776{
3777 int err;
3778 u32 tsk_portid = cb->args[0];
3779 u32 last_publ = cb->args[1];
3780 u32 done = cb->args[2];
3781 struct net *net = sock_net(skb->sk);
3782 struct tipc_sock *tsk;
3783
3784 if (!tsk_portid) {
3785 struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs;
3786 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3787
3788 if (!attrs[TIPC_NLA_SOCK])
3789 return -EINVAL;
3790
3791 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3792 attrs[TIPC_NLA_SOCK],
3793 tipc_nl_sock_policy, NULL);
3794 if (err)
3795 return err;
3796
3797 if (!sock[TIPC_NLA_SOCK_REF])
3798 return -EINVAL;
3799
3800 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3801 }
3802
3803 if (done)
3804 return 0;
3805
3806 tsk = tipc_sk_lookup(net, tsk_portid);
3807 if (!tsk)
3808 return -EINVAL;
3809
3810 lock_sock(&tsk->sk);
3811 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3812 if (!err)
3813 done = 1;
3814 release_sock(&tsk->sk);
3815 sock_put(&tsk->sk);
3816
3817 cb->args[0] = tsk_portid;
3818 cb->args[1] = last_publ;
3819 cb->args[2] = done;
3820
3821 return skb->len;
3822}
3823
3824/**
3825 * tipc_sk_filtering - check if a socket should be traced
3826 * @sk: the socket to be examined
3827 *
3828 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3829 * (portid, sock type, name type, name lower, name upper)
3830 *
3831 * Return: true if the socket meets the socket tuple data
3832 * (value 0 = 'any') or when there is no tuple set (all = 0),
3833 * otherwise false
3834 */
3835bool tipc_sk_filtering(struct sock *sk)
3836{
3837 struct tipc_sock *tsk;
3838 struct publication *p;
3839 u32 _port, _sktype, _type, _lower, _upper;
3840 u32 type = 0, lower = 0, upper = 0;
3841
3842 if (!sk)
3843 return true;
3844
3845 tsk = tipc_sk(sk);
3846
3847 _port = sysctl_tipc_sk_filter[0];
3848 _sktype = sysctl_tipc_sk_filter[1];
3849 _type = sysctl_tipc_sk_filter[2];
3850 _lower = sysctl_tipc_sk_filter[3];
3851 _upper = sysctl_tipc_sk_filter[4];
3852
3853 if (!_port && !_sktype && !_type && !_lower && !_upper)
3854 return true;
3855
3856 if (_port)
3857 return (_port == tsk->portid);
3858
3859 if (_sktype && _sktype != sk->sk_type)
3860 return false;
3861
3862 if (tsk->published) {
3863 p = list_first_entry_or_null(&tsk->publications,
3864 struct publication, binding_sock);
3865 if (p) {
3866 type = p->sr.type;
3867 lower = p->sr.lower;
3868 upper = p->sr.upper;
3869 }
3870 }
3871
3872 if (!tipc_sk_type_connectionless(sk)) {
3873 type = msg_nametype(&tsk->phdr);
3874 lower = msg_nameinst(&tsk->phdr);
3875 upper = lower;
3876 }
3877
3878 if ((_type && _type != type) || (_lower && _lower != lower) ||
3879 (_upper && _upper != upper))
3880 return false;
3881
3882 return true;
3883}
3884
3885u32 tipc_sock_get_portid(struct sock *sk)
3886{
3887 return (sk) ? (tipc_sk(sk))->portid : 0;
3888}
3889
3890/**
3891 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3892 * both the rcv and backlog queues are considered
3893 * @sk: tipc sk to be checked
3894 * @skb: tipc msg to be checked
3895 *
3896 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3897 */
3898
3899bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3900{
3901 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3902 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3903 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3904
3905 return (qsize > lim * 90 / 100);
3906}
3907
3908/**
3909 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3910 * only the rcv queue is considered
3911 * @sk: tipc sk to be checked
3912 * @skb: tipc msg to be checked
3913 *
3914 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3915 */
3916
3917bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3918{
3919 unsigned int lim = rcvbuf_limit(sk, skb);
3920 unsigned int qsize = sk_rmem_alloc_get(sk);
3921
3922 return (qsize > lim * 90 / 100);
3923}
3924
3925/**
3926 * tipc_sk_dump - dump TIPC socket
3927 * @sk: tipc sk to be dumped
3928 * @dqueues: bitmask to decide if any socket queue to be dumped?
3929 * - TIPC_DUMP_NONE: don't dump socket queues
3930 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3931 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3932 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3933 * - TIPC_DUMP_ALL: dump all the socket queues above
3934 * @buf: returned buffer of dump data in format
3935 */
3936int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3937{
3938 int i = 0;
3939 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3940 u32 conn_type, conn_instance;
3941 struct tipc_sock *tsk;
3942 struct publication *p;
3943 bool tsk_connected;
3944
3945 if (!sk) {
3946 i += scnprintf(buf, sz, "sk data: (null)\n");
3947 return i;
3948 }
3949
3950 tsk = tipc_sk(sk);
3951 tsk_connected = !tipc_sk_type_connectionless(sk);
3952
3953 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3954 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3955 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3956 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3957 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3958 if (tsk_connected) {
3959 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3960 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3961 conn_type = msg_nametype(&tsk->phdr);
3962 conn_instance = msg_nameinst(&tsk->phdr);
3963 i += scnprintf(buf + i, sz - i, " %u", conn_type);
3964 i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3965 }
3966 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3967 if (tsk->published) {
3968 p = list_first_entry_or_null(&tsk->publications,
3969 struct publication, binding_sock);
3970 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3971 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3972 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3973 }
3974 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3975 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3976 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3977 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3978 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3979 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3980 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3981 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3982 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3983 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3984 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3985 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3986 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3987 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3988
3989 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3990 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3991 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3992 }
3993
3994 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3995 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3996 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3997 }
3998
3999 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4000 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
4001 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4002 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4003 i += scnprintf(buf + i, sz - i, " tail ");
4004 i += tipc_skb_dump(sk->sk_backlog.tail, false,
4005 buf + i);
4006 }
4007 }
4008
4009 return i;
4010}