Loading...
1/*
2 * net/tipc/socket.c: TIPC socket API
3 *
4 * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
19 *
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
35 */
36
37#include <linux/rhashtable.h>
38#include <linux/sched/signal.h>
39
40#include "core.h"
41#include "name_table.h"
42#include "node.h"
43#include "link.h"
44#include "name_distr.h"
45#include "socket.h"
46#include "bcast.h"
47#include "netlink.h"
48#include "group.h"
49#include "trace.h"
50
51#define NAGLE_START_INIT 4
52#define NAGLE_START_MAX 1024
53#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
54#define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
55#define TIPC_FWD_MSG 1
56#define TIPC_MAX_PORT 0xffffffff
57#define TIPC_MIN_PORT 1
58#define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
59
60enum {
61 TIPC_LISTEN = TCP_LISTEN,
62 TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 TIPC_OPEN = TCP_CLOSE,
64 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 TIPC_CONNECTING = TCP_SYN_SENT,
66};
67
68struct sockaddr_pair {
69 struct sockaddr_tipc sock;
70 struct sockaddr_tipc member;
71};
72
73/**
74 * struct tipc_sock - TIPC socket structure
75 * @sk: socket - interacts with 'port' and with user via the socket API
76 * @conn_type: TIPC type used when connection was established
77 * @conn_instance: TIPC instance used when connection was established
78 * @published: non-zero if port has one or more associated names
79 * @max_pkt: maximum packet size "hint" used when building messages sent by port
80 * @maxnagle: maximum size of msg which can be subject to nagle
81 * @portid: unique port identity in TIPC socket hash table
82 * @phdr: preformatted message header used when sending messages
83 * #cong_links: list of congested links
84 * @publications: list of publications for port
85 * @blocking_link: address of the congested link we are currently sleeping on
86 * @pub_count: total # of publications port has made during its lifetime
87 * @conn_timeout: the time we can wait for an unresponded setup request
88 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
89 * @cong_link_cnt: number of congested links
90 * @snt_unacked: # messages sent by socket, and not yet acked by peer
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @peer: 'connected' peer for dgram/rdm
93 * @node: hash table node
94 * @mc_method: cookie for use between socket and broadcast layer
95 * @rcu: rcu struct for tipc_sock
96 */
97struct tipc_sock {
98 struct sock sk;
99 u32 conn_type;
100 u32 conn_instance;
101 int published;
102 u32 max_pkt;
103 u32 maxnagle;
104 u32 portid;
105 struct tipc_msg phdr;
106 struct list_head cong_links;
107 struct list_head publications;
108 u32 pub_count;
109 atomic_t dupl_rcvcnt;
110 u16 conn_timeout;
111 bool probe_unacked;
112 u16 cong_link_cnt;
113 u16 snt_unacked;
114 u16 snd_win;
115 u16 peer_caps;
116 u16 rcv_unacked;
117 u16 rcv_win;
118 struct sockaddr_tipc peer;
119 struct rhash_head node;
120 struct tipc_mc_method mc_method;
121 struct rcu_head rcu;
122 struct tipc_group *group;
123 u32 oneway;
124 u32 nagle_start;
125 u16 snd_backlog;
126 u16 msg_acc;
127 u16 pkt_cnt;
128 bool expect_ack;
129 bool nodelay;
130 bool group_is_open;
131};
132
133static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
134static void tipc_data_ready(struct sock *sk);
135static void tipc_write_space(struct sock *sk);
136static void tipc_sock_destruct(struct sock *sk);
137static int tipc_release(struct socket *sock);
138static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
139 bool kern);
140static void tipc_sk_timeout(struct timer_list *t);
141static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
142 struct tipc_name_seq const *seq);
143static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
144 struct tipc_name_seq const *seq);
145static int tipc_sk_leave(struct tipc_sock *tsk);
146static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
147static int tipc_sk_insert(struct tipc_sock *tsk);
148static void tipc_sk_remove(struct tipc_sock *tsk);
149static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
150static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
151static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
152
153static const struct proto_ops packet_ops;
154static const struct proto_ops stream_ops;
155static const struct proto_ops msg_ops;
156static struct proto tipc_proto;
157static const struct rhashtable_params tsk_rht_params;
158
159static u32 tsk_own_node(struct tipc_sock *tsk)
160{
161 return msg_prevnode(&tsk->phdr);
162}
163
164static u32 tsk_peer_node(struct tipc_sock *tsk)
165{
166 return msg_destnode(&tsk->phdr);
167}
168
169static u32 tsk_peer_port(struct tipc_sock *tsk)
170{
171 return msg_destport(&tsk->phdr);
172}
173
174static bool tsk_unreliable(struct tipc_sock *tsk)
175{
176 return msg_src_droppable(&tsk->phdr) != 0;
177}
178
179static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
180{
181 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
182}
183
184static bool tsk_unreturnable(struct tipc_sock *tsk)
185{
186 return msg_dest_droppable(&tsk->phdr) != 0;
187}
188
189static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
190{
191 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
192}
193
194static int tsk_importance(struct tipc_sock *tsk)
195{
196 return msg_importance(&tsk->phdr);
197}
198
199static struct tipc_sock *tipc_sk(const struct sock *sk)
200{
201 return container_of(sk, struct tipc_sock, sk);
202}
203
204int tsk_set_importance(struct sock *sk, int imp)
205{
206 if (imp > TIPC_CRITICAL_IMPORTANCE)
207 return -EINVAL;
208 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
209 return 0;
210}
211
212static bool tsk_conn_cong(struct tipc_sock *tsk)
213{
214 return tsk->snt_unacked > tsk->snd_win;
215}
216
217static u16 tsk_blocks(int len)
218{
219 return ((len / FLOWCTL_BLK_SZ) + 1);
220}
221
222/* tsk_blocks(): translate a buffer size in bytes to number of
223 * advertisable blocks, taking into account the ratio truesize(len)/len
224 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
225 */
226static u16 tsk_adv_blocks(int len)
227{
228 return len / FLOWCTL_BLK_SZ / 4;
229}
230
231/* tsk_inc(): increment counter for sent or received data
232 * - If block based flow control is not supported by peer we
233 * fall back to message based ditto, incrementing the counter
234 */
235static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
236{
237 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
238 return ((msglen / FLOWCTL_BLK_SZ) + 1);
239 return 1;
240}
241
242/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
243 */
244static void tsk_set_nagle(struct tipc_sock *tsk)
245{
246 struct sock *sk = &tsk->sk;
247
248 tsk->maxnagle = 0;
249 if (sk->sk_type != SOCK_STREAM)
250 return;
251 if (tsk->nodelay)
252 return;
253 if (!(tsk->peer_caps & TIPC_NAGLE))
254 return;
255 /* Limit node local buffer size to avoid receive queue overflow */
256 if (tsk->max_pkt == MAX_MSG_SIZE)
257 tsk->maxnagle = 1500;
258 else
259 tsk->maxnagle = tsk->max_pkt;
260}
261
262/**
263 * tsk_advance_rx_queue - discard first buffer in socket receive queue
264 *
265 * Caller must hold socket lock
266 */
267static void tsk_advance_rx_queue(struct sock *sk)
268{
269 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
270 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
271}
272
273/* tipc_sk_respond() : send response message back to sender
274 */
275static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
276{
277 u32 selector;
278 u32 dnode;
279 u32 onode = tipc_own_addr(sock_net(sk));
280
281 if (!tipc_msg_reverse(onode, &skb, err))
282 return;
283
284 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
285 dnode = msg_destnode(buf_msg(skb));
286 selector = msg_origport(buf_msg(skb));
287 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
288}
289
290/**
291 * tsk_rej_rx_queue - reject all buffers in socket receive queue
292 *
293 * Caller must hold socket lock
294 */
295static void tsk_rej_rx_queue(struct sock *sk, int error)
296{
297 struct sk_buff *skb;
298
299 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
300 tipc_sk_respond(sk, skb, error);
301}
302
303static bool tipc_sk_connected(struct sock *sk)
304{
305 return sk->sk_state == TIPC_ESTABLISHED;
306}
307
308/* tipc_sk_type_connectionless - check if the socket is datagram socket
309 * @sk: socket
310 *
311 * Returns true if connection less, false otherwise
312 */
313static bool tipc_sk_type_connectionless(struct sock *sk)
314{
315 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
316}
317
318/* tsk_peer_msg - verify if message was sent by connected port's peer
319 *
320 * Handles cases where the node's network address has changed from
321 * the default of <0.0.0> to its configured setting.
322 */
323static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
324{
325 struct sock *sk = &tsk->sk;
326 u32 self = tipc_own_addr(sock_net(sk));
327 u32 peer_port = tsk_peer_port(tsk);
328 u32 orig_node, peer_node;
329
330 if (unlikely(!tipc_sk_connected(sk)))
331 return false;
332
333 if (unlikely(msg_origport(msg) != peer_port))
334 return false;
335
336 orig_node = msg_orignode(msg);
337 peer_node = tsk_peer_node(tsk);
338
339 if (likely(orig_node == peer_node))
340 return true;
341
342 if (!orig_node && peer_node == self)
343 return true;
344
345 if (!peer_node && orig_node == self)
346 return true;
347
348 return false;
349}
350
351/* tipc_set_sk_state - set the sk_state of the socket
352 * @sk: socket
353 *
354 * Caller must hold socket lock
355 *
356 * Returns 0 on success, errno otherwise
357 */
358static int tipc_set_sk_state(struct sock *sk, int state)
359{
360 int oldsk_state = sk->sk_state;
361 int res = -EINVAL;
362
363 switch (state) {
364 case TIPC_OPEN:
365 res = 0;
366 break;
367 case TIPC_LISTEN:
368 case TIPC_CONNECTING:
369 if (oldsk_state == TIPC_OPEN)
370 res = 0;
371 break;
372 case TIPC_ESTABLISHED:
373 if (oldsk_state == TIPC_CONNECTING ||
374 oldsk_state == TIPC_OPEN)
375 res = 0;
376 break;
377 case TIPC_DISCONNECTING:
378 if (oldsk_state == TIPC_CONNECTING ||
379 oldsk_state == TIPC_ESTABLISHED)
380 res = 0;
381 break;
382 }
383
384 if (!res)
385 sk->sk_state = state;
386
387 return res;
388}
389
390static int tipc_sk_sock_err(struct socket *sock, long *timeout)
391{
392 struct sock *sk = sock->sk;
393 int err = sock_error(sk);
394 int typ = sock->type;
395
396 if (err)
397 return err;
398 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
399 if (sk->sk_state == TIPC_DISCONNECTING)
400 return -EPIPE;
401 else if (!tipc_sk_connected(sk))
402 return -ENOTCONN;
403 }
404 if (!*timeout)
405 return -EAGAIN;
406 if (signal_pending(current))
407 return sock_intr_errno(*timeout);
408
409 return 0;
410}
411
412#define tipc_wait_for_cond(sock_, timeo_, condition_) \
413({ \
414 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
415 struct sock *sk_; \
416 int rc_; \
417 \
418 while ((rc_ = !(condition_))) { \
419 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
420 smp_rmb(); \
421 sk_ = (sock_)->sk; \
422 rc_ = tipc_sk_sock_err((sock_), timeo_); \
423 if (rc_) \
424 break; \
425 add_wait_queue(sk_sleep(sk_), &wait_); \
426 release_sock(sk_); \
427 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
428 sched_annotate_sleep(); \
429 lock_sock(sk_); \
430 remove_wait_queue(sk_sleep(sk_), &wait_); \
431 } \
432 rc_; \
433})
434
435/**
436 * tipc_sk_create - create a TIPC socket
437 * @net: network namespace (must be default network)
438 * @sock: pre-allocated socket structure
439 * @protocol: protocol indicator (must be 0)
440 * @kern: caused by kernel or by userspace?
441 *
442 * This routine creates additional data structures used by the TIPC socket,
443 * initializes them, and links them together.
444 *
445 * Returns 0 on success, errno otherwise
446 */
447static int tipc_sk_create(struct net *net, struct socket *sock,
448 int protocol, int kern)
449{
450 const struct proto_ops *ops;
451 struct sock *sk;
452 struct tipc_sock *tsk;
453 struct tipc_msg *msg;
454
455 /* Validate arguments */
456 if (unlikely(protocol != 0))
457 return -EPROTONOSUPPORT;
458
459 switch (sock->type) {
460 case SOCK_STREAM:
461 ops = &stream_ops;
462 break;
463 case SOCK_SEQPACKET:
464 ops = &packet_ops;
465 break;
466 case SOCK_DGRAM:
467 case SOCK_RDM:
468 ops = &msg_ops;
469 break;
470 default:
471 return -EPROTOTYPE;
472 }
473
474 /* Allocate socket's protocol area */
475 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
476 if (sk == NULL)
477 return -ENOMEM;
478
479 tsk = tipc_sk(sk);
480 tsk->max_pkt = MAX_PKT_DEFAULT;
481 tsk->maxnagle = 0;
482 tsk->nagle_start = NAGLE_START_INIT;
483 INIT_LIST_HEAD(&tsk->publications);
484 INIT_LIST_HEAD(&tsk->cong_links);
485 msg = &tsk->phdr;
486
487 /* Finish initializing socket data structures */
488 sock->ops = ops;
489 sock_init_data(sock, sk);
490 tipc_set_sk_state(sk, TIPC_OPEN);
491 if (tipc_sk_insert(tsk)) {
492 pr_warn("Socket create failed; port number exhausted\n");
493 return -EINVAL;
494 }
495
496 /* Ensure tsk is visible before we read own_addr. */
497 smp_mb();
498
499 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
500 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
501
502 msg_set_origport(msg, tsk->portid);
503 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
504 sk->sk_shutdown = 0;
505 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
506 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
507 sk->sk_data_ready = tipc_data_ready;
508 sk->sk_write_space = tipc_write_space;
509 sk->sk_destruct = tipc_sock_destruct;
510 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
511 tsk->group_is_open = true;
512 atomic_set(&tsk->dupl_rcvcnt, 0);
513
514 /* Start out with safe limits until we receive an advertised window */
515 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
516 tsk->rcv_win = tsk->snd_win;
517
518 if (tipc_sk_type_connectionless(sk)) {
519 tsk_set_unreturnable(tsk, true);
520 if (sock->type == SOCK_DGRAM)
521 tsk_set_unreliable(tsk, true);
522 }
523 __skb_queue_head_init(&tsk->mc_method.deferredq);
524 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
525 return 0;
526}
527
528static void tipc_sk_callback(struct rcu_head *head)
529{
530 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
531
532 sock_put(&tsk->sk);
533}
534
535/* Caller should hold socket lock for the socket. */
536static void __tipc_shutdown(struct socket *sock, int error)
537{
538 struct sock *sk = sock->sk;
539 struct tipc_sock *tsk = tipc_sk(sk);
540 struct net *net = sock_net(sk);
541 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
542 u32 dnode = tsk_peer_node(tsk);
543 struct sk_buff *skb;
544
545 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
546 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
547 !tsk_conn_cong(tsk)));
548
549 /* Push out delayed messages if in Nagle mode */
550 tipc_sk_push_backlog(tsk, false);
551 /* Remove pending SYN */
552 __skb_queue_purge(&sk->sk_write_queue);
553
554 /* Remove partially received buffer if any */
555 skb = skb_peek(&sk->sk_receive_queue);
556 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
557 __skb_unlink(skb, &sk->sk_receive_queue);
558 kfree_skb(skb);
559 }
560
561 /* Reject all unreceived messages if connectionless */
562 if (tipc_sk_type_connectionless(sk)) {
563 tsk_rej_rx_queue(sk, error);
564 return;
565 }
566
567 switch (sk->sk_state) {
568 case TIPC_CONNECTING:
569 case TIPC_ESTABLISHED:
570 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
571 tipc_node_remove_conn(net, dnode, tsk->portid);
572 /* Send a FIN+/- to its peer */
573 skb = __skb_dequeue(&sk->sk_receive_queue);
574 if (skb) {
575 __skb_queue_purge(&sk->sk_receive_queue);
576 tipc_sk_respond(sk, skb, error);
577 break;
578 }
579 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
580 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
581 tsk_own_node(tsk), tsk_peer_port(tsk),
582 tsk->portid, error);
583 if (skb)
584 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
585 break;
586 case TIPC_LISTEN:
587 /* Reject all SYN messages */
588 tsk_rej_rx_queue(sk, error);
589 break;
590 default:
591 __skb_queue_purge(&sk->sk_receive_queue);
592 break;
593 }
594}
595
596/**
597 * tipc_release - destroy a TIPC socket
598 * @sock: socket to destroy
599 *
600 * This routine cleans up any messages that are still queued on the socket.
601 * For DGRAM and RDM socket types, all queued messages are rejected.
602 * For SEQPACKET and STREAM socket types, the first message is rejected
603 * and any others are discarded. (If the first message on a STREAM socket
604 * is partially-read, it is discarded and the next one is rejected instead.)
605 *
606 * NOTE: Rejected messages are not necessarily returned to the sender! They
607 * are returned or discarded according to the "destination droppable" setting
608 * specified for the message by the sender.
609 *
610 * Returns 0 on success, errno otherwise
611 */
612static int tipc_release(struct socket *sock)
613{
614 struct sock *sk = sock->sk;
615 struct tipc_sock *tsk;
616
617 /*
618 * Exit if socket isn't fully initialized (occurs when a failed accept()
619 * releases a pre-allocated child socket that was never used)
620 */
621 if (sk == NULL)
622 return 0;
623
624 tsk = tipc_sk(sk);
625 lock_sock(sk);
626
627 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
628 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
629 sk->sk_shutdown = SHUTDOWN_MASK;
630 tipc_sk_leave(tsk);
631 tipc_sk_withdraw(tsk, 0, NULL);
632 __skb_queue_purge(&tsk->mc_method.deferredq);
633 sk_stop_timer(sk, &sk->sk_timer);
634 tipc_sk_remove(tsk);
635
636 sock_orphan(sk);
637 /* Reject any messages that accumulated in backlog queue */
638 release_sock(sk);
639 tipc_dest_list_purge(&tsk->cong_links);
640 tsk->cong_link_cnt = 0;
641 call_rcu(&tsk->rcu, tipc_sk_callback);
642 sock->sk = NULL;
643
644 return 0;
645}
646
647/**
648 * tipc_bind - associate or disassocate TIPC name(s) with a socket
649 * @sock: socket structure
650 * @uaddr: socket address describing name(s) and desired operation
651 * @uaddr_len: size of socket address data structure
652 *
653 * Name and name sequence binding is indicated using a positive scope value;
654 * a negative scope value unbinds the specified name. Specifying no name
655 * (i.e. a socket address length of 0) unbinds all names from the socket.
656 *
657 * Returns 0 on success, errno otherwise
658 *
659 * NOTE: This routine doesn't need to take the socket lock since it doesn't
660 * access any non-constant socket information.
661 */
662static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
663 int uaddr_len)
664{
665 struct sock *sk = sock->sk;
666 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
667 struct tipc_sock *tsk = tipc_sk(sk);
668 int res = -EINVAL;
669
670 lock_sock(sk);
671 if (unlikely(!uaddr_len)) {
672 res = tipc_sk_withdraw(tsk, 0, NULL);
673 goto exit;
674 }
675 if (tsk->group) {
676 res = -EACCES;
677 goto exit;
678 }
679 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
680 res = -EINVAL;
681 goto exit;
682 }
683 if (addr->family != AF_TIPC) {
684 res = -EAFNOSUPPORT;
685 goto exit;
686 }
687
688 if (addr->addrtype == TIPC_ADDR_NAME)
689 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
690 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
691 res = -EAFNOSUPPORT;
692 goto exit;
693 }
694
695 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
696 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
697 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
698 res = -EACCES;
699 goto exit;
700 }
701
702 res = (addr->scope >= 0) ?
703 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
704 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
705exit:
706 release_sock(sk);
707 return res;
708}
709
710/**
711 * tipc_getname - get port ID of socket or peer socket
712 * @sock: socket structure
713 * @uaddr: area for returned socket address
714 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
715 *
716 * Returns 0 on success, errno otherwise
717 *
718 * NOTE: This routine doesn't need to take the socket lock since it only
719 * accesses socket information that is unchanging (or which changes in
720 * a completely predictable manner).
721 */
722static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
723 int peer)
724{
725 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
726 struct sock *sk = sock->sk;
727 struct tipc_sock *tsk = tipc_sk(sk);
728
729 memset(addr, 0, sizeof(*addr));
730 if (peer) {
731 if ((!tipc_sk_connected(sk)) &&
732 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
733 return -ENOTCONN;
734 addr->addr.id.ref = tsk_peer_port(tsk);
735 addr->addr.id.node = tsk_peer_node(tsk);
736 } else {
737 addr->addr.id.ref = tsk->portid;
738 addr->addr.id.node = tipc_own_addr(sock_net(sk));
739 }
740
741 addr->addrtype = TIPC_ADDR_ID;
742 addr->family = AF_TIPC;
743 addr->scope = 0;
744 addr->addr.name.domain = 0;
745
746 return sizeof(*addr);
747}
748
749/**
750 * tipc_poll - read and possibly block on pollmask
751 * @file: file structure associated with the socket
752 * @sock: socket for which to calculate the poll bits
753 * @wait: ???
754 *
755 * Returns pollmask value
756 *
757 * COMMENTARY:
758 * It appears that the usual socket locking mechanisms are not useful here
759 * since the pollmask info is potentially out-of-date the moment this routine
760 * exits. TCP and other protocols seem to rely on higher level poll routines
761 * to handle any preventable race conditions, so TIPC will do the same ...
762 *
763 * IMPORTANT: The fact that a read or write operation is indicated does NOT
764 * imply that the operation will succeed, merely that it should be performed
765 * and will not block.
766 */
767static __poll_t tipc_poll(struct file *file, struct socket *sock,
768 poll_table *wait)
769{
770 struct sock *sk = sock->sk;
771 struct tipc_sock *tsk = tipc_sk(sk);
772 __poll_t revents = 0;
773
774 sock_poll_wait(file, sock, wait);
775 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
776
777 if (sk->sk_shutdown & RCV_SHUTDOWN)
778 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
779 if (sk->sk_shutdown == SHUTDOWN_MASK)
780 revents |= EPOLLHUP;
781
782 switch (sk->sk_state) {
783 case TIPC_ESTABLISHED:
784 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
785 revents |= EPOLLOUT;
786 fallthrough;
787 case TIPC_LISTEN:
788 case TIPC_CONNECTING:
789 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
790 revents |= EPOLLIN | EPOLLRDNORM;
791 break;
792 case TIPC_OPEN:
793 if (tsk->group_is_open && !tsk->cong_link_cnt)
794 revents |= EPOLLOUT;
795 if (!tipc_sk_type_connectionless(sk))
796 break;
797 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
798 break;
799 revents |= EPOLLIN | EPOLLRDNORM;
800 break;
801 case TIPC_DISCONNECTING:
802 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
803 break;
804 }
805 return revents;
806}
807
808/**
809 * tipc_sendmcast - send multicast message
810 * @sock: socket structure
811 * @seq: destination address
812 * @msg: message to send
813 * @dlen: length of data to send
814 * @timeout: timeout to wait for wakeup
815 *
816 * Called from function tipc_sendmsg(), which has done all sanity checks
817 * Returns the number of bytes sent on success, or errno
818 */
819static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
820 struct msghdr *msg, size_t dlen, long timeout)
821{
822 struct sock *sk = sock->sk;
823 struct tipc_sock *tsk = tipc_sk(sk);
824 struct tipc_msg *hdr = &tsk->phdr;
825 struct net *net = sock_net(sk);
826 int mtu = tipc_bcast_get_mtu(net);
827 struct tipc_mc_method *method = &tsk->mc_method;
828 struct sk_buff_head pkts;
829 struct tipc_nlist dsts;
830 int rc;
831
832 if (tsk->group)
833 return -EACCES;
834
835 /* Block or return if any destination link is congested */
836 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
837 if (unlikely(rc))
838 return rc;
839
840 /* Lookup destination nodes */
841 tipc_nlist_init(&dsts, tipc_own_addr(net));
842 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
843 seq->upper, &dsts);
844 if (!dsts.local && !dsts.remote)
845 return -EHOSTUNREACH;
846
847 /* Build message header */
848 msg_set_type(hdr, TIPC_MCAST_MSG);
849 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
850 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
851 msg_set_destport(hdr, 0);
852 msg_set_destnode(hdr, 0);
853 msg_set_nametype(hdr, seq->type);
854 msg_set_namelower(hdr, seq->lower);
855 msg_set_nameupper(hdr, seq->upper);
856
857 /* Build message as chain of buffers */
858 __skb_queue_head_init(&pkts);
859 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
860
861 /* Send message if build was successful */
862 if (unlikely(rc == dlen)) {
863 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
864 TIPC_DUMP_SK_SNDQ, " ");
865 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
866 &tsk->cong_link_cnt);
867 }
868
869 tipc_nlist_purge(&dsts);
870
871 return rc ? rc : dlen;
872}
873
874/**
875 * tipc_send_group_msg - send a message to a member in the group
876 * @net: network namespace
877 * @m: message to send
878 * @mb: group member
879 * @dnode: destination node
880 * @dport: destination port
881 * @dlen: total length of message data
882 */
883static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
884 struct msghdr *m, struct tipc_member *mb,
885 u32 dnode, u32 dport, int dlen)
886{
887 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
888 struct tipc_mc_method *method = &tsk->mc_method;
889 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
890 struct tipc_msg *hdr = &tsk->phdr;
891 struct sk_buff_head pkts;
892 int mtu, rc;
893
894 /* Complete message header */
895 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
896 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
897 msg_set_destport(hdr, dport);
898 msg_set_destnode(hdr, dnode);
899 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
900
901 /* Build message as chain of buffers */
902 __skb_queue_head_init(&pkts);
903 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
904 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
905 if (unlikely(rc != dlen))
906 return rc;
907
908 /* Send message */
909 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
910 if (unlikely(rc == -ELINKCONG)) {
911 tipc_dest_push(&tsk->cong_links, dnode, 0);
912 tsk->cong_link_cnt++;
913 }
914
915 /* Update send window */
916 tipc_group_update_member(mb, blks);
917
918 /* A broadcast sent within next EXPIRE period must follow same path */
919 method->rcast = true;
920 method->mandatory = true;
921 return dlen;
922}
923
924/**
925 * tipc_send_group_unicast - send message to a member in the group
926 * @sock: socket structure
927 * @m: message to send
928 * @dlen: total length of message data
929 * @timeout: timeout to wait for wakeup
930 *
931 * Called from function tipc_sendmsg(), which has done all sanity checks
932 * Returns the number of bytes sent on success, or errno
933 */
934static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
935 int dlen, long timeout)
936{
937 struct sock *sk = sock->sk;
938 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
939 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
940 struct tipc_sock *tsk = tipc_sk(sk);
941 struct net *net = sock_net(sk);
942 struct tipc_member *mb = NULL;
943 u32 node, port;
944 int rc;
945
946 node = dest->addr.id.node;
947 port = dest->addr.id.ref;
948 if (!port && !node)
949 return -EHOSTUNREACH;
950
951 /* Block or return if destination link or member is congested */
952 rc = tipc_wait_for_cond(sock, &timeout,
953 !tipc_dest_find(&tsk->cong_links, node, 0) &&
954 tsk->group &&
955 !tipc_group_cong(tsk->group, node, port, blks,
956 &mb));
957 if (unlikely(rc))
958 return rc;
959
960 if (unlikely(!mb))
961 return -EHOSTUNREACH;
962
963 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
964
965 return rc ? rc : dlen;
966}
967
968/**
969 * tipc_send_group_anycast - send message to any member with given identity
970 * @sock: socket structure
971 * @m: message to send
972 * @dlen: total length of message data
973 * @timeout: timeout to wait for wakeup
974 *
975 * Called from function tipc_sendmsg(), which has done all sanity checks
976 * Returns the number of bytes sent on success, or errno
977 */
978static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
979 int dlen, long timeout)
980{
981 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
982 struct sock *sk = sock->sk;
983 struct tipc_sock *tsk = tipc_sk(sk);
984 struct list_head *cong_links = &tsk->cong_links;
985 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
986 struct tipc_msg *hdr = &tsk->phdr;
987 struct tipc_member *first = NULL;
988 struct tipc_member *mbr = NULL;
989 struct net *net = sock_net(sk);
990 u32 node, port, exclude;
991 struct list_head dsts;
992 u32 type, inst, scope;
993 int lookups = 0;
994 int dstcnt, rc;
995 bool cong;
996
997 INIT_LIST_HEAD(&dsts);
998
999 type = msg_nametype(hdr);
1000 inst = dest->addr.name.name.instance;
1001 scope = msg_lookup_scope(hdr);
1002
1003 while (++lookups < 4) {
1004 exclude = tipc_group_exclude(tsk->group);
1005
1006 first = NULL;
1007
1008 /* Look for a non-congested destination member, if any */
1009 while (1) {
1010 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1011 &dstcnt, exclude, false))
1012 return -EHOSTUNREACH;
1013 tipc_dest_pop(&dsts, &node, &port);
1014 cong = tipc_group_cong(tsk->group, node, port, blks,
1015 &mbr);
1016 if (!cong)
1017 break;
1018 if (mbr == first)
1019 break;
1020 if (!first)
1021 first = mbr;
1022 }
1023
1024 /* Start over if destination was not in member list */
1025 if (unlikely(!mbr))
1026 continue;
1027
1028 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1029 break;
1030
1031 /* Block or return if destination link or member is congested */
1032 rc = tipc_wait_for_cond(sock, &timeout,
1033 !tipc_dest_find(cong_links, node, 0) &&
1034 tsk->group &&
1035 !tipc_group_cong(tsk->group, node, port,
1036 blks, &mbr));
1037 if (unlikely(rc))
1038 return rc;
1039
1040 /* Send, unless destination disappeared while waiting */
1041 if (likely(mbr))
1042 break;
1043 }
1044
1045 if (unlikely(lookups >= 4))
1046 return -EHOSTUNREACH;
1047
1048 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1049
1050 return rc ? rc : dlen;
1051}
1052
1053/**
1054 * tipc_send_group_bcast - send message to all members in communication group
1055 * @sock: socket structure
1056 * @m: message to send
1057 * @dlen: total length of message data
1058 * @timeout: timeout to wait for wakeup
1059 *
1060 * Called from function tipc_sendmsg(), which has done all sanity checks
1061 * Returns the number of bytes sent on success, or errno
1062 */
1063static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1064 int dlen, long timeout)
1065{
1066 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1067 struct sock *sk = sock->sk;
1068 struct net *net = sock_net(sk);
1069 struct tipc_sock *tsk = tipc_sk(sk);
1070 struct tipc_nlist *dsts;
1071 struct tipc_mc_method *method = &tsk->mc_method;
1072 bool ack = method->mandatory && method->rcast;
1073 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1074 struct tipc_msg *hdr = &tsk->phdr;
1075 int mtu = tipc_bcast_get_mtu(net);
1076 struct sk_buff_head pkts;
1077 int rc = -EHOSTUNREACH;
1078
1079 /* Block or return if any destination link or member is congested */
1080 rc = tipc_wait_for_cond(sock, &timeout,
1081 !tsk->cong_link_cnt && tsk->group &&
1082 !tipc_group_bc_cong(tsk->group, blks));
1083 if (unlikely(rc))
1084 return rc;
1085
1086 dsts = tipc_group_dests(tsk->group);
1087 if (!dsts->local && !dsts->remote)
1088 return -EHOSTUNREACH;
1089
1090 /* Complete message header */
1091 if (dest) {
1092 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1093 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1094 } else {
1095 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1096 msg_set_nameinst(hdr, 0);
1097 }
1098 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1099 msg_set_destport(hdr, 0);
1100 msg_set_destnode(hdr, 0);
1101 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1102
1103 /* Avoid getting stuck with repeated forced replicasts */
1104 msg_set_grp_bc_ack_req(hdr, ack);
1105
1106 /* Build message as chain of buffers */
1107 __skb_queue_head_init(&pkts);
1108 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1109 if (unlikely(rc != dlen))
1110 return rc;
1111
1112 /* Send message */
1113 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1114 if (unlikely(rc))
1115 return rc;
1116
1117 /* Update broadcast sequence number and send windows */
1118 tipc_group_update_bc_members(tsk->group, blks, ack);
1119
1120 /* Broadcast link is now free to choose method for next broadcast */
1121 method->mandatory = false;
1122 method->expires = jiffies;
1123
1124 return dlen;
1125}
1126
1127/**
1128 * tipc_send_group_mcast - send message to all members with given identity
1129 * @sock: socket structure
1130 * @m: message to send
1131 * @dlen: total length of message data
1132 * @timeout: timeout to wait for wakeup
1133 *
1134 * Called from function tipc_sendmsg(), which has done all sanity checks
1135 * Returns the number of bytes sent on success, or errno
1136 */
1137static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1138 int dlen, long timeout)
1139{
1140 struct sock *sk = sock->sk;
1141 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1142 struct tipc_sock *tsk = tipc_sk(sk);
1143 struct tipc_group *grp = tsk->group;
1144 struct tipc_msg *hdr = &tsk->phdr;
1145 struct net *net = sock_net(sk);
1146 u32 type, inst, scope, exclude;
1147 struct list_head dsts;
1148 u32 dstcnt;
1149
1150 INIT_LIST_HEAD(&dsts);
1151
1152 type = msg_nametype(hdr);
1153 inst = dest->addr.name.name.instance;
1154 scope = msg_lookup_scope(hdr);
1155 exclude = tipc_group_exclude(grp);
1156
1157 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1158 &dstcnt, exclude, true))
1159 return -EHOSTUNREACH;
1160
1161 if (dstcnt == 1) {
1162 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1163 return tipc_send_group_unicast(sock, m, dlen, timeout);
1164 }
1165
1166 tipc_dest_list_purge(&dsts);
1167 return tipc_send_group_bcast(sock, m, dlen, timeout);
1168}
1169
1170/**
1171 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1172 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1173 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1174 *
1175 * Multi-threaded: parallel calls with reference to same queues may occur
1176 */
1177void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1178 struct sk_buff_head *inputq)
1179{
1180 u32 self = tipc_own_addr(net);
1181 u32 type, lower, upper, scope;
1182 struct sk_buff *skb, *_skb;
1183 u32 portid, onode;
1184 struct sk_buff_head tmpq;
1185 struct list_head dports;
1186 struct tipc_msg *hdr;
1187 int user, mtyp, hlen;
1188 bool exact;
1189
1190 __skb_queue_head_init(&tmpq);
1191 INIT_LIST_HEAD(&dports);
1192
1193 skb = tipc_skb_peek(arrvq, &inputq->lock);
1194 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1195 hdr = buf_msg(skb);
1196 user = msg_user(hdr);
1197 mtyp = msg_type(hdr);
1198 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1199 onode = msg_orignode(hdr);
1200 type = msg_nametype(hdr);
1201
1202 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1203 spin_lock_bh(&inputq->lock);
1204 if (skb_peek(arrvq) == skb) {
1205 __skb_dequeue(arrvq);
1206 __skb_queue_tail(inputq, skb);
1207 }
1208 kfree_skb(skb);
1209 spin_unlock_bh(&inputq->lock);
1210 continue;
1211 }
1212
1213 /* Group messages require exact scope match */
1214 if (msg_in_group(hdr)) {
1215 lower = 0;
1216 upper = ~0;
1217 scope = msg_lookup_scope(hdr);
1218 exact = true;
1219 } else {
1220 /* TIPC_NODE_SCOPE means "any scope" in this context */
1221 if (onode == self)
1222 scope = TIPC_NODE_SCOPE;
1223 else
1224 scope = TIPC_CLUSTER_SCOPE;
1225 exact = false;
1226 lower = msg_namelower(hdr);
1227 upper = msg_nameupper(hdr);
1228 }
1229
1230 /* Create destination port list: */
1231 tipc_nametbl_mc_lookup(net, type, lower, upper,
1232 scope, exact, &dports);
1233
1234 /* Clone message per destination */
1235 while (tipc_dest_pop(&dports, NULL, &portid)) {
1236 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1237 if (_skb) {
1238 msg_set_destport(buf_msg(_skb), portid);
1239 __skb_queue_tail(&tmpq, _skb);
1240 continue;
1241 }
1242 pr_warn("Failed to clone mcast rcv buffer\n");
1243 }
1244 /* Append to inputq if not already done by other thread */
1245 spin_lock_bh(&inputq->lock);
1246 if (skb_peek(arrvq) == skb) {
1247 skb_queue_splice_tail_init(&tmpq, inputq);
1248 kfree_skb(__skb_dequeue(arrvq));
1249 }
1250 spin_unlock_bh(&inputq->lock);
1251 __skb_queue_purge(&tmpq);
1252 kfree_skb(skb);
1253 }
1254 tipc_sk_rcv(net, inputq);
1255}
1256
1257/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1258 * when socket is in Nagle mode
1259 */
1260static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1261{
1262 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1263 struct sk_buff *skb = skb_peek_tail(txq);
1264 struct net *net = sock_net(&tsk->sk);
1265 u32 dnode = tsk_peer_node(tsk);
1266 int rc;
1267
1268 if (nagle_ack) {
1269 tsk->pkt_cnt += skb_queue_len(txq);
1270 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1271 tsk->oneway = 0;
1272 if (tsk->nagle_start < NAGLE_START_MAX)
1273 tsk->nagle_start *= 2;
1274 tsk->expect_ack = false;
1275 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1276 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1277 tsk->nagle_start);
1278 } else {
1279 tsk->nagle_start = NAGLE_START_INIT;
1280 if (skb) {
1281 msg_set_ack_required(buf_msg(skb));
1282 tsk->expect_ack = true;
1283 } else {
1284 tsk->expect_ack = false;
1285 }
1286 }
1287 tsk->msg_acc = 0;
1288 tsk->pkt_cnt = 0;
1289 }
1290
1291 if (!skb || tsk->cong_link_cnt)
1292 return;
1293
1294 /* Do not send SYN again after congestion */
1295 if (msg_is_syn(buf_msg(skb)))
1296 return;
1297
1298 if (tsk->msg_acc)
1299 tsk->pkt_cnt += skb_queue_len(txq);
1300 tsk->snt_unacked += tsk->snd_backlog;
1301 tsk->snd_backlog = 0;
1302 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1303 if (rc == -ELINKCONG)
1304 tsk->cong_link_cnt = 1;
1305}
1306
1307/**
1308 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1309 * @tsk: receiving socket
1310 * @skb: pointer to message buffer.
1311 */
1312static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1313 struct sk_buff_head *inputq,
1314 struct sk_buff_head *xmitq)
1315{
1316 struct tipc_msg *hdr = buf_msg(skb);
1317 u32 onode = tsk_own_node(tsk);
1318 struct sock *sk = &tsk->sk;
1319 int mtyp = msg_type(hdr);
1320 bool was_cong;
1321
1322 /* Ignore if connection cannot be validated: */
1323 if (!tsk_peer_msg(tsk, hdr)) {
1324 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1325 goto exit;
1326 }
1327
1328 if (unlikely(msg_errcode(hdr))) {
1329 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1330 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1331 tsk_peer_port(tsk));
1332 sk->sk_state_change(sk);
1333
1334 /* State change is ignored if socket already awake,
1335 * - convert msg to abort msg and add to inqueue
1336 */
1337 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1338 msg_set_type(hdr, TIPC_CONN_MSG);
1339 msg_set_size(hdr, BASIC_H_SIZE);
1340 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1341 __skb_queue_tail(inputq, skb);
1342 return;
1343 }
1344
1345 tsk->probe_unacked = false;
1346
1347 if (mtyp == CONN_PROBE) {
1348 msg_set_type(hdr, CONN_PROBE_REPLY);
1349 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1350 __skb_queue_tail(xmitq, skb);
1351 return;
1352 } else if (mtyp == CONN_ACK) {
1353 was_cong = tsk_conn_cong(tsk);
1354 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1355 tsk->snt_unacked -= msg_conn_ack(hdr);
1356 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1357 tsk->snd_win = msg_adv_win(hdr);
1358 if (was_cong && !tsk_conn_cong(tsk))
1359 sk->sk_write_space(sk);
1360 } else if (mtyp != CONN_PROBE_REPLY) {
1361 pr_warn("Received unknown CONN_PROTO msg\n");
1362 }
1363exit:
1364 kfree_skb(skb);
1365}
1366
1367/**
1368 * tipc_sendmsg - send message in connectionless manner
1369 * @sock: socket structure
1370 * @m: message to send
1371 * @dsz: amount of user data to be sent
1372 *
1373 * Message must have an destination specified explicitly.
1374 * Used for SOCK_RDM and SOCK_DGRAM messages,
1375 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1376 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1377 *
1378 * Returns the number of bytes sent on success, or errno otherwise
1379 */
1380static int tipc_sendmsg(struct socket *sock,
1381 struct msghdr *m, size_t dsz)
1382{
1383 struct sock *sk = sock->sk;
1384 int ret;
1385
1386 lock_sock(sk);
1387 ret = __tipc_sendmsg(sock, m, dsz);
1388 release_sock(sk);
1389
1390 return ret;
1391}
1392
1393static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1394{
1395 struct sock *sk = sock->sk;
1396 struct net *net = sock_net(sk);
1397 struct tipc_sock *tsk = tipc_sk(sk);
1398 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1399 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1400 struct list_head *clinks = &tsk->cong_links;
1401 bool syn = !tipc_sk_type_connectionless(sk);
1402 struct tipc_group *grp = tsk->group;
1403 struct tipc_msg *hdr = &tsk->phdr;
1404 struct tipc_name_seq *seq;
1405 struct sk_buff_head pkts;
1406 u32 dport = 0, dnode = 0;
1407 u32 type = 0, inst = 0;
1408 int mtu, rc;
1409
1410 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1411 return -EMSGSIZE;
1412
1413 if (likely(dest)) {
1414 if (unlikely(m->msg_namelen < sizeof(*dest)))
1415 return -EINVAL;
1416 if (unlikely(dest->family != AF_TIPC))
1417 return -EINVAL;
1418 }
1419
1420 if (grp) {
1421 if (!dest)
1422 return tipc_send_group_bcast(sock, m, dlen, timeout);
1423 if (dest->addrtype == TIPC_ADDR_NAME)
1424 return tipc_send_group_anycast(sock, m, dlen, timeout);
1425 if (dest->addrtype == TIPC_ADDR_ID)
1426 return tipc_send_group_unicast(sock, m, dlen, timeout);
1427 if (dest->addrtype == TIPC_ADDR_MCAST)
1428 return tipc_send_group_mcast(sock, m, dlen, timeout);
1429 return -EINVAL;
1430 }
1431
1432 if (unlikely(!dest)) {
1433 dest = &tsk->peer;
1434 if (!syn && dest->family != AF_TIPC)
1435 return -EDESTADDRREQ;
1436 }
1437
1438 if (unlikely(syn)) {
1439 if (sk->sk_state == TIPC_LISTEN)
1440 return -EPIPE;
1441 if (sk->sk_state != TIPC_OPEN)
1442 return -EISCONN;
1443 if (tsk->published)
1444 return -EOPNOTSUPP;
1445 if (dest->addrtype == TIPC_ADDR_NAME) {
1446 tsk->conn_type = dest->addr.name.name.type;
1447 tsk->conn_instance = dest->addr.name.name.instance;
1448 }
1449 msg_set_syn(hdr, 1);
1450 }
1451
1452 seq = &dest->addr.nameseq;
1453 if (dest->addrtype == TIPC_ADDR_MCAST)
1454 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1455
1456 if (dest->addrtype == TIPC_ADDR_NAME) {
1457 type = dest->addr.name.name.type;
1458 inst = dest->addr.name.name.instance;
1459 dnode = dest->addr.name.domain;
1460 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1461 if (unlikely(!dport && !dnode))
1462 return -EHOSTUNREACH;
1463 } else if (dest->addrtype == TIPC_ADDR_ID) {
1464 dnode = dest->addr.id.node;
1465 } else {
1466 return -EINVAL;
1467 }
1468
1469 /* Block or return if destination link is congested */
1470 rc = tipc_wait_for_cond(sock, &timeout,
1471 !tipc_dest_find(clinks, dnode, 0));
1472 if (unlikely(rc))
1473 return rc;
1474
1475 if (dest->addrtype == TIPC_ADDR_NAME) {
1476 msg_set_type(hdr, TIPC_NAMED_MSG);
1477 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1478 msg_set_nametype(hdr, type);
1479 msg_set_nameinst(hdr, inst);
1480 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1481 msg_set_destnode(hdr, dnode);
1482 msg_set_destport(hdr, dport);
1483 } else { /* TIPC_ADDR_ID */
1484 msg_set_type(hdr, TIPC_DIRECT_MSG);
1485 msg_set_lookup_scope(hdr, 0);
1486 msg_set_destnode(hdr, dnode);
1487 msg_set_destport(hdr, dest->addr.id.ref);
1488 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1489 }
1490
1491 __skb_queue_head_init(&pkts);
1492 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1493 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1494 if (unlikely(rc != dlen))
1495 return rc;
1496 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1497 __skb_queue_purge(&pkts);
1498 return -ENOMEM;
1499 }
1500
1501 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1502 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1503 if (unlikely(rc == -ELINKCONG)) {
1504 tipc_dest_push(clinks, dnode, 0);
1505 tsk->cong_link_cnt++;
1506 rc = 0;
1507 }
1508
1509 if (unlikely(syn && !rc))
1510 tipc_set_sk_state(sk, TIPC_CONNECTING);
1511
1512 return rc ? rc : dlen;
1513}
1514
1515/**
1516 * tipc_sendstream - send stream-oriented data
1517 * @sock: socket structure
1518 * @m: data to send
1519 * @dsz: total length of data to be transmitted
1520 *
1521 * Used for SOCK_STREAM data.
1522 *
1523 * Returns the number of bytes sent on success (or partial success),
1524 * or errno if no data sent
1525 */
1526static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1527{
1528 struct sock *sk = sock->sk;
1529 int ret;
1530
1531 lock_sock(sk);
1532 ret = __tipc_sendstream(sock, m, dsz);
1533 release_sock(sk);
1534
1535 return ret;
1536}
1537
1538static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1539{
1540 struct sock *sk = sock->sk;
1541 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1542 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1543 struct sk_buff_head *txq = &sk->sk_write_queue;
1544 struct tipc_sock *tsk = tipc_sk(sk);
1545 struct tipc_msg *hdr = &tsk->phdr;
1546 struct net *net = sock_net(sk);
1547 struct sk_buff *skb;
1548 u32 dnode = tsk_peer_node(tsk);
1549 int maxnagle = tsk->maxnagle;
1550 int maxpkt = tsk->max_pkt;
1551 int send, sent = 0;
1552 int blocks, rc = 0;
1553
1554 if (unlikely(dlen > INT_MAX))
1555 return -EMSGSIZE;
1556
1557 /* Handle implicit connection setup */
1558 if (unlikely(dest)) {
1559 rc = __tipc_sendmsg(sock, m, dlen);
1560 if (dlen && dlen == rc) {
1561 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1562 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1563 }
1564 return rc;
1565 }
1566
1567 do {
1568 rc = tipc_wait_for_cond(sock, &timeout,
1569 (!tsk->cong_link_cnt &&
1570 !tsk_conn_cong(tsk) &&
1571 tipc_sk_connected(sk)));
1572 if (unlikely(rc))
1573 break;
1574 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1575 blocks = tsk->snd_backlog;
1576 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1577 send <= maxnagle) {
1578 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1579 if (unlikely(rc < 0))
1580 break;
1581 blocks += rc;
1582 tsk->msg_acc++;
1583 if (blocks <= 64 && tsk->expect_ack) {
1584 tsk->snd_backlog = blocks;
1585 sent += send;
1586 break;
1587 } else if (blocks > 64) {
1588 tsk->pkt_cnt += skb_queue_len(txq);
1589 } else {
1590 skb = skb_peek_tail(txq);
1591 if (skb) {
1592 msg_set_ack_required(buf_msg(skb));
1593 tsk->expect_ack = true;
1594 } else {
1595 tsk->expect_ack = false;
1596 }
1597 tsk->msg_acc = 0;
1598 tsk->pkt_cnt = 0;
1599 }
1600 } else {
1601 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1602 if (unlikely(rc != send))
1603 break;
1604 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1605 }
1606 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1607 TIPC_DUMP_SK_SNDQ, " ");
1608 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1609 if (unlikely(rc == -ELINKCONG)) {
1610 tsk->cong_link_cnt = 1;
1611 rc = 0;
1612 }
1613 if (likely(!rc)) {
1614 tsk->snt_unacked += blocks;
1615 tsk->snd_backlog = 0;
1616 sent += send;
1617 }
1618 } while (sent < dlen && !rc);
1619
1620 return sent ? sent : rc;
1621}
1622
1623/**
1624 * tipc_send_packet - send a connection-oriented message
1625 * @sock: socket structure
1626 * @m: message to send
1627 * @dsz: length of data to be transmitted
1628 *
1629 * Used for SOCK_SEQPACKET messages.
1630 *
1631 * Returns the number of bytes sent on success, or errno otherwise
1632 */
1633static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1634{
1635 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1636 return -EMSGSIZE;
1637
1638 return tipc_sendstream(sock, m, dsz);
1639}
1640
1641/* tipc_sk_finish_conn - complete the setup of a connection
1642 */
1643static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1644 u32 peer_node)
1645{
1646 struct sock *sk = &tsk->sk;
1647 struct net *net = sock_net(sk);
1648 struct tipc_msg *msg = &tsk->phdr;
1649
1650 msg_set_syn(msg, 0);
1651 msg_set_destnode(msg, peer_node);
1652 msg_set_destport(msg, peer_port);
1653 msg_set_type(msg, TIPC_CONN_MSG);
1654 msg_set_lookup_scope(msg, 0);
1655 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1656
1657 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1658 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1659 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1660 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1661 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1662 tsk_set_nagle(tsk);
1663 __skb_queue_purge(&sk->sk_write_queue);
1664 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1665 return;
1666
1667 /* Fall back to message based flow control */
1668 tsk->rcv_win = FLOWCTL_MSG_WIN;
1669 tsk->snd_win = FLOWCTL_MSG_WIN;
1670}
1671
1672/**
1673 * tipc_sk_set_orig_addr - capture sender's address for received message
1674 * @m: descriptor for message info
1675 * @skb: received message
1676 *
1677 * Note: Address is not captured if not requested by receiver.
1678 */
1679static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1680{
1681 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1682 struct tipc_msg *hdr = buf_msg(skb);
1683
1684 if (!srcaddr)
1685 return;
1686
1687 srcaddr->sock.family = AF_TIPC;
1688 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1689 srcaddr->sock.scope = 0;
1690 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1691 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1692 srcaddr->sock.addr.name.domain = 0;
1693 m->msg_namelen = sizeof(struct sockaddr_tipc);
1694
1695 if (!msg_in_group(hdr))
1696 return;
1697
1698 /* Group message users may also want to know sending member's id */
1699 srcaddr->member.family = AF_TIPC;
1700 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1701 srcaddr->member.scope = 0;
1702 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1703 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1704 srcaddr->member.addr.name.domain = 0;
1705 m->msg_namelen = sizeof(*srcaddr);
1706}
1707
1708/**
1709 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1710 * @m: descriptor for message info
1711 * @skb: received message buffer
1712 * @tsk: TIPC port associated with message
1713 *
1714 * Note: Ancillary data is not captured if not requested by receiver.
1715 *
1716 * Returns 0 if successful, otherwise errno
1717 */
1718static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1719 struct tipc_sock *tsk)
1720{
1721 struct tipc_msg *msg;
1722 u32 anc_data[3];
1723 u32 err;
1724 u32 dest_type;
1725 int has_name;
1726 int res;
1727
1728 if (likely(m->msg_controllen == 0))
1729 return 0;
1730 msg = buf_msg(skb);
1731
1732 /* Optionally capture errored message object(s) */
1733 err = msg ? msg_errcode(msg) : 0;
1734 if (unlikely(err)) {
1735 anc_data[0] = err;
1736 anc_data[1] = msg_data_sz(msg);
1737 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1738 if (res)
1739 return res;
1740 if (anc_data[1]) {
1741 if (skb_linearize(skb))
1742 return -ENOMEM;
1743 msg = buf_msg(skb);
1744 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1745 msg_data(msg));
1746 if (res)
1747 return res;
1748 }
1749 }
1750
1751 /* Optionally capture message destination object */
1752 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1753 switch (dest_type) {
1754 case TIPC_NAMED_MSG:
1755 has_name = 1;
1756 anc_data[0] = msg_nametype(msg);
1757 anc_data[1] = msg_namelower(msg);
1758 anc_data[2] = msg_namelower(msg);
1759 break;
1760 case TIPC_MCAST_MSG:
1761 has_name = 1;
1762 anc_data[0] = msg_nametype(msg);
1763 anc_data[1] = msg_namelower(msg);
1764 anc_data[2] = msg_nameupper(msg);
1765 break;
1766 case TIPC_CONN_MSG:
1767 has_name = (tsk->conn_type != 0);
1768 anc_data[0] = tsk->conn_type;
1769 anc_data[1] = tsk->conn_instance;
1770 anc_data[2] = tsk->conn_instance;
1771 break;
1772 default:
1773 has_name = 0;
1774 }
1775 if (has_name) {
1776 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1777 if (res)
1778 return res;
1779 }
1780
1781 return 0;
1782}
1783
1784static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1785{
1786 struct sock *sk = &tsk->sk;
1787 struct sk_buff *skb = NULL;
1788 struct tipc_msg *msg;
1789 u32 peer_port = tsk_peer_port(tsk);
1790 u32 dnode = tsk_peer_node(tsk);
1791
1792 if (!tipc_sk_connected(sk))
1793 return NULL;
1794 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1795 dnode, tsk_own_node(tsk), peer_port,
1796 tsk->portid, TIPC_OK);
1797 if (!skb)
1798 return NULL;
1799 msg = buf_msg(skb);
1800 msg_set_conn_ack(msg, tsk->rcv_unacked);
1801 tsk->rcv_unacked = 0;
1802
1803 /* Adjust to and advertize the correct window limit */
1804 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1805 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1806 msg_set_adv_win(msg, tsk->rcv_win);
1807 }
1808 return skb;
1809}
1810
1811static void tipc_sk_send_ack(struct tipc_sock *tsk)
1812{
1813 struct sk_buff *skb;
1814
1815 skb = tipc_sk_build_ack(tsk);
1816 if (!skb)
1817 return;
1818
1819 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1820 msg_link_selector(buf_msg(skb)));
1821}
1822
1823static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1824{
1825 struct sock *sk = sock->sk;
1826 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1827 long timeo = *timeop;
1828 int err = sock_error(sk);
1829
1830 if (err)
1831 return err;
1832
1833 for (;;) {
1834 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1835 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1836 err = -ENOTCONN;
1837 break;
1838 }
1839 add_wait_queue(sk_sleep(sk), &wait);
1840 release_sock(sk);
1841 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1842 sched_annotate_sleep();
1843 lock_sock(sk);
1844 remove_wait_queue(sk_sleep(sk), &wait);
1845 }
1846 err = 0;
1847 if (!skb_queue_empty(&sk->sk_receive_queue))
1848 break;
1849 err = -EAGAIN;
1850 if (!timeo)
1851 break;
1852 err = sock_intr_errno(timeo);
1853 if (signal_pending(current))
1854 break;
1855
1856 err = sock_error(sk);
1857 if (err)
1858 break;
1859 }
1860 *timeop = timeo;
1861 return err;
1862}
1863
1864/**
1865 * tipc_recvmsg - receive packet-oriented message
1866 * @m: descriptor for message info
1867 * @buflen: length of user buffer area
1868 * @flags: receive flags
1869 *
1870 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1871 * If the complete message doesn't fit in user area, truncate it.
1872 *
1873 * Returns size of returned message data, errno otherwise
1874 */
1875static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1876 size_t buflen, int flags)
1877{
1878 struct sock *sk = sock->sk;
1879 bool connected = !tipc_sk_type_connectionless(sk);
1880 struct tipc_sock *tsk = tipc_sk(sk);
1881 int rc, err, hlen, dlen, copy;
1882 struct sk_buff_head xmitq;
1883 struct tipc_msg *hdr;
1884 struct sk_buff *skb;
1885 bool grp_evt;
1886 long timeout;
1887
1888 /* Catch invalid receive requests */
1889 if (unlikely(!buflen))
1890 return -EINVAL;
1891
1892 lock_sock(sk);
1893 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1894 rc = -ENOTCONN;
1895 goto exit;
1896 }
1897 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1898
1899 /* Step rcv queue to first msg with data or error; wait if necessary */
1900 do {
1901 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1902 if (unlikely(rc))
1903 goto exit;
1904 skb = skb_peek(&sk->sk_receive_queue);
1905 hdr = buf_msg(skb);
1906 dlen = msg_data_sz(hdr);
1907 hlen = msg_hdr_sz(hdr);
1908 err = msg_errcode(hdr);
1909 grp_evt = msg_is_grp_evt(hdr);
1910 if (likely(dlen || err))
1911 break;
1912 tsk_advance_rx_queue(sk);
1913 } while (1);
1914
1915 /* Collect msg meta data, including error code and rejected data */
1916 tipc_sk_set_orig_addr(m, skb);
1917 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1918 if (unlikely(rc))
1919 goto exit;
1920 hdr = buf_msg(skb);
1921
1922 /* Capture data if non-error msg, otherwise just set return value */
1923 if (likely(!err)) {
1924 copy = min_t(int, dlen, buflen);
1925 if (unlikely(copy != dlen))
1926 m->msg_flags |= MSG_TRUNC;
1927 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1928 } else {
1929 copy = 0;
1930 rc = 0;
1931 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1932 rc = -ECONNRESET;
1933 }
1934 if (unlikely(rc))
1935 goto exit;
1936
1937 /* Mark message as group event if applicable */
1938 if (unlikely(grp_evt)) {
1939 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1940 m->msg_flags |= MSG_EOR;
1941 m->msg_flags |= MSG_OOB;
1942 copy = 0;
1943 }
1944
1945 /* Caption of data or error code/rejected data was successful */
1946 if (unlikely(flags & MSG_PEEK))
1947 goto exit;
1948
1949 /* Send group flow control advertisement when applicable */
1950 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1951 __skb_queue_head_init(&xmitq);
1952 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1953 msg_orignode(hdr), msg_origport(hdr),
1954 &xmitq);
1955 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1956 }
1957
1958 tsk_advance_rx_queue(sk);
1959
1960 if (likely(!connected))
1961 goto exit;
1962
1963 /* Send connection flow control advertisement when applicable */
1964 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1965 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1966 tipc_sk_send_ack(tsk);
1967exit:
1968 release_sock(sk);
1969 return rc ? rc : copy;
1970}
1971
1972/**
1973 * tipc_recvstream - receive stream-oriented data
1974 * @m: descriptor for message info
1975 * @buflen: total size of user buffer area
1976 * @flags: receive flags
1977 *
1978 * Used for SOCK_STREAM messages only. If not enough data is available
1979 * will optionally wait for more; never truncates data.
1980 *
1981 * Returns size of returned message data, errno otherwise
1982 */
1983static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1984 size_t buflen, int flags)
1985{
1986 struct sock *sk = sock->sk;
1987 struct tipc_sock *tsk = tipc_sk(sk);
1988 struct sk_buff *skb;
1989 struct tipc_msg *hdr;
1990 struct tipc_skb_cb *skb_cb;
1991 bool peek = flags & MSG_PEEK;
1992 int offset, required, copy, copied = 0;
1993 int hlen, dlen, err, rc;
1994 long timeout;
1995
1996 /* Catch invalid receive attempts */
1997 if (unlikely(!buflen))
1998 return -EINVAL;
1999
2000 lock_sock(sk);
2001
2002 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2003 rc = -ENOTCONN;
2004 goto exit;
2005 }
2006 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2007 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2008
2009 do {
2010 /* Look at first msg in receive queue; wait if necessary */
2011 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2012 if (unlikely(rc))
2013 break;
2014 skb = skb_peek(&sk->sk_receive_queue);
2015 skb_cb = TIPC_SKB_CB(skb);
2016 hdr = buf_msg(skb);
2017 dlen = msg_data_sz(hdr);
2018 hlen = msg_hdr_sz(hdr);
2019 err = msg_errcode(hdr);
2020
2021 /* Discard any empty non-errored (SYN-) message */
2022 if (unlikely(!dlen && !err)) {
2023 tsk_advance_rx_queue(sk);
2024 continue;
2025 }
2026
2027 /* Collect msg meta data, incl. error code and rejected data */
2028 if (!copied) {
2029 tipc_sk_set_orig_addr(m, skb);
2030 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2031 if (rc)
2032 break;
2033 hdr = buf_msg(skb);
2034 }
2035
2036 /* Copy data if msg ok, otherwise return error/partial data */
2037 if (likely(!err)) {
2038 offset = skb_cb->bytes_read;
2039 copy = min_t(int, dlen - offset, buflen - copied);
2040 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2041 if (unlikely(rc))
2042 break;
2043 copied += copy;
2044 offset += copy;
2045 if (unlikely(offset < dlen)) {
2046 if (!peek)
2047 skb_cb->bytes_read = offset;
2048 break;
2049 }
2050 } else {
2051 rc = 0;
2052 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2053 rc = -ECONNRESET;
2054 if (copied || rc)
2055 break;
2056 }
2057
2058 if (unlikely(peek))
2059 break;
2060
2061 tsk_advance_rx_queue(sk);
2062
2063 /* Send connection flow control advertisement when applicable */
2064 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2065 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2066 tipc_sk_send_ack(tsk);
2067
2068 /* Exit if all requested data or FIN/error received */
2069 if (copied == buflen || err)
2070 break;
2071
2072 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2073exit:
2074 release_sock(sk);
2075 return copied ? copied : rc;
2076}
2077
2078/**
2079 * tipc_write_space - wake up thread if port congestion is released
2080 * @sk: socket
2081 */
2082static void tipc_write_space(struct sock *sk)
2083{
2084 struct socket_wq *wq;
2085
2086 rcu_read_lock();
2087 wq = rcu_dereference(sk->sk_wq);
2088 if (skwq_has_sleeper(wq))
2089 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2090 EPOLLWRNORM | EPOLLWRBAND);
2091 rcu_read_unlock();
2092}
2093
2094/**
2095 * tipc_data_ready - wake up threads to indicate messages have been received
2096 * @sk: socket
2097 */
2098static void tipc_data_ready(struct sock *sk)
2099{
2100 struct socket_wq *wq;
2101
2102 rcu_read_lock();
2103 wq = rcu_dereference(sk->sk_wq);
2104 if (skwq_has_sleeper(wq))
2105 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2106 EPOLLRDNORM | EPOLLRDBAND);
2107 rcu_read_unlock();
2108}
2109
2110static void tipc_sock_destruct(struct sock *sk)
2111{
2112 __skb_queue_purge(&sk->sk_receive_queue);
2113}
2114
2115static void tipc_sk_proto_rcv(struct sock *sk,
2116 struct sk_buff_head *inputq,
2117 struct sk_buff_head *xmitq)
2118{
2119 struct sk_buff *skb = __skb_dequeue(inputq);
2120 struct tipc_sock *tsk = tipc_sk(sk);
2121 struct tipc_msg *hdr = buf_msg(skb);
2122 struct tipc_group *grp = tsk->group;
2123 bool wakeup = false;
2124
2125 switch (msg_user(hdr)) {
2126 case CONN_MANAGER:
2127 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2128 return;
2129 case SOCK_WAKEUP:
2130 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2131 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2132 smp_wmb();
2133 tsk->cong_link_cnt--;
2134 wakeup = true;
2135 tipc_sk_push_backlog(tsk, false);
2136 break;
2137 case GROUP_PROTOCOL:
2138 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2139 break;
2140 case TOP_SRV:
2141 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2142 hdr, inputq, xmitq);
2143 break;
2144 default:
2145 break;
2146 }
2147
2148 if (wakeup)
2149 sk->sk_write_space(sk);
2150
2151 kfree_skb(skb);
2152}
2153
2154/**
2155 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2156 * @tsk: TIPC socket
2157 * @skb: pointer to message buffer.
2158 * @xmitq: for Nagle ACK if any
2159 * Returns true if message should be added to receive queue, false otherwise
2160 */
2161static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2162 struct sk_buff_head *xmitq)
2163{
2164 struct sock *sk = &tsk->sk;
2165 struct net *net = sock_net(sk);
2166 struct tipc_msg *hdr = buf_msg(skb);
2167 bool con_msg = msg_connected(hdr);
2168 u32 pport = tsk_peer_port(tsk);
2169 u32 pnode = tsk_peer_node(tsk);
2170 u32 oport = msg_origport(hdr);
2171 u32 onode = msg_orignode(hdr);
2172 int err = msg_errcode(hdr);
2173 unsigned long delay;
2174
2175 if (unlikely(msg_mcast(hdr)))
2176 return false;
2177 tsk->oneway = 0;
2178
2179 switch (sk->sk_state) {
2180 case TIPC_CONNECTING:
2181 /* Setup ACK */
2182 if (likely(con_msg)) {
2183 if (err)
2184 break;
2185 tipc_sk_finish_conn(tsk, oport, onode);
2186 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2187 /* ACK+ message with data is added to receive queue */
2188 if (msg_data_sz(hdr))
2189 return true;
2190 /* Empty ACK-, - wake up sleeping connect() and drop */
2191 sk->sk_state_change(sk);
2192 msg_set_dest_droppable(hdr, 1);
2193 return false;
2194 }
2195 /* Ignore connectionless message if not from listening socket */
2196 if (oport != pport || onode != pnode)
2197 return false;
2198
2199 /* Rejected SYN */
2200 if (err != TIPC_ERR_OVERLOAD)
2201 break;
2202
2203 /* Prepare for new setup attempt if we have a SYN clone */
2204 if (skb_queue_empty(&sk->sk_write_queue))
2205 break;
2206 get_random_bytes(&delay, 2);
2207 delay %= (tsk->conn_timeout / 4);
2208 delay = msecs_to_jiffies(delay + 100);
2209 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2210 return false;
2211 case TIPC_OPEN:
2212 case TIPC_DISCONNECTING:
2213 return false;
2214 case TIPC_LISTEN:
2215 /* Accept only SYN message */
2216 if (!msg_is_syn(hdr) &&
2217 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2218 return false;
2219 if (!con_msg && !err)
2220 return true;
2221 return false;
2222 case TIPC_ESTABLISHED:
2223 if (!skb_queue_empty(&sk->sk_write_queue))
2224 tipc_sk_push_backlog(tsk, false);
2225 /* Accept only connection-based messages sent by peer */
2226 if (likely(con_msg && !err && pport == oport &&
2227 pnode == onode)) {
2228 if (msg_ack_required(hdr)) {
2229 struct sk_buff *skb;
2230
2231 skb = tipc_sk_build_ack(tsk);
2232 if (skb) {
2233 msg_set_nagle_ack(buf_msg(skb));
2234 __skb_queue_tail(xmitq, skb);
2235 }
2236 }
2237 return true;
2238 }
2239 if (!tsk_peer_msg(tsk, hdr))
2240 return false;
2241 if (!err)
2242 return true;
2243 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2244 tipc_node_remove_conn(net, pnode, tsk->portid);
2245 sk->sk_state_change(sk);
2246 return true;
2247 default:
2248 pr_err("Unknown sk_state %u\n", sk->sk_state);
2249 }
2250 /* Abort connection setup attempt */
2251 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2252 sk->sk_err = ECONNREFUSED;
2253 sk->sk_state_change(sk);
2254 return true;
2255}
2256
2257/**
2258 * rcvbuf_limit - get proper overload limit of socket receive queue
2259 * @sk: socket
2260 * @skb: message
2261 *
2262 * For connection oriented messages, irrespective of importance,
2263 * default queue limit is 2 MB.
2264 *
2265 * For connectionless messages, queue limits are based on message
2266 * importance as follows:
2267 *
2268 * TIPC_LOW_IMPORTANCE (2 MB)
2269 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2270 * TIPC_HIGH_IMPORTANCE (8 MB)
2271 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2272 *
2273 * Returns overload limit according to corresponding message importance
2274 */
2275static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2276{
2277 struct tipc_sock *tsk = tipc_sk(sk);
2278 struct tipc_msg *hdr = buf_msg(skb);
2279
2280 if (unlikely(msg_in_group(hdr)))
2281 return READ_ONCE(sk->sk_rcvbuf);
2282
2283 if (unlikely(!msg_connected(hdr)))
2284 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2285
2286 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2287 return READ_ONCE(sk->sk_rcvbuf);
2288
2289 return FLOWCTL_MSG_LIM;
2290}
2291
2292/**
2293 * tipc_sk_filter_rcv - validate incoming message
2294 * @sk: socket
2295 * @skb: pointer to message.
2296 *
2297 * Enqueues message on receive queue if acceptable; optionally handles
2298 * disconnect indication for a connected socket.
2299 *
2300 * Called with socket lock already taken
2301 *
2302 */
2303static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2304 struct sk_buff_head *xmitq)
2305{
2306 bool sk_conn = !tipc_sk_type_connectionless(sk);
2307 struct tipc_sock *tsk = tipc_sk(sk);
2308 struct tipc_group *grp = tsk->group;
2309 struct tipc_msg *hdr = buf_msg(skb);
2310 struct net *net = sock_net(sk);
2311 struct sk_buff_head inputq;
2312 int mtyp = msg_type(hdr);
2313 int limit, err = TIPC_OK;
2314
2315 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2316 TIPC_SKB_CB(skb)->bytes_read = 0;
2317 __skb_queue_head_init(&inputq);
2318 __skb_queue_tail(&inputq, skb);
2319
2320 if (unlikely(!msg_isdata(hdr)))
2321 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2322
2323 if (unlikely(grp))
2324 tipc_group_filter_msg(grp, &inputq, xmitq);
2325
2326 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2327 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2328
2329 /* Validate and add to receive buffer if there is space */
2330 while ((skb = __skb_dequeue(&inputq))) {
2331 hdr = buf_msg(skb);
2332 limit = rcvbuf_limit(sk, skb);
2333 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2334 (!sk_conn && msg_connected(hdr)) ||
2335 (!grp && msg_in_group(hdr)))
2336 err = TIPC_ERR_NO_PORT;
2337 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2338 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2339 "err_overload2!");
2340 atomic_inc(&sk->sk_drops);
2341 err = TIPC_ERR_OVERLOAD;
2342 }
2343
2344 if (unlikely(err)) {
2345 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2346 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2347 "@filter_rcv!");
2348 __skb_queue_tail(xmitq, skb);
2349 }
2350 err = TIPC_OK;
2351 continue;
2352 }
2353 __skb_queue_tail(&sk->sk_receive_queue, skb);
2354 skb_set_owner_r(skb, sk);
2355 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2356 "rcvq >90% allocated!");
2357 sk->sk_data_ready(sk);
2358 }
2359}
2360
2361/**
2362 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2363 * @sk: socket
2364 * @skb: message
2365 *
2366 * Caller must hold socket lock
2367 */
2368static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2369{
2370 unsigned int before = sk_rmem_alloc_get(sk);
2371 struct sk_buff_head xmitq;
2372 unsigned int added;
2373
2374 __skb_queue_head_init(&xmitq);
2375
2376 tipc_sk_filter_rcv(sk, skb, &xmitq);
2377 added = sk_rmem_alloc_get(sk) - before;
2378 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2379
2380 /* Send pending response/rejected messages, if any */
2381 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2382 return 0;
2383}
2384
2385/**
2386 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2387 * inputq and try adding them to socket or backlog queue
2388 * @inputq: list of incoming buffers with potentially different destinations
2389 * @sk: socket where the buffers should be enqueued
2390 * @dport: port number for the socket
2391 *
2392 * Caller must hold socket lock
2393 */
2394static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2395 u32 dport, struct sk_buff_head *xmitq)
2396{
2397 unsigned long time_limit = jiffies + 2;
2398 struct sk_buff *skb;
2399 unsigned int lim;
2400 atomic_t *dcnt;
2401 u32 onode;
2402
2403 while (skb_queue_len(inputq)) {
2404 if (unlikely(time_after_eq(jiffies, time_limit)))
2405 return;
2406
2407 skb = tipc_skb_dequeue(inputq, dport);
2408 if (unlikely(!skb))
2409 return;
2410
2411 /* Add message directly to receive queue if possible */
2412 if (!sock_owned_by_user(sk)) {
2413 tipc_sk_filter_rcv(sk, skb, xmitq);
2414 continue;
2415 }
2416
2417 /* Try backlog, compensating for double-counted bytes */
2418 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2419 if (!sk->sk_backlog.len)
2420 atomic_set(dcnt, 0);
2421 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2422 if (likely(!sk_add_backlog(sk, skb, lim))) {
2423 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2424 "bklg & rcvq >90% allocated!");
2425 continue;
2426 }
2427
2428 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2429 /* Overload => reject message back to sender */
2430 onode = tipc_own_addr(sock_net(sk));
2431 atomic_inc(&sk->sk_drops);
2432 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2433 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2434 "@sk_enqueue!");
2435 __skb_queue_tail(xmitq, skb);
2436 }
2437 break;
2438 }
2439}
2440
2441/**
2442 * tipc_sk_rcv - handle a chain of incoming buffers
2443 * @inputq: buffer list containing the buffers
2444 * Consumes all buffers in list until inputq is empty
2445 * Note: may be called in multiple threads referring to the same queue
2446 */
2447void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2448{
2449 struct sk_buff_head xmitq;
2450 u32 dnode, dport = 0;
2451 int err;
2452 struct tipc_sock *tsk;
2453 struct sock *sk;
2454 struct sk_buff *skb;
2455
2456 __skb_queue_head_init(&xmitq);
2457 while (skb_queue_len(inputq)) {
2458 dport = tipc_skb_peek_port(inputq, dport);
2459 tsk = tipc_sk_lookup(net, dport);
2460
2461 if (likely(tsk)) {
2462 sk = &tsk->sk;
2463 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2464 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2465 spin_unlock_bh(&sk->sk_lock.slock);
2466 }
2467 /* Send pending response/rejected messages, if any */
2468 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2469 sock_put(sk);
2470 continue;
2471 }
2472 /* No destination socket => dequeue skb if still there */
2473 skb = tipc_skb_dequeue(inputq, dport);
2474 if (!skb)
2475 return;
2476
2477 /* Try secondary lookup if unresolved named message */
2478 err = TIPC_ERR_NO_PORT;
2479 if (tipc_msg_lookup_dest(net, skb, &err))
2480 goto xmit;
2481
2482 /* Prepare for message rejection */
2483 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2484 continue;
2485
2486 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2487xmit:
2488 dnode = msg_destnode(buf_msg(skb));
2489 tipc_node_xmit_skb(net, skb, dnode, dport);
2490 }
2491}
2492
2493static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2494{
2495 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2496 struct sock *sk = sock->sk;
2497 int done;
2498
2499 do {
2500 int err = sock_error(sk);
2501 if (err)
2502 return err;
2503 if (!*timeo_p)
2504 return -ETIMEDOUT;
2505 if (signal_pending(current))
2506 return sock_intr_errno(*timeo_p);
2507 if (sk->sk_state == TIPC_DISCONNECTING)
2508 break;
2509
2510 add_wait_queue(sk_sleep(sk), &wait);
2511 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2512 &wait);
2513 remove_wait_queue(sk_sleep(sk), &wait);
2514 } while (!done);
2515 return 0;
2516}
2517
2518static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2519{
2520 if (addr->family != AF_TIPC)
2521 return false;
2522 if (addr->addrtype == TIPC_SERVICE_RANGE)
2523 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2524 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2525 addr->addrtype == TIPC_SOCKET_ADDR);
2526}
2527
2528/**
2529 * tipc_connect - establish a connection to another TIPC port
2530 * @sock: socket structure
2531 * @dest: socket address for destination port
2532 * @destlen: size of socket address data structure
2533 * @flags: file-related flags associated with socket
2534 *
2535 * Returns 0 on success, errno otherwise
2536 */
2537static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2538 int destlen, int flags)
2539{
2540 struct sock *sk = sock->sk;
2541 struct tipc_sock *tsk = tipc_sk(sk);
2542 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2543 struct msghdr m = {NULL,};
2544 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2545 int previous;
2546 int res = 0;
2547
2548 if (destlen != sizeof(struct sockaddr_tipc))
2549 return -EINVAL;
2550
2551 lock_sock(sk);
2552
2553 if (tsk->group) {
2554 res = -EINVAL;
2555 goto exit;
2556 }
2557
2558 if (dst->family == AF_UNSPEC) {
2559 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2560 if (!tipc_sk_type_connectionless(sk))
2561 res = -EINVAL;
2562 goto exit;
2563 }
2564 if (!tipc_sockaddr_is_sane(dst)) {
2565 res = -EINVAL;
2566 goto exit;
2567 }
2568 /* DGRAM/RDM connect(), just save the destaddr */
2569 if (tipc_sk_type_connectionless(sk)) {
2570 memcpy(&tsk->peer, dest, destlen);
2571 goto exit;
2572 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2573 res = -EINVAL;
2574 goto exit;
2575 }
2576
2577 previous = sk->sk_state;
2578
2579 switch (sk->sk_state) {
2580 case TIPC_OPEN:
2581 /* Send a 'SYN-' to destination */
2582 m.msg_name = dest;
2583 m.msg_namelen = destlen;
2584
2585 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2586 * indicate send_msg() is never blocked.
2587 */
2588 if (!timeout)
2589 m.msg_flags = MSG_DONTWAIT;
2590
2591 res = __tipc_sendmsg(sock, &m, 0);
2592 if ((res < 0) && (res != -EWOULDBLOCK))
2593 goto exit;
2594
2595 /* Just entered TIPC_CONNECTING state; the only
2596 * difference is that return value in non-blocking
2597 * case is EINPROGRESS, rather than EALREADY.
2598 */
2599 res = -EINPROGRESS;
2600 fallthrough;
2601 case TIPC_CONNECTING:
2602 if (!timeout) {
2603 if (previous == TIPC_CONNECTING)
2604 res = -EALREADY;
2605 goto exit;
2606 }
2607 timeout = msecs_to_jiffies(timeout);
2608 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2609 res = tipc_wait_for_connect(sock, &timeout);
2610 break;
2611 case TIPC_ESTABLISHED:
2612 res = -EISCONN;
2613 break;
2614 default:
2615 res = -EINVAL;
2616 }
2617
2618exit:
2619 release_sock(sk);
2620 return res;
2621}
2622
2623/**
2624 * tipc_listen - allow socket to listen for incoming connections
2625 * @sock: socket structure
2626 * @len: (unused)
2627 *
2628 * Returns 0 on success, errno otherwise
2629 */
2630static int tipc_listen(struct socket *sock, int len)
2631{
2632 struct sock *sk = sock->sk;
2633 int res;
2634
2635 lock_sock(sk);
2636 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2637 release_sock(sk);
2638
2639 return res;
2640}
2641
2642static int tipc_wait_for_accept(struct socket *sock, long timeo)
2643{
2644 struct sock *sk = sock->sk;
2645 DEFINE_WAIT(wait);
2646 int err;
2647
2648 /* True wake-one mechanism for incoming connections: only
2649 * one process gets woken up, not the 'whole herd'.
2650 * Since we do not 'race & poll' for established sockets
2651 * anymore, the common case will execute the loop only once.
2652 */
2653 for (;;) {
2654 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2655 TASK_INTERRUPTIBLE);
2656 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2657 release_sock(sk);
2658 timeo = schedule_timeout(timeo);
2659 lock_sock(sk);
2660 }
2661 err = 0;
2662 if (!skb_queue_empty(&sk->sk_receive_queue))
2663 break;
2664 err = -EAGAIN;
2665 if (!timeo)
2666 break;
2667 err = sock_intr_errno(timeo);
2668 if (signal_pending(current))
2669 break;
2670 }
2671 finish_wait(sk_sleep(sk), &wait);
2672 return err;
2673}
2674
2675/**
2676 * tipc_accept - wait for connection request
2677 * @sock: listening socket
2678 * @new_sock: new socket that is to be connected
2679 * @flags: file-related flags associated with socket
2680 *
2681 * Returns 0 on success, errno otherwise
2682 */
2683static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2684 bool kern)
2685{
2686 struct sock *new_sk, *sk = sock->sk;
2687 struct sk_buff *buf;
2688 struct tipc_sock *new_tsock;
2689 struct tipc_msg *msg;
2690 long timeo;
2691 int res;
2692
2693 lock_sock(sk);
2694
2695 if (sk->sk_state != TIPC_LISTEN) {
2696 res = -EINVAL;
2697 goto exit;
2698 }
2699 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2700 res = tipc_wait_for_accept(sock, timeo);
2701 if (res)
2702 goto exit;
2703
2704 buf = skb_peek(&sk->sk_receive_queue);
2705
2706 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2707 if (res)
2708 goto exit;
2709 security_sk_clone(sock->sk, new_sock->sk);
2710
2711 new_sk = new_sock->sk;
2712 new_tsock = tipc_sk(new_sk);
2713 msg = buf_msg(buf);
2714
2715 /* we lock on new_sk; but lockdep sees the lock on sk */
2716 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2717
2718 /*
2719 * Reject any stray messages received by new socket
2720 * before the socket lock was taken (very, very unlikely)
2721 */
2722 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2723
2724 /* Connect new socket to it's peer */
2725 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2726
2727 tsk_set_importance(new_sk, msg_importance(msg));
2728 if (msg_named(msg)) {
2729 new_tsock->conn_type = msg_nametype(msg);
2730 new_tsock->conn_instance = msg_nameinst(msg);
2731 }
2732
2733 /*
2734 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2735 * Respond to 'SYN+' by queuing it on new socket.
2736 */
2737 if (!msg_data_sz(msg)) {
2738 struct msghdr m = {NULL,};
2739
2740 tsk_advance_rx_queue(sk);
2741 __tipc_sendstream(new_sock, &m, 0);
2742 } else {
2743 __skb_dequeue(&sk->sk_receive_queue);
2744 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2745 skb_set_owner_r(buf, new_sk);
2746 }
2747 release_sock(new_sk);
2748exit:
2749 release_sock(sk);
2750 return res;
2751}
2752
2753/**
2754 * tipc_shutdown - shutdown socket connection
2755 * @sock: socket structure
2756 * @how: direction to close (must be SHUT_RDWR)
2757 *
2758 * Terminates connection (if necessary), then purges socket's receive queue.
2759 *
2760 * Returns 0 on success, errno otherwise
2761 */
2762static int tipc_shutdown(struct socket *sock, int how)
2763{
2764 struct sock *sk = sock->sk;
2765 int res;
2766
2767 if (how != SHUT_RDWR)
2768 return -EINVAL;
2769
2770 lock_sock(sk);
2771
2772 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2773 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2774 sk->sk_shutdown = SHUTDOWN_MASK;
2775
2776 if (sk->sk_state == TIPC_DISCONNECTING) {
2777 /* Discard any unreceived messages */
2778 __skb_queue_purge(&sk->sk_receive_queue);
2779
2780 res = 0;
2781 } else {
2782 res = -ENOTCONN;
2783 }
2784 /* Wake up anyone sleeping in poll. */
2785 sk->sk_state_change(sk);
2786
2787 release_sock(sk);
2788 return res;
2789}
2790
2791static void tipc_sk_check_probing_state(struct sock *sk,
2792 struct sk_buff_head *list)
2793{
2794 struct tipc_sock *tsk = tipc_sk(sk);
2795 u32 pnode = tsk_peer_node(tsk);
2796 u32 pport = tsk_peer_port(tsk);
2797 u32 self = tsk_own_node(tsk);
2798 u32 oport = tsk->portid;
2799 struct sk_buff *skb;
2800
2801 if (tsk->probe_unacked) {
2802 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2803 sk->sk_err = ECONNABORTED;
2804 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2805 sk->sk_state_change(sk);
2806 return;
2807 }
2808 /* Prepare new probe */
2809 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2810 pnode, self, pport, oport, TIPC_OK);
2811 if (skb)
2812 __skb_queue_tail(list, skb);
2813 tsk->probe_unacked = true;
2814 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2815}
2816
2817static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2818{
2819 struct tipc_sock *tsk = tipc_sk(sk);
2820
2821 /* Try again later if dest link is congested */
2822 if (tsk->cong_link_cnt) {
2823 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2824 return;
2825 }
2826 /* Prepare SYN for retransmit */
2827 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2828}
2829
2830static void tipc_sk_timeout(struct timer_list *t)
2831{
2832 struct sock *sk = from_timer(sk, t, sk_timer);
2833 struct tipc_sock *tsk = tipc_sk(sk);
2834 u32 pnode = tsk_peer_node(tsk);
2835 struct sk_buff_head list;
2836 int rc = 0;
2837
2838 __skb_queue_head_init(&list);
2839 bh_lock_sock(sk);
2840
2841 /* Try again later if socket is busy */
2842 if (sock_owned_by_user(sk)) {
2843 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2844 bh_unlock_sock(sk);
2845 sock_put(sk);
2846 return;
2847 }
2848
2849 if (sk->sk_state == TIPC_ESTABLISHED)
2850 tipc_sk_check_probing_state(sk, &list);
2851 else if (sk->sk_state == TIPC_CONNECTING)
2852 tipc_sk_retry_connect(sk, &list);
2853
2854 bh_unlock_sock(sk);
2855
2856 if (!skb_queue_empty(&list))
2857 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2858
2859 /* SYN messages may cause link congestion */
2860 if (rc == -ELINKCONG) {
2861 tipc_dest_push(&tsk->cong_links, pnode, 0);
2862 tsk->cong_link_cnt = 1;
2863 }
2864 sock_put(sk);
2865}
2866
2867static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2868 struct tipc_name_seq const *seq)
2869{
2870 struct sock *sk = &tsk->sk;
2871 struct net *net = sock_net(sk);
2872 struct publication *publ;
2873 u32 key;
2874
2875 if (scope != TIPC_NODE_SCOPE)
2876 scope = TIPC_CLUSTER_SCOPE;
2877
2878 if (tipc_sk_connected(sk))
2879 return -EINVAL;
2880 key = tsk->portid + tsk->pub_count + 1;
2881 if (key == tsk->portid)
2882 return -EADDRINUSE;
2883
2884 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2885 scope, tsk->portid, key);
2886 if (unlikely(!publ))
2887 return -EINVAL;
2888
2889 list_add(&publ->binding_sock, &tsk->publications);
2890 tsk->pub_count++;
2891 tsk->published = 1;
2892 return 0;
2893}
2894
2895static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2896 struct tipc_name_seq const *seq)
2897{
2898 struct net *net = sock_net(&tsk->sk);
2899 struct publication *publ;
2900 struct publication *safe;
2901 int rc = -EINVAL;
2902
2903 if (scope != TIPC_NODE_SCOPE)
2904 scope = TIPC_CLUSTER_SCOPE;
2905
2906 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2907 if (seq) {
2908 if (publ->scope != scope)
2909 continue;
2910 if (publ->type != seq->type)
2911 continue;
2912 if (publ->lower != seq->lower)
2913 continue;
2914 if (publ->upper != seq->upper)
2915 break;
2916 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2917 publ->upper, publ->key);
2918 rc = 0;
2919 break;
2920 }
2921 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2922 publ->upper, publ->key);
2923 rc = 0;
2924 }
2925 if (list_empty(&tsk->publications))
2926 tsk->published = 0;
2927 return rc;
2928}
2929
2930/* tipc_sk_reinit: set non-zero address in all existing sockets
2931 * when we go from standalone to network mode.
2932 */
2933void tipc_sk_reinit(struct net *net)
2934{
2935 struct tipc_net *tn = net_generic(net, tipc_net_id);
2936 struct rhashtable_iter iter;
2937 struct tipc_sock *tsk;
2938 struct tipc_msg *msg;
2939
2940 rhashtable_walk_enter(&tn->sk_rht, &iter);
2941
2942 do {
2943 rhashtable_walk_start(&iter);
2944
2945 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2946 sock_hold(&tsk->sk);
2947 rhashtable_walk_stop(&iter);
2948 lock_sock(&tsk->sk);
2949 msg = &tsk->phdr;
2950 msg_set_prevnode(msg, tipc_own_addr(net));
2951 msg_set_orignode(msg, tipc_own_addr(net));
2952 release_sock(&tsk->sk);
2953 rhashtable_walk_start(&iter);
2954 sock_put(&tsk->sk);
2955 }
2956
2957 rhashtable_walk_stop(&iter);
2958 } while (tsk == ERR_PTR(-EAGAIN));
2959
2960 rhashtable_walk_exit(&iter);
2961}
2962
2963static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2964{
2965 struct tipc_net *tn = net_generic(net, tipc_net_id);
2966 struct tipc_sock *tsk;
2967
2968 rcu_read_lock();
2969 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2970 if (tsk)
2971 sock_hold(&tsk->sk);
2972 rcu_read_unlock();
2973
2974 return tsk;
2975}
2976
2977static int tipc_sk_insert(struct tipc_sock *tsk)
2978{
2979 struct sock *sk = &tsk->sk;
2980 struct net *net = sock_net(sk);
2981 struct tipc_net *tn = net_generic(net, tipc_net_id);
2982 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2983 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2984
2985 while (remaining--) {
2986 portid++;
2987 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2988 portid = TIPC_MIN_PORT;
2989 tsk->portid = portid;
2990 sock_hold(&tsk->sk);
2991 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2992 tsk_rht_params))
2993 return 0;
2994 sock_put(&tsk->sk);
2995 }
2996
2997 return -1;
2998}
2999
3000static void tipc_sk_remove(struct tipc_sock *tsk)
3001{
3002 struct sock *sk = &tsk->sk;
3003 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3004
3005 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3006 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3007 __sock_put(sk);
3008 }
3009}
3010
3011static const struct rhashtable_params tsk_rht_params = {
3012 .nelem_hint = 192,
3013 .head_offset = offsetof(struct tipc_sock, node),
3014 .key_offset = offsetof(struct tipc_sock, portid),
3015 .key_len = sizeof(u32), /* portid */
3016 .max_size = 1048576,
3017 .min_size = 256,
3018 .automatic_shrinking = true,
3019};
3020
3021int tipc_sk_rht_init(struct net *net)
3022{
3023 struct tipc_net *tn = net_generic(net, tipc_net_id);
3024
3025 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3026}
3027
3028void tipc_sk_rht_destroy(struct net *net)
3029{
3030 struct tipc_net *tn = net_generic(net, tipc_net_id);
3031
3032 /* Wait for socket readers to complete */
3033 synchronize_net();
3034
3035 rhashtable_destroy(&tn->sk_rht);
3036}
3037
3038static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3039{
3040 struct net *net = sock_net(&tsk->sk);
3041 struct tipc_group *grp = tsk->group;
3042 struct tipc_msg *hdr = &tsk->phdr;
3043 struct tipc_name_seq seq;
3044 int rc;
3045
3046 if (mreq->type < TIPC_RESERVED_TYPES)
3047 return -EACCES;
3048 if (mreq->scope > TIPC_NODE_SCOPE)
3049 return -EINVAL;
3050 if (grp)
3051 return -EACCES;
3052 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3053 if (!grp)
3054 return -ENOMEM;
3055 tsk->group = grp;
3056 msg_set_lookup_scope(hdr, mreq->scope);
3057 msg_set_nametype(hdr, mreq->type);
3058 msg_set_dest_droppable(hdr, true);
3059 seq.type = mreq->type;
3060 seq.lower = mreq->instance;
3061 seq.upper = seq.lower;
3062 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3063 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3064 if (rc) {
3065 tipc_group_delete(net, grp);
3066 tsk->group = NULL;
3067 return rc;
3068 }
3069 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3070 tsk->mc_method.rcast = true;
3071 tsk->mc_method.mandatory = true;
3072 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3073 return rc;
3074}
3075
3076static int tipc_sk_leave(struct tipc_sock *tsk)
3077{
3078 struct net *net = sock_net(&tsk->sk);
3079 struct tipc_group *grp = tsk->group;
3080 struct tipc_name_seq seq;
3081 int scope;
3082
3083 if (!grp)
3084 return -EINVAL;
3085 tipc_group_self(grp, &seq, &scope);
3086 tipc_group_delete(net, grp);
3087 tsk->group = NULL;
3088 tipc_sk_withdraw(tsk, scope, &seq);
3089 return 0;
3090}
3091
3092/**
3093 * tipc_setsockopt - set socket option
3094 * @sock: socket structure
3095 * @lvl: option level
3096 * @opt: option identifier
3097 * @ov: pointer to new option value
3098 * @ol: length of option value
3099 *
3100 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3101 * (to ease compatibility).
3102 *
3103 * Returns 0 on success, errno otherwise
3104 */
3105static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3106 sockptr_t ov, unsigned int ol)
3107{
3108 struct sock *sk = sock->sk;
3109 struct tipc_sock *tsk = tipc_sk(sk);
3110 struct tipc_group_req mreq;
3111 u32 value = 0;
3112 int res = 0;
3113
3114 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3115 return 0;
3116 if (lvl != SOL_TIPC)
3117 return -ENOPROTOOPT;
3118
3119 switch (opt) {
3120 case TIPC_IMPORTANCE:
3121 case TIPC_SRC_DROPPABLE:
3122 case TIPC_DEST_DROPPABLE:
3123 case TIPC_CONN_TIMEOUT:
3124 case TIPC_NODELAY:
3125 if (ol < sizeof(value))
3126 return -EINVAL;
3127 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3128 return -EFAULT;
3129 break;
3130 case TIPC_GROUP_JOIN:
3131 if (ol < sizeof(mreq))
3132 return -EINVAL;
3133 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3134 return -EFAULT;
3135 break;
3136 default:
3137 if (!sockptr_is_null(ov) || ol)
3138 return -EINVAL;
3139 }
3140
3141 lock_sock(sk);
3142
3143 switch (opt) {
3144 case TIPC_IMPORTANCE:
3145 res = tsk_set_importance(sk, value);
3146 break;
3147 case TIPC_SRC_DROPPABLE:
3148 if (sock->type != SOCK_STREAM)
3149 tsk_set_unreliable(tsk, value);
3150 else
3151 res = -ENOPROTOOPT;
3152 break;
3153 case TIPC_DEST_DROPPABLE:
3154 tsk_set_unreturnable(tsk, value);
3155 break;
3156 case TIPC_CONN_TIMEOUT:
3157 tipc_sk(sk)->conn_timeout = value;
3158 break;
3159 case TIPC_MCAST_BROADCAST:
3160 tsk->mc_method.rcast = false;
3161 tsk->mc_method.mandatory = true;
3162 break;
3163 case TIPC_MCAST_REPLICAST:
3164 tsk->mc_method.rcast = true;
3165 tsk->mc_method.mandatory = true;
3166 break;
3167 case TIPC_GROUP_JOIN:
3168 res = tipc_sk_join(tsk, &mreq);
3169 break;
3170 case TIPC_GROUP_LEAVE:
3171 res = tipc_sk_leave(tsk);
3172 break;
3173 case TIPC_NODELAY:
3174 tsk->nodelay = !!value;
3175 tsk_set_nagle(tsk);
3176 break;
3177 default:
3178 res = -EINVAL;
3179 }
3180
3181 release_sock(sk);
3182
3183 return res;
3184}
3185
3186/**
3187 * tipc_getsockopt - get socket option
3188 * @sock: socket structure
3189 * @lvl: option level
3190 * @opt: option identifier
3191 * @ov: receptacle for option value
3192 * @ol: receptacle for length of option value
3193 *
3194 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3195 * (to ease compatibility).
3196 *
3197 * Returns 0 on success, errno otherwise
3198 */
3199static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3200 char __user *ov, int __user *ol)
3201{
3202 struct sock *sk = sock->sk;
3203 struct tipc_sock *tsk = tipc_sk(sk);
3204 struct tipc_name_seq seq;
3205 int len, scope;
3206 u32 value;
3207 int res;
3208
3209 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3210 return put_user(0, ol);
3211 if (lvl != SOL_TIPC)
3212 return -ENOPROTOOPT;
3213 res = get_user(len, ol);
3214 if (res)
3215 return res;
3216
3217 lock_sock(sk);
3218
3219 switch (opt) {
3220 case TIPC_IMPORTANCE:
3221 value = tsk_importance(tsk);
3222 break;
3223 case TIPC_SRC_DROPPABLE:
3224 value = tsk_unreliable(tsk);
3225 break;
3226 case TIPC_DEST_DROPPABLE:
3227 value = tsk_unreturnable(tsk);
3228 break;
3229 case TIPC_CONN_TIMEOUT:
3230 value = tsk->conn_timeout;
3231 /* no need to set "res", since already 0 at this point */
3232 break;
3233 case TIPC_NODE_RECVQ_DEPTH:
3234 value = 0; /* was tipc_queue_size, now obsolete */
3235 break;
3236 case TIPC_SOCK_RECVQ_DEPTH:
3237 value = skb_queue_len(&sk->sk_receive_queue);
3238 break;
3239 case TIPC_SOCK_RECVQ_USED:
3240 value = sk_rmem_alloc_get(sk);
3241 break;
3242 case TIPC_GROUP_JOIN:
3243 seq.type = 0;
3244 if (tsk->group)
3245 tipc_group_self(tsk->group, &seq, &scope);
3246 value = seq.type;
3247 break;
3248 default:
3249 res = -EINVAL;
3250 }
3251
3252 release_sock(sk);
3253
3254 if (res)
3255 return res; /* "get" failed */
3256
3257 if (len < sizeof(value))
3258 return -EINVAL;
3259
3260 if (copy_to_user(ov, &value, sizeof(value)))
3261 return -EFAULT;
3262
3263 return put_user(sizeof(value), ol);
3264}
3265
3266static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3267{
3268 struct net *net = sock_net(sock->sk);
3269 struct tipc_sioc_nodeid_req nr = {0};
3270 struct tipc_sioc_ln_req lnr;
3271 void __user *argp = (void __user *)arg;
3272
3273 switch (cmd) {
3274 case SIOCGETLINKNAME:
3275 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3276 return -EFAULT;
3277 if (!tipc_node_get_linkname(net,
3278 lnr.bearer_id & 0xffff, lnr.peer,
3279 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3280 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3281 return -EFAULT;
3282 return 0;
3283 }
3284 return -EADDRNOTAVAIL;
3285 case SIOCGETNODEID:
3286 if (copy_from_user(&nr, argp, sizeof(nr)))
3287 return -EFAULT;
3288 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3289 return -EADDRNOTAVAIL;
3290 if (copy_to_user(argp, &nr, sizeof(nr)))
3291 return -EFAULT;
3292 return 0;
3293 default:
3294 return -ENOIOCTLCMD;
3295 }
3296}
3297
3298static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3299{
3300 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3301 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3302 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3303
3304 tsk1->peer.family = AF_TIPC;
3305 tsk1->peer.addrtype = TIPC_ADDR_ID;
3306 tsk1->peer.scope = TIPC_NODE_SCOPE;
3307 tsk1->peer.addr.id.ref = tsk2->portid;
3308 tsk1->peer.addr.id.node = onode;
3309 tsk2->peer.family = AF_TIPC;
3310 tsk2->peer.addrtype = TIPC_ADDR_ID;
3311 tsk2->peer.scope = TIPC_NODE_SCOPE;
3312 tsk2->peer.addr.id.ref = tsk1->portid;
3313 tsk2->peer.addr.id.node = onode;
3314
3315 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3316 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3317 return 0;
3318}
3319
3320/* Protocol switches for the various types of TIPC sockets */
3321
3322static const struct proto_ops msg_ops = {
3323 .owner = THIS_MODULE,
3324 .family = AF_TIPC,
3325 .release = tipc_release,
3326 .bind = tipc_bind,
3327 .connect = tipc_connect,
3328 .socketpair = tipc_socketpair,
3329 .accept = sock_no_accept,
3330 .getname = tipc_getname,
3331 .poll = tipc_poll,
3332 .ioctl = tipc_ioctl,
3333 .listen = sock_no_listen,
3334 .shutdown = tipc_shutdown,
3335 .setsockopt = tipc_setsockopt,
3336 .getsockopt = tipc_getsockopt,
3337 .sendmsg = tipc_sendmsg,
3338 .recvmsg = tipc_recvmsg,
3339 .mmap = sock_no_mmap,
3340 .sendpage = sock_no_sendpage
3341};
3342
3343static const struct proto_ops packet_ops = {
3344 .owner = THIS_MODULE,
3345 .family = AF_TIPC,
3346 .release = tipc_release,
3347 .bind = tipc_bind,
3348 .connect = tipc_connect,
3349 .socketpair = tipc_socketpair,
3350 .accept = tipc_accept,
3351 .getname = tipc_getname,
3352 .poll = tipc_poll,
3353 .ioctl = tipc_ioctl,
3354 .listen = tipc_listen,
3355 .shutdown = tipc_shutdown,
3356 .setsockopt = tipc_setsockopt,
3357 .getsockopt = tipc_getsockopt,
3358 .sendmsg = tipc_send_packet,
3359 .recvmsg = tipc_recvmsg,
3360 .mmap = sock_no_mmap,
3361 .sendpage = sock_no_sendpage
3362};
3363
3364static const struct proto_ops stream_ops = {
3365 .owner = THIS_MODULE,
3366 .family = AF_TIPC,
3367 .release = tipc_release,
3368 .bind = tipc_bind,
3369 .connect = tipc_connect,
3370 .socketpair = tipc_socketpair,
3371 .accept = tipc_accept,
3372 .getname = tipc_getname,
3373 .poll = tipc_poll,
3374 .ioctl = tipc_ioctl,
3375 .listen = tipc_listen,
3376 .shutdown = tipc_shutdown,
3377 .setsockopt = tipc_setsockopt,
3378 .getsockopt = tipc_getsockopt,
3379 .sendmsg = tipc_sendstream,
3380 .recvmsg = tipc_recvstream,
3381 .mmap = sock_no_mmap,
3382 .sendpage = sock_no_sendpage
3383};
3384
3385static const struct net_proto_family tipc_family_ops = {
3386 .owner = THIS_MODULE,
3387 .family = AF_TIPC,
3388 .create = tipc_sk_create
3389};
3390
3391static struct proto tipc_proto = {
3392 .name = "TIPC",
3393 .owner = THIS_MODULE,
3394 .obj_size = sizeof(struct tipc_sock),
3395 .sysctl_rmem = sysctl_tipc_rmem
3396};
3397
3398/**
3399 * tipc_socket_init - initialize TIPC socket interface
3400 *
3401 * Returns 0 on success, errno otherwise
3402 */
3403int tipc_socket_init(void)
3404{
3405 int res;
3406
3407 res = proto_register(&tipc_proto, 1);
3408 if (res) {
3409 pr_err("Failed to register TIPC protocol type\n");
3410 goto out;
3411 }
3412
3413 res = sock_register(&tipc_family_ops);
3414 if (res) {
3415 pr_err("Failed to register TIPC socket type\n");
3416 proto_unregister(&tipc_proto);
3417 goto out;
3418 }
3419 out:
3420 return res;
3421}
3422
3423/**
3424 * tipc_socket_stop - stop TIPC socket interface
3425 */
3426void tipc_socket_stop(void)
3427{
3428 sock_unregister(tipc_family_ops.family);
3429 proto_unregister(&tipc_proto);
3430}
3431
3432/* Caller should hold socket lock for the passed tipc socket. */
3433static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3434{
3435 u32 peer_node;
3436 u32 peer_port;
3437 struct nlattr *nest;
3438
3439 peer_node = tsk_peer_node(tsk);
3440 peer_port = tsk_peer_port(tsk);
3441
3442 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3443 if (!nest)
3444 return -EMSGSIZE;
3445
3446 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3447 goto msg_full;
3448 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3449 goto msg_full;
3450
3451 if (tsk->conn_type != 0) {
3452 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3453 goto msg_full;
3454 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3455 goto msg_full;
3456 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3457 goto msg_full;
3458 }
3459 nla_nest_end(skb, nest);
3460
3461 return 0;
3462
3463msg_full:
3464 nla_nest_cancel(skb, nest);
3465
3466 return -EMSGSIZE;
3467}
3468
3469static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3470 *tsk)
3471{
3472 struct net *net = sock_net(skb->sk);
3473 struct sock *sk = &tsk->sk;
3474
3475 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3476 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3477 return -EMSGSIZE;
3478
3479 if (tipc_sk_connected(sk)) {
3480 if (__tipc_nl_add_sk_con(skb, tsk))
3481 return -EMSGSIZE;
3482 } else if (!list_empty(&tsk->publications)) {
3483 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3484 return -EMSGSIZE;
3485 }
3486 return 0;
3487}
3488
3489/* Caller should hold socket lock for the passed tipc socket. */
3490static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3491 struct tipc_sock *tsk)
3492{
3493 struct nlattr *attrs;
3494 void *hdr;
3495
3496 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3497 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3498 if (!hdr)
3499 goto msg_cancel;
3500
3501 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3502 if (!attrs)
3503 goto genlmsg_cancel;
3504
3505 if (__tipc_nl_add_sk_info(skb, tsk))
3506 goto attr_msg_cancel;
3507
3508 nla_nest_end(skb, attrs);
3509 genlmsg_end(skb, hdr);
3510
3511 return 0;
3512
3513attr_msg_cancel:
3514 nla_nest_cancel(skb, attrs);
3515genlmsg_cancel:
3516 genlmsg_cancel(skb, hdr);
3517msg_cancel:
3518 return -EMSGSIZE;
3519}
3520
3521int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3522 int (*skb_handler)(struct sk_buff *skb,
3523 struct netlink_callback *cb,
3524 struct tipc_sock *tsk))
3525{
3526 struct rhashtable_iter *iter = (void *)cb->args[4];
3527 struct tipc_sock *tsk;
3528 int err;
3529
3530 rhashtable_walk_start(iter);
3531 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3532 if (IS_ERR(tsk)) {
3533 err = PTR_ERR(tsk);
3534 if (err == -EAGAIN) {
3535 err = 0;
3536 continue;
3537 }
3538 break;
3539 }
3540
3541 sock_hold(&tsk->sk);
3542 rhashtable_walk_stop(iter);
3543 lock_sock(&tsk->sk);
3544 err = skb_handler(skb, cb, tsk);
3545 if (err) {
3546 release_sock(&tsk->sk);
3547 sock_put(&tsk->sk);
3548 goto out;
3549 }
3550 release_sock(&tsk->sk);
3551 rhashtable_walk_start(iter);
3552 sock_put(&tsk->sk);
3553 }
3554 rhashtable_walk_stop(iter);
3555out:
3556 return skb->len;
3557}
3558EXPORT_SYMBOL(tipc_nl_sk_walk);
3559
3560int tipc_dump_start(struct netlink_callback *cb)
3561{
3562 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3563}
3564EXPORT_SYMBOL(tipc_dump_start);
3565
3566int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3567{
3568 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3569 struct rhashtable_iter *iter = (void *)cb->args[4];
3570 struct tipc_net *tn = tipc_net(net);
3571
3572 if (!iter) {
3573 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3574 if (!iter)
3575 return -ENOMEM;
3576
3577 cb->args[4] = (long)iter;
3578 }
3579
3580 rhashtable_walk_enter(&tn->sk_rht, iter);
3581 return 0;
3582}
3583
3584int tipc_dump_done(struct netlink_callback *cb)
3585{
3586 struct rhashtable_iter *hti = (void *)cb->args[4];
3587
3588 rhashtable_walk_exit(hti);
3589 kfree(hti);
3590 return 0;
3591}
3592EXPORT_SYMBOL(tipc_dump_done);
3593
3594int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3595 struct tipc_sock *tsk, u32 sk_filter_state,
3596 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3597{
3598 struct sock *sk = &tsk->sk;
3599 struct nlattr *attrs;
3600 struct nlattr *stat;
3601
3602 /*filter response w.r.t sk_state*/
3603 if (!(sk_filter_state & (1 << sk->sk_state)))
3604 return 0;
3605
3606 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3607 if (!attrs)
3608 goto msg_cancel;
3609
3610 if (__tipc_nl_add_sk_info(skb, tsk))
3611 goto attr_msg_cancel;
3612
3613 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3614 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3615 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3616 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3617 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3618 sock_i_uid(sk))) ||
3619 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3620 tipc_diag_gen_cookie(sk),
3621 TIPC_NLA_SOCK_PAD))
3622 goto attr_msg_cancel;
3623
3624 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3625 if (!stat)
3626 goto attr_msg_cancel;
3627
3628 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3629 skb_queue_len(&sk->sk_receive_queue)) ||
3630 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3631 skb_queue_len(&sk->sk_write_queue)) ||
3632 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3633 atomic_read(&sk->sk_drops)))
3634 goto stat_msg_cancel;
3635
3636 if (tsk->cong_link_cnt &&
3637 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3638 goto stat_msg_cancel;
3639
3640 if (tsk_conn_cong(tsk) &&
3641 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3642 goto stat_msg_cancel;
3643
3644 nla_nest_end(skb, stat);
3645
3646 if (tsk->group)
3647 if (tipc_group_fill_sock_diag(tsk->group, skb))
3648 goto stat_msg_cancel;
3649
3650 nla_nest_end(skb, attrs);
3651
3652 return 0;
3653
3654stat_msg_cancel:
3655 nla_nest_cancel(skb, stat);
3656attr_msg_cancel:
3657 nla_nest_cancel(skb, attrs);
3658msg_cancel:
3659 return -EMSGSIZE;
3660}
3661EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3662
3663int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3664{
3665 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3666}
3667
3668/* Caller should hold socket lock for the passed tipc socket. */
3669static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3670 struct netlink_callback *cb,
3671 struct publication *publ)
3672{
3673 void *hdr;
3674 struct nlattr *attrs;
3675
3676 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3677 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3678 if (!hdr)
3679 goto msg_cancel;
3680
3681 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3682 if (!attrs)
3683 goto genlmsg_cancel;
3684
3685 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3686 goto attr_msg_cancel;
3687 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3688 goto attr_msg_cancel;
3689 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3690 goto attr_msg_cancel;
3691 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3692 goto attr_msg_cancel;
3693
3694 nla_nest_end(skb, attrs);
3695 genlmsg_end(skb, hdr);
3696
3697 return 0;
3698
3699attr_msg_cancel:
3700 nla_nest_cancel(skb, attrs);
3701genlmsg_cancel:
3702 genlmsg_cancel(skb, hdr);
3703msg_cancel:
3704 return -EMSGSIZE;
3705}
3706
3707/* Caller should hold socket lock for the passed tipc socket. */
3708static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3709 struct netlink_callback *cb,
3710 struct tipc_sock *tsk, u32 *last_publ)
3711{
3712 int err;
3713 struct publication *p;
3714
3715 if (*last_publ) {
3716 list_for_each_entry(p, &tsk->publications, binding_sock) {
3717 if (p->key == *last_publ)
3718 break;
3719 }
3720 if (p->key != *last_publ) {
3721 /* We never set seq or call nl_dump_check_consistent()
3722 * this means that setting prev_seq here will cause the
3723 * consistence check to fail in the netlink callback
3724 * handler. Resulting in the last NLMSG_DONE message
3725 * having the NLM_F_DUMP_INTR flag set.
3726 */
3727 cb->prev_seq = 1;
3728 *last_publ = 0;
3729 return -EPIPE;
3730 }
3731 } else {
3732 p = list_first_entry(&tsk->publications, struct publication,
3733 binding_sock);
3734 }
3735
3736 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3737 err = __tipc_nl_add_sk_publ(skb, cb, p);
3738 if (err) {
3739 *last_publ = p->key;
3740 return err;
3741 }
3742 }
3743 *last_publ = 0;
3744
3745 return 0;
3746}
3747
3748int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3749{
3750 int err;
3751 u32 tsk_portid = cb->args[0];
3752 u32 last_publ = cb->args[1];
3753 u32 done = cb->args[2];
3754 struct net *net = sock_net(skb->sk);
3755 struct tipc_sock *tsk;
3756
3757 if (!tsk_portid) {
3758 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3759 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3760
3761 if (!attrs[TIPC_NLA_SOCK])
3762 return -EINVAL;
3763
3764 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3765 attrs[TIPC_NLA_SOCK],
3766 tipc_nl_sock_policy, NULL);
3767 if (err)
3768 return err;
3769
3770 if (!sock[TIPC_NLA_SOCK_REF])
3771 return -EINVAL;
3772
3773 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3774 }
3775
3776 if (done)
3777 return 0;
3778
3779 tsk = tipc_sk_lookup(net, tsk_portid);
3780 if (!tsk)
3781 return -EINVAL;
3782
3783 lock_sock(&tsk->sk);
3784 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3785 if (!err)
3786 done = 1;
3787 release_sock(&tsk->sk);
3788 sock_put(&tsk->sk);
3789
3790 cb->args[0] = tsk_portid;
3791 cb->args[1] = last_publ;
3792 cb->args[2] = done;
3793
3794 return skb->len;
3795}
3796
3797/**
3798 * tipc_sk_filtering - check if a socket should be traced
3799 * @sk: the socket to be examined
3800 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3801 * (portid, sock type, name type, name lower, name upper)
3802 *
3803 * Returns true if the socket meets the socket tuple data
3804 * (value 0 = 'any') or when there is no tuple set (all = 0),
3805 * otherwise false
3806 */
3807bool tipc_sk_filtering(struct sock *sk)
3808{
3809 struct tipc_sock *tsk;
3810 struct publication *p;
3811 u32 _port, _sktype, _type, _lower, _upper;
3812 u32 type = 0, lower = 0, upper = 0;
3813
3814 if (!sk)
3815 return true;
3816
3817 tsk = tipc_sk(sk);
3818
3819 _port = sysctl_tipc_sk_filter[0];
3820 _sktype = sysctl_tipc_sk_filter[1];
3821 _type = sysctl_tipc_sk_filter[2];
3822 _lower = sysctl_tipc_sk_filter[3];
3823 _upper = sysctl_tipc_sk_filter[4];
3824
3825 if (!_port && !_sktype && !_type && !_lower && !_upper)
3826 return true;
3827
3828 if (_port)
3829 return (_port == tsk->portid);
3830
3831 if (_sktype && _sktype != sk->sk_type)
3832 return false;
3833
3834 if (tsk->published) {
3835 p = list_first_entry_or_null(&tsk->publications,
3836 struct publication, binding_sock);
3837 if (p) {
3838 type = p->type;
3839 lower = p->lower;
3840 upper = p->upper;
3841 }
3842 }
3843
3844 if (!tipc_sk_type_connectionless(sk)) {
3845 type = tsk->conn_type;
3846 lower = tsk->conn_instance;
3847 upper = tsk->conn_instance;
3848 }
3849
3850 if ((_type && _type != type) || (_lower && _lower != lower) ||
3851 (_upper && _upper != upper))
3852 return false;
3853
3854 return true;
3855}
3856
3857u32 tipc_sock_get_portid(struct sock *sk)
3858{
3859 return (sk) ? (tipc_sk(sk))->portid : 0;
3860}
3861
3862/**
3863 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3864 * both the rcv and backlog queues are considered
3865 * @sk: tipc sk to be checked
3866 * @skb: tipc msg to be checked
3867 *
3868 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3869 */
3870
3871bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3872{
3873 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3874 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3875 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3876
3877 return (qsize > lim * 90 / 100);
3878}
3879
3880/**
3881 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3882 * only the rcv queue is considered
3883 * @sk: tipc sk to be checked
3884 * @skb: tipc msg to be checked
3885 *
3886 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3887 */
3888
3889bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3890{
3891 unsigned int lim = rcvbuf_limit(sk, skb);
3892 unsigned int qsize = sk_rmem_alloc_get(sk);
3893
3894 return (qsize > lim * 90 / 100);
3895}
3896
3897/**
3898 * tipc_sk_dump - dump TIPC socket
3899 * @sk: tipc sk to be dumped
3900 * @dqueues: bitmask to decide if any socket queue to be dumped?
3901 * - TIPC_DUMP_NONE: don't dump socket queues
3902 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3903 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3904 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3905 * - TIPC_DUMP_ALL: dump all the socket queues above
3906 * @buf: returned buffer of dump data in format
3907 */
3908int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3909{
3910 int i = 0;
3911 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3912 struct tipc_sock *tsk;
3913 struct publication *p;
3914 bool tsk_connected;
3915
3916 if (!sk) {
3917 i += scnprintf(buf, sz, "sk data: (null)\n");
3918 return i;
3919 }
3920
3921 tsk = tipc_sk(sk);
3922 tsk_connected = !tipc_sk_type_connectionless(sk);
3923
3924 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3925 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3926 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3927 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3928 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3929 if (tsk_connected) {
3930 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3931 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3932 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3933 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3934 }
3935 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3936 if (tsk->published) {
3937 p = list_first_entry_or_null(&tsk->publications,
3938 struct publication, binding_sock);
3939 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3940 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3941 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3942 }
3943 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3944 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3945 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3946 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3947 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3948 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3949 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3950 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3951 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3952 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3953 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3954 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3955 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3956 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3957
3958 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3959 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3960 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3961 }
3962
3963 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3964 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3965 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3966 }
3967
3968 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3969 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3970 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3971 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3972 i += scnprintf(buf + i, sz - i, " tail ");
3973 i += tipc_skb_dump(sk->sk_backlog.tail, false,
3974 buf + i);
3975 }
3976 }
3977
3978 return i;
3979}
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 int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
150 bool kern);
151static void tipc_sk_timeout(struct timer_list *t);
152static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
153static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
154static int tipc_sk_leave(struct tipc_sock *tsk);
155static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
156static int tipc_sk_insert(struct tipc_sock *tsk);
157static void tipc_sk_remove(struct tipc_sock *tsk);
158static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
159static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
160static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
161static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
162
163static const struct proto_ops packet_ops;
164static const struct proto_ops stream_ops;
165static const struct proto_ops msg_ops;
166static struct proto tipc_proto;
167static const struct rhashtable_params tsk_rht_params;
168
169static u32 tsk_own_node(struct tipc_sock *tsk)
170{
171 return msg_prevnode(&tsk->phdr);
172}
173
174static u32 tsk_peer_node(struct tipc_sock *tsk)
175{
176 return msg_destnode(&tsk->phdr);
177}
178
179static u32 tsk_peer_port(struct tipc_sock *tsk)
180{
181 return msg_destport(&tsk->phdr);
182}
183
184static bool tsk_unreliable(struct tipc_sock *tsk)
185{
186 return msg_src_droppable(&tsk->phdr) != 0;
187}
188
189static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
190{
191 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
192}
193
194static bool tsk_unreturnable(struct tipc_sock *tsk)
195{
196 return msg_dest_droppable(&tsk->phdr) != 0;
197}
198
199static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
200{
201 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
202}
203
204static int tsk_importance(struct tipc_sock *tsk)
205{
206 return msg_importance(&tsk->phdr);
207}
208
209static struct tipc_sock *tipc_sk(const struct sock *sk)
210{
211 return container_of(sk, struct tipc_sock, sk);
212}
213
214int tsk_set_importance(struct sock *sk, int imp)
215{
216 if (imp > TIPC_CRITICAL_IMPORTANCE)
217 return -EINVAL;
218 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
219 return 0;
220}
221
222static bool tsk_conn_cong(struct tipc_sock *tsk)
223{
224 return tsk->snt_unacked > tsk->snd_win;
225}
226
227static u16 tsk_blocks(int len)
228{
229 return ((len / FLOWCTL_BLK_SZ) + 1);
230}
231
232/* tsk_blocks(): translate a buffer size in bytes to number of
233 * advertisable blocks, taking into account the ratio truesize(len)/len
234 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
235 */
236static u16 tsk_adv_blocks(int len)
237{
238 return len / FLOWCTL_BLK_SZ / 4;
239}
240
241/* tsk_inc(): increment counter for sent or received data
242 * - If block based flow control is not supported by peer we
243 * fall back to message based ditto, incrementing the counter
244 */
245static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
246{
247 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
248 return ((msglen / FLOWCTL_BLK_SZ) + 1);
249 return 1;
250}
251
252/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
253 */
254static void tsk_set_nagle(struct tipc_sock *tsk)
255{
256 struct sock *sk = &tsk->sk;
257
258 tsk->maxnagle = 0;
259 if (sk->sk_type != SOCK_STREAM)
260 return;
261 if (tsk->nodelay)
262 return;
263 if (!(tsk->peer_caps & TIPC_NAGLE))
264 return;
265 /* Limit node local buffer size to avoid receive queue overflow */
266 if (tsk->max_pkt == MAX_MSG_SIZE)
267 tsk->maxnagle = 1500;
268 else
269 tsk->maxnagle = tsk->max_pkt;
270}
271
272/**
273 * tsk_advance_rx_queue - discard first buffer in socket receive queue
274 * @sk: network socket
275 *
276 * Caller must hold socket lock
277 */
278static void tsk_advance_rx_queue(struct sock *sk)
279{
280 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
281 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
282}
283
284/* tipc_sk_respond() : send response message back to sender
285 */
286static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
287{
288 u32 selector;
289 u32 dnode;
290 u32 onode = tipc_own_addr(sock_net(sk));
291
292 if (!tipc_msg_reverse(onode, &skb, err))
293 return;
294
295 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
296 dnode = msg_destnode(buf_msg(skb));
297 selector = msg_origport(buf_msg(skb));
298 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
299}
300
301/**
302 * tsk_rej_rx_queue - reject all buffers in socket receive queue
303 * @sk: network socket
304 * @error: response error code
305 *
306 * Caller must hold socket lock
307 */
308static void tsk_rej_rx_queue(struct sock *sk, int error)
309{
310 struct sk_buff *skb;
311
312 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
313 tipc_sk_respond(sk, skb, error);
314}
315
316static bool tipc_sk_connected(const struct sock *sk)
317{
318 return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED;
319}
320
321/* tipc_sk_type_connectionless - check if the socket is datagram socket
322 * @sk: socket
323 *
324 * Returns true if connection less, false otherwise
325 */
326static bool tipc_sk_type_connectionless(struct sock *sk)
327{
328 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
329}
330
331/* tsk_peer_msg - verify if message was sent by connected port's peer
332 *
333 * Handles cases where the node's network address has changed from
334 * the default of <0.0.0> to its configured setting.
335 */
336static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
337{
338 struct sock *sk = &tsk->sk;
339 u32 self = tipc_own_addr(sock_net(sk));
340 u32 peer_port = tsk_peer_port(tsk);
341 u32 orig_node, peer_node;
342
343 if (unlikely(!tipc_sk_connected(sk)))
344 return false;
345
346 if (unlikely(msg_origport(msg) != peer_port))
347 return false;
348
349 orig_node = msg_orignode(msg);
350 peer_node = tsk_peer_node(tsk);
351
352 if (likely(orig_node == peer_node))
353 return true;
354
355 if (!orig_node && peer_node == self)
356 return true;
357
358 if (!peer_node && orig_node == self)
359 return true;
360
361 return false;
362}
363
364/* tipc_set_sk_state - set the sk_state of the socket
365 * @sk: socket
366 *
367 * Caller must hold socket lock
368 *
369 * Returns 0 on success, errno otherwise
370 */
371static int tipc_set_sk_state(struct sock *sk, int state)
372{
373 int oldsk_state = sk->sk_state;
374 int res = -EINVAL;
375
376 switch (state) {
377 case TIPC_OPEN:
378 res = 0;
379 break;
380 case TIPC_LISTEN:
381 case TIPC_CONNECTING:
382 if (oldsk_state == TIPC_OPEN)
383 res = 0;
384 break;
385 case TIPC_ESTABLISHED:
386 if (oldsk_state == TIPC_CONNECTING ||
387 oldsk_state == TIPC_OPEN)
388 res = 0;
389 break;
390 case TIPC_DISCONNECTING:
391 if (oldsk_state == TIPC_CONNECTING ||
392 oldsk_state == TIPC_ESTABLISHED)
393 res = 0;
394 break;
395 }
396
397 if (!res)
398 sk->sk_state = state;
399
400 return res;
401}
402
403static int tipc_sk_sock_err(struct socket *sock, long *timeout)
404{
405 struct sock *sk = sock->sk;
406 int err = sock_error(sk);
407 int typ = sock->type;
408
409 if (err)
410 return err;
411 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
412 if (sk->sk_state == TIPC_DISCONNECTING)
413 return -EPIPE;
414 else if (!tipc_sk_connected(sk))
415 return -ENOTCONN;
416 }
417 if (!*timeout)
418 return -EAGAIN;
419 if (signal_pending(current))
420 return sock_intr_errno(*timeout);
421
422 return 0;
423}
424
425#define tipc_wait_for_cond(sock_, timeo_, condition_) \
426({ \
427 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
428 struct sock *sk_; \
429 int rc_; \
430 \
431 while ((rc_ = !(condition_))) { \
432 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
433 smp_rmb(); \
434 sk_ = (sock_)->sk; \
435 rc_ = tipc_sk_sock_err((sock_), timeo_); \
436 if (rc_) \
437 break; \
438 add_wait_queue(sk_sleep(sk_), &wait_); \
439 release_sock(sk_); \
440 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
441 sched_annotate_sleep(); \
442 lock_sock(sk_); \
443 remove_wait_queue(sk_sleep(sk_), &wait_); \
444 } \
445 rc_; \
446})
447
448/**
449 * tipc_sk_create - create a TIPC socket
450 * @net: network namespace (must be default network)
451 * @sock: pre-allocated socket structure
452 * @protocol: protocol indicator (must be 0)
453 * @kern: caused by kernel or by userspace?
454 *
455 * This routine creates additional data structures used by the TIPC socket,
456 * initializes them, and links them together.
457 *
458 * Return: 0 on success, errno otherwise
459 */
460static int tipc_sk_create(struct net *net, struct socket *sock,
461 int protocol, int kern)
462{
463 const struct proto_ops *ops;
464 struct sock *sk;
465 struct tipc_sock *tsk;
466 struct tipc_msg *msg;
467
468 /* Validate arguments */
469 if (unlikely(protocol != 0))
470 return -EPROTONOSUPPORT;
471
472 switch (sock->type) {
473 case SOCK_STREAM:
474 ops = &stream_ops;
475 break;
476 case SOCK_SEQPACKET:
477 ops = &packet_ops;
478 break;
479 case SOCK_DGRAM:
480 case SOCK_RDM:
481 ops = &msg_ops;
482 break;
483 default:
484 return -EPROTOTYPE;
485 }
486
487 /* Allocate socket's protocol area */
488 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
489 if (sk == NULL)
490 return -ENOMEM;
491
492 tsk = tipc_sk(sk);
493 tsk->max_pkt = MAX_PKT_DEFAULT;
494 tsk->maxnagle = 0;
495 tsk->nagle_start = NAGLE_START_INIT;
496 INIT_LIST_HEAD(&tsk->publications);
497 INIT_LIST_HEAD(&tsk->cong_links);
498 msg = &tsk->phdr;
499
500 /* Finish initializing socket data structures */
501 sock->ops = ops;
502 sock_init_data(sock, sk);
503 tipc_set_sk_state(sk, TIPC_OPEN);
504 if (tipc_sk_insert(tsk)) {
505 sk_free(sk);
506 pr_warn("Socket create failed; port number exhausted\n");
507 return -EINVAL;
508 }
509
510 /* Ensure tsk is visible before we read own_addr. */
511 smp_mb();
512
513 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
514 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
515
516 msg_set_origport(msg, tsk->portid);
517 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
518 sk->sk_shutdown = 0;
519 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
520 sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
521 sk->sk_data_ready = tipc_data_ready;
522 sk->sk_write_space = tipc_write_space;
523 sk->sk_destruct = tipc_sock_destruct;
524 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
525 tsk->group_is_open = true;
526 atomic_set(&tsk->dupl_rcvcnt, 0);
527
528 /* Start out with safe limits until we receive an advertised window */
529 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
530 tsk->rcv_win = tsk->snd_win;
531
532 if (tipc_sk_type_connectionless(sk)) {
533 tsk_set_unreturnable(tsk, true);
534 if (sock->type == SOCK_DGRAM)
535 tsk_set_unreliable(tsk, true);
536 }
537 __skb_queue_head_init(&tsk->mc_method.deferredq);
538 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
539 return 0;
540}
541
542static void tipc_sk_callback(struct rcu_head *head)
543{
544 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
545
546 sock_put(&tsk->sk);
547}
548
549/* Caller should hold socket lock for the socket. */
550static void __tipc_shutdown(struct socket *sock, int error)
551{
552 struct sock *sk = sock->sk;
553 struct tipc_sock *tsk = tipc_sk(sk);
554 struct net *net = sock_net(sk);
555 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
556 u32 dnode = tsk_peer_node(tsk);
557 struct sk_buff *skb;
558
559 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
560 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
561 !tsk_conn_cong(tsk)));
562
563 /* Push out delayed messages if in Nagle mode */
564 tipc_sk_push_backlog(tsk, false);
565 /* Remove pending SYN */
566 __skb_queue_purge(&sk->sk_write_queue);
567
568 /* Remove partially received buffer if any */
569 skb = skb_peek(&sk->sk_receive_queue);
570 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
571 __skb_unlink(skb, &sk->sk_receive_queue);
572 kfree_skb(skb);
573 }
574
575 /* Reject all unreceived messages if connectionless */
576 if (tipc_sk_type_connectionless(sk)) {
577 tsk_rej_rx_queue(sk, error);
578 return;
579 }
580
581 switch (sk->sk_state) {
582 case TIPC_CONNECTING:
583 case TIPC_ESTABLISHED:
584 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
585 tipc_node_remove_conn(net, dnode, tsk->portid);
586 /* Send a FIN+/- to its peer */
587 skb = __skb_dequeue(&sk->sk_receive_queue);
588 if (skb) {
589 __skb_queue_purge(&sk->sk_receive_queue);
590 tipc_sk_respond(sk, skb, error);
591 break;
592 }
593 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
594 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
595 tsk_own_node(tsk), tsk_peer_port(tsk),
596 tsk->portid, error);
597 if (skb)
598 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
599 break;
600 case TIPC_LISTEN:
601 /* Reject all SYN messages */
602 tsk_rej_rx_queue(sk, error);
603 break;
604 default:
605 __skb_queue_purge(&sk->sk_receive_queue);
606 break;
607 }
608}
609
610/**
611 * tipc_release - destroy a TIPC socket
612 * @sock: socket to destroy
613 *
614 * This routine cleans up any messages that are still queued on the socket.
615 * For DGRAM and RDM socket types, all queued messages are rejected.
616 * For SEQPACKET and STREAM socket types, the first message is rejected
617 * and any others are discarded. (If the first message on a STREAM socket
618 * is partially-read, it is discarded and the next one is rejected instead.)
619 *
620 * NOTE: Rejected messages are not necessarily returned to the sender! They
621 * are returned or discarded according to the "destination droppable" setting
622 * specified for the message by the sender.
623 *
624 * Return: 0 on success, errno otherwise
625 */
626static int tipc_release(struct socket *sock)
627{
628 struct sock *sk = sock->sk;
629 struct tipc_sock *tsk;
630
631 /*
632 * Exit if socket isn't fully initialized (occurs when a failed accept()
633 * releases a pre-allocated child socket that was never used)
634 */
635 if (sk == NULL)
636 return 0;
637
638 tsk = tipc_sk(sk);
639 lock_sock(sk);
640
641 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
642 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
643 sk->sk_shutdown = SHUTDOWN_MASK;
644 tipc_sk_leave(tsk);
645 tipc_sk_withdraw(tsk, NULL);
646 __skb_queue_purge(&tsk->mc_method.deferredq);
647 sk_stop_timer(sk, &sk->sk_timer);
648 tipc_sk_remove(tsk);
649
650 sock_orphan(sk);
651 /* Reject any messages that accumulated in backlog queue */
652 release_sock(sk);
653 tipc_dest_list_purge(&tsk->cong_links);
654 tsk->cong_link_cnt = 0;
655 call_rcu(&tsk->rcu, tipc_sk_callback);
656 sock->sk = NULL;
657
658 return 0;
659}
660
661/**
662 * __tipc_bind - associate or disassocate TIPC name(s) with a socket
663 * @sock: socket structure
664 * @skaddr: socket address describing name(s) and desired operation
665 * @alen: size of socket address data structure
666 *
667 * Name and name sequence binding are indicated using a positive scope value;
668 * a negative scope value unbinds the specified name. Specifying no name
669 * (i.e. a socket address length of 0) unbinds all names from the socket.
670 *
671 * Return: 0 on success, errno otherwise
672 *
673 * NOTE: This routine doesn't need to take the socket lock since it doesn't
674 * access any non-constant socket information.
675 */
676static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
677{
678 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
679 struct tipc_sock *tsk = tipc_sk(sock->sk);
680 bool unbind = false;
681
682 if (unlikely(!alen))
683 return tipc_sk_withdraw(tsk, NULL);
684
685 if (ua->addrtype == TIPC_SERVICE_ADDR) {
686 ua->addrtype = TIPC_SERVICE_RANGE;
687 ua->sr.upper = ua->sr.lower;
688 }
689 if (ua->scope < 0) {
690 unbind = true;
691 ua->scope = -ua->scope;
692 }
693 /* Users may still use deprecated TIPC_ZONE_SCOPE */
694 if (ua->scope != TIPC_NODE_SCOPE)
695 ua->scope = TIPC_CLUSTER_SCOPE;
696
697 if (tsk->group)
698 return -EACCES;
699
700 if (unbind)
701 return tipc_sk_withdraw(tsk, ua);
702 return tipc_sk_publish(tsk, ua);
703}
704
705int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
706{
707 int res;
708
709 lock_sock(sock->sk);
710 res = __tipc_bind(sock, skaddr, alen);
711 release_sock(sock->sk);
712 return res;
713}
714
715static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
716{
717 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
718 u32 atype = ua->addrtype;
719
720 if (alen) {
721 if (!tipc_uaddr_valid(ua, alen))
722 return -EINVAL;
723 if (atype == TIPC_SOCKET_ADDR)
724 return -EAFNOSUPPORT;
725 if (ua->sr.type < TIPC_RESERVED_TYPES) {
726 pr_warn_once("Can't bind to reserved service type %u\n",
727 ua->sr.type);
728 return -EACCES;
729 }
730 }
731 return tipc_sk_bind(sock, skaddr, alen);
732}
733
734/**
735 * tipc_getname - get port ID of socket or peer socket
736 * @sock: socket structure
737 * @uaddr: area for returned socket address
738 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
739 *
740 * Return: 0 on success, errno otherwise
741 *
742 * NOTE: This routine doesn't need to take the socket lock since it only
743 * accesses socket information that is unchanging (or which changes in
744 * a completely predictable manner).
745 */
746static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
747 int peer)
748{
749 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
750 struct sock *sk = sock->sk;
751 struct tipc_sock *tsk = tipc_sk(sk);
752
753 memset(addr, 0, sizeof(*addr));
754 if (peer) {
755 if ((!tipc_sk_connected(sk)) &&
756 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
757 return -ENOTCONN;
758 addr->addr.id.ref = tsk_peer_port(tsk);
759 addr->addr.id.node = tsk_peer_node(tsk);
760 } else {
761 addr->addr.id.ref = tsk->portid;
762 addr->addr.id.node = tipc_own_addr(sock_net(sk));
763 }
764
765 addr->addrtype = TIPC_SOCKET_ADDR;
766 addr->family = AF_TIPC;
767 addr->scope = 0;
768 addr->addr.name.domain = 0;
769
770 return sizeof(*addr);
771}
772
773/**
774 * tipc_poll - read and possibly block on pollmask
775 * @file: file structure associated with the socket
776 * @sock: socket for which to calculate the poll bits
777 * @wait: ???
778 *
779 * Return: pollmask value
780 *
781 * COMMENTARY:
782 * It appears that the usual socket locking mechanisms are not useful here
783 * since the pollmask info is potentially out-of-date the moment this routine
784 * exits. TCP and other protocols seem to rely on higher level poll routines
785 * to handle any preventable race conditions, so TIPC will do the same ...
786 *
787 * IMPORTANT: The fact that a read or write operation is indicated does NOT
788 * imply that the operation will succeed, merely that it should be performed
789 * and will not block.
790 */
791static __poll_t tipc_poll(struct file *file, struct socket *sock,
792 poll_table *wait)
793{
794 struct sock *sk = sock->sk;
795 struct tipc_sock *tsk = tipc_sk(sk);
796 __poll_t revents = 0;
797
798 sock_poll_wait(file, sock, wait);
799 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
800
801 if (sk->sk_shutdown & RCV_SHUTDOWN)
802 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
803 if (sk->sk_shutdown == SHUTDOWN_MASK)
804 revents |= EPOLLHUP;
805
806 switch (sk->sk_state) {
807 case TIPC_ESTABLISHED:
808 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
809 revents |= EPOLLOUT;
810 fallthrough;
811 case TIPC_LISTEN:
812 case TIPC_CONNECTING:
813 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
814 revents |= EPOLLIN | EPOLLRDNORM;
815 break;
816 case TIPC_OPEN:
817 if (tsk->group_is_open && !tsk->cong_link_cnt)
818 revents |= EPOLLOUT;
819 if (!tipc_sk_type_connectionless(sk))
820 break;
821 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
822 break;
823 revents |= EPOLLIN | EPOLLRDNORM;
824 break;
825 case TIPC_DISCONNECTING:
826 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
827 break;
828 }
829 return revents;
830}
831
832/**
833 * tipc_sendmcast - send multicast message
834 * @sock: socket structure
835 * @ua: destination address struct
836 * @msg: message to send
837 * @dlen: length of data to send
838 * @timeout: timeout to wait for wakeup
839 *
840 * Called from function tipc_sendmsg(), which has done all sanity checks
841 * Return: the number of bytes sent on success, or errno
842 */
843static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
844 struct msghdr *msg, size_t dlen, long timeout)
845{
846 struct sock *sk = sock->sk;
847 struct tipc_sock *tsk = tipc_sk(sk);
848 struct tipc_msg *hdr = &tsk->phdr;
849 struct net *net = sock_net(sk);
850 int mtu = tipc_bcast_get_mtu(net);
851 struct sk_buff_head pkts;
852 struct tipc_nlist dsts;
853 int rc;
854
855 if (tsk->group)
856 return -EACCES;
857
858 /* Block or return if any destination link is congested */
859 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
860 if (unlikely(rc))
861 return rc;
862
863 /* Lookup destination nodes */
864 tipc_nlist_init(&dsts, tipc_own_addr(net));
865 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
866 if (!dsts.local && !dsts.remote)
867 return -EHOSTUNREACH;
868
869 /* Build message header */
870 msg_set_type(hdr, TIPC_MCAST_MSG);
871 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
872 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
873 msg_set_destport(hdr, 0);
874 msg_set_destnode(hdr, 0);
875 msg_set_nametype(hdr, ua->sr.type);
876 msg_set_namelower(hdr, ua->sr.lower);
877 msg_set_nameupper(hdr, ua->sr.upper);
878
879 /* Build message as chain of buffers */
880 __skb_queue_head_init(&pkts);
881 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
882
883 /* Send message if build was successful */
884 if (unlikely(rc == dlen)) {
885 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
886 TIPC_DUMP_SK_SNDQ, " ");
887 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
888 &tsk->cong_link_cnt);
889 }
890
891 tipc_nlist_purge(&dsts);
892
893 return rc ? rc : dlen;
894}
895
896/**
897 * tipc_send_group_msg - send a message to a member in the group
898 * @net: network namespace
899 * @tsk: tipc socket
900 * @m: message to send
901 * @mb: group member
902 * @dnode: destination node
903 * @dport: destination port
904 * @dlen: total length of message data
905 */
906static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
907 struct msghdr *m, struct tipc_member *mb,
908 u32 dnode, u32 dport, int dlen)
909{
910 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
911 struct tipc_mc_method *method = &tsk->mc_method;
912 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
913 struct tipc_msg *hdr = &tsk->phdr;
914 struct sk_buff_head pkts;
915 int mtu, rc;
916
917 /* Complete message header */
918 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
919 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
920 msg_set_destport(hdr, dport);
921 msg_set_destnode(hdr, dnode);
922 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
923
924 /* Build message as chain of buffers */
925 __skb_queue_head_init(&pkts);
926 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
927 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
928 if (unlikely(rc != dlen))
929 return rc;
930
931 /* Send message */
932 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
933 if (unlikely(rc == -ELINKCONG)) {
934 tipc_dest_push(&tsk->cong_links, dnode, 0);
935 tsk->cong_link_cnt++;
936 }
937
938 /* Update send window */
939 tipc_group_update_member(mb, blks);
940
941 /* A broadcast sent within next EXPIRE period must follow same path */
942 method->rcast = true;
943 method->mandatory = true;
944 return dlen;
945}
946
947/**
948 * tipc_send_group_unicast - send message to a member in the group
949 * @sock: socket structure
950 * @m: message to send
951 * @dlen: total length of message data
952 * @timeout: timeout to wait for wakeup
953 *
954 * Called from function tipc_sendmsg(), which has done all sanity checks
955 * Return: the number of bytes sent on success, or errno
956 */
957static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
958 int dlen, long timeout)
959{
960 struct sock *sk = sock->sk;
961 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
962 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
963 struct tipc_sock *tsk = tipc_sk(sk);
964 struct net *net = sock_net(sk);
965 struct tipc_member *mb = NULL;
966 u32 node, port;
967 int rc;
968
969 node = ua->sk.node;
970 port = ua->sk.ref;
971 if (!port && !node)
972 return -EHOSTUNREACH;
973
974 /* Block or return if destination link or member is congested */
975 rc = tipc_wait_for_cond(sock, &timeout,
976 !tipc_dest_find(&tsk->cong_links, node, 0) &&
977 tsk->group &&
978 !tipc_group_cong(tsk->group, node, port, blks,
979 &mb));
980 if (unlikely(rc))
981 return rc;
982
983 if (unlikely(!mb))
984 return -EHOSTUNREACH;
985
986 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
987
988 return rc ? rc : dlen;
989}
990
991/**
992 * tipc_send_group_anycast - send message to any member with given identity
993 * @sock: socket structure
994 * @m: message to send
995 * @dlen: total length of message data
996 * @timeout: timeout to wait for wakeup
997 *
998 * Called from function tipc_sendmsg(), which has done all sanity checks
999 * Return: the number of bytes sent on success, or errno
1000 */
1001static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1002 int dlen, long timeout)
1003{
1004 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1005 struct sock *sk = sock->sk;
1006 struct tipc_sock *tsk = tipc_sk(sk);
1007 struct list_head *cong_links = &tsk->cong_links;
1008 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1009 struct tipc_msg *hdr = &tsk->phdr;
1010 struct tipc_member *first = NULL;
1011 struct tipc_member *mbr = NULL;
1012 struct net *net = sock_net(sk);
1013 u32 node, port, exclude;
1014 struct list_head dsts;
1015 int lookups = 0;
1016 int dstcnt, rc;
1017 bool cong;
1018
1019 INIT_LIST_HEAD(&dsts);
1020 ua->sa.type = msg_nametype(hdr);
1021 ua->scope = msg_lookup_scope(hdr);
1022
1023 while (++lookups < 4) {
1024 exclude = tipc_group_exclude(tsk->group);
1025
1026 first = NULL;
1027
1028 /* Look for a non-congested destination member, if any */
1029 while (1) {
1030 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1031 exclude, false))
1032 return -EHOSTUNREACH;
1033 tipc_dest_pop(&dsts, &node, &port);
1034 cong = tipc_group_cong(tsk->group, node, port, blks,
1035 &mbr);
1036 if (!cong)
1037 break;
1038 if (mbr == first)
1039 break;
1040 if (!first)
1041 first = mbr;
1042 }
1043
1044 /* Start over if destination was not in member list */
1045 if (unlikely(!mbr))
1046 continue;
1047
1048 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1049 break;
1050
1051 /* Block or return if destination link or member is congested */
1052 rc = tipc_wait_for_cond(sock, &timeout,
1053 !tipc_dest_find(cong_links, node, 0) &&
1054 tsk->group &&
1055 !tipc_group_cong(tsk->group, node, port,
1056 blks, &mbr));
1057 if (unlikely(rc))
1058 return rc;
1059
1060 /* Send, unless destination disappeared while waiting */
1061 if (likely(mbr))
1062 break;
1063 }
1064
1065 if (unlikely(lookups >= 4))
1066 return -EHOSTUNREACH;
1067
1068 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1069
1070 return rc ? rc : dlen;
1071}
1072
1073/**
1074 * tipc_send_group_bcast - send message to all members in communication group
1075 * @sock: socket structure
1076 * @m: message to send
1077 * @dlen: total length of message data
1078 * @timeout: timeout to wait for wakeup
1079 *
1080 * Called from function tipc_sendmsg(), which has done all sanity checks
1081 * Return: the number of bytes sent on success, or errno
1082 */
1083static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1084 int dlen, long timeout)
1085{
1086 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1087 struct sock *sk = sock->sk;
1088 struct net *net = sock_net(sk);
1089 struct tipc_sock *tsk = tipc_sk(sk);
1090 struct tipc_nlist *dsts;
1091 struct tipc_mc_method *method = &tsk->mc_method;
1092 bool ack = method->mandatory && method->rcast;
1093 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1094 struct tipc_msg *hdr = &tsk->phdr;
1095 int mtu = tipc_bcast_get_mtu(net);
1096 struct sk_buff_head pkts;
1097 int rc = -EHOSTUNREACH;
1098
1099 /* Block or return if any destination link or member is congested */
1100 rc = tipc_wait_for_cond(sock, &timeout,
1101 !tsk->cong_link_cnt && tsk->group &&
1102 !tipc_group_bc_cong(tsk->group, blks));
1103 if (unlikely(rc))
1104 return rc;
1105
1106 dsts = tipc_group_dests(tsk->group);
1107 if (!dsts->local && !dsts->remote)
1108 return -EHOSTUNREACH;
1109
1110 /* Complete message header */
1111 if (ua) {
1112 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1113 msg_set_nameinst(hdr, ua->sa.instance);
1114 } else {
1115 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1116 msg_set_nameinst(hdr, 0);
1117 }
1118 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1119 msg_set_destport(hdr, 0);
1120 msg_set_destnode(hdr, 0);
1121 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1122
1123 /* Avoid getting stuck with repeated forced replicasts */
1124 msg_set_grp_bc_ack_req(hdr, ack);
1125
1126 /* Build message as chain of buffers */
1127 __skb_queue_head_init(&pkts);
1128 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1129 if (unlikely(rc != dlen))
1130 return rc;
1131
1132 /* Send message */
1133 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1134 if (unlikely(rc))
1135 return rc;
1136
1137 /* Update broadcast sequence number and send windows */
1138 tipc_group_update_bc_members(tsk->group, blks, ack);
1139
1140 /* Broadcast link is now free to choose method for next broadcast */
1141 method->mandatory = false;
1142 method->expires = jiffies;
1143
1144 return dlen;
1145}
1146
1147/**
1148 * tipc_send_group_mcast - send message to all members with given identity
1149 * @sock: socket structure
1150 * @m: message to send
1151 * @dlen: total length of message data
1152 * @timeout: timeout to wait for wakeup
1153 *
1154 * Called from function tipc_sendmsg(), which has done all sanity checks
1155 * Return: the number of bytes sent on success, or errno
1156 */
1157static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1158 int dlen, long timeout)
1159{
1160 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1161 struct sock *sk = sock->sk;
1162 struct tipc_sock *tsk = tipc_sk(sk);
1163 struct tipc_group *grp = tsk->group;
1164 struct tipc_msg *hdr = &tsk->phdr;
1165 struct net *net = sock_net(sk);
1166 struct list_head dsts;
1167 u32 dstcnt, exclude;
1168
1169 INIT_LIST_HEAD(&dsts);
1170 ua->sa.type = msg_nametype(hdr);
1171 ua->scope = msg_lookup_scope(hdr);
1172 exclude = tipc_group_exclude(grp);
1173
1174 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1175 return -EHOSTUNREACH;
1176
1177 if (dstcnt == 1) {
1178 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1179 return tipc_send_group_unicast(sock, m, dlen, timeout);
1180 }
1181
1182 tipc_dest_list_purge(&dsts);
1183 return tipc_send_group_bcast(sock, m, dlen, timeout);
1184}
1185
1186/**
1187 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1188 * @net: the associated network namespace
1189 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1190 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1191 *
1192 * Multi-threaded: parallel calls with reference to same queues may occur
1193 */
1194void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1195 struct sk_buff_head *inputq)
1196{
1197 u32 self = tipc_own_addr(net);
1198 struct sk_buff *skb, *_skb;
1199 u32 portid, onode;
1200 struct sk_buff_head tmpq;
1201 struct list_head dports;
1202 struct tipc_msg *hdr;
1203 struct tipc_uaddr ua;
1204 int user, mtyp, hlen;
1205
1206 __skb_queue_head_init(&tmpq);
1207 INIT_LIST_HEAD(&dports);
1208 ua.addrtype = TIPC_SERVICE_RANGE;
1209
1210 /* tipc_skb_peek() increments the head skb's reference counter */
1211 skb = tipc_skb_peek(arrvq, &inputq->lock);
1212 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1213 hdr = buf_msg(skb);
1214 user = msg_user(hdr);
1215 mtyp = msg_type(hdr);
1216 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1217 onode = msg_orignode(hdr);
1218 ua.sr.type = msg_nametype(hdr);
1219 ua.sr.lower = msg_namelower(hdr);
1220 ua.sr.upper = msg_nameupper(hdr);
1221 if (onode == self)
1222 ua.scope = TIPC_ANY_SCOPE;
1223 else
1224 ua.scope = TIPC_CLUSTER_SCOPE;
1225
1226 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1227 spin_lock_bh(&inputq->lock);
1228 if (skb_peek(arrvq) == skb) {
1229 __skb_dequeue(arrvq);
1230 __skb_queue_tail(inputq, skb);
1231 }
1232 kfree_skb(skb);
1233 spin_unlock_bh(&inputq->lock);
1234 continue;
1235 }
1236
1237 /* Group messages require exact scope match */
1238 if (msg_in_group(hdr)) {
1239 ua.sr.lower = 0;
1240 ua.sr.upper = ~0;
1241 ua.scope = msg_lookup_scope(hdr);
1242 }
1243
1244 /* Create destination port list: */
1245 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1246
1247 /* Clone message per destination */
1248 while (tipc_dest_pop(&dports, NULL, &portid)) {
1249 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1250 if (_skb) {
1251 msg_set_destport(buf_msg(_skb), portid);
1252 __skb_queue_tail(&tmpq, _skb);
1253 continue;
1254 }
1255 pr_warn("Failed to clone mcast rcv buffer\n");
1256 }
1257 /* Append clones to inputq only if skb is still head of arrvq */
1258 spin_lock_bh(&inputq->lock);
1259 if (skb_peek(arrvq) == skb) {
1260 skb_queue_splice_tail_init(&tmpq, inputq);
1261 /* Decrement the skb's refcnt */
1262 kfree_skb(__skb_dequeue(arrvq));
1263 }
1264 spin_unlock_bh(&inputq->lock);
1265 __skb_queue_purge(&tmpq);
1266 kfree_skb(skb);
1267 }
1268 tipc_sk_rcv(net, inputq);
1269}
1270
1271/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1272 * when socket is in Nagle mode
1273 */
1274static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1275{
1276 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1277 struct sk_buff *skb = skb_peek_tail(txq);
1278 struct net *net = sock_net(&tsk->sk);
1279 u32 dnode = tsk_peer_node(tsk);
1280 int rc;
1281
1282 if (nagle_ack) {
1283 tsk->pkt_cnt += skb_queue_len(txq);
1284 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1285 tsk->oneway = 0;
1286 if (tsk->nagle_start < NAGLE_START_MAX)
1287 tsk->nagle_start *= 2;
1288 tsk->expect_ack = false;
1289 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1290 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1291 tsk->nagle_start);
1292 } else {
1293 tsk->nagle_start = NAGLE_START_INIT;
1294 if (skb) {
1295 msg_set_ack_required(buf_msg(skb));
1296 tsk->expect_ack = true;
1297 } else {
1298 tsk->expect_ack = false;
1299 }
1300 }
1301 tsk->msg_acc = 0;
1302 tsk->pkt_cnt = 0;
1303 }
1304
1305 if (!skb || tsk->cong_link_cnt)
1306 return;
1307
1308 /* Do not send SYN again after congestion */
1309 if (msg_is_syn(buf_msg(skb)))
1310 return;
1311
1312 if (tsk->msg_acc)
1313 tsk->pkt_cnt += skb_queue_len(txq);
1314 tsk->snt_unacked += tsk->snd_backlog;
1315 tsk->snd_backlog = 0;
1316 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1317 if (rc == -ELINKCONG)
1318 tsk->cong_link_cnt = 1;
1319}
1320
1321/**
1322 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1323 * @tsk: receiving socket
1324 * @skb: pointer to message buffer.
1325 * @inputq: buffer list containing the buffers
1326 * @xmitq: output message area
1327 */
1328static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1329 struct sk_buff_head *inputq,
1330 struct sk_buff_head *xmitq)
1331{
1332 struct tipc_msg *hdr = buf_msg(skb);
1333 u32 onode = tsk_own_node(tsk);
1334 struct sock *sk = &tsk->sk;
1335 int mtyp = msg_type(hdr);
1336 bool was_cong;
1337
1338 /* Ignore if connection cannot be validated: */
1339 if (!tsk_peer_msg(tsk, hdr)) {
1340 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1341 goto exit;
1342 }
1343
1344 if (unlikely(msg_errcode(hdr))) {
1345 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1346 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1347 tsk_peer_port(tsk));
1348 sk->sk_state_change(sk);
1349
1350 /* State change is ignored if socket already awake,
1351 * - convert msg to abort msg and add to inqueue
1352 */
1353 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1354 msg_set_type(hdr, TIPC_CONN_MSG);
1355 msg_set_size(hdr, BASIC_H_SIZE);
1356 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1357 __skb_queue_tail(inputq, skb);
1358 return;
1359 }
1360
1361 tsk->probe_unacked = false;
1362
1363 if (mtyp == CONN_PROBE) {
1364 msg_set_type(hdr, CONN_PROBE_REPLY);
1365 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1366 __skb_queue_tail(xmitq, skb);
1367 return;
1368 } else if (mtyp == CONN_ACK) {
1369 was_cong = tsk_conn_cong(tsk);
1370 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1371 tsk->snt_unacked -= msg_conn_ack(hdr);
1372 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1373 tsk->snd_win = msg_adv_win(hdr);
1374 if (was_cong && !tsk_conn_cong(tsk))
1375 sk->sk_write_space(sk);
1376 } else if (mtyp != CONN_PROBE_REPLY) {
1377 pr_warn("Received unknown CONN_PROTO msg\n");
1378 }
1379exit:
1380 kfree_skb(skb);
1381}
1382
1383/**
1384 * tipc_sendmsg - send message in connectionless manner
1385 * @sock: socket structure
1386 * @m: message to send
1387 * @dsz: amount of user data to be sent
1388 *
1389 * Message must have an destination specified explicitly.
1390 * Used for SOCK_RDM and SOCK_DGRAM messages,
1391 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1392 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1393 *
1394 * Return: the number of bytes sent on success, or errno otherwise
1395 */
1396static int tipc_sendmsg(struct socket *sock,
1397 struct msghdr *m, size_t dsz)
1398{
1399 struct sock *sk = sock->sk;
1400 int ret;
1401
1402 lock_sock(sk);
1403 ret = __tipc_sendmsg(sock, m, dsz);
1404 release_sock(sk);
1405
1406 return ret;
1407}
1408
1409static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1410{
1411 struct sock *sk = sock->sk;
1412 struct net *net = sock_net(sk);
1413 struct tipc_sock *tsk = tipc_sk(sk);
1414 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1415 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1416 struct list_head *clinks = &tsk->cong_links;
1417 bool syn = !tipc_sk_type_connectionless(sk);
1418 struct tipc_group *grp = tsk->group;
1419 struct tipc_msg *hdr = &tsk->phdr;
1420 struct tipc_socket_addr skaddr;
1421 struct sk_buff_head pkts;
1422 int atype, mtu, rc;
1423
1424 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1425 return -EMSGSIZE;
1426
1427 if (ua) {
1428 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1429 return -EINVAL;
1430 atype = ua->addrtype;
1431 }
1432
1433 /* If socket belongs to a communication group follow other paths */
1434 if (grp) {
1435 if (!ua)
1436 return tipc_send_group_bcast(sock, m, dlen, timeout);
1437 if (atype == TIPC_SERVICE_ADDR)
1438 return tipc_send_group_anycast(sock, m, dlen, timeout);
1439 if (atype == TIPC_SOCKET_ADDR)
1440 return tipc_send_group_unicast(sock, m, dlen, timeout);
1441 if (atype == TIPC_SERVICE_RANGE)
1442 return tipc_send_group_mcast(sock, m, dlen, timeout);
1443 return -EINVAL;
1444 }
1445
1446 if (!ua) {
1447 ua = (struct tipc_uaddr *)&tsk->peer;
1448 if (!syn && ua->family != AF_TIPC)
1449 return -EDESTADDRREQ;
1450 atype = ua->addrtype;
1451 }
1452
1453 if (unlikely(syn)) {
1454 if (sk->sk_state == TIPC_LISTEN)
1455 return -EPIPE;
1456 if (sk->sk_state != TIPC_OPEN)
1457 return -EISCONN;
1458 if (tsk->published)
1459 return -EOPNOTSUPP;
1460 if (atype == TIPC_SERVICE_ADDR)
1461 tsk->conn_addrtype = atype;
1462 msg_set_syn(hdr, 1);
1463 }
1464
1465 memset(&skaddr, 0, sizeof(skaddr));
1466
1467 /* Determine destination */
1468 if (atype == TIPC_SERVICE_RANGE) {
1469 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1470 } else if (atype == TIPC_SERVICE_ADDR) {
1471 skaddr.node = ua->lookup_node;
1472 ua->scope = tipc_node2scope(skaddr.node);
1473 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1474 return -EHOSTUNREACH;
1475 } else if (atype == TIPC_SOCKET_ADDR) {
1476 skaddr = ua->sk;
1477 } else {
1478 return -EINVAL;
1479 }
1480
1481 /* Block or return if destination link is congested */
1482 rc = tipc_wait_for_cond(sock, &timeout,
1483 !tipc_dest_find(clinks, skaddr.node, 0));
1484 if (unlikely(rc))
1485 return rc;
1486
1487 /* Finally build message header */
1488 msg_set_destnode(hdr, skaddr.node);
1489 msg_set_destport(hdr, skaddr.ref);
1490 if (atype == TIPC_SERVICE_ADDR) {
1491 msg_set_type(hdr, TIPC_NAMED_MSG);
1492 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1493 msg_set_nametype(hdr, ua->sa.type);
1494 msg_set_nameinst(hdr, ua->sa.instance);
1495 msg_set_lookup_scope(hdr, ua->scope);
1496 } else { /* TIPC_SOCKET_ADDR */
1497 msg_set_type(hdr, TIPC_DIRECT_MSG);
1498 msg_set_lookup_scope(hdr, 0);
1499 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1500 }
1501
1502 /* Add message body */
1503 __skb_queue_head_init(&pkts);
1504 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1505 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1506 if (unlikely(rc != dlen))
1507 return rc;
1508 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1509 __skb_queue_purge(&pkts);
1510 return -ENOMEM;
1511 }
1512
1513 /* Send message */
1514 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1515 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1516 if (unlikely(rc == -ELINKCONG)) {
1517 tipc_dest_push(clinks, skaddr.node, 0);
1518 tsk->cong_link_cnt++;
1519 rc = 0;
1520 }
1521
1522 if (unlikely(syn && !rc)) {
1523 tipc_set_sk_state(sk, TIPC_CONNECTING);
1524 if (dlen && timeout) {
1525 timeout = msecs_to_jiffies(timeout);
1526 tipc_wait_for_connect(sock, &timeout);
1527 }
1528 }
1529
1530 return rc ? rc : dlen;
1531}
1532
1533/**
1534 * tipc_sendstream - send stream-oriented data
1535 * @sock: socket structure
1536 * @m: data to send
1537 * @dsz: total length of data to be transmitted
1538 *
1539 * Used for SOCK_STREAM data.
1540 *
1541 * Return: the number of bytes sent on success (or partial success),
1542 * or errno if no data sent
1543 */
1544static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1545{
1546 struct sock *sk = sock->sk;
1547 int ret;
1548
1549 lock_sock(sk);
1550 ret = __tipc_sendstream(sock, m, dsz);
1551 release_sock(sk);
1552
1553 return ret;
1554}
1555
1556static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1557{
1558 struct sock *sk = sock->sk;
1559 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1560 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1561 struct sk_buff_head *txq = &sk->sk_write_queue;
1562 struct tipc_sock *tsk = tipc_sk(sk);
1563 struct tipc_msg *hdr = &tsk->phdr;
1564 struct net *net = sock_net(sk);
1565 struct sk_buff *skb;
1566 u32 dnode = tsk_peer_node(tsk);
1567 int maxnagle = tsk->maxnagle;
1568 int maxpkt = tsk->max_pkt;
1569 int send, sent = 0;
1570 int blocks, rc = 0;
1571
1572 if (unlikely(dlen > INT_MAX))
1573 return -EMSGSIZE;
1574
1575 /* Handle implicit connection setup */
1576 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1577 rc = __tipc_sendmsg(sock, m, dlen);
1578 if (dlen && dlen == rc) {
1579 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1580 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1581 }
1582 return rc;
1583 }
1584
1585 do {
1586 rc = tipc_wait_for_cond(sock, &timeout,
1587 (!tsk->cong_link_cnt &&
1588 !tsk_conn_cong(tsk) &&
1589 tipc_sk_connected(sk)));
1590 if (unlikely(rc))
1591 break;
1592 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1593 blocks = tsk->snd_backlog;
1594 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1595 send <= maxnagle) {
1596 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1597 if (unlikely(rc < 0))
1598 break;
1599 blocks += rc;
1600 tsk->msg_acc++;
1601 if (blocks <= 64 && tsk->expect_ack) {
1602 tsk->snd_backlog = blocks;
1603 sent += send;
1604 break;
1605 } else if (blocks > 64) {
1606 tsk->pkt_cnt += skb_queue_len(txq);
1607 } else {
1608 skb = skb_peek_tail(txq);
1609 if (skb) {
1610 msg_set_ack_required(buf_msg(skb));
1611 tsk->expect_ack = true;
1612 } else {
1613 tsk->expect_ack = false;
1614 }
1615 tsk->msg_acc = 0;
1616 tsk->pkt_cnt = 0;
1617 }
1618 } else {
1619 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1620 if (unlikely(rc != send))
1621 break;
1622 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1623 }
1624 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1625 TIPC_DUMP_SK_SNDQ, " ");
1626 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1627 if (unlikely(rc == -ELINKCONG)) {
1628 tsk->cong_link_cnt = 1;
1629 rc = 0;
1630 }
1631 if (likely(!rc)) {
1632 tsk->snt_unacked += blocks;
1633 tsk->snd_backlog = 0;
1634 sent += send;
1635 }
1636 } while (sent < dlen && !rc);
1637
1638 return sent ? sent : rc;
1639}
1640
1641/**
1642 * tipc_send_packet - send a connection-oriented message
1643 * @sock: socket structure
1644 * @m: message to send
1645 * @dsz: length of data to be transmitted
1646 *
1647 * Used for SOCK_SEQPACKET messages.
1648 *
1649 * Return: the number of bytes sent on success, or errno otherwise
1650 */
1651static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1652{
1653 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1654 return -EMSGSIZE;
1655
1656 return tipc_sendstream(sock, m, dsz);
1657}
1658
1659/* tipc_sk_finish_conn - complete the setup of a connection
1660 */
1661static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1662 u32 peer_node)
1663{
1664 struct sock *sk = &tsk->sk;
1665 struct net *net = sock_net(sk);
1666 struct tipc_msg *msg = &tsk->phdr;
1667
1668 msg_set_syn(msg, 0);
1669 msg_set_destnode(msg, peer_node);
1670 msg_set_destport(msg, peer_port);
1671 msg_set_type(msg, TIPC_CONN_MSG);
1672 msg_set_lookup_scope(msg, 0);
1673 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1674
1675 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1676 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1677 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1678 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1679 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1680 tsk_set_nagle(tsk);
1681 __skb_queue_purge(&sk->sk_write_queue);
1682 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1683 return;
1684
1685 /* Fall back to message based flow control */
1686 tsk->rcv_win = FLOWCTL_MSG_WIN;
1687 tsk->snd_win = FLOWCTL_MSG_WIN;
1688}
1689
1690/**
1691 * tipc_sk_set_orig_addr - capture sender's address for received message
1692 * @m: descriptor for message info
1693 * @skb: received message
1694 *
1695 * Note: Address is not captured if not requested by receiver.
1696 */
1697static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1698{
1699 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1700 struct tipc_msg *hdr = buf_msg(skb);
1701
1702 if (!srcaddr)
1703 return;
1704
1705 srcaddr->sock.family = AF_TIPC;
1706 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1707 srcaddr->sock.scope = 0;
1708 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1709 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1710 srcaddr->sock.addr.name.domain = 0;
1711 m->msg_namelen = sizeof(struct sockaddr_tipc);
1712
1713 if (!msg_in_group(hdr))
1714 return;
1715
1716 /* Group message users may also want to know sending member's id */
1717 srcaddr->member.family = AF_TIPC;
1718 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1719 srcaddr->member.scope = 0;
1720 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1721 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1722 srcaddr->member.addr.name.domain = 0;
1723 m->msg_namelen = sizeof(*srcaddr);
1724}
1725
1726/**
1727 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1728 * @m: descriptor for message info
1729 * @skb: received message buffer
1730 * @tsk: TIPC port associated with message
1731 *
1732 * Note: Ancillary data is not captured if not requested by receiver.
1733 *
1734 * Return: 0 if successful, otherwise errno
1735 */
1736static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1737 struct tipc_sock *tsk)
1738{
1739 struct tipc_msg *hdr;
1740 u32 data[3] = {0,};
1741 bool has_addr;
1742 int dlen, rc;
1743
1744 if (likely(m->msg_controllen == 0))
1745 return 0;
1746
1747 hdr = buf_msg(skb);
1748 dlen = msg_data_sz(hdr);
1749
1750 /* Capture errored message object, if any */
1751 if (msg_errcode(hdr)) {
1752 if (skb_linearize(skb))
1753 return -ENOMEM;
1754 hdr = buf_msg(skb);
1755 data[0] = msg_errcode(hdr);
1756 data[1] = dlen;
1757 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1758 if (rc || !dlen)
1759 return rc;
1760 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1761 if (rc)
1762 return rc;
1763 }
1764
1765 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1766 switch (msg_type(hdr)) {
1767 case TIPC_NAMED_MSG:
1768 has_addr = true;
1769 data[0] = msg_nametype(hdr);
1770 data[1] = msg_namelower(hdr);
1771 data[2] = data[1];
1772 break;
1773 case TIPC_MCAST_MSG:
1774 has_addr = true;
1775 data[0] = msg_nametype(hdr);
1776 data[1] = msg_namelower(hdr);
1777 data[2] = msg_nameupper(hdr);
1778 break;
1779 case TIPC_CONN_MSG:
1780 has_addr = !!tsk->conn_addrtype;
1781 data[0] = msg_nametype(&tsk->phdr);
1782 data[1] = msg_nameinst(&tsk->phdr);
1783 data[2] = data[1];
1784 break;
1785 default:
1786 has_addr = false;
1787 }
1788 if (!has_addr)
1789 return 0;
1790 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1791}
1792
1793static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1794{
1795 struct sock *sk = &tsk->sk;
1796 struct sk_buff *skb = NULL;
1797 struct tipc_msg *msg;
1798 u32 peer_port = tsk_peer_port(tsk);
1799 u32 dnode = tsk_peer_node(tsk);
1800
1801 if (!tipc_sk_connected(sk))
1802 return NULL;
1803 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1804 dnode, tsk_own_node(tsk), peer_port,
1805 tsk->portid, TIPC_OK);
1806 if (!skb)
1807 return NULL;
1808 msg = buf_msg(skb);
1809 msg_set_conn_ack(msg, tsk->rcv_unacked);
1810 tsk->rcv_unacked = 0;
1811
1812 /* Adjust to and advertize the correct window limit */
1813 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1814 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1815 msg_set_adv_win(msg, tsk->rcv_win);
1816 }
1817 return skb;
1818}
1819
1820static void tipc_sk_send_ack(struct tipc_sock *tsk)
1821{
1822 struct sk_buff *skb;
1823
1824 skb = tipc_sk_build_ack(tsk);
1825 if (!skb)
1826 return;
1827
1828 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1829 msg_link_selector(buf_msg(skb)));
1830}
1831
1832static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1833{
1834 struct sock *sk = sock->sk;
1835 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1836 long timeo = *timeop;
1837 int err = sock_error(sk);
1838
1839 if (err)
1840 return err;
1841
1842 for (;;) {
1843 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1844 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1845 err = -ENOTCONN;
1846 break;
1847 }
1848 add_wait_queue(sk_sleep(sk), &wait);
1849 release_sock(sk);
1850 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1851 sched_annotate_sleep();
1852 lock_sock(sk);
1853 remove_wait_queue(sk_sleep(sk), &wait);
1854 }
1855 err = 0;
1856 if (!skb_queue_empty(&sk->sk_receive_queue))
1857 break;
1858 err = -EAGAIN;
1859 if (!timeo)
1860 break;
1861 err = sock_intr_errno(timeo);
1862 if (signal_pending(current))
1863 break;
1864
1865 err = sock_error(sk);
1866 if (err)
1867 break;
1868 }
1869 *timeop = timeo;
1870 return err;
1871}
1872
1873/**
1874 * tipc_recvmsg - receive packet-oriented message
1875 * @sock: network socket
1876 * @m: descriptor for message info
1877 * @buflen: length of user buffer area
1878 * @flags: receive flags
1879 *
1880 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1881 * If the complete message doesn't fit in user area, truncate it.
1882 *
1883 * Return: size of returned message data, errno otherwise
1884 */
1885static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1886 size_t buflen, int flags)
1887{
1888 struct sock *sk = sock->sk;
1889 bool connected = !tipc_sk_type_connectionless(sk);
1890 struct tipc_sock *tsk = tipc_sk(sk);
1891 int rc, err, hlen, dlen, copy;
1892 struct tipc_skb_cb *skb_cb;
1893 struct sk_buff_head xmitq;
1894 struct tipc_msg *hdr;
1895 struct sk_buff *skb;
1896 bool grp_evt;
1897 long timeout;
1898
1899 /* Catch invalid receive requests */
1900 if (unlikely(!buflen))
1901 return -EINVAL;
1902
1903 lock_sock(sk);
1904 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1905 rc = -ENOTCONN;
1906 goto exit;
1907 }
1908 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1909
1910 /* Step rcv queue to first msg with data or error; wait if necessary */
1911 do {
1912 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1913 if (unlikely(rc))
1914 goto exit;
1915 skb = skb_peek(&sk->sk_receive_queue);
1916 skb_cb = TIPC_SKB_CB(skb);
1917 hdr = buf_msg(skb);
1918 dlen = msg_data_sz(hdr);
1919 hlen = msg_hdr_sz(hdr);
1920 err = msg_errcode(hdr);
1921 grp_evt = msg_is_grp_evt(hdr);
1922 if (likely(dlen || err))
1923 break;
1924 tsk_advance_rx_queue(sk);
1925 } while (1);
1926
1927 /* Collect msg meta data, including error code and rejected data */
1928 tipc_sk_set_orig_addr(m, skb);
1929 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1930 if (unlikely(rc))
1931 goto exit;
1932 hdr = buf_msg(skb);
1933
1934 /* Capture data if non-error msg, otherwise just set return value */
1935 if (likely(!err)) {
1936 int offset = skb_cb->bytes_read;
1937
1938 copy = min_t(int, dlen - offset, buflen);
1939 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1940 if (unlikely(rc))
1941 goto exit;
1942 if (unlikely(offset + copy < dlen)) {
1943 if (flags & MSG_EOR) {
1944 if (!(flags & MSG_PEEK))
1945 skb_cb->bytes_read = offset + copy;
1946 } else {
1947 m->msg_flags |= MSG_TRUNC;
1948 skb_cb->bytes_read = 0;
1949 }
1950 } else {
1951 if (flags & MSG_EOR)
1952 m->msg_flags |= MSG_EOR;
1953 skb_cb->bytes_read = 0;
1954 }
1955 } else {
1956 copy = 0;
1957 rc = 0;
1958 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1959 rc = -ECONNRESET;
1960 goto exit;
1961 }
1962 }
1963
1964 /* Mark message as group event if applicable */
1965 if (unlikely(grp_evt)) {
1966 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1967 m->msg_flags |= MSG_EOR;
1968 m->msg_flags |= MSG_OOB;
1969 copy = 0;
1970 }
1971
1972 /* Caption of data or error code/rejected data was successful */
1973 if (unlikely(flags & MSG_PEEK))
1974 goto exit;
1975
1976 /* Send group flow control advertisement when applicable */
1977 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1978 __skb_queue_head_init(&xmitq);
1979 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1980 msg_orignode(hdr), msg_origport(hdr),
1981 &xmitq);
1982 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1983 }
1984
1985 if (skb_cb->bytes_read)
1986 goto exit;
1987
1988 tsk_advance_rx_queue(sk);
1989
1990 if (likely(!connected))
1991 goto exit;
1992
1993 /* Send connection flow control advertisement when applicable */
1994 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1995 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1996 tipc_sk_send_ack(tsk);
1997exit:
1998 release_sock(sk);
1999 return rc ? rc : copy;
2000}
2001
2002/**
2003 * tipc_recvstream - receive stream-oriented data
2004 * @sock: network socket
2005 * @m: descriptor for message info
2006 * @buflen: total size of user buffer area
2007 * @flags: receive flags
2008 *
2009 * Used for SOCK_STREAM messages only. If not enough data is available
2010 * will optionally wait for more; never truncates data.
2011 *
2012 * Return: size of returned message data, errno otherwise
2013 */
2014static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2015 size_t buflen, int flags)
2016{
2017 struct sock *sk = sock->sk;
2018 struct tipc_sock *tsk = tipc_sk(sk);
2019 struct sk_buff *skb;
2020 struct tipc_msg *hdr;
2021 struct tipc_skb_cb *skb_cb;
2022 bool peek = flags & MSG_PEEK;
2023 int offset, required, copy, copied = 0;
2024 int hlen, dlen, err, rc;
2025 long timeout;
2026
2027 /* Catch invalid receive attempts */
2028 if (unlikely(!buflen))
2029 return -EINVAL;
2030
2031 lock_sock(sk);
2032
2033 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2034 rc = -ENOTCONN;
2035 goto exit;
2036 }
2037 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2038 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2039
2040 do {
2041 /* Look at first msg in receive queue; wait if necessary */
2042 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2043 if (unlikely(rc))
2044 break;
2045 skb = skb_peek(&sk->sk_receive_queue);
2046 skb_cb = TIPC_SKB_CB(skb);
2047 hdr = buf_msg(skb);
2048 dlen = msg_data_sz(hdr);
2049 hlen = msg_hdr_sz(hdr);
2050 err = msg_errcode(hdr);
2051
2052 /* Discard any empty non-errored (SYN-) message */
2053 if (unlikely(!dlen && !err)) {
2054 tsk_advance_rx_queue(sk);
2055 continue;
2056 }
2057
2058 /* Collect msg meta data, incl. error code and rejected data */
2059 if (!copied) {
2060 tipc_sk_set_orig_addr(m, skb);
2061 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2062 if (rc)
2063 break;
2064 hdr = buf_msg(skb);
2065 }
2066
2067 /* Copy data if msg ok, otherwise return error/partial data */
2068 if (likely(!err)) {
2069 offset = skb_cb->bytes_read;
2070 copy = min_t(int, dlen - offset, buflen - copied);
2071 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2072 if (unlikely(rc))
2073 break;
2074 copied += copy;
2075 offset += copy;
2076 if (unlikely(offset < dlen)) {
2077 if (!peek)
2078 skb_cb->bytes_read = offset;
2079 break;
2080 }
2081 } else {
2082 rc = 0;
2083 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2084 rc = -ECONNRESET;
2085 if (copied || rc)
2086 break;
2087 }
2088
2089 if (unlikely(peek))
2090 break;
2091
2092 tsk_advance_rx_queue(sk);
2093
2094 /* Send connection flow control advertisement when applicable */
2095 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2096 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2097 tipc_sk_send_ack(tsk);
2098
2099 /* Exit if all requested data or FIN/error received */
2100 if (copied == buflen || err)
2101 break;
2102
2103 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2104exit:
2105 release_sock(sk);
2106 return copied ? copied : rc;
2107}
2108
2109/**
2110 * tipc_write_space - wake up thread if port congestion is released
2111 * @sk: socket
2112 */
2113static void tipc_write_space(struct sock *sk)
2114{
2115 struct socket_wq *wq;
2116
2117 rcu_read_lock();
2118 wq = rcu_dereference(sk->sk_wq);
2119 if (skwq_has_sleeper(wq))
2120 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2121 EPOLLWRNORM | EPOLLWRBAND);
2122 rcu_read_unlock();
2123}
2124
2125/**
2126 * tipc_data_ready - wake up threads to indicate messages have been received
2127 * @sk: socket
2128 */
2129static void tipc_data_ready(struct sock *sk)
2130{
2131 struct socket_wq *wq;
2132
2133 trace_sk_data_ready(sk);
2134
2135 rcu_read_lock();
2136 wq = rcu_dereference(sk->sk_wq);
2137 if (skwq_has_sleeper(wq))
2138 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2139 EPOLLRDNORM | EPOLLRDBAND);
2140 rcu_read_unlock();
2141}
2142
2143static void tipc_sock_destruct(struct sock *sk)
2144{
2145 __skb_queue_purge(&sk->sk_receive_queue);
2146}
2147
2148static void tipc_sk_proto_rcv(struct sock *sk,
2149 struct sk_buff_head *inputq,
2150 struct sk_buff_head *xmitq)
2151{
2152 struct sk_buff *skb = __skb_dequeue(inputq);
2153 struct tipc_sock *tsk = tipc_sk(sk);
2154 struct tipc_msg *hdr = buf_msg(skb);
2155 struct tipc_group *grp = tsk->group;
2156 bool wakeup = false;
2157
2158 switch (msg_user(hdr)) {
2159 case CONN_MANAGER:
2160 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2161 return;
2162 case SOCK_WAKEUP:
2163 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2164 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2165 smp_wmb();
2166 tsk->cong_link_cnt--;
2167 wakeup = true;
2168 tipc_sk_push_backlog(tsk, false);
2169 break;
2170 case GROUP_PROTOCOL:
2171 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2172 break;
2173 case TOP_SRV:
2174 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2175 hdr, inputq, xmitq);
2176 break;
2177 default:
2178 break;
2179 }
2180
2181 if (wakeup)
2182 sk->sk_write_space(sk);
2183
2184 kfree_skb(skb);
2185}
2186
2187/**
2188 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2189 * @tsk: TIPC socket
2190 * @skb: pointer to message buffer.
2191 * @xmitq: for Nagle ACK if any
2192 * Return: true if message should be added to receive queue, false otherwise
2193 */
2194static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2195 struct sk_buff_head *xmitq)
2196{
2197 struct sock *sk = &tsk->sk;
2198 struct net *net = sock_net(sk);
2199 struct tipc_msg *hdr = buf_msg(skb);
2200 bool con_msg = msg_connected(hdr);
2201 u32 pport = tsk_peer_port(tsk);
2202 u32 pnode = tsk_peer_node(tsk);
2203 u32 oport = msg_origport(hdr);
2204 u32 onode = msg_orignode(hdr);
2205 int err = msg_errcode(hdr);
2206 unsigned long delay;
2207
2208 if (unlikely(msg_mcast(hdr)))
2209 return false;
2210 tsk->oneway = 0;
2211
2212 switch (sk->sk_state) {
2213 case TIPC_CONNECTING:
2214 /* Setup ACK */
2215 if (likely(con_msg)) {
2216 if (err)
2217 break;
2218 tipc_sk_finish_conn(tsk, oport, onode);
2219 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2220 /* ACK+ message with data is added to receive queue */
2221 if (msg_data_sz(hdr))
2222 return true;
2223 /* Empty ACK-, - wake up sleeping connect() and drop */
2224 sk->sk_state_change(sk);
2225 msg_set_dest_droppable(hdr, 1);
2226 return false;
2227 }
2228 /* Ignore connectionless message if not from listening socket */
2229 if (oport != pport || onode != pnode)
2230 return false;
2231
2232 /* Rejected SYN */
2233 if (err != TIPC_ERR_OVERLOAD)
2234 break;
2235
2236 /* Prepare for new setup attempt if we have a SYN clone */
2237 if (skb_queue_empty(&sk->sk_write_queue))
2238 break;
2239 get_random_bytes(&delay, 2);
2240 delay %= (tsk->conn_timeout / 4);
2241 delay = msecs_to_jiffies(delay + 100);
2242 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2243 return false;
2244 case TIPC_OPEN:
2245 case TIPC_DISCONNECTING:
2246 return false;
2247 case TIPC_LISTEN:
2248 /* Accept only SYN message */
2249 if (!msg_is_syn(hdr) &&
2250 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2251 return false;
2252 if (!con_msg && !err)
2253 return true;
2254 return false;
2255 case TIPC_ESTABLISHED:
2256 if (!skb_queue_empty(&sk->sk_write_queue))
2257 tipc_sk_push_backlog(tsk, false);
2258 /* Accept only connection-based messages sent by peer */
2259 if (likely(con_msg && !err && pport == oport &&
2260 pnode == onode)) {
2261 if (msg_ack_required(hdr)) {
2262 struct sk_buff *skb;
2263
2264 skb = tipc_sk_build_ack(tsk);
2265 if (skb) {
2266 msg_set_nagle_ack(buf_msg(skb));
2267 __skb_queue_tail(xmitq, skb);
2268 }
2269 }
2270 return true;
2271 }
2272 if (!tsk_peer_msg(tsk, hdr))
2273 return false;
2274 if (!err)
2275 return true;
2276 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2277 tipc_node_remove_conn(net, pnode, tsk->portid);
2278 sk->sk_state_change(sk);
2279 return true;
2280 default:
2281 pr_err("Unknown sk_state %u\n", sk->sk_state);
2282 }
2283 /* Abort connection setup attempt */
2284 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2285 sk->sk_err = ECONNREFUSED;
2286 sk->sk_state_change(sk);
2287 return true;
2288}
2289
2290/**
2291 * rcvbuf_limit - get proper overload limit of socket receive queue
2292 * @sk: socket
2293 * @skb: message
2294 *
2295 * For connection oriented messages, irrespective of importance,
2296 * default queue limit is 2 MB.
2297 *
2298 * For connectionless messages, queue limits are based on message
2299 * importance as follows:
2300 *
2301 * TIPC_LOW_IMPORTANCE (2 MB)
2302 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2303 * TIPC_HIGH_IMPORTANCE (8 MB)
2304 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2305 *
2306 * Return: overload limit according to corresponding message importance
2307 */
2308static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2309{
2310 struct tipc_sock *tsk = tipc_sk(sk);
2311 struct tipc_msg *hdr = buf_msg(skb);
2312
2313 if (unlikely(msg_in_group(hdr)))
2314 return READ_ONCE(sk->sk_rcvbuf);
2315
2316 if (unlikely(!msg_connected(hdr)))
2317 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2318
2319 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2320 return READ_ONCE(sk->sk_rcvbuf);
2321
2322 return FLOWCTL_MSG_LIM;
2323}
2324
2325/**
2326 * tipc_sk_filter_rcv - validate incoming message
2327 * @sk: socket
2328 * @skb: pointer to message.
2329 * @xmitq: output message area (FIXME)
2330 *
2331 * Enqueues message on receive queue if acceptable; optionally handles
2332 * disconnect indication for a connected socket.
2333 *
2334 * Called with socket lock already taken
2335 */
2336static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2337 struct sk_buff_head *xmitq)
2338{
2339 bool sk_conn = !tipc_sk_type_connectionless(sk);
2340 struct tipc_sock *tsk = tipc_sk(sk);
2341 struct tipc_group *grp = tsk->group;
2342 struct tipc_msg *hdr = buf_msg(skb);
2343 struct net *net = sock_net(sk);
2344 struct sk_buff_head inputq;
2345 int mtyp = msg_type(hdr);
2346 int limit, err = TIPC_OK;
2347
2348 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2349 TIPC_SKB_CB(skb)->bytes_read = 0;
2350 __skb_queue_head_init(&inputq);
2351 __skb_queue_tail(&inputq, skb);
2352
2353 if (unlikely(!msg_isdata(hdr)))
2354 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2355
2356 if (unlikely(grp))
2357 tipc_group_filter_msg(grp, &inputq, xmitq);
2358
2359 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2360 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2361
2362 /* Validate and add to receive buffer if there is space */
2363 while ((skb = __skb_dequeue(&inputq))) {
2364 hdr = buf_msg(skb);
2365 limit = rcvbuf_limit(sk, skb);
2366 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2367 (!sk_conn && msg_connected(hdr)) ||
2368 (!grp && msg_in_group(hdr)))
2369 err = TIPC_ERR_NO_PORT;
2370 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2371 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2372 "err_overload2!");
2373 atomic_inc(&sk->sk_drops);
2374 err = TIPC_ERR_OVERLOAD;
2375 }
2376
2377 if (unlikely(err)) {
2378 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2379 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2380 "@filter_rcv!");
2381 __skb_queue_tail(xmitq, skb);
2382 }
2383 err = TIPC_OK;
2384 continue;
2385 }
2386 __skb_queue_tail(&sk->sk_receive_queue, skb);
2387 skb_set_owner_r(skb, sk);
2388 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2389 "rcvq >90% allocated!");
2390 sk->sk_data_ready(sk);
2391 }
2392}
2393
2394/**
2395 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2396 * @sk: socket
2397 * @skb: message
2398 *
2399 * Caller must hold socket lock
2400 */
2401static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2402{
2403 unsigned int before = sk_rmem_alloc_get(sk);
2404 struct sk_buff_head xmitq;
2405 unsigned int added;
2406
2407 __skb_queue_head_init(&xmitq);
2408
2409 tipc_sk_filter_rcv(sk, skb, &xmitq);
2410 added = sk_rmem_alloc_get(sk) - before;
2411 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2412
2413 /* Send pending response/rejected messages, if any */
2414 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2415 return 0;
2416}
2417
2418/**
2419 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2420 * inputq and try adding them to socket or backlog queue
2421 * @inputq: list of incoming buffers with potentially different destinations
2422 * @sk: socket where the buffers should be enqueued
2423 * @dport: port number for the socket
2424 * @xmitq: output queue
2425 *
2426 * Caller must hold socket lock
2427 */
2428static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2429 u32 dport, struct sk_buff_head *xmitq)
2430{
2431 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2432 struct sk_buff *skb;
2433 unsigned int lim;
2434 atomic_t *dcnt;
2435 u32 onode;
2436
2437 while (skb_queue_len(inputq)) {
2438 if (unlikely(time_after_eq(jiffies, time_limit)))
2439 return;
2440
2441 skb = tipc_skb_dequeue(inputq, dport);
2442 if (unlikely(!skb))
2443 return;
2444
2445 /* Add message directly to receive queue if possible */
2446 if (!sock_owned_by_user(sk)) {
2447 tipc_sk_filter_rcv(sk, skb, xmitq);
2448 continue;
2449 }
2450
2451 /* Try backlog, compensating for double-counted bytes */
2452 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2453 if (!sk->sk_backlog.len)
2454 atomic_set(dcnt, 0);
2455 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2456 if (likely(!sk_add_backlog(sk, skb, lim))) {
2457 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2458 "bklg & rcvq >90% allocated!");
2459 continue;
2460 }
2461
2462 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2463 /* Overload => reject message back to sender */
2464 onode = tipc_own_addr(sock_net(sk));
2465 atomic_inc(&sk->sk_drops);
2466 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2467 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2468 "@sk_enqueue!");
2469 __skb_queue_tail(xmitq, skb);
2470 }
2471 break;
2472 }
2473}
2474
2475/**
2476 * tipc_sk_rcv - handle a chain of incoming buffers
2477 * @net: the associated network namespace
2478 * @inputq: buffer list containing the buffers
2479 * Consumes all buffers in list until inputq is empty
2480 * Note: may be called in multiple threads referring to the same queue
2481 */
2482void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2483{
2484 struct sk_buff_head xmitq;
2485 u32 dnode, dport = 0;
2486 int err;
2487 struct tipc_sock *tsk;
2488 struct sock *sk;
2489 struct sk_buff *skb;
2490
2491 __skb_queue_head_init(&xmitq);
2492 while (skb_queue_len(inputq)) {
2493 dport = tipc_skb_peek_port(inputq, dport);
2494 tsk = tipc_sk_lookup(net, dport);
2495
2496 if (likely(tsk)) {
2497 sk = &tsk->sk;
2498 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2499 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2500 spin_unlock_bh(&sk->sk_lock.slock);
2501 }
2502 /* Send pending response/rejected messages, if any */
2503 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2504 sock_put(sk);
2505 continue;
2506 }
2507 /* No destination socket => dequeue skb if still there */
2508 skb = tipc_skb_dequeue(inputq, dport);
2509 if (!skb)
2510 return;
2511
2512 /* Try secondary lookup if unresolved named message */
2513 err = TIPC_ERR_NO_PORT;
2514 if (tipc_msg_lookup_dest(net, skb, &err))
2515 goto xmit;
2516
2517 /* Prepare for message rejection */
2518 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2519 continue;
2520
2521 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2522xmit:
2523 dnode = msg_destnode(buf_msg(skb));
2524 tipc_node_xmit_skb(net, skb, dnode, dport);
2525 }
2526}
2527
2528static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2529{
2530 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2531 struct sock *sk = sock->sk;
2532 int done;
2533
2534 do {
2535 int err = sock_error(sk);
2536 if (err)
2537 return err;
2538 if (!*timeo_p)
2539 return -ETIMEDOUT;
2540 if (signal_pending(current))
2541 return sock_intr_errno(*timeo_p);
2542 if (sk->sk_state == TIPC_DISCONNECTING)
2543 break;
2544
2545 add_wait_queue(sk_sleep(sk), &wait);
2546 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2547 &wait);
2548 remove_wait_queue(sk_sleep(sk), &wait);
2549 } while (!done);
2550 return 0;
2551}
2552
2553static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2554{
2555 if (addr->family != AF_TIPC)
2556 return false;
2557 if (addr->addrtype == TIPC_SERVICE_RANGE)
2558 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2559 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2560 addr->addrtype == TIPC_SOCKET_ADDR);
2561}
2562
2563/**
2564 * tipc_connect - establish a connection to another TIPC port
2565 * @sock: socket structure
2566 * @dest: socket address for destination port
2567 * @destlen: size of socket address data structure
2568 * @flags: file-related flags associated with socket
2569 *
2570 * Return: 0 on success, errno otherwise
2571 */
2572static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2573 int destlen, int flags)
2574{
2575 struct sock *sk = sock->sk;
2576 struct tipc_sock *tsk = tipc_sk(sk);
2577 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2578 struct msghdr m = {NULL,};
2579 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2580 int previous;
2581 int res = 0;
2582
2583 if (destlen != sizeof(struct sockaddr_tipc))
2584 return -EINVAL;
2585
2586 lock_sock(sk);
2587
2588 if (tsk->group) {
2589 res = -EINVAL;
2590 goto exit;
2591 }
2592
2593 if (dst->family == AF_UNSPEC) {
2594 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2595 if (!tipc_sk_type_connectionless(sk))
2596 res = -EINVAL;
2597 goto exit;
2598 }
2599 if (!tipc_sockaddr_is_sane(dst)) {
2600 res = -EINVAL;
2601 goto exit;
2602 }
2603 /* DGRAM/RDM connect(), just save the destaddr */
2604 if (tipc_sk_type_connectionless(sk)) {
2605 memcpy(&tsk->peer, dest, destlen);
2606 goto exit;
2607 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2608 res = -EINVAL;
2609 goto exit;
2610 }
2611
2612 previous = sk->sk_state;
2613
2614 switch (sk->sk_state) {
2615 case TIPC_OPEN:
2616 /* Send a 'SYN-' to destination */
2617 m.msg_name = dest;
2618 m.msg_namelen = destlen;
2619 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2620
2621 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2622 * indicate send_msg() is never blocked.
2623 */
2624 if (!timeout)
2625 m.msg_flags = MSG_DONTWAIT;
2626
2627 res = __tipc_sendmsg(sock, &m, 0);
2628 if ((res < 0) && (res != -EWOULDBLOCK))
2629 goto exit;
2630
2631 /* Just entered TIPC_CONNECTING state; the only
2632 * difference is that return value in non-blocking
2633 * case is EINPROGRESS, rather than EALREADY.
2634 */
2635 res = -EINPROGRESS;
2636 fallthrough;
2637 case TIPC_CONNECTING:
2638 if (!timeout) {
2639 if (previous == TIPC_CONNECTING)
2640 res = -EALREADY;
2641 goto exit;
2642 }
2643 timeout = msecs_to_jiffies(timeout);
2644 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2645 res = tipc_wait_for_connect(sock, &timeout);
2646 break;
2647 case TIPC_ESTABLISHED:
2648 res = -EISCONN;
2649 break;
2650 default:
2651 res = -EINVAL;
2652 }
2653
2654exit:
2655 release_sock(sk);
2656 return res;
2657}
2658
2659/**
2660 * tipc_listen - allow socket to listen for incoming connections
2661 * @sock: socket structure
2662 * @len: (unused)
2663 *
2664 * Return: 0 on success, errno otherwise
2665 */
2666static int tipc_listen(struct socket *sock, int len)
2667{
2668 struct sock *sk = sock->sk;
2669 int res;
2670
2671 lock_sock(sk);
2672 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2673 release_sock(sk);
2674
2675 return res;
2676}
2677
2678static int tipc_wait_for_accept(struct socket *sock, long timeo)
2679{
2680 struct sock *sk = sock->sk;
2681 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2682 int err;
2683
2684 /* True wake-one mechanism for incoming connections: only
2685 * one process gets woken up, not the 'whole herd'.
2686 * Since we do not 'race & poll' for established sockets
2687 * anymore, the common case will execute the loop only once.
2688 */
2689 for (;;) {
2690 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2691 add_wait_queue(sk_sleep(sk), &wait);
2692 release_sock(sk);
2693 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2694 lock_sock(sk);
2695 remove_wait_queue(sk_sleep(sk), &wait);
2696 }
2697 err = 0;
2698 if (!skb_queue_empty(&sk->sk_receive_queue))
2699 break;
2700 err = -EAGAIN;
2701 if (!timeo)
2702 break;
2703 err = sock_intr_errno(timeo);
2704 if (signal_pending(current))
2705 break;
2706 }
2707 return err;
2708}
2709
2710/**
2711 * tipc_accept - wait for connection request
2712 * @sock: listening socket
2713 * @new_sock: new socket that is to be connected
2714 * @flags: file-related flags associated with socket
2715 * @kern: caused by kernel or by userspace?
2716 *
2717 * Return: 0 on success, errno otherwise
2718 */
2719static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2720 bool kern)
2721{
2722 struct sock *new_sk, *sk = sock->sk;
2723 struct tipc_sock *new_tsock;
2724 struct msghdr m = {NULL,};
2725 struct tipc_msg *msg;
2726 struct sk_buff *buf;
2727 long timeo;
2728 int res;
2729
2730 lock_sock(sk);
2731
2732 if (sk->sk_state != TIPC_LISTEN) {
2733 res = -EINVAL;
2734 goto exit;
2735 }
2736 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2737 res = tipc_wait_for_accept(sock, timeo);
2738 if (res)
2739 goto exit;
2740
2741 buf = skb_peek(&sk->sk_receive_queue);
2742
2743 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2744 if (res)
2745 goto exit;
2746 security_sk_clone(sock->sk, new_sock->sk);
2747
2748 new_sk = new_sock->sk;
2749 new_tsock = tipc_sk(new_sk);
2750 msg = buf_msg(buf);
2751
2752 /* we lock on new_sk; but lockdep sees the lock on sk */
2753 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2754
2755 /*
2756 * Reject any stray messages received by new socket
2757 * before the socket lock was taken (very, very unlikely)
2758 */
2759 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2760
2761 /* Connect new socket to it's peer */
2762 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2763
2764 tsk_set_importance(new_sk, msg_importance(msg));
2765 if (msg_named(msg)) {
2766 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2767 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2768 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2769 }
2770
2771 /*
2772 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2773 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2774 */
2775 if (!msg_data_sz(msg)) {
2776 tsk_advance_rx_queue(sk);
2777 } else {
2778 __skb_dequeue(&sk->sk_receive_queue);
2779 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2780 skb_set_owner_r(buf, new_sk);
2781 }
2782 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2783 __tipc_sendstream(new_sock, &m, 0);
2784 release_sock(new_sk);
2785exit:
2786 release_sock(sk);
2787 return res;
2788}
2789
2790/**
2791 * tipc_shutdown - shutdown socket connection
2792 * @sock: socket structure
2793 * @how: direction to close (must be SHUT_RDWR)
2794 *
2795 * Terminates connection (if necessary), then purges socket's receive queue.
2796 *
2797 * Return: 0 on success, errno otherwise
2798 */
2799static int tipc_shutdown(struct socket *sock, int how)
2800{
2801 struct sock *sk = sock->sk;
2802 int res;
2803
2804 if (how != SHUT_RDWR)
2805 return -EINVAL;
2806
2807 lock_sock(sk);
2808
2809 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2810 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2811 sk->sk_shutdown = SHUTDOWN_MASK;
2812
2813 if (sk->sk_state == TIPC_DISCONNECTING) {
2814 /* Discard any unreceived messages */
2815 __skb_queue_purge(&sk->sk_receive_queue);
2816
2817 res = 0;
2818 } else {
2819 res = -ENOTCONN;
2820 }
2821 /* Wake up anyone sleeping in poll. */
2822 sk->sk_state_change(sk);
2823
2824 release_sock(sk);
2825 return res;
2826}
2827
2828static void tipc_sk_check_probing_state(struct sock *sk,
2829 struct sk_buff_head *list)
2830{
2831 struct tipc_sock *tsk = tipc_sk(sk);
2832 u32 pnode = tsk_peer_node(tsk);
2833 u32 pport = tsk_peer_port(tsk);
2834 u32 self = tsk_own_node(tsk);
2835 u32 oport = tsk->portid;
2836 struct sk_buff *skb;
2837
2838 if (tsk->probe_unacked) {
2839 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2840 sk->sk_err = ECONNABORTED;
2841 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2842 sk->sk_state_change(sk);
2843 return;
2844 }
2845 /* Prepare new probe */
2846 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2847 pnode, self, pport, oport, TIPC_OK);
2848 if (skb)
2849 __skb_queue_tail(list, skb);
2850 tsk->probe_unacked = true;
2851 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2852}
2853
2854static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2855{
2856 struct tipc_sock *tsk = tipc_sk(sk);
2857
2858 /* Try again later if dest link is congested */
2859 if (tsk->cong_link_cnt) {
2860 sk_reset_timer(sk, &sk->sk_timer,
2861 jiffies + msecs_to_jiffies(100));
2862 return;
2863 }
2864 /* Prepare SYN for retransmit */
2865 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2866}
2867
2868static void tipc_sk_timeout(struct timer_list *t)
2869{
2870 struct sock *sk = from_timer(sk, t, sk_timer);
2871 struct tipc_sock *tsk = tipc_sk(sk);
2872 u32 pnode = tsk_peer_node(tsk);
2873 struct sk_buff_head list;
2874 int rc = 0;
2875
2876 __skb_queue_head_init(&list);
2877 bh_lock_sock(sk);
2878
2879 /* Try again later if socket is busy */
2880 if (sock_owned_by_user(sk)) {
2881 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2882 bh_unlock_sock(sk);
2883 sock_put(sk);
2884 return;
2885 }
2886
2887 if (sk->sk_state == TIPC_ESTABLISHED)
2888 tipc_sk_check_probing_state(sk, &list);
2889 else if (sk->sk_state == TIPC_CONNECTING)
2890 tipc_sk_retry_connect(sk, &list);
2891
2892 bh_unlock_sock(sk);
2893
2894 if (!skb_queue_empty(&list))
2895 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2896
2897 /* SYN messages may cause link congestion */
2898 if (rc == -ELINKCONG) {
2899 tipc_dest_push(&tsk->cong_links, pnode, 0);
2900 tsk->cong_link_cnt = 1;
2901 }
2902 sock_put(sk);
2903}
2904
2905static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2906{
2907 struct sock *sk = &tsk->sk;
2908 struct net *net = sock_net(sk);
2909 struct tipc_socket_addr skaddr;
2910 struct publication *p;
2911 u32 key;
2912
2913 if (tipc_sk_connected(sk))
2914 return -EINVAL;
2915 key = tsk->portid + tsk->pub_count + 1;
2916 if (key == tsk->portid)
2917 return -EADDRINUSE;
2918 skaddr.ref = tsk->portid;
2919 skaddr.node = tipc_own_addr(net);
2920 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2921 if (unlikely(!p))
2922 return -EINVAL;
2923
2924 list_add(&p->binding_sock, &tsk->publications);
2925 tsk->pub_count++;
2926 tsk->published = true;
2927 return 0;
2928}
2929
2930static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2931{
2932 struct net *net = sock_net(&tsk->sk);
2933 struct publication *safe, *p;
2934 struct tipc_uaddr _ua;
2935 int rc = -EINVAL;
2936
2937 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2938 if (!ua) {
2939 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2940 p->sr.type, p->sr.lower, p->sr.upper);
2941 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2942 continue;
2943 }
2944 /* Unbind specific publication */
2945 if (p->scope != ua->scope)
2946 continue;
2947 if (p->sr.type != ua->sr.type)
2948 continue;
2949 if (p->sr.lower != ua->sr.lower)
2950 continue;
2951 if (p->sr.upper != ua->sr.upper)
2952 break;
2953 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2954 rc = 0;
2955 break;
2956 }
2957 if (list_empty(&tsk->publications)) {
2958 tsk->published = 0;
2959 rc = 0;
2960 }
2961 return rc;
2962}
2963
2964/* tipc_sk_reinit: set non-zero address in all existing sockets
2965 * when we go from standalone to network mode.
2966 */
2967void tipc_sk_reinit(struct net *net)
2968{
2969 struct tipc_net *tn = net_generic(net, tipc_net_id);
2970 struct rhashtable_iter iter;
2971 struct tipc_sock *tsk;
2972 struct tipc_msg *msg;
2973
2974 rhashtable_walk_enter(&tn->sk_rht, &iter);
2975
2976 do {
2977 rhashtable_walk_start(&iter);
2978
2979 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2980 sock_hold(&tsk->sk);
2981 rhashtable_walk_stop(&iter);
2982 lock_sock(&tsk->sk);
2983 msg = &tsk->phdr;
2984 msg_set_prevnode(msg, tipc_own_addr(net));
2985 msg_set_orignode(msg, tipc_own_addr(net));
2986 release_sock(&tsk->sk);
2987 rhashtable_walk_start(&iter);
2988 sock_put(&tsk->sk);
2989 }
2990
2991 rhashtable_walk_stop(&iter);
2992 } while (tsk == ERR_PTR(-EAGAIN));
2993
2994 rhashtable_walk_exit(&iter);
2995}
2996
2997static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2998{
2999 struct tipc_net *tn = net_generic(net, tipc_net_id);
3000 struct tipc_sock *tsk;
3001
3002 rcu_read_lock();
3003 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
3004 if (tsk)
3005 sock_hold(&tsk->sk);
3006 rcu_read_unlock();
3007
3008 return tsk;
3009}
3010
3011static int tipc_sk_insert(struct tipc_sock *tsk)
3012{
3013 struct sock *sk = &tsk->sk;
3014 struct net *net = sock_net(sk);
3015 struct tipc_net *tn = net_generic(net, tipc_net_id);
3016 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3017 u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT;
3018
3019 while (remaining--) {
3020 portid++;
3021 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3022 portid = TIPC_MIN_PORT;
3023 tsk->portid = portid;
3024 sock_hold(&tsk->sk);
3025 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3026 tsk_rht_params))
3027 return 0;
3028 sock_put(&tsk->sk);
3029 }
3030
3031 return -1;
3032}
3033
3034static void tipc_sk_remove(struct tipc_sock *tsk)
3035{
3036 struct sock *sk = &tsk->sk;
3037 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3038
3039 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3040 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3041 __sock_put(sk);
3042 }
3043}
3044
3045static const struct rhashtable_params tsk_rht_params = {
3046 .nelem_hint = 192,
3047 .head_offset = offsetof(struct tipc_sock, node),
3048 .key_offset = offsetof(struct tipc_sock, portid),
3049 .key_len = sizeof(u32), /* portid */
3050 .max_size = 1048576,
3051 .min_size = 256,
3052 .automatic_shrinking = true,
3053};
3054
3055int tipc_sk_rht_init(struct net *net)
3056{
3057 struct tipc_net *tn = net_generic(net, tipc_net_id);
3058
3059 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3060}
3061
3062void tipc_sk_rht_destroy(struct net *net)
3063{
3064 struct tipc_net *tn = net_generic(net, tipc_net_id);
3065
3066 /* Wait for socket readers to complete */
3067 synchronize_net();
3068
3069 rhashtable_destroy(&tn->sk_rht);
3070}
3071
3072static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3073{
3074 struct net *net = sock_net(&tsk->sk);
3075 struct tipc_group *grp = tsk->group;
3076 struct tipc_msg *hdr = &tsk->phdr;
3077 struct tipc_uaddr ua;
3078 int rc;
3079
3080 if (mreq->type < TIPC_RESERVED_TYPES)
3081 return -EACCES;
3082 if (mreq->scope > TIPC_NODE_SCOPE)
3083 return -EINVAL;
3084 if (mreq->scope != TIPC_NODE_SCOPE)
3085 mreq->scope = TIPC_CLUSTER_SCOPE;
3086 if (grp)
3087 return -EACCES;
3088 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3089 if (!grp)
3090 return -ENOMEM;
3091 tsk->group = grp;
3092 msg_set_lookup_scope(hdr, mreq->scope);
3093 msg_set_nametype(hdr, mreq->type);
3094 msg_set_dest_droppable(hdr, true);
3095 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3096 mreq->type, mreq->instance, mreq->instance);
3097 tipc_nametbl_build_group(net, grp, &ua);
3098 rc = tipc_sk_publish(tsk, &ua);
3099 if (rc) {
3100 tipc_group_delete(net, grp);
3101 tsk->group = NULL;
3102 return rc;
3103 }
3104 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3105 tsk->mc_method.rcast = true;
3106 tsk->mc_method.mandatory = true;
3107 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3108 return rc;
3109}
3110
3111static int tipc_sk_leave(struct tipc_sock *tsk)
3112{
3113 struct net *net = sock_net(&tsk->sk);
3114 struct tipc_group *grp = tsk->group;
3115 struct tipc_uaddr ua;
3116 int scope;
3117
3118 if (!grp)
3119 return -EINVAL;
3120 ua.addrtype = TIPC_SERVICE_RANGE;
3121 tipc_group_self(grp, &ua.sr, &scope);
3122 ua.scope = scope;
3123 tipc_group_delete(net, grp);
3124 tsk->group = NULL;
3125 tipc_sk_withdraw(tsk, &ua);
3126 return 0;
3127}
3128
3129/**
3130 * tipc_setsockopt - set socket option
3131 * @sock: socket structure
3132 * @lvl: option level
3133 * @opt: option identifier
3134 * @ov: pointer to new option value
3135 * @ol: length of option value
3136 *
3137 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3138 * (to ease compatibility).
3139 *
3140 * Return: 0 on success, errno otherwise
3141 */
3142static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3143 sockptr_t ov, unsigned int ol)
3144{
3145 struct sock *sk = sock->sk;
3146 struct tipc_sock *tsk = tipc_sk(sk);
3147 struct tipc_group_req mreq;
3148 u32 value = 0;
3149 int res = 0;
3150
3151 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3152 return 0;
3153 if (lvl != SOL_TIPC)
3154 return -ENOPROTOOPT;
3155
3156 switch (opt) {
3157 case TIPC_IMPORTANCE:
3158 case TIPC_SRC_DROPPABLE:
3159 case TIPC_DEST_DROPPABLE:
3160 case TIPC_CONN_TIMEOUT:
3161 case TIPC_NODELAY:
3162 if (ol < sizeof(value))
3163 return -EINVAL;
3164 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3165 return -EFAULT;
3166 break;
3167 case TIPC_GROUP_JOIN:
3168 if (ol < sizeof(mreq))
3169 return -EINVAL;
3170 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3171 return -EFAULT;
3172 break;
3173 default:
3174 if (!sockptr_is_null(ov) || ol)
3175 return -EINVAL;
3176 }
3177
3178 lock_sock(sk);
3179
3180 switch (opt) {
3181 case TIPC_IMPORTANCE:
3182 res = tsk_set_importance(sk, value);
3183 break;
3184 case TIPC_SRC_DROPPABLE:
3185 if (sock->type != SOCK_STREAM)
3186 tsk_set_unreliable(tsk, value);
3187 else
3188 res = -ENOPROTOOPT;
3189 break;
3190 case TIPC_DEST_DROPPABLE:
3191 tsk_set_unreturnable(tsk, value);
3192 break;
3193 case TIPC_CONN_TIMEOUT:
3194 tipc_sk(sk)->conn_timeout = value;
3195 break;
3196 case TIPC_MCAST_BROADCAST:
3197 tsk->mc_method.rcast = false;
3198 tsk->mc_method.mandatory = true;
3199 break;
3200 case TIPC_MCAST_REPLICAST:
3201 tsk->mc_method.rcast = true;
3202 tsk->mc_method.mandatory = true;
3203 break;
3204 case TIPC_GROUP_JOIN:
3205 res = tipc_sk_join(tsk, &mreq);
3206 break;
3207 case TIPC_GROUP_LEAVE:
3208 res = tipc_sk_leave(tsk);
3209 break;
3210 case TIPC_NODELAY:
3211 tsk->nodelay = !!value;
3212 tsk_set_nagle(tsk);
3213 break;
3214 default:
3215 res = -EINVAL;
3216 }
3217
3218 release_sock(sk);
3219
3220 return res;
3221}
3222
3223/**
3224 * tipc_getsockopt - get socket option
3225 * @sock: socket structure
3226 * @lvl: option level
3227 * @opt: option identifier
3228 * @ov: receptacle for option value
3229 * @ol: receptacle for length of option value
3230 *
3231 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3232 * (to ease compatibility).
3233 *
3234 * Return: 0 on success, errno otherwise
3235 */
3236static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3237 char __user *ov, int __user *ol)
3238{
3239 struct sock *sk = sock->sk;
3240 struct tipc_sock *tsk = tipc_sk(sk);
3241 struct tipc_service_range seq;
3242 int len, scope;
3243 u32 value;
3244 int res;
3245
3246 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3247 return put_user(0, ol);
3248 if (lvl != SOL_TIPC)
3249 return -ENOPROTOOPT;
3250 res = get_user(len, ol);
3251 if (res)
3252 return res;
3253
3254 lock_sock(sk);
3255
3256 switch (opt) {
3257 case TIPC_IMPORTANCE:
3258 value = tsk_importance(tsk);
3259 break;
3260 case TIPC_SRC_DROPPABLE:
3261 value = tsk_unreliable(tsk);
3262 break;
3263 case TIPC_DEST_DROPPABLE:
3264 value = tsk_unreturnable(tsk);
3265 break;
3266 case TIPC_CONN_TIMEOUT:
3267 value = tsk->conn_timeout;
3268 /* no need to set "res", since already 0 at this point */
3269 break;
3270 case TIPC_NODE_RECVQ_DEPTH:
3271 value = 0; /* was tipc_queue_size, now obsolete */
3272 break;
3273 case TIPC_SOCK_RECVQ_DEPTH:
3274 value = skb_queue_len(&sk->sk_receive_queue);
3275 break;
3276 case TIPC_SOCK_RECVQ_USED:
3277 value = sk_rmem_alloc_get(sk);
3278 break;
3279 case TIPC_GROUP_JOIN:
3280 seq.type = 0;
3281 if (tsk->group)
3282 tipc_group_self(tsk->group, &seq, &scope);
3283 value = seq.type;
3284 break;
3285 default:
3286 res = -EINVAL;
3287 }
3288
3289 release_sock(sk);
3290
3291 if (res)
3292 return res; /* "get" failed */
3293
3294 if (len < sizeof(value))
3295 return -EINVAL;
3296
3297 if (copy_to_user(ov, &value, sizeof(value)))
3298 return -EFAULT;
3299
3300 return put_user(sizeof(value), ol);
3301}
3302
3303static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3304{
3305 struct net *net = sock_net(sock->sk);
3306 struct tipc_sioc_nodeid_req nr = {0};
3307 struct tipc_sioc_ln_req lnr;
3308 void __user *argp = (void __user *)arg;
3309
3310 switch (cmd) {
3311 case SIOCGETLINKNAME:
3312 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3313 return -EFAULT;
3314 if (!tipc_node_get_linkname(net,
3315 lnr.bearer_id & 0xffff, lnr.peer,
3316 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3317 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3318 return -EFAULT;
3319 return 0;
3320 }
3321 return -EADDRNOTAVAIL;
3322 case SIOCGETNODEID:
3323 if (copy_from_user(&nr, argp, sizeof(nr)))
3324 return -EFAULT;
3325 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3326 return -EADDRNOTAVAIL;
3327 if (copy_to_user(argp, &nr, sizeof(nr)))
3328 return -EFAULT;
3329 return 0;
3330 default:
3331 return -ENOIOCTLCMD;
3332 }
3333}
3334
3335static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3336{
3337 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3338 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3339 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3340
3341 tsk1->peer.family = AF_TIPC;
3342 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3343 tsk1->peer.scope = TIPC_NODE_SCOPE;
3344 tsk1->peer.addr.id.ref = tsk2->portid;
3345 tsk1->peer.addr.id.node = onode;
3346 tsk2->peer.family = AF_TIPC;
3347 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3348 tsk2->peer.scope = TIPC_NODE_SCOPE;
3349 tsk2->peer.addr.id.ref = tsk1->portid;
3350 tsk2->peer.addr.id.node = onode;
3351
3352 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3353 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3354 return 0;
3355}
3356
3357/* Protocol switches for the various types of TIPC sockets */
3358
3359static const struct proto_ops msg_ops = {
3360 .owner = THIS_MODULE,
3361 .family = AF_TIPC,
3362 .release = tipc_release,
3363 .bind = tipc_bind,
3364 .connect = tipc_connect,
3365 .socketpair = tipc_socketpair,
3366 .accept = sock_no_accept,
3367 .getname = tipc_getname,
3368 .poll = tipc_poll,
3369 .ioctl = tipc_ioctl,
3370 .listen = sock_no_listen,
3371 .shutdown = tipc_shutdown,
3372 .setsockopt = tipc_setsockopt,
3373 .getsockopt = tipc_getsockopt,
3374 .sendmsg = tipc_sendmsg,
3375 .recvmsg = tipc_recvmsg,
3376 .mmap = sock_no_mmap,
3377};
3378
3379static const struct proto_ops packet_ops = {
3380 .owner = THIS_MODULE,
3381 .family = AF_TIPC,
3382 .release = tipc_release,
3383 .bind = tipc_bind,
3384 .connect = tipc_connect,
3385 .socketpair = tipc_socketpair,
3386 .accept = tipc_accept,
3387 .getname = tipc_getname,
3388 .poll = tipc_poll,
3389 .ioctl = tipc_ioctl,
3390 .listen = tipc_listen,
3391 .shutdown = tipc_shutdown,
3392 .setsockopt = tipc_setsockopt,
3393 .getsockopt = tipc_getsockopt,
3394 .sendmsg = tipc_send_packet,
3395 .recvmsg = tipc_recvmsg,
3396 .mmap = sock_no_mmap,
3397};
3398
3399static const struct proto_ops stream_ops = {
3400 .owner = THIS_MODULE,
3401 .family = AF_TIPC,
3402 .release = tipc_release,
3403 .bind = tipc_bind,
3404 .connect = tipc_connect,
3405 .socketpair = tipc_socketpair,
3406 .accept = tipc_accept,
3407 .getname = tipc_getname,
3408 .poll = tipc_poll,
3409 .ioctl = tipc_ioctl,
3410 .listen = tipc_listen,
3411 .shutdown = tipc_shutdown,
3412 .setsockopt = tipc_setsockopt,
3413 .getsockopt = tipc_getsockopt,
3414 .sendmsg = tipc_sendstream,
3415 .recvmsg = tipc_recvstream,
3416 .mmap = sock_no_mmap,
3417};
3418
3419static const struct net_proto_family tipc_family_ops = {
3420 .owner = THIS_MODULE,
3421 .family = AF_TIPC,
3422 .create = tipc_sk_create
3423};
3424
3425static struct proto tipc_proto = {
3426 .name = "TIPC",
3427 .owner = THIS_MODULE,
3428 .obj_size = sizeof(struct tipc_sock),
3429 .sysctl_rmem = sysctl_tipc_rmem
3430};
3431
3432/**
3433 * tipc_socket_init - initialize TIPC socket interface
3434 *
3435 * Return: 0 on success, errno otherwise
3436 */
3437int tipc_socket_init(void)
3438{
3439 int res;
3440
3441 res = proto_register(&tipc_proto, 1);
3442 if (res) {
3443 pr_err("Failed to register TIPC protocol type\n");
3444 goto out;
3445 }
3446
3447 res = sock_register(&tipc_family_ops);
3448 if (res) {
3449 pr_err("Failed to register TIPC socket type\n");
3450 proto_unregister(&tipc_proto);
3451 goto out;
3452 }
3453 out:
3454 return res;
3455}
3456
3457/**
3458 * tipc_socket_stop - stop TIPC socket interface
3459 */
3460void tipc_socket_stop(void)
3461{
3462 sock_unregister(tipc_family_ops.family);
3463 proto_unregister(&tipc_proto);
3464}
3465
3466/* Caller should hold socket lock for the passed tipc socket. */
3467static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3468{
3469 u32 peer_node, peer_port;
3470 u32 conn_type, conn_instance;
3471 struct nlattr *nest;
3472
3473 peer_node = tsk_peer_node(tsk);
3474 peer_port = tsk_peer_port(tsk);
3475 conn_type = msg_nametype(&tsk->phdr);
3476 conn_instance = msg_nameinst(&tsk->phdr);
3477 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3478 if (!nest)
3479 return -EMSGSIZE;
3480
3481 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3482 goto msg_full;
3483 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3484 goto msg_full;
3485
3486 if (tsk->conn_addrtype != 0) {
3487 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3488 goto msg_full;
3489 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3490 goto msg_full;
3491 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3492 goto msg_full;
3493 }
3494 nla_nest_end(skb, nest);
3495
3496 return 0;
3497
3498msg_full:
3499 nla_nest_cancel(skb, nest);
3500
3501 return -EMSGSIZE;
3502}
3503
3504static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3505 *tsk)
3506{
3507 struct net *net = sock_net(skb->sk);
3508 struct sock *sk = &tsk->sk;
3509
3510 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3511 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3512 return -EMSGSIZE;
3513
3514 if (tipc_sk_connected(sk)) {
3515 if (__tipc_nl_add_sk_con(skb, tsk))
3516 return -EMSGSIZE;
3517 } else if (!list_empty(&tsk->publications)) {
3518 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3519 return -EMSGSIZE;
3520 }
3521 return 0;
3522}
3523
3524/* Caller should hold socket lock for the passed tipc socket. */
3525static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3526 struct tipc_sock *tsk)
3527{
3528 struct nlattr *attrs;
3529 void *hdr;
3530
3531 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3532 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3533 if (!hdr)
3534 goto msg_cancel;
3535
3536 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3537 if (!attrs)
3538 goto genlmsg_cancel;
3539
3540 if (__tipc_nl_add_sk_info(skb, tsk))
3541 goto attr_msg_cancel;
3542
3543 nla_nest_end(skb, attrs);
3544 genlmsg_end(skb, hdr);
3545
3546 return 0;
3547
3548attr_msg_cancel:
3549 nla_nest_cancel(skb, attrs);
3550genlmsg_cancel:
3551 genlmsg_cancel(skb, hdr);
3552msg_cancel:
3553 return -EMSGSIZE;
3554}
3555
3556int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3557 int (*skb_handler)(struct sk_buff *skb,
3558 struct netlink_callback *cb,
3559 struct tipc_sock *tsk))
3560{
3561 struct rhashtable_iter *iter = (void *)cb->args[4];
3562 struct tipc_sock *tsk;
3563 int err;
3564
3565 rhashtable_walk_start(iter);
3566 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3567 if (IS_ERR(tsk)) {
3568 err = PTR_ERR(tsk);
3569 if (err == -EAGAIN) {
3570 err = 0;
3571 continue;
3572 }
3573 break;
3574 }
3575
3576 sock_hold(&tsk->sk);
3577 rhashtable_walk_stop(iter);
3578 lock_sock(&tsk->sk);
3579 err = skb_handler(skb, cb, tsk);
3580 if (err) {
3581 release_sock(&tsk->sk);
3582 sock_put(&tsk->sk);
3583 goto out;
3584 }
3585 release_sock(&tsk->sk);
3586 rhashtable_walk_start(iter);
3587 sock_put(&tsk->sk);
3588 }
3589 rhashtable_walk_stop(iter);
3590out:
3591 return skb->len;
3592}
3593EXPORT_SYMBOL(tipc_nl_sk_walk);
3594
3595int tipc_dump_start(struct netlink_callback *cb)
3596{
3597 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3598}
3599EXPORT_SYMBOL(tipc_dump_start);
3600
3601int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3602{
3603 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3604 struct rhashtable_iter *iter = (void *)cb->args[4];
3605 struct tipc_net *tn = tipc_net(net);
3606
3607 if (!iter) {
3608 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3609 if (!iter)
3610 return -ENOMEM;
3611
3612 cb->args[4] = (long)iter;
3613 }
3614
3615 rhashtable_walk_enter(&tn->sk_rht, iter);
3616 return 0;
3617}
3618
3619int tipc_dump_done(struct netlink_callback *cb)
3620{
3621 struct rhashtable_iter *hti = (void *)cb->args[4];
3622
3623 rhashtable_walk_exit(hti);
3624 kfree(hti);
3625 return 0;
3626}
3627EXPORT_SYMBOL(tipc_dump_done);
3628
3629int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3630 struct tipc_sock *tsk, u32 sk_filter_state,
3631 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3632{
3633 struct sock *sk = &tsk->sk;
3634 struct nlattr *attrs;
3635 struct nlattr *stat;
3636
3637 /*filter response w.r.t sk_state*/
3638 if (!(sk_filter_state & (1 << sk->sk_state)))
3639 return 0;
3640
3641 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3642 if (!attrs)
3643 goto msg_cancel;
3644
3645 if (__tipc_nl_add_sk_info(skb, tsk))
3646 goto attr_msg_cancel;
3647
3648 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3649 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3650 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3651 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3652 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3653 sock_i_uid(sk))) ||
3654 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3655 tipc_diag_gen_cookie(sk),
3656 TIPC_NLA_SOCK_PAD))
3657 goto attr_msg_cancel;
3658
3659 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3660 if (!stat)
3661 goto attr_msg_cancel;
3662
3663 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3664 skb_queue_len(&sk->sk_receive_queue)) ||
3665 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3666 skb_queue_len(&sk->sk_write_queue)) ||
3667 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3668 atomic_read(&sk->sk_drops)))
3669 goto stat_msg_cancel;
3670
3671 if (tsk->cong_link_cnt &&
3672 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3673 goto stat_msg_cancel;
3674
3675 if (tsk_conn_cong(tsk) &&
3676 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3677 goto stat_msg_cancel;
3678
3679 nla_nest_end(skb, stat);
3680
3681 if (tsk->group)
3682 if (tipc_group_fill_sock_diag(tsk->group, skb))
3683 goto stat_msg_cancel;
3684
3685 nla_nest_end(skb, attrs);
3686
3687 return 0;
3688
3689stat_msg_cancel:
3690 nla_nest_cancel(skb, stat);
3691attr_msg_cancel:
3692 nla_nest_cancel(skb, attrs);
3693msg_cancel:
3694 return -EMSGSIZE;
3695}
3696EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3697
3698int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3699{
3700 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3701}
3702
3703/* Caller should hold socket lock for the passed tipc socket. */
3704static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3705 struct netlink_callback *cb,
3706 struct publication *publ)
3707{
3708 void *hdr;
3709 struct nlattr *attrs;
3710
3711 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3712 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3713 if (!hdr)
3714 goto msg_cancel;
3715
3716 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3717 if (!attrs)
3718 goto genlmsg_cancel;
3719
3720 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3721 goto attr_msg_cancel;
3722 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3723 goto attr_msg_cancel;
3724 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3725 goto attr_msg_cancel;
3726 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3727 goto attr_msg_cancel;
3728
3729 nla_nest_end(skb, attrs);
3730 genlmsg_end(skb, hdr);
3731
3732 return 0;
3733
3734attr_msg_cancel:
3735 nla_nest_cancel(skb, attrs);
3736genlmsg_cancel:
3737 genlmsg_cancel(skb, hdr);
3738msg_cancel:
3739 return -EMSGSIZE;
3740}
3741
3742/* Caller should hold socket lock for the passed tipc socket. */
3743static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3744 struct netlink_callback *cb,
3745 struct tipc_sock *tsk, u32 *last_publ)
3746{
3747 int err;
3748 struct publication *p;
3749
3750 if (*last_publ) {
3751 list_for_each_entry(p, &tsk->publications, binding_sock) {
3752 if (p->key == *last_publ)
3753 break;
3754 }
3755 if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3756 /* We never set seq or call nl_dump_check_consistent()
3757 * this means that setting prev_seq here will cause the
3758 * consistence check to fail in the netlink callback
3759 * handler. Resulting in the last NLMSG_DONE message
3760 * having the NLM_F_DUMP_INTR flag set.
3761 */
3762 cb->prev_seq = 1;
3763 *last_publ = 0;
3764 return -EPIPE;
3765 }
3766 } else {
3767 p = list_first_entry(&tsk->publications, struct publication,
3768 binding_sock);
3769 }
3770
3771 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3772 err = __tipc_nl_add_sk_publ(skb, cb, p);
3773 if (err) {
3774 *last_publ = p->key;
3775 return err;
3776 }
3777 }
3778 *last_publ = 0;
3779
3780 return 0;
3781}
3782
3783int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3784{
3785 int err;
3786 u32 tsk_portid = cb->args[0];
3787 u32 last_publ = cb->args[1];
3788 u32 done = cb->args[2];
3789 struct net *net = sock_net(skb->sk);
3790 struct tipc_sock *tsk;
3791
3792 if (!tsk_portid) {
3793 struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs;
3794 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3795
3796 if (!attrs[TIPC_NLA_SOCK])
3797 return -EINVAL;
3798
3799 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3800 attrs[TIPC_NLA_SOCK],
3801 tipc_nl_sock_policy, NULL);
3802 if (err)
3803 return err;
3804
3805 if (!sock[TIPC_NLA_SOCK_REF])
3806 return -EINVAL;
3807
3808 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3809 }
3810
3811 if (done)
3812 return 0;
3813
3814 tsk = tipc_sk_lookup(net, tsk_portid);
3815 if (!tsk)
3816 return -EINVAL;
3817
3818 lock_sock(&tsk->sk);
3819 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3820 if (!err)
3821 done = 1;
3822 release_sock(&tsk->sk);
3823 sock_put(&tsk->sk);
3824
3825 cb->args[0] = tsk_portid;
3826 cb->args[1] = last_publ;
3827 cb->args[2] = done;
3828
3829 return skb->len;
3830}
3831
3832/**
3833 * tipc_sk_filtering - check if a socket should be traced
3834 * @sk: the socket to be examined
3835 *
3836 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3837 * (portid, sock type, name type, name lower, name upper)
3838 *
3839 * Return: true if the socket meets the socket tuple data
3840 * (value 0 = 'any') or when there is no tuple set (all = 0),
3841 * otherwise false
3842 */
3843bool tipc_sk_filtering(struct sock *sk)
3844{
3845 struct tipc_sock *tsk;
3846 struct publication *p;
3847 u32 _port, _sktype, _type, _lower, _upper;
3848 u32 type = 0, lower = 0, upper = 0;
3849
3850 if (!sk)
3851 return true;
3852
3853 tsk = tipc_sk(sk);
3854
3855 _port = sysctl_tipc_sk_filter[0];
3856 _sktype = sysctl_tipc_sk_filter[1];
3857 _type = sysctl_tipc_sk_filter[2];
3858 _lower = sysctl_tipc_sk_filter[3];
3859 _upper = sysctl_tipc_sk_filter[4];
3860
3861 if (!_port && !_sktype && !_type && !_lower && !_upper)
3862 return true;
3863
3864 if (_port)
3865 return (_port == tsk->portid);
3866
3867 if (_sktype && _sktype != sk->sk_type)
3868 return false;
3869
3870 if (tsk->published) {
3871 p = list_first_entry_or_null(&tsk->publications,
3872 struct publication, binding_sock);
3873 if (p) {
3874 type = p->sr.type;
3875 lower = p->sr.lower;
3876 upper = p->sr.upper;
3877 }
3878 }
3879
3880 if (!tipc_sk_type_connectionless(sk)) {
3881 type = msg_nametype(&tsk->phdr);
3882 lower = msg_nameinst(&tsk->phdr);
3883 upper = lower;
3884 }
3885
3886 if ((_type && _type != type) || (_lower && _lower != lower) ||
3887 (_upper && _upper != upper))
3888 return false;
3889
3890 return true;
3891}
3892
3893u32 tipc_sock_get_portid(struct sock *sk)
3894{
3895 return (sk) ? (tipc_sk(sk))->portid : 0;
3896}
3897
3898/**
3899 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3900 * both the rcv and backlog queues are considered
3901 * @sk: tipc sk to be checked
3902 * @skb: tipc msg to be checked
3903 *
3904 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3905 */
3906
3907bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3908{
3909 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3910 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3911 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3912
3913 return (qsize > lim * 90 / 100);
3914}
3915
3916/**
3917 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3918 * only the rcv queue is considered
3919 * @sk: tipc sk to be checked
3920 * @skb: tipc msg to be checked
3921 *
3922 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3923 */
3924
3925bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3926{
3927 unsigned int lim = rcvbuf_limit(sk, skb);
3928 unsigned int qsize = sk_rmem_alloc_get(sk);
3929
3930 return (qsize > lim * 90 / 100);
3931}
3932
3933/**
3934 * tipc_sk_dump - dump TIPC socket
3935 * @sk: tipc sk to be dumped
3936 * @dqueues: bitmask to decide if any socket queue to be dumped?
3937 * - TIPC_DUMP_NONE: don't dump socket queues
3938 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3939 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3940 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3941 * - TIPC_DUMP_ALL: dump all the socket queues above
3942 * @buf: returned buffer of dump data in format
3943 */
3944int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3945{
3946 int i = 0;
3947 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3948 u32 conn_type, conn_instance;
3949 struct tipc_sock *tsk;
3950 struct publication *p;
3951 bool tsk_connected;
3952
3953 if (!sk) {
3954 i += scnprintf(buf, sz, "sk data: (null)\n");
3955 return i;
3956 }
3957
3958 tsk = tipc_sk(sk);
3959 tsk_connected = !tipc_sk_type_connectionless(sk);
3960
3961 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3962 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3963 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3964 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3965 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3966 if (tsk_connected) {
3967 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3968 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3969 conn_type = msg_nametype(&tsk->phdr);
3970 conn_instance = msg_nameinst(&tsk->phdr);
3971 i += scnprintf(buf + i, sz - i, " %u", conn_type);
3972 i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3973 }
3974 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3975 if (tsk->published) {
3976 p = list_first_entry_or_null(&tsk->publications,
3977 struct publication, binding_sock);
3978 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3979 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3980 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3981 }
3982 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3983 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3984 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3985 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3986 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3987 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3988 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3989 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3990 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3991 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3992 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3993 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3994 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3995 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3996
3997 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3998 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3999 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
4000 }
4001
4002 if (dqueues & TIPC_DUMP_SK_RCVQ) {
4003 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
4004 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
4005 }
4006
4007 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4008 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
4009 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4010 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4011 i += scnprintf(buf + i, sz - i, " tail ");
4012 i += tipc_skb_dump(sk->sk_backlog.tail, false,
4013 buf + i);
4014 }
4015 }
4016
4017 return i;
4018}