Linux 5.7.7
[linux/fpc-iii.git] / net / tipc / socket.c
blobe370ad0edd768c6826be94d20cd37b1a9e2f66e0
1 /*
2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
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.
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.
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.
37 #include <linux/rhashtable.h>
38 #include <linux/sched/signal.h>
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"
51 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
52 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
53 #define TIPC_FWD_MSG 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
56 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
58 enum {
59 TIPC_LISTEN = TCP_LISTEN,
60 TIPC_ESTABLISHED = TCP_ESTABLISHED,
61 TIPC_OPEN = TCP_CLOSE,
62 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
63 TIPC_CONNECTING = TCP_SYN_SENT,
66 struct sockaddr_pair {
67 struct sockaddr_tipc sock;
68 struct sockaddr_tipc member;
71 /**
72 * struct tipc_sock - TIPC socket structure
73 * @sk: socket - interacts with 'port' and with user via the socket API
74 * @conn_type: TIPC type used when connection was established
75 * @conn_instance: TIPC instance used when connection was established
76 * @published: non-zero if port has one or more associated names
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 * @blocking_link: address of the congested link we are currently sleeping on
84 * @pub_count: total # of publications port has made during its lifetime
85 * @conn_timeout: the time we can wait for an unresponded setup request
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 * @rcv_unacked: # messages read by user, but not yet acked back to peer
90 * @peer: 'connected' peer for dgram/rdm
91 * @node: hash table node
92 * @mc_method: cookie for use between socket and broadcast layer
93 * @rcu: rcu struct for tipc_sock
95 struct tipc_sock {
96 struct sock sk;
97 u32 conn_type;
98 u32 conn_instance;
99 int published;
100 u32 max_pkt;
101 u32 maxnagle;
102 u32 portid;
103 struct tipc_msg phdr;
104 struct list_head cong_links;
105 struct list_head publications;
106 u32 pub_count;
107 atomic_t dupl_rcvcnt;
108 u16 conn_timeout;
109 bool probe_unacked;
110 u16 cong_link_cnt;
111 u16 snt_unacked;
112 u16 snd_win;
113 u16 peer_caps;
114 u16 rcv_unacked;
115 u16 rcv_win;
116 struct sockaddr_tipc peer;
117 struct rhash_head node;
118 struct tipc_mc_method mc_method;
119 struct rcu_head rcu;
120 struct tipc_group *group;
121 u32 oneway;
122 u16 snd_backlog;
123 bool expect_ack;
124 bool nodelay;
125 bool group_is_open;
128 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
129 static void tipc_data_ready(struct sock *sk);
130 static void tipc_write_space(struct sock *sk);
131 static void tipc_sock_destruct(struct sock *sk);
132 static int tipc_release(struct socket *sock);
133 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
134 bool kern);
135 static void tipc_sk_timeout(struct timer_list *t);
136 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
137 struct tipc_name_seq const *seq);
138 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
139 struct tipc_name_seq const *seq);
140 static int tipc_sk_leave(struct tipc_sock *tsk);
141 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
142 static int tipc_sk_insert(struct tipc_sock *tsk);
143 static void tipc_sk_remove(struct tipc_sock *tsk);
144 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
145 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
146 static void tipc_sk_push_backlog(struct tipc_sock *tsk);
148 static const struct proto_ops packet_ops;
149 static const struct proto_ops stream_ops;
150 static const struct proto_ops msg_ops;
151 static struct proto tipc_proto;
152 static const struct rhashtable_params tsk_rht_params;
154 static u32 tsk_own_node(struct tipc_sock *tsk)
156 return msg_prevnode(&tsk->phdr);
159 static u32 tsk_peer_node(struct tipc_sock *tsk)
161 return msg_destnode(&tsk->phdr);
164 static u32 tsk_peer_port(struct tipc_sock *tsk)
166 return msg_destport(&tsk->phdr);
169 static bool tsk_unreliable(struct tipc_sock *tsk)
171 return msg_src_droppable(&tsk->phdr) != 0;
174 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
176 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
179 static bool tsk_unreturnable(struct tipc_sock *tsk)
181 return msg_dest_droppable(&tsk->phdr) != 0;
184 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
186 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
189 static int tsk_importance(struct tipc_sock *tsk)
191 return msg_importance(&tsk->phdr);
194 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
196 if (imp > TIPC_CRITICAL_IMPORTANCE)
197 return -EINVAL;
198 msg_set_importance(&tsk->phdr, (u32)imp);
199 return 0;
202 static struct tipc_sock *tipc_sk(const struct sock *sk)
204 return container_of(sk, struct tipc_sock, sk);
207 static bool tsk_conn_cong(struct tipc_sock *tsk)
209 return tsk->snt_unacked > tsk->snd_win;
212 static u16 tsk_blocks(int len)
214 return ((len / FLOWCTL_BLK_SZ) + 1);
217 /* tsk_blocks(): translate a buffer size in bytes to number of
218 * advertisable blocks, taking into account the ratio truesize(len)/len
219 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
221 static u16 tsk_adv_blocks(int len)
223 return len / FLOWCTL_BLK_SZ / 4;
226 /* tsk_inc(): increment counter for sent or received data
227 * - If block based flow control is not supported by peer we
228 * fall back to message based ditto, incrementing the counter
230 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
232 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
233 return ((msglen / FLOWCTL_BLK_SZ) + 1);
234 return 1;
237 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
239 static void tsk_set_nagle(struct tipc_sock *tsk)
241 struct sock *sk = &tsk->sk;
243 tsk->maxnagle = 0;
244 if (sk->sk_type != SOCK_STREAM)
245 return;
246 if (tsk->nodelay)
247 return;
248 if (!(tsk->peer_caps & TIPC_NAGLE))
249 return;
250 /* Limit node local buffer size to avoid receive queue overflow */
251 if (tsk->max_pkt == MAX_MSG_SIZE)
252 tsk->maxnagle = 1500;
253 else
254 tsk->maxnagle = tsk->max_pkt;
258 * tsk_advance_rx_queue - discard first buffer in socket receive queue
260 * Caller must hold socket lock
262 static void tsk_advance_rx_queue(struct sock *sk)
264 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
265 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
268 /* tipc_sk_respond() : send response message back to sender
270 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
272 u32 selector;
273 u32 dnode;
274 u32 onode = tipc_own_addr(sock_net(sk));
276 if (!tipc_msg_reverse(onode, &skb, err))
277 return;
279 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
280 dnode = msg_destnode(buf_msg(skb));
281 selector = msg_origport(buf_msg(skb));
282 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
286 * tsk_rej_rx_queue - reject all buffers in socket receive queue
288 * Caller must hold socket lock
290 static void tsk_rej_rx_queue(struct sock *sk, int error)
292 struct sk_buff *skb;
294 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
295 tipc_sk_respond(sk, skb, error);
298 static bool tipc_sk_connected(struct sock *sk)
300 return sk->sk_state == TIPC_ESTABLISHED;
303 /* tipc_sk_type_connectionless - check if the socket is datagram socket
304 * @sk: socket
306 * Returns true if connection less, false otherwise
308 static bool tipc_sk_type_connectionless(struct sock *sk)
310 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
313 /* tsk_peer_msg - verify if message was sent by connected port's peer
315 * Handles cases where the node's network address has changed from
316 * the default of <0.0.0> to its configured setting.
318 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
320 struct sock *sk = &tsk->sk;
321 u32 self = tipc_own_addr(sock_net(sk));
322 u32 peer_port = tsk_peer_port(tsk);
323 u32 orig_node, peer_node;
325 if (unlikely(!tipc_sk_connected(sk)))
326 return false;
328 if (unlikely(msg_origport(msg) != peer_port))
329 return false;
331 orig_node = msg_orignode(msg);
332 peer_node = tsk_peer_node(tsk);
334 if (likely(orig_node == peer_node))
335 return true;
337 if (!orig_node && peer_node == self)
338 return true;
340 if (!peer_node && orig_node == self)
341 return true;
343 return false;
346 /* tipc_set_sk_state - set the sk_state of the socket
347 * @sk: socket
349 * Caller must hold socket lock
351 * Returns 0 on success, errno otherwise
353 static int tipc_set_sk_state(struct sock *sk, int state)
355 int oldsk_state = sk->sk_state;
356 int res = -EINVAL;
358 switch (state) {
359 case TIPC_OPEN:
360 res = 0;
361 break;
362 case TIPC_LISTEN:
363 case TIPC_CONNECTING:
364 if (oldsk_state == TIPC_OPEN)
365 res = 0;
366 break;
367 case TIPC_ESTABLISHED:
368 if (oldsk_state == TIPC_CONNECTING ||
369 oldsk_state == TIPC_OPEN)
370 res = 0;
371 break;
372 case TIPC_DISCONNECTING:
373 if (oldsk_state == TIPC_CONNECTING ||
374 oldsk_state == TIPC_ESTABLISHED)
375 res = 0;
376 break;
379 if (!res)
380 sk->sk_state = state;
382 return res;
385 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
387 struct sock *sk = sock->sk;
388 int err = sock_error(sk);
389 int typ = sock->type;
391 if (err)
392 return err;
393 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
394 if (sk->sk_state == TIPC_DISCONNECTING)
395 return -EPIPE;
396 else if (!tipc_sk_connected(sk))
397 return -ENOTCONN;
399 if (!*timeout)
400 return -EAGAIN;
401 if (signal_pending(current))
402 return sock_intr_errno(*timeout);
404 return 0;
407 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
408 ({ \
409 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
410 struct sock *sk_; \
411 int rc_; \
413 while ((rc_ = !(condition_))) { \
414 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
415 smp_rmb(); \
416 sk_ = (sock_)->sk; \
417 rc_ = tipc_sk_sock_err((sock_), timeo_); \
418 if (rc_) \
419 break; \
420 add_wait_queue(sk_sleep(sk_), &wait_); \
421 release_sock(sk_); \
422 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
423 sched_annotate_sleep(); \
424 lock_sock(sk_); \
425 remove_wait_queue(sk_sleep(sk_), &wait_); \
427 rc_; \
431 * tipc_sk_create - create a TIPC socket
432 * @net: network namespace (must be default network)
433 * @sock: pre-allocated socket structure
434 * @protocol: protocol indicator (must be 0)
435 * @kern: caused by kernel or by userspace?
437 * This routine creates additional data structures used by the TIPC socket,
438 * initializes them, and links them together.
440 * Returns 0 on success, errno otherwise
442 static int tipc_sk_create(struct net *net, struct socket *sock,
443 int protocol, int kern)
445 const struct proto_ops *ops;
446 struct sock *sk;
447 struct tipc_sock *tsk;
448 struct tipc_msg *msg;
450 /* Validate arguments */
451 if (unlikely(protocol != 0))
452 return -EPROTONOSUPPORT;
454 switch (sock->type) {
455 case SOCK_STREAM:
456 ops = &stream_ops;
457 break;
458 case SOCK_SEQPACKET:
459 ops = &packet_ops;
460 break;
461 case SOCK_DGRAM:
462 case SOCK_RDM:
463 ops = &msg_ops;
464 break;
465 default:
466 return -EPROTOTYPE;
469 /* Allocate socket's protocol area */
470 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
471 if (sk == NULL)
472 return -ENOMEM;
474 tsk = tipc_sk(sk);
475 tsk->max_pkt = MAX_PKT_DEFAULT;
476 tsk->maxnagle = 0;
477 INIT_LIST_HEAD(&tsk->publications);
478 INIT_LIST_HEAD(&tsk->cong_links);
479 msg = &tsk->phdr;
481 /* Finish initializing socket data structures */
482 sock->ops = ops;
483 sock_init_data(sock, sk);
484 tipc_set_sk_state(sk, TIPC_OPEN);
485 if (tipc_sk_insert(tsk)) {
486 pr_warn("Socket create failed; port number exhausted\n");
487 return -EINVAL;
490 /* Ensure tsk is visible before we read own_addr. */
491 smp_mb();
493 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
494 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
496 msg_set_origport(msg, tsk->portid);
497 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
498 sk->sk_shutdown = 0;
499 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
500 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
501 sk->sk_data_ready = tipc_data_ready;
502 sk->sk_write_space = tipc_write_space;
503 sk->sk_destruct = tipc_sock_destruct;
504 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
505 tsk->group_is_open = true;
506 atomic_set(&tsk->dupl_rcvcnt, 0);
508 /* Start out with safe limits until we receive an advertised window */
509 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
510 tsk->rcv_win = tsk->snd_win;
512 if (tipc_sk_type_connectionless(sk)) {
513 tsk_set_unreturnable(tsk, true);
514 if (sock->type == SOCK_DGRAM)
515 tsk_set_unreliable(tsk, true);
517 __skb_queue_head_init(&tsk->mc_method.deferredq);
518 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
519 return 0;
522 static void tipc_sk_callback(struct rcu_head *head)
524 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
526 sock_put(&tsk->sk);
529 /* Caller should hold socket lock for the socket. */
530 static void __tipc_shutdown(struct socket *sock, int error)
532 struct sock *sk = sock->sk;
533 struct tipc_sock *tsk = tipc_sk(sk);
534 struct net *net = sock_net(sk);
535 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
536 u32 dnode = tsk_peer_node(tsk);
537 struct sk_buff *skb;
539 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
540 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
541 !tsk_conn_cong(tsk)));
543 /* Push out delayed messages if in Nagle mode */
544 tipc_sk_push_backlog(tsk);
545 /* Remove pending SYN */
546 __skb_queue_purge(&sk->sk_write_queue);
548 /* Remove partially received buffer if any */
549 skb = skb_peek(&sk->sk_receive_queue);
550 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
551 __skb_unlink(skb, &sk->sk_receive_queue);
552 kfree_skb(skb);
555 /* Reject all unreceived messages if connectionless */
556 if (tipc_sk_type_connectionless(sk)) {
557 tsk_rej_rx_queue(sk, error);
558 return;
561 switch (sk->sk_state) {
562 case TIPC_CONNECTING:
563 case TIPC_ESTABLISHED:
564 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
565 tipc_node_remove_conn(net, dnode, tsk->portid);
566 /* Send a FIN+/- to its peer */
567 skb = __skb_dequeue(&sk->sk_receive_queue);
568 if (skb) {
569 __skb_queue_purge(&sk->sk_receive_queue);
570 tipc_sk_respond(sk, skb, error);
571 break;
573 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
574 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
575 tsk_own_node(tsk), tsk_peer_port(tsk),
576 tsk->portid, error);
577 if (skb)
578 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
579 break;
580 case TIPC_LISTEN:
581 /* Reject all SYN messages */
582 tsk_rej_rx_queue(sk, error);
583 break;
584 default:
585 __skb_queue_purge(&sk->sk_receive_queue);
586 break;
591 * tipc_release - destroy a TIPC socket
592 * @sock: socket to destroy
594 * This routine cleans up any messages that are still queued on the socket.
595 * For DGRAM and RDM socket types, all queued messages are rejected.
596 * For SEQPACKET and STREAM socket types, the first message is rejected
597 * and any others are discarded. (If the first message on a STREAM socket
598 * is partially-read, it is discarded and the next one is rejected instead.)
600 * NOTE: Rejected messages are not necessarily returned to the sender! They
601 * are returned or discarded according to the "destination droppable" setting
602 * specified for the message by the sender.
604 * Returns 0 on success, errno otherwise
606 static int tipc_release(struct socket *sock)
608 struct sock *sk = sock->sk;
609 struct tipc_sock *tsk;
612 * Exit if socket isn't fully initialized (occurs when a failed accept()
613 * releases a pre-allocated child socket that was never used)
615 if (sk == NULL)
616 return 0;
618 tsk = tipc_sk(sk);
619 lock_sock(sk);
621 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
622 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
623 sk->sk_shutdown = SHUTDOWN_MASK;
624 tipc_sk_leave(tsk);
625 tipc_sk_withdraw(tsk, 0, NULL);
626 __skb_queue_purge(&tsk->mc_method.deferredq);
627 sk_stop_timer(sk, &sk->sk_timer);
628 tipc_sk_remove(tsk);
630 sock_orphan(sk);
631 /* Reject any messages that accumulated in backlog queue */
632 release_sock(sk);
633 tipc_dest_list_purge(&tsk->cong_links);
634 tsk->cong_link_cnt = 0;
635 call_rcu(&tsk->rcu, tipc_sk_callback);
636 sock->sk = NULL;
638 return 0;
642 * tipc_bind - associate or disassocate TIPC name(s) with a socket
643 * @sock: socket structure
644 * @uaddr: socket address describing name(s) and desired operation
645 * @uaddr_len: size of socket address data structure
647 * Name and name sequence binding is indicated using a positive scope value;
648 * a negative scope value unbinds the specified name. Specifying no name
649 * (i.e. a socket address length of 0) unbinds all names from the socket.
651 * Returns 0 on success, errno otherwise
653 * NOTE: This routine doesn't need to take the socket lock since it doesn't
654 * access any non-constant socket information.
656 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
657 int uaddr_len)
659 struct sock *sk = sock->sk;
660 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
661 struct tipc_sock *tsk = tipc_sk(sk);
662 int res = -EINVAL;
664 lock_sock(sk);
665 if (unlikely(!uaddr_len)) {
666 res = tipc_sk_withdraw(tsk, 0, NULL);
667 goto exit;
669 if (tsk->group) {
670 res = -EACCES;
671 goto exit;
673 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
674 res = -EINVAL;
675 goto exit;
677 if (addr->family != AF_TIPC) {
678 res = -EAFNOSUPPORT;
679 goto exit;
682 if (addr->addrtype == TIPC_ADDR_NAME)
683 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
684 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
685 res = -EAFNOSUPPORT;
686 goto exit;
689 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
690 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
691 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
692 res = -EACCES;
693 goto exit;
696 res = (addr->scope >= 0) ?
697 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
698 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
699 exit:
700 release_sock(sk);
701 return res;
705 * tipc_getname - get port ID of socket or peer socket
706 * @sock: socket structure
707 * @uaddr: area for returned socket address
708 * @uaddr_len: area for returned length of socket address
709 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
711 * Returns 0 on success, errno otherwise
713 * NOTE: This routine doesn't need to take the socket lock since it only
714 * accesses socket information that is unchanging (or which changes in
715 * a completely predictable manner).
717 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
718 int peer)
720 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
721 struct sock *sk = sock->sk;
722 struct tipc_sock *tsk = tipc_sk(sk);
724 memset(addr, 0, sizeof(*addr));
725 if (peer) {
726 if ((!tipc_sk_connected(sk)) &&
727 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
728 return -ENOTCONN;
729 addr->addr.id.ref = tsk_peer_port(tsk);
730 addr->addr.id.node = tsk_peer_node(tsk);
731 } else {
732 addr->addr.id.ref = tsk->portid;
733 addr->addr.id.node = tipc_own_addr(sock_net(sk));
736 addr->addrtype = TIPC_ADDR_ID;
737 addr->family = AF_TIPC;
738 addr->scope = 0;
739 addr->addr.name.domain = 0;
741 return sizeof(*addr);
745 * tipc_poll - read and possibly block on pollmask
746 * @file: file structure associated with the socket
747 * @sock: socket for which to calculate the poll bits
748 * @wait: ???
750 * Returns pollmask value
752 * COMMENTARY:
753 * It appears that the usual socket locking mechanisms are not useful here
754 * since the pollmask info is potentially out-of-date the moment this routine
755 * exits. TCP and other protocols seem to rely on higher level poll routines
756 * to handle any preventable race conditions, so TIPC will do the same ...
758 * IMPORTANT: The fact that a read or write operation is indicated does NOT
759 * imply that the operation will succeed, merely that it should be performed
760 * and will not block.
762 static __poll_t tipc_poll(struct file *file, struct socket *sock,
763 poll_table *wait)
765 struct sock *sk = sock->sk;
766 struct tipc_sock *tsk = tipc_sk(sk);
767 __poll_t revents = 0;
769 sock_poll_wait(file, sock, wait);
770 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
772 if (sk->sk_shutdown & RCV_SHUTDOWN)
773 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
774 if (sk->sk_shutdown == SHUTDOWN_MASK)
775 revents |= EPOLLHUP;
777 switch (sk->sk_state) {
778 case TIPC_ESTABLISHED:
779 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
780 revents |= EPOLLOUT;
781 /* fall through */
782 case TIPC_LISTEN:
783 case TIPC_CONNECTING:
784 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
785 revents |= EPOLLIN | EPOLLRDNORM;
786 break;
787 case TIPC_OPEN:
788 if (tsk->group_is_open && !tsk->cong_link_cnt)
789 revents |= EPOLLOUT;
790 if (!tipc_sk_type_connectionless(sk))
791 break;
792 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
793 break;
794 revents |= EPOLLIN | EPOLLRDNORM;
795 break;
796 case TIPC_DISCONNECTING:
797 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
798 break;
800 return revents;
804 * tipc_sendmcast - send multicast message
805 * @sock: socket structure
806 * @seq: destination address
807 * @msg: message to send
808 * @dlen: length of data to send
809 * @timeout: timeout to wait for wakeup
811 * Called from function tipc_sendmsg(), which has done all sanity checks
812 * Returns the number of bytes sent on success, or errno
814 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
815 struct msghdr *msg, size_t dlen, long timeout)
817 struct sock *sk = sock->sk;
818 struct tipc_sock *tsk = tipc_sk(sk);
819 struct tipc_msg *hdr = &tsk->phdr;
820 struct net *net = sock_net(sk);
821 int mtu = tipc_bcast_get_mtu(net);
822 struct tipc_mc_method *method = &tsk->mc_method;
823 struct sk_buff_head pkts;
824 struct tipc_nlist dsts;
825 int rc;
827 if (tsk->group)
828 return -EACCES;
830 /* Block or return if any destination link is congested */
831 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
832 if (unlikely(rc))
833 return rc;
835 /* Lookup destination nodes */
836 tipc_nlist_init(&dsts, tipc_own_addr(net));
837 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
838 seq->upper, &dsts);
839 if (!dsts.local && !dsts.remote)
840 return -EHOSTUNREACH;
842 /* Build message header */
843 msg_set_type(hdr, TIPC_MCAST_MSG);
844 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
845 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
846 msg_set_destport(hdr, 0);
847 msg_set_destnode(hdr, 0);
848 msg_set_nametype(hdr, seq->type);
849 msg_set_namelower(hdr, seq->lower);
850 msg_set_nameupper(hdr, seq->upper);
852 /* Build message as chain of buffers */
853 __skb_queue_head_init(&pkts);
854 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
856 /* Send message if build was successful */
857 if (unlikely(rc == dlen)) {
858 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
859 TIPC_DUMP_SK_SNDQ, " ");
860 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
861 &tsk->cong_link_cnt);
864 tipc_nlist_purge(&dsts);
866 return rc ? rc : dlen;
870 * tipc_send_group_msg - send a message to a member in the group
871 * @net: network namespace
872 * @m: message to send
873 * @mb: group member
874 * @dnode: destination node
875 * @dport: destination port
876 * @dlen: total length of message data
878 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
879 struct msghdr *m, struct tipc_member *mb,
880 u32 dnode, u32 dport, int dlen)
882 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
883 struct tipc_mc_method *method = &tsk->mc_method;
884 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
885 struct tipc_msg *hdr = &tsk->phdr;
886 struct sk_buff_head pkts;
887 int mtu, rc;
889 /* Complete message header */
890 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
891 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
892 msg_set_destport(hdr, dport);
893 msg_set_destnode(hdr, dnode);
894 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
896 /* Build message as chain of buffers */
897 __skb_queue_head_init(&pkts);
898 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
899 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
900 if (unlikely(rc != dlen))
901 return rc;
903 /* Send message */
904 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
905 if (unlikely(rc == -ELINKCONG)) {
906 tipc_dest_push(&tsk->cong_links, dnode, 0);
907 tsk->cong_link_cnt++;
910 /* Update send window */
911 tipc_group_update_member(mb, blks);
913 /* A broadcast sent within next EXPIRE period must follow same path */
914 method->rcast = true;
915 method->mandatory = true;
916 return dlen;
920 * tipc_send_group_unicast - send message to a member in the group
921 * @sock: socket structure
922 * @m: message to send
923 * @dlen: total length of message data
924 * @timeout: timeout to wait for wakeup
926 * Called from function tipc_sendmsg(), which has done all sanity checks
927 * Returns the number of bytes sent on success, or errno
929 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
930 int dlen, long timeout)
932 struct sock *sk = sock->sk;
933 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
934 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
935 struct tipc_sock *tsk = tipc_sk(sk);
936 struct net *net = sock_net(sk);
937 struct tipc_member *mb = NULL;
938 u32 node, port;
939 int rc;
941 node = dest->addr.id.node;
942 port = dest->addr.id.ref;
943 if (!port && !node)
944 return -EHOSTUNREACH;
946 /* Block or return if destination link or member is congested */
947 rc = tipc_wait_for_cond(sock, &timeout,
948 !tipc_dest_find(&tsk->cong_links, node, 0) &&
949 tsk->group &&
950 !tipc_group_cong(tsk->group, node, port, blks,
951 &mb));
952 if (unlikely(rc))
953 return rc;
955 if (unlikely(!mb))
956 return -EHOSTUNREACH;
958 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
960 return rc ? rc : dlen;
964 * tipc_send_group_anycast - send message to any member with given identity
965 * @sock: socket structure
966 * @m: message to send
967 * @dlen: total length of message data
968 * @timeout: timeout to wait for wakeup
970 * Called from function tipc_sendmsg(), which has done all sanity checks
971 * Returns the number of bytes sent on success, or errno
973 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
974 int dlen, long timeout)
976 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
977 struct sock *sk = sock->sk;
978 struct tipc_sock *tsk = tipc_sk(sk);
979 struct list_head *cong_links = &tsk->cong_links;
980 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
981 struct tipc_msg *hdr = &tsk->phdr;
982 struct tipc_member *first = NULL;
983 struct tipc_member *mbr = NULL;
984 struct net *net = sock_net(sk);
985 u32 node, port, exclude;
986 struct list_head dsts;
987 u32 type, inst, scope;
988 int lookups = 0;
989 int dstcnt, rc;
990 bool cong;
992 INIT_LIST_HEAD(&dsts);
994 type = msg_nametype(hdr);
995 inst = dest->addr.name.name.instance;
996 scope = msg_lookup_scope(hdr);
998 while (++lookups < 4) {
999 exclude = tipc_group_exclude(tsk->group);
1001 first = NULL;
1003 /* Look for a non-congested destination member, if any */
1004 while (1) {
1005 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1006 &dstcnt, exclude, false))
1007 return -EHOSTUNREACH;
1008 tipc_dest_pop(&dsts, &node, &port);
1009 cong = tipc_group_cong(tsk->group, node, port, blks,
1010 &mbr);
1011 if (!cong)
1012 break;
1013 if (mbr == first)
1014 break;
1015 if (!first)
1016 first = mbr;
1019 /* Start over if destination was not in member list */
1020 if (unlikely(!mbr))
1021 continue;
1023 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1024 break;
1026 /* Block or return if destination link or member is congested */
1027 rc = tipc_wait_for_cond(sock, &timeout,
1028 !tipc_dest_find(cong_links, node, 0) &&
1029 tsk->group &&
1030 !tipc_group_cong(tsk->group, node, port,
1031 blks, &mbr));
1032 if (unlikely(rc))
1033 return rc;
1035 /* Send, unless destination disappeared while waiting */
1036 if (likely(mbr))
1037 break;
1040 if (unlikely(lookups >= 4))
1041 return -EHOSTUNREACH;
1043 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1045 return rc ? rc : dlen;
1049 * tipc_send_group_bcast - send message to all members in communication group
1050 * @sk: socket structure
1051 * @m: message to send
1052 * @dlen: total length of message data
1053 * @timeout: timeout to wait for wakeup
1055 * Called from function tipc_sendmsg(), which has done all sanity checks
1056 * Returns the number of bytes sent on success, or errno
1058 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1059 int dlen, long timeout)
1061 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1062 struct sock *sk = sock->sk;
1063 struct net *net = sock_net(sk);
1064 struct tipc_sock *tsk = tipc_sk(sk);
1065 struct tipc_nlist *dsts;
1066 struct tipc_mc_method *method = &tsk->mc_method;
1067 bool ack = method->mandatory && method->rcast;
1068 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1069 struct tipc_msg *hdr = &tsk->phdr;
1070 int mtu = tipc_bcast_get_mtu(net);
1071 struct sk_buff_head pkts;
1072 int rc = -EHOSTUNREACH;
1074 /* Block or return if any destination link or member is congested */
1075 rc = tipc_wait_for_cond(sock, &timeout,
1076 !tsk->cong_link_cnt && tsk->group &&
1077 !tipc_group_bc_cong(tsk->group, blks));
1078 if (unlikely(rc))
1079 return rc;
1081 dsts = tipc_group_dests(tsk->group);
1082 if (!dsts->local && !dsts->remote)
1083 return -EHOSTUNREACH;
1085 /* Complete message header */
1086 if (dest) {
1087 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1088 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1089 } else {
1090 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1091 msg_set_nameinst(hdr, 0);
1093 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1094 msg_set_destport(hdr, 0);
1095 msg_set_destnode(hdr, 0);
1096 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1098 /* Avoid getting stuck with repeated forced replicasts */
1099 msg_set_grp_bc_ack_req(hdr, ack);
1101 /* Build message as chain of buffers */
1102 __skb_queue_head_init(&pkts);
1103 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1104 if (unlikely(rc != dlen))
1105 return rc;
1107 /* Send message */
1108 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1109 if (unlikely(rc))
1110 return rc;
1112 /* Update broadcast sequence number and send windows */
1113 tipc_group_update_bc_members(tsk->group, blks, ack);
1115 /* Broadcast link is now free to choose method for next broadcast */
1116 method->mandatory = false;
1117 method->expires = jiffies;
1119 return dlen;
1123 * tipc_send_group_mcast - send message to all members with given identity
1124 * @sock: socket structure
1125 * @m: message to send
1126 * @dlen: total length of message data
1127 * @timeout: timeout to wait for wakeup
1129 * Called from function tipc_sendmsg(), which has done all sanity checks
1130 * Returns the number of bytes sent on success, or errno
1132 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1133 int dlen, long timeout)
1135 struct sock *sk = sock->sk;
1136 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1137 struct tipc_sock *tsk = tipc_sk(sk);
1138 struct tipc_group *grp = tsk->group;
1139 struct tipc_msg *hdr = &tsk->phdr;
1140 struct net *net = sock_net(sk);
1141 u32 type, inst, scope, exclude;
1142 struct list_head dsts;
1143 u32 dstcnt;
1145 INIT_LIST_HEAD(&dsts);
1147 type = msg_nametype(hdr);
1148 inst = dest->addr.name.name.instance;
1149 scope = msg_lookup_scope(hdr);
1150 exclude = tipc_group_exclude(grp);
1152 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1153 &dstcnt, exclude, true))
1154 return -EHOSTUNREACH;
1156 if (dstcnt == 1) {
1157 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1158 return tipc_send_group_unicast(sock, m, dlen, timeout);
1161 tipc_dest_list_purge(&dsts);
1162 return tipc_send_group_bcast(sock, m, dlen, timeout);
1166 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1167 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1168 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1170 * Multi-threaded: parallel calls with reference to same queues may occur
1172 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1173 struct sk_buff_head *inputq)
1175 u32 self = tipc_own_addr(net);
1176 u32 type, lower, upper, scope;
1177 struct sk_buff *skb, *_skb;
1178 u32 portid, onode;
1179 struct sk_buff_head tmpq;
1180 struct list_head dports;
1181 struct tipc_msg *hdr;
1182 int user, mtyp, hlen;
1183 bool exact;
1185 __skb_queue_head_init(&tmpq);
1186 INIT_LIST_HEAD(&dports);
1188 skb = tipc_skb_peek(arrvq, &inputq->lock);
1189 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1190 hdr = buf_msg(skb);
1191 user = msg_user(hdr);
1192 mtyp = msg_type(hdr);
1193 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1194 onode = msg_orignode(hdr);
1195 type = msg_nametype(hdr);
1197 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1198 spin_lock_bh(&inputq->lock);
1199 if (skb_peek(arrvq) == skb) {
1200 __skb_dequeue(arrvq);
1201 __skb_queue_tail(inputq, skb);
1203 kfree_skb(skb);
1204 spin_unlock_bh(&inputq->lock);
1205 continue;
1208 /* Group messages require exact scope match */
1209 if (msg_in_group(hdr)) {
1210 lower = 0;
1211 upper = ~0;
1212 scope = msg_lookup_scope(hdr);
1213 exact = true;
1214 } else {
1215 /* TIPC_NODE_SCOPE means "any scope" in this context */
1216 if (onode == self)
1217 scope = TIPC_NODE_SCOPE;
1218 else
1219 scope = TIPC_CLUSTER_SCOPE;
1220 exact = false;
1221 lower = msg_namelower(hdr);
1222 upper = msg_nameupper(hdr);
1225 /* Create destination port list: */
1226 tipc_nametbl_mc_lookup(net, type, lower, upper,
1227 scope, exact, &dports);
1229 /* Clone message per destination */
1230 while (tipc_dest_pop(&dports, NULL, &portid)) {
1231 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1232 if (_skb) {
1233 msg_set_destport(buf_msg(_skb), portid);
1234 __skb_queue_tail(&tmpq, _skb);
1235 continue;
1237 pr_warn("Failed to clone mcast rcv buffer\n");
1239 /* Append to inputq if not already done by other thread */
1240 spin_lock_bh(&inputq->lock);
1241 if (skb_peek(arrvq) == skb) {
1242 skb_queue_splice_tail_init(&tmpq, inputq);
1243 kfree_skb(__skb_dequeue(arrvq));
1245 spin_unlock_bh(&inputq->lock);
1246 __skb_queue_purge(&tmpq);
1247 kfree_skb(skb);
1249 tipc_sk_rcv(net, inputq);
1252 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1253 * when socket is in Nagle mode
1255 static void tipc_sk_push_backlog(struct tipc_sock *tsk)
1257 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1258 struct net *net = sock_net(&tsk->sk);
1259 u32 dnode = tsk_peer_node(tsk);
1260 struct sk_buff *skb = skb_peek(txq);
1261 int rc;
1263 if (!skb || tsk->cong_link_cnt)
1264 return;
1266 /* Do not send SYN again after congestion */
1267 if (msg_is_syn(buf_msg(skb)))
1268 return;
1270 tsk->snt_unacked += tsk->snd_backlog;
1271 tsk->snd_backlog = 0;
1272 tsk->expect_ack = true;
1273 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1274 if (rc == -ELINKCONG)
1275 tsk->cong_link_cnt = 1;
1279 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1280 * @tsk: receiving socket
1281 * @skb: pointer to message buffer.
1283 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1284 struct sk_buff_head *inputq,
1285 struct sk_buff_head *xmitq)
1287 struct tipc_msg *hdr = buf_msg(skb);
1288 u32 onode = tsk_own_node(tsk);
1289 struct sock *sk = &tsk->sk;
1290 int mtyp = msg_type(hdr);
1291 bool was_cong;
1293 /* Ignore if connection cannot be validated: */
1294 if (!tsk_peer_msg(tsk, hdr)) {
1295 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1296 goto exit;
1299 if (unlikely(msg_errcode(hdr))) {
1300 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1301 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1302 tsk_peer_port(tsk));
1303 sk->sk_state_change(sk);
1305 /* State change is ignored if socket already awake,
1306 * - convert msg to abort msg and add to inqueue
1308 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1309 msg_set_type(hdr, TIPC_CONN_MSG);
1310 msg_set_size(hdr, BASIC_H_SIZE);
1311 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1312 __skb_queue_tail(inputq, skb);
1313 return;
1316 tsk->probe_unacked = false;
1318 if (mtyp == CONN_PROBE) {
1319 msg_set_type(hdr, CONN_PROBE_REPLY);
1320 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1321 __skb_queue_tail(xmitq, skb);
1322 return;
1323 } else if (mtyp == CONN_ACK) {
1324 was_cong = tsk_conn_cong(tsk);
1325 tsk->expect_ack = false;
1326 tipc_sk_push_backlog(tsk);
1327 tsk->snt_unacked -= msg_conn_ack(hdr);
1328 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1329 tsk->snd_win = msg_adv_win(hdr);
1330 if (was_cong && !tsk_conn_cong(tsk))
1331 sk->sk_write_space(sk);
1332 } else if (mtyp != CONN_PROBE_REPLY) {
1333 pr_warn("Received unknown CONN_PROTO msg\n");
1335 exit:
1336 kfree_skb(skb);
1340 * tipc_sendmsg - send message in connectionless manner
1341 * @sock: socket structure
1342 * @m: message to send
1343 * @dsz: amount of user data to be sent
1345 * Message must have an destination specified explicitly.
1346 * Used for SOCK_RDM and SOCK_DGRAM messages,
1347 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1348 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1350 * Returns the number of bytes sent on success, or errno otherwise
1352 static int tipc_sendmsg(struct socket *sock,
1353 struct msghdr *m, size_t dsz)
1355 struct sock *sk = sock->sk;
1356 int ret;
1358 lock_sock(sk);
1359 ret = __tipc_sendmsg(sock, m, dsz);
1360 release_sock(sk);
1362 return ret;
1365 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1367 struct sock *sk = sock->sk;
1368 struct net *net = sock_net(sk);
1369 struct tipc_sock *tsk = tipc_sk(sk);
1370 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1371 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1372 struct list_head *clinks = &tsk->cong_links;
1373 bool syn = !tipc_sk_type_connectionless(sk);
1374 struct tipc_group *grp = tsk->group;
1375 struct tipc_msg *hdr = &tsk->phdr;
1376 struct tipc_name_seq *seq;
1377 struct sk_buff_head pkts;
1378 u32 dport = 0, dnode = 0;
1379 u32 type = 0, inst = 0;
1380 int mtu, rc;
1382 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1383 return -EMSGSIZE;
1385 if (likely(dest)) {
1386 if (unlikely(m->msg_namelen < sizeof(*dest)))
1387 return -EINVAL;
1388 if (unlikely(dest->family != AF_TIPC))
1389 return -EINVAL;
1392 if (grp) {
1393 if (!dest)
1394 return tipc_send_group_bcast(sock, m, dlen, timeout);
1395 if (dest->addrtype == TIPC_ADDR_NAME)
1396 return tipc_send_group_anycast(sock, m, dlen, timeout);
1397 if (dest->addrtype == TIPC_ADDR_ID)
1398 return tipc_send_group_unicast(sock, m, dlen, timeout);
1399 if (dest->addrtype == TIPC_ADDR_MCAST)
1400 return tipc_send_group_mcast(sock, m, dlen, timeout);
1401 return -EINVAL;
1404 if (unlikely(!dest)) {
1405 dest = &tsk->peer;
1406 if (!syn && dest->family != AF_TIPC)
1407 return -EDESTADDRREQ;
1410 if (unlikely(syn)) {
1411 if (sk->sk_state == TIPC_LISTEN)
1412 return -EPIPE;
1413 if (sk->sk_state != TIPC_OPEN)
1414 return -EISCONN;
1415 if (tsk->published)
1416 return -EOPNOTSUPP;
1417 if (dest->addrtype == TIPC_ADDR_NAME) {
1418 tsk->conn_type = dest->addr.name.name.type;
1419 tsk->conn_instance = dest->addr.name.name.instance;
1421 msg_set_syn(hdr, 1);
1424 seq = &dest->addr.nameseq;
1425 if (dest->addrtype == TIPC_ADDR_MCAST)
1426 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1428 if (dest->addrtype == TIPC_ADDR_NAME) {
1429 type = dest->addr.name.name.type;
1430 inst = dest->addr.name.name.instance;
1431 dnode = dest->addr.name.domain;
1432 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1433 if (unlikely(!dport && !dnode))
1434 return -EHOSTUNREACH;
1435 } else if (dest->addrtype == TIPC_ADDR_ID) {
1436 dnode = dest->addr.id.node;
1437 } else {
1438 return -EINVAL;
1441 /* Block or return if destination link is congested */
1442 rc = tipc_wait_for_cond(sock, &timeout,
1443 !tipc_dest_find(clinks, dnode, 0));
1444 if (unlikely(rc))
1445 return rc;
1447 if (dest->addrtype == TIPC_ADDR_NAME) {
1448 msg_set_type(hdr, TIPC_NAMED_MSG);
1449 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1450 msg_set_nametype(hdr, type);
1451 msg_set_nameinst(hdr, inst);
1452 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1453 msg_set_destnode(hdr, dnode);
1454 msg_set_destport(hdr, dport);
1455 } else { /* TIPC_ADDR_ID */
1456 msg_set_type(hdr, TIPC_DIRECT_MSG);
1457 msg_set_lookup_scope(hdr, 0);
1458 msg_set_destnode(hdr, dnode);
1459 msg_set_destport(hdr, dest->addr.id.ref);
1460 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1463 __skb_queue_head_init(&pkts);
1464 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1465 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1466 if (unlikely(rc != dlen))
1467 return rc;
1468 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1469 __skb_queue_purge(&pkts);
1470 return -ENOMEM;
1473 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1474 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1475 if (unlikely(rc == -ELINKCONG)) {
1476 tipc_dest_push(clinks, dnode, 0);
1477 tsk->cong_link_cnt++;
1478 rc = 0;
1481 if (unlikely(syn && !rc))
1482 tipc_set_sk_state(sk, TIPC_CONNECTING);
1484 return rc ? rc : dlen;
1488 * tipc_sendstream - send stream-oriented data
1489 * @sock: socket structure
1490 * @m: data to send
1491 * @dsz: total length of data to be transmitted
1493 * Used for SOCK_STREAM data.
1495 * Returns the number of bytes sent on success (or partial success),
1496 * or errno if no data sent
1498 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1500 struct sock *sk = sock->sk;
1501 int ret;
1503 lock_sock(sk);
1504 ret = __tipc_sendstream(sock, m, dsz);
1505 release_sock(sk);
1507 return ret;
1510 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1512 struct sock *sk = sock->sk;
1513 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1514 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1515 struct sk_buff_head *txq = &sk->sk_write_queue;
1516 struct tipc_sock *tsk = tipc_sk(sk);
1517 struct tipc_msg *hdr = &tsk->phdr;
1518 struct net *net = sock_net(sk);
1519 u32 dnode = tsk_peer_node(tsk);
1520 int maxnagle = tsk->maxnagle;
1521 int maxpkt = tsk->max_pkt;
1522 int send, sent = 0;
1523 int blocks, rc = 0;
1525 if (unlikely(dlen > INT_MAX))
1526 return -EMSGSIZE;
1528 /* Handle implicit connection setup */
1529 if (unlikely(dest)) {
1530 rc = __tipc_sendmsg(sock, m, dlen);
1531 if (dlen && dlen == rc) {
1532 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1533 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1535 return rc;
1538 do {
1539 rc = tipc_wait_for_cond(sock, &timeout,
1540 (!tsk->cong_link_cnt &&
1541 !tsk_conn_cong(tsk) &&
1542 tipc_sk_connected(sk)));
1543 if (unlikely(rc))
1544 break;
1545 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1546 blocks = tsk->snd_backlog;
1547 if (tsk->oneway++ >= 4 && send <= maxnagle) {
1548 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1549 if (unlikely(rc < 0))
1550 break;
1551 blocks += rc;
1552 if (blocks <= 64 && tsk->expect_ack) {
1553 tsk->snd_backlog = blocks;
1554 sent += send;
1555 break;
1557 tsk->expect_ack = true;
1558 } else {
1559 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1560 if (unlikely(rc != send))
1561 break;
1562 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1564 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1565 TIPC_DUMP_SK_SNDQ, " ");
1566 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1567 if (unlikely(rc == -ELINKCONG)) {
1568 tsk->cong_link_cnt = 1;
1569 rc = 0;
1571 if (likely(!rc)) {
1572 tsk->snt_unacked += blocks;
1573 tsk->snd_backlog = 0;
1574 sent += send;
1576 } while (sent < dlen && !rc);
1578 return sent ? sent : rc;
1582 * tipc_send_packet - send a connection-oriented message
1583 * @sock: socket structure
1584 * @m: message to send
1585 * @dsz: length of data to be transmitted
1587 * Used for SOCK_SEQPACKET messages.
1589 * Returns the number of bytes sent on success, or errno otherwise
1591 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1593 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1594 return -EMSGSIZE;
1596 return tipc_sendstream(sock, m, dsz);
1599 /* tipc_sk_finish_conn - complete the setup of a connection
1601 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1602 u32 peer_node)
1604 struct sock *sk = &tsk->sk;
1605 struct net *net = sock_net(sk);
1606 struct tipc_msg *msg = &tsk->phdr;
1608 msg_set_syn(msg, 0);
1609 msg_set_destnode(msg, peer_node);
1610 msg_set_destport(msg, peer_port);
1611 msg_set_type(msg, TIPC_CONN_MSG);
1612 msg_set_lookup_scope(msg, 0);
1613 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1615 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1616 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1617 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1618 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1619 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1620 tsk_set_nagle(tsk);
1621 __skb_queue_purge(&sk->sk_write_queue);
1622 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1623 return;
1625 /* Fall back to message based flow control */
1626 tsk->rcv_win = FLOWCTL_MSG_WIN;
1627 tsk->snd_win = FLOWCTL_MSG_WIN;
1631 * tipc_sk_set_orig_addr - capture sender's address for received message
1632 * @m: descriptor for message info
1633 * @hdr: received message header
1635 * Note: Address is not captured if not requested by receiver.
1637 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1639 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1640 struct tipc_msg *hdr = buf_msg(skb);
1642 if (!srcaddr)
1643 return;
1645 srcaddr->sock.family = AF_TIPC;
1646 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1647 srcaddr->sock.scope = 0;
1648 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1649 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1650 srcaddr->sock.addr.name.domain = 0;
1651 m->msg_namelen = sizeof(struct sockaddr_tipc);
1653 if (!msg_in_group(hdr))
1654 return;
1656 /* Group message users may also want to know sending member's id */
1657 srcaddr->member.family = AF_TIPC;
1658 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1659 srcaddr->member.scope = 0;
1660 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1661 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1662 srcaddr->member.addr.name.domain = 0;
1663 m->msg_namelen = sizeof(*srcaddr);
1667 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1668 * @m: descriptor for message info
1669 * @skb: received message buffer
1670 * @tsk: TIPC port associated with message
1672 * Note: Ancillary data is not captured if not requested by receiver.
1674 * Returns 0 if successful, otherwise errno
1676 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1677 struct tipc_sock *tsk)
1679 struct tipc_msg *msg;
1680 u32 anc_data[3];
1681 u32 err;
1682 u32 dest_type;
1683 int has_name;
1684 int res;
1686 if (likely(m->msg_controllen == 0))
1687 return 0;
1688 msg = buf_msg(skb);
1690 /* Optionally capture errored message object(s) */
1691 err = msg ? msg_errcode(msg) : 0;
1692 if (unlikely(err)) {
1693 anc_data[0] = err;
1694 anc_data[1] = msg_data_sz(msg);
1695 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1696 if (res)
1697 return res;
1698 if (anc_data[1]) {
1699 if (skb_linearize(skb))
1700 return -ENOMEM;
1701 msg = buf_msg(skb);
1702 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1703 msg_data(msg));
1704 if (res)
1705 return res;
1709 /* Optionally capture message destination object */
1710 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1711 switch (dest_type) {
1712 case TIPC_NAMED_MSG:
1713 has_name = 1;
1714 anc_data[0] = msg_nametype(msg);
1715 anc_data[1] = msg_namelower(msg);
1716 anc_data[2] = msg_namelower(msg);
1717 break;
1718 case TIPC_MCAST_MSG:
1719 has_name = 1;
1720 anc_data[0] = msg_nametype(msg);
1721 anc_data[1] = msg_namelower(msg);
1722 anc_data[2] = msg_nameupper(msg);
1723 break;
1724 case TIPC_CONN_MSG:
1725 has_name = (tsk->conn_type != 0);
1726 anc_data[0] = tsk->conn_type;
1727 anc_data[1] = tsk->conn_instance;
1728 anc_data[2] = tsk->conn_instance;
1729 break;
1730 default:
1731 has_name = 0;
1733 if (has_name) {
1734 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1735 if (res)
1736 return res;
1739 return 0;
1742 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1744 struct sock *sk = &tsk->sk;
1745 struct sk_buff *skb = NULL;
1746 struct tipc_msg *msg;
1747 u32 peer_port = tsk_peer_port(tsk);
1748 u32 dnode = tsk_peer_node(tsk);
1750 if (!tipc_sk_connected(sk))
1751 return NULL;
1752 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1753 dnode, tsk_own_node(tsk), peer_port,
1754 tsk->portid, TIPC_OK);
1755 if (!skb)
1756 return NULL;
1757 msg = buf_msg(skb);
1758 msg_set_conn_ack(msg, tsk->rcv_unacked);
1759 tsk->rcv_unacked = 0;
1761 /* Adjust to and advertize the correct window limit */
1762 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1763 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1764 msg_set_adv_win(msg, tsk->rcv_win);
1766 return skb;
1769 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1771 struct sk_buff *skb;
1773 skb = tipc_sk_build_ack(tsk);
1774 if (!skb)
1775 return;
1777 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1778 msg_link_selector(buf_msg(skb)));
1781 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1783 struct sock *sk = sock->sk;
1784 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1785 long timeo = *timeop;
1786 int err = sock_error(sk);
1788 if (err)
1789 return err;
1791 for (;;) {
1792 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1793 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1794 err = -ENOTCONN;
1795 break;
1797 add_wait_queue(sk_sleep(sk), &wait);
1798 release_sock(sk);
1799 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1800 sched_annotate_sleep();
1801 lock_sock(sk);
1802 remove_wait_queue(sk_sleep(sk), &wait);
1804 err = 0;
1805 if (!skb_queue_empty(&sk->sk_receive_queue))
1806 break;
1807 err = -EAGAIN;
1808 if (!timeo)
1809 break;
1810 err = sock_intr_errno(timeo);
1811 if (signal_pending(current))
1812 break;
1814 err = sock_error(sk);
1815 if (err)
1816 break;
1818 *timeop = timeo;
1819 return err;
1823 * tipc_recvmsg - receive packet-oriented message
1824 * @m: descriptor for message info
1825 * @buflen: length of user buffer area
1826 * @flags: receive flags
1828 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1829 * If the complete message doesn't fit in user area, truncate it.
1831 * Returns size of returned message data, errno otherwise
1833 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1834 size_t buflen, int flags)
1836 struct sock *sk = sock->sk;
1837 bool connected = !tipc_sk_type_connectionless(sk);
1838 struct tipc_sock *tsk = tipc_sk(sk);
1839 int rc, err, hlen, dlen, copy;
1840 struct sk_buff_head xmitq;
1841 struct tipc_msg *hdr;
1842 struct sk_buff *skb;
1843 bool grp_evt;
1844 long timeout;
1846 /* Catch invalid receive requests */
1847 if (unlikely(!buflen))
1848 return -EINVAL;
1850 lock_sock(sk);
1851 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1852 rc = -ENOTCONN;
1853 goto exit;
1855 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1857 /* Step rcv queue to first msg with data or error; wait if necessary */
1858 do {
1859 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1860 if (unlikely(rc))
1861 goto exit;
1862 skb = skb_peek(&sk->sk_receive_queue);
1863 hdr = buf_msg(skb);
1864 dlen = msg_data_sz(hdr);
1865 hlen = msg_hdr_sz(hdr);
1866 err = msg_errcode(hdr);
1867 grp_evt = msg_is_grp_evt(hdr);
1868 if (likely(dlen || err))
1869 break;
1870 tsk_advance_rx_queue(sk);
1871 } while (1);
1873 /* Collect msg meta data, including error code and rejected data */
1874 tipc_sk_set_orig_addr(m, skb);
1875 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1876 if (unlikely(rc))
1877 goto exit;
1878 hdr = buf_msg(skb);
1880 /* Capture data if non-error msg, otherwise just set return value */
1881 if (likely(!err)) {
1882 copy = min_t(int, dlen, buflen);
1883 if (unlikely(copy != dlen))
1884 m->msg_flags |= MSG_TRUNC;
1885 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1886 } else {
1887 copy = 0;
1888 rc = 0;
1889 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1890 rc = -ECONNRESET;
1892 if (unlikely(rc))
1893 goto exit;
1895 /* Mark message as group event if applicable */
1896 if (unlikely(grp_evt)) {
1897 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1898 m->msg_flags |= MSG_EOR;
1899 m->msg_flags |= MSG_OOB;
1900 copy = 0;
1903 /* Caption of data or error code/rejected data was successful */
1904 if (unlikely(flags & MSG_PEEK))
1905 goto exit;
1907 /* Send group flow control advertisement when applicable */
1908 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1909 __skb_queue_head_init(&xmitq);
1910 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1911 msg_orignode(hdr), msg_origport(hdr),
1912 &xmitq);
1913 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1916 tsk_advance_rx_queue(sk);
1918 if (likely(!connected))
1919 goto exit;
1921 /* Send connection flow control advertisement when applicable */
1922 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1923 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1924 tipc_sk_send_ack(tsk);
1925 exit:
1926 release_sock(sk);
1927 return rc ? rc : copy;
1931 * tipc_recvstream - receive stream-oriented data
1932 * @m: descriptor for message info
1933 * @buflen: total size of user buffer area
1934 * @flags: receive flags
1936 * Used for SOCK_STREAM messages only. If not enough data is available
1937 * will optionally wait for more; never truncates data.
1939 * Returns size of returned message data, errno otherwise
1941 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1942 size_t buflen, int flags)
1944 struct sock *sk = sock->sk;
1945 struct tipc_sock *tsk = tipc_sk(sk);
1946 struct sk_buff *skb;
1947 struct tipc_msg *hdr;
1948 struct tipc_skb_cb *skb_cb;
1949 bool peek = flags & MSG_PEEK;
1950 int offset, required, copy, copied = 0;
1951 int hlen, dlen, err, rc;
1952 long timeout;
1954 /* Catch invalid receive attempts */
1955 if (unlikely(!buflen))
1956 return -EINVAL;
1958 lock_sock(sk);
1960 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1961 rc = -ENOTCONN;
1962 goto exit;
1964 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1965 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1967 do {
1968 /* Look at first msg in receive queue; wait if necessary */
1969 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1970 if (unlikely(rc))
1971 break;
1972 skb = skb_peek(&sk->sk_receive_queue);
1973 skb_cb = TIPC_SKB_CB(skb);
1974 hdr = buf_msg(skb);
1975 dlen = msg_data_sz(hdr);
1976 hlen = msg_hdr_sz(hdr);
1977 err = msg_errcode(hdr);
1979 /* Discard any empty non-errored (SYN-) message */
1980 if (unlikely(!dlen && !err)) {
1981 tsk_advance_rx_queue(sk);
1982 continue;
1985 /* Collect msg meta data, incl. error code and rejected data */
1986 if (!copied) {
1987 tipc_sk_set_orig_addr(m, skb);
1988 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1989 if (rc)
1990 break;
1991 hdr = buf_msg(skb);
1994 /* Copy data if msg ok, otherwise return error/partial data */
1995 if (likely(!err)) {
1996 offset = skb_cb->bytes_read;
1997 copy = min_t(int, dlen - offset, buflen - copied);
1998 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1999 if (unlikely(rc))
2000 break;
2001 copied += copy;
2002 offset += copy;
2003 if (unlikely(offset < dlen)) {
2004 if (!peek)
2005 skb_cb->bytes_read = offset;
2006 break;
2008 } else {
2009 rc = 0;
2010 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2011 rc = -ECONNRESET;
2012 if (copied || rc)
2013 break;
2016 if (unlikely(peek))
2017 break;
2019 tsk_advance_rx_queue(sk);
2021 /* Send connection flow control advertisement when applicable */
2022 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2023 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2024 tipc_sk_send_ack(tsk);
2026 /* Exit if all requested data or FIN/error received */
2027 if (copied == buflen || err)
2028 break;
2030 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2031 exit:
2032 release_sock(sk);
2033 return copied ? copied : rc;
2037 * tipc_write_space - wake up thread if port congestion is released
2038 * @sk: socket
2040 static void tipc_write_space(struct sock *sk)
2042 struct socket_wq *wq;
2044 rcu_read_lock();
2045 wq = rcu_dereference(sk->sk_wq);
2046 if (skwq_has_sleeper(wq))
2047 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2048 EPOLLWRNORM | EPOLLWRBAND);
2049 rcu_read_unlock();
2053 * tipc_data_ready - wake up threads to indicate messages have been received
2054 * @sk: socket
2055 * @len: the length of messages
2057 static void tipc_data_ready(struct sock *sk)
2059 struct socket_wq *wq;
2061 rcu_read_lock();
2062 wq = rcu_dereference(sk->sk_wq);
2063 if (skwq_has_sleeper(wq))
2064 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2065 EPOLLRDNORM | EPOLLRDBAND);
2066 rcu_read_unlock();
2069 static void tipc_sock_destruct(struct sock *sk)
2071 __skb_queue_purge(&sk->sk_receive_queue);
2074 static void tipc_sk_proto_rcv(struct sock *sk,
2075 struct sk_buff_head *inputq,
2076 struct sk_buff_head *xmitq)
2078 struct sk_buff *skb = __skb_dequeue(inputq);
2079 struct tipc_sock *tsk = tipc_sk(sk);
2080 struct tipc_msg *hdr = buf_msg(skb);
2081 struct tipc_group *grp = tsk->group;
2082 bool wakeup = false;
2084 switch (msg_user(hdr)) {
2085 case CONN_MANAGER:
2086 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2087 return;
2088 case SOCK_WAKEUP:
2089 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2090 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2091 smp_wmb();
2092 tsk->cong_link_cnt--;
2093 wakeup = true;
2094 tipc_sk_push_backlog(tsk);
2095 break;
2096 case GROUP_PROTOCOL:
2097 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2098 break;
2099 case TOP_SRV:
2100 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2101 hdr, inputq, xmitq);
2102 break;
2103 default:
2104 break;
2107 if (wakeup)
2108 sk->sk_write_space(sk);
2110 kfree_skb(skb);
2114 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2115 * @tsk: TIPC socket
2116 * @skb: pointer to message buffer.
2117 * @xmitq: for Nagle ACK if any
2118 * Returns true if message should be added to receive queue, false otherwise
2120 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2121 struct sk_buff_head *xmitq)
2123 struct sock *sk = &tsk->sk;
2124 struct net *net = sock_net(sk);
2125 struct tipc_msg *hdr = buf_msg(skb);
2126 bool con_msg = msg_connected(hdr);
2127 u32 pport = tsk_peer_port(tsk);
2128 u32 pnode = tsk_peer_node(tsk);
2129 u32 oport = msg_origport(hdr);
2130 u32 onode = msg_orignode(hdr);
2131 int err = msg_errcode(hdr);
2132 unsigned long delay;
2134 if (unlikely(msg_mcast(hdr)))
2135 return false;
2136 tsk->oneway = 0;
2138 switch (sk->sk_state) {
2139 case TIPC_CONNECTING:
2140 /* Setup ACK */
2141 if (likely(con_msg)) {
2142 if (err)
2143 break;
2144 tipc_sk_finish_conn(tsk, oport, onode);
2145 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2146 /* ACK+ message with data is added to receive queue */
2147 if (msg_data_sz(hdr))
2148 return true;
2149 /* Empty ACK-, - wake up sleeping connect() and drop */
2150 sk->sk_state_change(sk);
2151 msg_set_dest_droppable(hdr, 1);
2152 return false;
2154 /* Ignore connectionless message if not from listening socket */
2155 if (oport != pport || onode != pnode)
2156 return false;
2158 /* Rejected SYN */
2159 if (err != TIPC_ERR_OVERLOAD)
2160 break;
2162 /* Prepare for new setup attempt if we have a SYN clone */
2163 if (skb_queue_empty(&sk->sk_write_queue))
2164 break;
2165 get_random_bytes(&delay, 2);
2166 delay %= (tsk->conn_timeout / 4);
2167 delay = msecs_to_jiffies(delay + 100);
2168 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2169 return false;
2170 case TIPC_OPEN:
2171 case TIPC_DISCONNECTING:
2172 return false;
2173 case TIPC_LISTEN:
2174 /* Accept only SYN message */
2175 if (!msg_is_syn(hdr) &&
2176 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2177 return false;
2178 if (!con_msg && !err)
2179 return true;
2180 return false;
2181 case TIPC_ESTABLISHED:
2182 if (!skb_queue_empty(&sk->sk_write_queue))
2183 tipc_sk_push_backlog(tsk);
2184 /* Accept only connection-based messages sent by peer */
2185 if (likely(con_msg && !err && pport == oport &&
2186 pnode == onode)) {
2187 if (msg_ack_required(hdr)) {
2188 struct sk_buff *skb;
2190 skb = tipc_sk_build_ack(tsk);
2191 if (skb)
2192 __skb_queue_tail(xmitq, skb);
2194 return true;
2196 if (!tsk_peer_msg(tsk, hdr))
2197 return false;
2198 if (!err)
2199 return true;
2200 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2201 tipc_node_remove_conn(net, pnode, tsk->portid);
2202 sk->sk_state_change(sk);
2203 return true;
2204 default:
2205 pr_err("Unknown sk_state %u\n", sk->sk_state);
2207 /* Abort connection setup attempt */
2208 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2209 sk->sk_err = ECONNREFUSED;
2210 sk->sk_state_change(sk);
2211 return true;
2215 * rcvbuf_limit - get proper overload limit of socket receive queue
2216 * @sk: socket
2217 * @skb: message
2219 * For connection oriented messages, irrespective of importance,
2220 * default queue limit is 2 MB.
2222 * For connectionless messages, queue limits are based on message
2223 * importance as follows:
2225 * TIPC_LOW_IMPORTANCE (2 MB)
2226 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2227 * TIPC_HIGH_IMPORTANCE (8 MB)
2228 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2230 * Returns overload limit according to corresponding message importance
2232 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2234 struct tipc_sock *tsk = tipc_sk(sk);
2235 struct tipc_msg *hdr = buf_msg(skb);
2237 if (unlikely(msg_in_group(hdr)))
2238 return READ_ONCE(sk->sk_rcvbuf);
2240 if (unlikely(!msg_connected(hdr)))
2241 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2243 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2244 return READ_ONCE(sk->sk_rcvbuf);
2246 return FLOWCTL_MSG_LIM;
2250 * tipc_sk_filter_rcv - validate incoming message
2251 * @sk: socket
2252 * @skb: pointer to message.
2254 * Enqueues message on receive queue if acceptable; optionally handles
2255 * disconnect indication for a connected socket.
2257 * Called with socket lock already taken
2260 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2261 struct sk_buff_head *xmitq)
2263 bool sk_conn = !tipc_sk_type_connectionless(sk);
2264 struct tipc_sock *tsk = tipc_sk(sk);
2265 struct tipc_group *grp = tsk->group;
2266 struct tipc_msg *hdr = buf_msg(skb);
2267 struct net *net = sock_net(sk);
2268 struct sk_buff_head inputq;
2269 int mtyp = msg_type(hdr);
2270 int limit, err = TIPC_OK;
2272 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2273 TIPC_SKB_CB(skb)->bytes_read = 0;
2274 __skb_queue_head_init(&inputq);
2275 __skb_queue_tail(&inputq, skb);
2277 if (unlikely(!msg_isdata(hdr)))
2278 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2280 if (unlikely(grp))
2281 tipc_group_filter_msg(grp, &inputq, xmitq);
2283 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2284 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2286 /* Validate and add to receive buffer if there is space */
2287 while ((skb = __skb_dequeue(&inputq))) {
2288 hdr = buf_msg(skb);
2289 limit = rcvbuf_limit(sk, skb);
2290 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2291 (!sk_conn && msg_connected(hdr)) ||
2292 (!grp && msg_in_group(hdr)))
2293 err = TIPC_ERR_NO_PORT;
2294 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2295 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2296 "err_overload2!");
2297 atomic_inc(&sk->sk_drops);
2298 err = TIPC_ERR_OVERLOAD;
2301 if (unlikely(err)) {
2302 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2303 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2304 "@filter_rcv!");
2305 __skb_queue_tail(xmitq, skb);
2307 err = TIPC_OK;
2308 continue;
2310 __skb_queue_tail(&sk->sk_receive_queue, skb);
2311 skb_set_owner_r(skb, sk);
2312 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2313 "rcvq >90% allocated!");
2314 sk->sk_data_ready(sk);
2319 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2320 * @sk: socket
2321 * @skb: message
2323 * Caller must hold socket lock
2325 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2327 unsigned int before = sk_rmem_alloc_get(sk);
2328 struct sk_buff_head xmitq;
2329 unsigned int added;
2331 __skb_queue_head_init(&xmitq);
2333 tipc_sk_filter_rcv(sk, skb, &xmitq);
2334 added = sk_rmem_alloc_get(sk) - before;
2335 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2337 /* Send pending response/rejected messages, if any */
2338 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2339 return 0;
2343 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2344 * inputq and try adding them to socket or backlog queue
2345 * @inputq: list of incoming buffers with potentially different destinations
2346 * @sk: socket where the buffers should be enqueued
2347 * @dport: port number for the socket
2349 * Caller must hold socket lock
2351 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2352 u32 dport, struct sk_buff_head *xmitq)
2354 unsigned long time_limit = jiffies + 2;
2355 struct sk_buff *skb;
2356 unsigned int lim;
2357 atomic_t *dcnt;
2358 u32 onode;
2360 while (skb_queue_len(inputq)) {
2361 if (unlikely(time_after_eq(jiffies, time_limit)))
2362 return;
2364 skb = tipc_skb_dequeue(inputq, dport);
2365 if (unlikely(!skb))
2366 return;
2368 /* Add message directly to receive queue if possible */
2369 if (!sock_owned_by_user(sk)) {
2370 tipc_sk_filter_rcv(sk, skb, xmitq);
2371 continue;
2374 /* Try backlog, compensating for double-counted bytes */
2375 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2376 if (!sk->sk_backlog.len)
2377 atomic_set(dcnt, 0);
2378 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2379 if (likely(!sk_add_backlog(sk, skb, lim))) {
2380 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2381 "bklg & rcvq >90% allocated!");
2382 continue;
2385 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2386 /* Overload => reject message back to sender */
2387 onode = tipc_own_addr(sock_net(sk));
2388 atomic_inc(&sk->sk_drops);
2389 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2390 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2391 "@sk_enqueue!");
2392 __skb_queue_tail(xmitq, skb);
2394 break;
2399 * tipc_sk_rcv - handle a chain of incoming buffers
2400 * @inputq: buffer list containing the buffers
2401 * Consumes all buffers in list until inputq is empty
2402 * Note: may be called in multiple threads referring to the same queue
2404 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2406 struct sk_buff_head xmitq;
2407 u32 dnode, dport = 0;
2408 int err;
2409 struct tipc_sock *tsk;
2410 struct sock *sk;
2411 struct sk_buff *skb;
2413 __skb_queue_head_init(&xmitq);
2414 while (skb_queue_len(inputq)) {
2415 dport = tipc_skb_peek_port(inputq, dport);
2416 tsk = tipc_sk_lookup(net, dport);
2418 if (likely(tsk)) {
2419 sk = &tsk->sk;
2420 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2421 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2422 spin_unlock_bh(&sk->sk_lock.slock);
2424 /* Send pending response/rejected messages, if any */
2425 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2426 sock_put(sk);
2427 continue;
2429 /* No destination socket => dequeue skb if still there */
2430 skb = tipc_skb_dequeue(inputq, dport);
2431 if (!skb)
2432 return;
2434 /* Try secondary lookup if unresolved named message */
2435 err = TIPC_ERR_NO_PORT;
2436 if (tipc_msg_lookup_dest(net, skb, &err))
2437 goto xmit;
2439 /* Prepare for message rejection */
2440 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2441 continue;
2443 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2444 xmit:
2445 dnode = msg_destnode(buf_msg(skb));
2446 tipc_node_xmit_skb(net, skb, dnode, dport);
2450 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2452 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2453 struct sock *sk = sock->sk;
2454 int done;
2456 do {
2457 int err = sock_error(sk);
2458 if (err)
2459 return err;
2460 if (!*timeo_p)
2461 return -ETIMEDOUT;
2462 if (signal_pending(current))
2463 return sock_intr_errno(*timeo_p);
2464 if (sk->sk_state == TIPC_DISCONNECTING)
2465 break;
2467 add_wait_queue(sk_sleep(sk), &wait);
2468 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2469 &wait);
2470 remove_wait_queue(sk_sleep(sk), &wait);
2471 } while (!done);
2472 return 0;
2475 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2477 if (addr->family != AF_TIPC)
2478 return false;
2479 if (addr->addrtype == TIPC_SERVICE_RANGE)
2480 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2481 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2482 addr->addrtype == TIPC_SOCKET_ADDR);
2486 * tipc_connect - establish a connection to another TIPC port
2487 * @sock: socket structure
2488 * @dest: socket address for destination port
2489 * @destlen: size of socket address data structure
2490 * @flags: file-related flags associated with socket
2492 * Returns 0 on success, errno otherwise
2494 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2495 int destlen, int flags)
2497 struct sock *sk = sock->sk;
2498 struct tipc_sock *tsk = tipc_sk(sk);
2499 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2500 struct msghdr m = {NULL,};
2501 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2502 int previous;
2503 int res = 0;
2505 if (destlen != sizeof(struct sockaddr_tipc))
2506 return -EINVAL;
2508 lock_sock(sk);
2510 if (tsk->group) {
2511 res = -EINVAL;
2512 goto exit;
2515 if (dst->family == AF_UNSPEC) {
2516 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2517 if (!tipc_sk_type_connectionless(sk))
2518 res = -EINVAL;
2519 goto exit;
2521 if (!tipc_sockaddr_is_sane(dst)) {
2522 res = -EINVAL;
2523 goto exit;
2525 /* DGRAM/RDM connect(), just save the destaddr */
2526 if (tipc_sk_type_connectionless(sk)) {
2527 memcpy(&tsk->peer, dest, destlen);
2528 goto exit;
2529 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2530 res = -EINVAL;
2531 goto exit;
2534 previous = sk->sk_state;
2536 switch (sk->sk_state) {
2537 case TIPC_OPEN:
2538 /* Send a 'SYN-' to destination */
2539 m.msg_name = dest;
2540 m.msg_namelen = destlen;
2542 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2543 * indicate send_msg() is never blocked.
2545 if (!timeout)
2546 m.msg_flags = MSG_DONTWAIT;
2548 res = __tipc_sendmsg(sock, &m, 0);
2549 if ((res < 0) && (res != -EWOULDBLOCK))
2550 goto exit;
2552 /* Just entered TIPC_CONNECTING state; the only
2553 * difference is that return value in non-blocking
2554 * case is EINPROGRESS, rather than EALREADY.
2556 res = -EINPROGRESS;
2557 /* fall through */
2558 case TIPC_CONNECTING:
2559 if (!timeout) {
2560 if (previous == TIPC_CONNECTING)
2561 res = -EALREADY;
2562 goto exit;
2564 timeout = msecs_to_jiffies(timeout);
2565 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2566 res = tipc_wait_for_connect(sock, &timeout);
2567 break;
2568 case TIPC_ESTABLISHED:
2569 res = -EISCONN;
2570 break;
2571 default:
2572 res = -EINVAL;
2575 exit:
2576 release_sock(sk);
2577 return res;
2581 * tipc_listen - allow socket to listen for incoming connections
2582 * @sock: socket structure
2583 * @len: (unused)
2585 * Returns 0 on success, errno otherwise
2587 static int tipc_listen(struct socket *sock, int len)
2589 struct sock *sk = sock->sk;
2590 int res;
2592 lock_sock(sk);
2593 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2594 release_sock(sk);
2596 return res;
2599 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2601 struct sock *sk = sock->sk;
2602 DEFINE_WAIT(wait);
2603 int err;
2605 /* True wake-one mechanism for incoming connections: only
2606 * one process gets woken up, not the 'whole herd'.
2607 * Since we do not 'race & poll' for established sockets
2608 * anymore, the common case will execute the loop only once.
2610 for (;;) {
2611 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2612 TASK_INTERRUPTIBLE);
2613 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2614 release_sock(sk);
2615 timeo = schedule_timeout(timeo);
2616 lock_sock(sk);
2618 err = 0;
2619 if (!skb_queue_empty(&sk->sk_receive_queue))
2620 break;
2621 err = -EAGAIN;
2622 if (!timeo)
2623 break;
2624 err = sock_intr_errno(timeo);
2625 if (signal_pending(current))
2626 break;
2628 finish_wait(sk_sleep(sk), &wait);
2629 return err;
2633 * tipc_accept - wait for connection request
2634 * @sock: listening socket
2635 * @newsock: new socket that is to be connected
2636 * @flags: file-related flags associated with socket
2638 * Returns 0 on success, errno otherwise
2640 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2641 bool kern)
2643 struct sock *new_sk, *sk = sock->sk;
2644 struct sk_buff *buf;
2645 struct tipc_sock *new_tsock;
2646 struct tipc_msg *msg;
2647 long timeo;
2648 int res;
2650 lock_sock(sk);
2652 if (sk->sk_state != TIPC_LISTEN) {
2653 res = -EINVAL;
2654 goto exit;
2656 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2657 res = tipc_wait_for_accept(sock, timeo);
2658 if (res)
2659 goto exit;
2661 buf = skb_peek(&sk->sk_receive_queue);
2663 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2664 if (res)
2665 goto exit;
2666 security_sk_clone(sock->sk, new_sock->sk);
2668 new_sk = new_sock->sk;
2669 new_tsock = tipc_sk(new_sk);
2670 msg = buf_msg(buf);
2672 /* we lock on new_sk; but lockdep sees the lock on sk */
2673 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2676 * Reject any stray messages received by new socket
2677 * before the socket lock was taken (very, very unlikely)
2679 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2681 /* Connect new socket to it's peer */
2682 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2684 tsk_set_importance(new_tsock, msg_importance(msg));
2685 if (msg_named(msg)) {
2686 new_tsock->conn_type = msg_nametype(msg);
2687 new_tsock->conn_instance = msg_nameinst(msg);
2691 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2692 * Respond to 'SYN+' by queuing it on new socket.
2694 if (!msg_data_sz(msg)) {
2695 struct msghdr m = {NULL,};
2697 tsk_advance_rx_queue(sk);
2698 __tipc_sendstream(new_sock, &m, 0);
2699 } else {
2700 __skb_dequeue(&sk->sk_receive_queue);
2701 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2702 skb_set_owner_r(buf, new_sk);
2704 release_sock(new_sk);
2705 exit:
2706 release_sock(sk);
2707 return res;
2711 * tipc_shutdown - shutdown socket connection
2712 * @sock: socket structure
2713 * @how: direction to close (must be SHUT_RDWR)
2715 * Terminates connection (if necessary), then purges socket's receive queue.
2717 * Returns 0 on success, errno otherwise
2719 static int tipc_shutdown(struct socket *sock, int how)
2721 struct sock *sk = sock->sk;
2722 int res;
2724 if (how != SHUT_RDWR)
2725 return -EINVAL;
2727 lock_sock(sk);
2729 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2730 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2731 sk->sk_shutdown = SEND_SHUTDOWN;
2733 if (sk->sk_state == TIPC_DISCONNECTING) {
2734 /* Discard any unreceived messages */
2735 __skb_queue_purge(&sk->sk_receive_queue);
2737 /* Wake up anyone sleeping in poll */
2738 sk->sk_state_change(sk);
2739 res = 0;
2740 } else {
2741 res = -ENOTCONN;
2744 release_sock(sk);
2745 return res;
2748 static void tipc_sk_check_probing_state(struct sock *sk,
2749 struct sk_buff_head *list)
2751 struct tipc_sock *tsk = tipc_sk(sk);
2752 u32 pnode = tsk_peer_node(tsk);
2753 u32 pport = tsk_peer_port(tsk);
2754 u32 self = tsk_own_node(tsk);
2755 u32 oport = tsk->portid;
2756 struct sk_buff *skb;
2758 if (tsk->probe_unacked) {
2759 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2760 sk->sk_err = ECONNABORTED;
2761 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2762 sk->sk_state_change(sk);
2763 return;
2765 /* Prepare new probe */
2766 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2767 pnode, self, pport, oport, TIPC_OK);
2768 if (skb)
2769 __skb_queue_tail(list, skb);
2770 tsk->probe_unacked = true;
2771 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2774 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2776 struct tipc_sock *tsk = tipc_sk(sk);
2778 /* Try again later if dest link is congested */
2779 if (tsk->cong_link_cnt) {
2780 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2781 return;
2783 /* Prepare SYN for retransmit */
2784 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2787 static void tipc_sk_timeout(struct timer_list *t)
2789 struct sock *sk = from_timer(sk, t, sk_timer);
2790 struct tipc_sock *tsk = tipc_sk(sk);
2791 u32 pnode = tsk_peer_node(tsk);
2792 struct sk_buff_head list;
2793 int rc = 0;
2795 __skb_queue_head_init(&list);
2796 bh_lock_sock(sk);
2798 /* Try again later if socket is busy */
2799 if (sock_owned_by_user(sk)) {
2800 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2801 bh_unlock_sock(sk);
2802 sock_put(sk);
2803 return;
2806 if (sk->sk_state == TIPC_ESTABLISHED)
2807 tipc_sk_check_probing_state(sk, &list);
2808 else if (sk->sk_state == TIPC_CONNECTING)
2809 tipc_sk_retry_connect(sk, &list);
2811 bh_unlock_sock(sk);
2813 if (!skb_queue_empty(&list))
2814 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2816 /* SYN messages may cause link congestion */
2817 if (rc == -ELINKCONG) {
2818 tipc_dest_push(&tsk->cong_links, pnode, 0);
2819 tsk->cong_link_cnt = 1;
2821 sock_put(sk);
2824 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2825 struct tipc_name_seq const *seq)
2827 struct sock *sk = &tsk->sk;
2828 struct net *net = sock_net(sk);
2829 struct publication *publ;
2830 u32 key;
2832 if (scope != TIPC_NODE_SCOPE)
2833 scope = TIPC_CLUSTER_SCOPE;
2835 if (tipc_sk_connected(sk))
2836 return -EINVAL;
2837 key = tsk->portid + tsk->pub_count + 1;
2838 if (key == tsk->portid)
2839 return -EADDRINUSE;
2841 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2842 scope, tsk->portid, key);
2843 if (unlikely(!publ))
2844 return -EINVAL;
2846 list_add(&publ->binding_sock, &tsk->publications);
2847 tsk->pub_count++;
2848 tsk->published = 1;
2849 return 0;
2852 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2853 struct tipc_name_seq const *seq)
2855 struct net *net = sock_net(&tsk->sk);
2856 struct publication *publ;
2857 struct publication *safe;
2858 int rc = -EINVAL;
2860 if (scope != TIPC_NODE_SCOPE)
2861 scope = TIPC_CLUSTER_SCOPE;
2863 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2864 if (seq) {
2865 if (publ->scope != scope)
2866 continue;
2867 if (publ->type != seq->type)
2868 continue;
2869 if (publ->lower != seq->lower)
2870 continue;
2871 if (publ->upper != seq->upper)
2872 break;
2873 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2874 publ->upper, publ->key);
2875 rc = 0;
2876 break;
2878 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2879 publ->upper, publ->key);
2880 rc = 0;
2882 if (list_empty(&tsk->publications))
2883 tsk->published = 0;
2884 return rc;
2887 /* tipc_sk_reinit: set non-zero address in all existing sockets
2888 * when we go from standalone to network mode.
2890 void tipc_sk_reinit(struct net *net)
2892 struct tipc_net *tn = net_generic(net, tipc_net_id);
2893 struct rhashtable_iter iter;
2894 struct tipc_sock *tsk;
2895 struct tipc_msg *msg;
2897 rhashtable_walk_enter(&tn->sk_rht, &iter);
2899 do {
2900 rhashtable_walk_start(&iter);
2902 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2903 sock_hold(&tsk->sk);
2904 rhashtable_walk_stop(&iter);
2905 lock_sock(&tsk->sk);
2906 msg = &tsk->phdr;
2907 msg_set_prevnode(msg, tipc_own_addr(net));
2908 msg_set_orignode(msg, tipc_own_addr(net));
2909 release_sock(&tsk->sk);
2910 rhashtable_walk_start(&iter);
2911 sock_put(&tsk->sk);
2914 rhashtable_walk_stop(&iter);
2915 } while (tsk == ERR_PTR(-EAGAIN));
2917 rhashtable_walk_exit(&iter);
2920 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2922 struct tipc_net *tn = net_generic(net, tipc_net_id);
2923 struct tipc_sock *tsk;
2925 rcu_read_lock();
2926 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2927 if (tsk)
2928 sock_hold(&tsk->sk);
2929 rcu_read_unlock();
2931 return tsk;
2934 static int tipc_sk_insert(struct tipc_sock *tsk)
2936 struct sock *sk = &tsk->sk;
2937 struct net *net = sock_net(sk);
2938 struct tipc_net *tn = net_generic(net, tipc_net_id);
2939 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2940 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2942 while (remaining--) {
2943 portid++;
2944 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2945 portid = TIPC_MIN_PORT;
2946 tsk->portid = portid;
2947 sock_hold(&tsk->sk);
2948 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2949 tsk_rht_params))
2950 return 0;
2951 sock_put(&tsk->sk);
2954 return -1;
2957 static void tipc_sk_remove(struct tipc_sock *tsk)
2959 struct sock *sk = &tsk->sk;
2960 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2962 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2963 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2964 __sock_put(sk);
2968 static const struct rhashtable_params tsk_rht_params = {
2969 .nelem_hint = 192,
2970 .head_offset = offsetof(struct tipc_sock, node),
2971 .key_offset = offsetof(struct tipc_sock, portid),
2972 .key_len = sizeof(u32), /* portid */
2973 .max_size = 1048576,
2974 .min_size = 256,
2975 .automatic_shrinking = true,
2978 int tipc_sk_rht_init(struct net *net)
2980 struct tipc_net *tn = net_generic(net, tipc_net_id);
2982 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2985 void tipc_sk_rht_destroy(struct net *net)
2987 struct tipc_net *tn = net_generic(net, tipc_net_id);
2989 /* Wait for socket readers to complete */
2990 synchronize_net();
2992 rhashtable_destroy(&tn->sk_rht);
2995 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2997 struct net *net = sock_net(&tsk->sk);
2998 struct tipc_group *grp = tsk->group;
2999 struct tipc_msg *hdr = &tsk->phdr;
3000 struct tipc_name_seq seq;
3001 int rc;
3003 if (mreq->type < TIPC_RESERVED_TYPES)
3004 return -EACCES;
3005 if (mreq->scope > TIPC_NODE_SCOPE)
3006 return -EINVAL;
3007 if (grp)
3008 return -EACCES;
3009 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3010 if (!grp)
3011 return -ENOMEM;
3012 tsk->group = grp;
3013 msg_set_lookup_scope(hdr, mreq->scope);
3014 msg_set_nametype(hdr, mreq->type);
3015 msg_set_dest_droppable(hdr, true);
3016 seq.type = mreq->type;
3017 seq.lower = mreq->instance;
3018 seq.upper = seq.lower;
3019 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3020 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3021 if (rc) {
3022 tipc_group_delete(net, grp);
3023 tsk->group = NULL;
3024 return rc;
3026 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3027 tsk->mc_method.rcast = true;
3028 tsk->mc_method.mandatory = true;
3029 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3030 return rc;
3033 static int tipc_sk_leave(struct tipc_sock *tsk)
3035 struct net *net = sock_net(&tsk->sk);
3036 struct tipc_group *grp = tsk->group;
3037 struct tipc_name_seq seq;
3038 int scope;
3040 if (!grp)
3041 return -EINVAL;
3042 tipc_group_self(grp, &seq, &scope);
3043 tipc_group_delete(net, grp);
3044 tsk->group = NULL;
3045 tipc_sk_withdraw(tsk, scope, &seq);
3046 return 0;
3050 * tipc_setsockopt - set socket option
3051 * @sock: socket structure
3052 * @lvl: option level
3053 * @opt: option identifier
3054 * @ov: pointer to new option value
3055 * @ol: length of option value
3057 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3058 * (to ease compatibility).
3060 * Returns 0 on success, errno otherwise
3062 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3063 char __user *ov, unsigned int ol)
3065 struct sock *sk = sock->sk;
3066 struct tipc_sock *tsk = tipc_sk(sk);
3067 struct tipc_group_req mreq;
3068 u32 value = 0;
3069 int res = 0;
3071 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3072 return 0;
3073 if (lvl != SOL_TIPC)
3074 return -ENOPROTOOPT;
3076 switch (opt) {
3077 case TIPC_IMPORTANCE:
3078 case TIPC_SRC_DROPPABLE:
3079 case TIPC_DEST_DROPPABLE:
3080 case TIPC_CONN_TIMEOUT:
3081 case TIPC_NODELAY:
3082 if (ol < sizeof(value))
3083 return -EINVAL;
3084 if (get_user(value, (u32 __user *)ov))
3085 return -EFAULT;
3086 break;
3087 case TIPC_GROUP_JOIN:
3088 if (ol < sizeof(mreq))
3089 return -EINVAL;
3090 if (copy_from_user(&mreq, ov, sizeof(mreq)))
3091 return -EFAULT;
3092 break;
3093 default:
3094 if (ov || ol)
3095 return -EINVAL;
3098 lock_sock(sk);
3100 switch (opt) {
3101 case TIPC_IMPORTANCE:
3102 res = tsk_set_importance(tsk, value);
3103 break;
3104 case TIPC_SRC_DROPPABLE:
3105 if (sock->type != SOCK_STREAM)
3106 tsk_set_unreliable(tsk, value);
3107 else
3108 res = -ENOPROTOOPT;
3109 break;
3110 case TIPC_DEST_DROPPABLE:
3111 tsk_set_unreturnable(tsk, value);
3112 break;
3113 case TIPC_CONN_TIMEOUT:
3114 tipc_sk(sk)->conn_timeout = value;
3115 break;
3116 case TIPC_MCAST_BROADCAST:
3117 tsk->mc_method.rcast = false;
3118 tsk->mc_method.mandatory = true;
3119 break;
3120 case TIPC_MCAST_REPLICAST:
3121 tsk->mc_method.rcast = true;
3122 tsk->mc_method.mandatory = true;
3123 break;
3124 case TIPC_GROUP_JOIN:
3125 res = tipc_sk_join(tsk, &mreq);
3126 break;
3127 case TIPC_GROUP_LEAVE:
3128 res = tipc_sk_leave(tsk);
3129 break;
3130 case TIPC_NODELAY:
3131 tsk->nodelay = !!value;
3132 tsk_set_nagle(tsk);
3133 break;
3134 default:
3135 res = -EINVAL;
3138 release_sock(sk);
3140 return res;
3144 * tipc_getsockopt - get socket option
3145 * @sock: socket structure
3146 * @lvl: option level
3147 * @opt: option identifier
3148 * @ov: receptacle for option value
3149 * @ol: receptacle for length of option value
3151 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3152 * (to ease compatibility).
3154 * Returns 0 on success, errno otherwise
3156 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3157 char __user *ov, int __user *ol)
3159 struct sock *sk = sock->sk;
3160 struct tipc_sock *tsk = tipc_sk(sk);
3161 struct tipc_name_seq seq;
3162 int len, scope;
3163 u32 value;
3164 int res;
3166 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3167 return put_user(0, ol);
3168 if (lvl != SOL_TIPC)
3169 return -ENOPROTOOPT;
3170 res = get_user(len, ol);
3171 if (res)
3172 return res;
3174 lock_sock(sk);
3176 switch (opt) {
3177 case TIPC_IMPORTANCE:
3178 value = tsk_importance(tsk);
3179 break;
3180 case TIPC_SRC_DROPPABLE:
3181 value = tsk_unreliable(tsk);
3182 break;
3183 case TIPC_DEST_DROPPABLE:
3184 value = tsk_unreturnable(tsk);
3185 break;
3186 case TIPC_CONN_TIMEOUT:
3187 value = tsk->conn_timeout;
3188 /* no need to set "res", since already 0 at this point */
3189 break;
3190 case TIPC_NODE_RECVQ_DEPTH:
3191 value = 0; /* was tipc_queue_size, now obsolete */
3192 break;
3193 case TIPC_SOCK_RECVQ_DEPTH:
3194 value = skb_queue_len(&sk->sk_receive_queue);
3195 break;
3196 case TIPC_SOCK_RECVQ_USED:
3197 value = sk_rmem_alloc_get(sk);
3198 break;
3199 case TIPC_GROUP_JOIN:
3200 seq.type = 0;
3201 if (tsk->group)
3202 tipc_group_self(tsk->group, &seq, &scope);
3203 value = seq.type;
3204 break;
3205 default:
3206 res = -EINVAL;
3209 release_sock(sk);
3211 if (res)
3212 return res; /* "get" failed */
3214 if (len < sizeof(value))
3215 return -EINVAL;
3217 if (copy_to_user(ov, &value, sizeof(value)))
3218 return -EFAULT;
3220 return put_user(sizeof(value), ol);
3223 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3225 struct net *net = sock_net(sock->sk);
3226 struct tipc_sioc_nodeid_req nr = {0};
3227 struct tipc_sioc_ln_req lnr;
3228 void __user *argp = (void __user *)arg;
3230 switch (cmd) {
3231 case SIOCGETLINKNAME:
3232 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3233 return -EFAULT;
3234 if (!tipc_node_get_linkname(net,
3235 lnr.bearer_id & 0xffff, lnr.peer,
3236 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3237 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3238 return -EFAULT;
3239 return 0;
3241 return -EADDRNOTAVAIL;
3242 case SIOCGETNODEID:
3243 if (copy_from_user(&nr, argp, sizeof(nr)))
3244 return -EFAULT;
3245 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3246 return -EADDRNOTAVAIL;
3247 if (copy_to_user(argp, &nr, sizeof(nr)))
3248 return -EFAULT;
3249 return 0;
3250 default:
3251 return -ENOIOCTLCMD;
3255 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3257 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3258 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3259 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3261 tsk1->peer.family = AF_TIPC;
3262 tsk1->peer.addrtype = TIPC_ADDR_ID;
3263 tsk1->peer.scope = TIPC_NODE_SCOPE;
3264 tsk1->peer.addr.id.ref = tsk2->portid;
3265 tsk1->peer.addr.id.node = onode;
3266 tsk2->peer.family = AF_TIPC;
3267 tsk2->peer.addrtype = TIPC_ADDR_ID;
3268 tsk2->peer.scope = TIPC_NODE_SCOPE;
3269 tsk2->peer.addr.id.ref = tsk1->portid;
3270 tsk2->peer.addr.id.node = onode;
3272 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3273 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3274 return 0;
3277 /* Protocol switches for the various types of TIPC sockets */
3279 static const struct proto_ops msg_ops = {
3280 .owner = THIS_MODULE,
3281 .family = AF_TIPC,
3282 .release = tipc_release,
3283 .bind = tipc_bind,
3284 .connect = tipc_connect,
3285 .socketpair = tipc_socketpair,
3286 .accept = sock_no_accept,
3287 .getname = tipc_getname,
3288 .poll = tipc_poll,
3289 .ioctl = tipc_ioctl,
3290 .listen = sock_no_listen,
3291 .shutdown = tipc_shutdown,
3292 .setsockopt = tipc_setsockopt,
3293 .getsockopt = tipc_getsockopt,
3294 .sendmsg = tipc_sendmsg,
3295 .recvmsg = tipc_recvmsg,
3296 .mmap = sock_no_mmap,
3297 .sendpage = sock_no_sendpage
3300 static const struct proto_ops packet_ops = {
3301 .owner = THIS_MODULE,
3302 .family = AF_TIPC,
3303 .release = tipc_release,
3304 .bind = tipc_bind,
3305 .connect = tipc_connect,
3306 .socketpair = tipc_socketpair,
3307 .accept = tipc_accept,
3308 .getname = tipc_getname,
3309 .poll = tipc_poll,
3310 .ioctl = tipc_ioctl,
3311 .listen = tipc_listen,
3312 .shutdown = tipc_shutdown,
3313 .setsockopt = tipc_setsockopt,
3314 .getsockopt = tipc_getsockopt,
3315 .sendmsg = tipc_send_packet,
3316 .recvmsg = tipc_recvmsg,
3317 .mmap = sock_no_mmap,
3318 .sendpage = sock_no_sendpage
3321 static const struct proto_ops stream_ops = {
3322 .owner = THIS_MODULE,
3323 .family = AF_TIPC,
3324 .release = tipc_release,
3325 .bind = tipc_bind,
3326 .connect = tipc_connect,
3327 .socketpair = tipc_socketpair,
3328 .accept = tipc_accept,
3329 .getname = tipc_getname,
3330 .poll = tipc_poll,
3331 .ioctl = tipc_ioctl,
3332 .listen = tipc_listen,
3333 .shutdown = tipc_shutdown,
3334 .setsockopt = tipc_setsockopt,
3335 .getsockopt = tipc_getsockopt,
3336 .sendmsg = tipc_sendstream,
3337 .recvmsg = tipc_recvstream,
3338 .mmap = sock_no_mmap,
3339 .sendpage = sock_no_sendpage
3342 static const struct net_proto_family tipc_family_ops = {
3343 .owner = THIS_MODULE,
3344 .family = AF_TIPC,
3345 .create = tipc_sk_create
3348 static struct proto tipc_proto = {
3349 .name = "TIPC",
3350 .owner = THIS_MODULE,
3351 .obj_size = sizeof(struct tipc_sock),
3352 .sysctl_rmem = sysctl_tipc_rmem
3356 * tipc_socket_init - initialize TIPC socket interface
3358 * Returns 0 on success, errno otherwise
3360 int tipc_socket_init(void)
3362 int res;
3364 res = proto_register(&tipc_proto, 1);
3365 if (res) {
3366 pr_err("Failed to register TIPC protocol type\n");
3367 goto out;
3370 res = sock_register(&tipc_family_ops);
3371 if (res) {
3372 pr_err("Failed to register TIPC socket type\n");
3373 proto_unregister(&tipc_proto);
3374 goto out;
3376 out:
3377 return res;
3381 * tipc_socket_stop - stop TIPC socket interface
3383 void tipc_socket_stop(void)
3385 sock_unregister(tipc_family_ops.family);
3386 proto_unregister(&tipc_proto);
3389 /* Caller should hold socket lock for the passed tipc socket. */
3390 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3392 u32 peer_node;
3393 u32 peer_port;
3394 struct nlattr *nest;
3396 peer_node = tsk_peer_node(tsk);
3397 peer_port = tsk_peer_port(tsk);
3399 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3400 if (!nest)
3401 return -EMSGSIZE;
3403 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3404 goto msg_full;
3405 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3406 goto msg_full;
3408 if (tsk->conn_type != 0) {
3409 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3410 goto msg_full;
3411 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3412 goto msg_full;
3413 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3414 goto msg_full;
3416 nla_nest_end(skb, nest);
3418 return 0;
3420 msg_full:
3421 nla_nest_cancel(skb, nest);
3423 return -EMSGSIZE;
3426 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3427 *tsk)
3429 struct net *net = sock_net(skb->sk);
3430 struct sock *sk = &tsk->sk;
3432 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3433 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3434 return -EMSGSIZE;
3436 if (tipc_sk_connected(sk)) {
3437 if (__tipc_nl_add_sk_con(skb, tsk))
3438 return -EMSGSIZE;
3439 } else if (!list_empty(&tsk->publications)) {
3440 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3441 return -EMSGSIZE;
3443 return 0;
3446 /* Caller should hold socket lock for the passed tipc socket. */
3447 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3448 struct tipc_sock *tsk)
3450 struct nlattr *attrs;
3451 void *hdr;
3453 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3454 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3455 if (!hdr)
3456 goto msg_cancel;
3458 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3459 if (!attrs)
3460 goto genlmsg_cancel;
3462 if (__tipc_nl_add_sk_info(skb, tsk))
3463 goto attr_msg_cancel;
3465 nla_nest_end(skb, attrs);
3466 genlmsg_end(skb, hdr);
3468 return 0;
3470 attr_msg_cancel:
3471 nla_nest_cancel(skb, attrs);
3472 genlmsg_cancel:
3473 genlmsg_cancel(skb, hdr);
3474 msg_cancel:
3475 return -EMSGSIZE;
3478 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3479 int (*skb_handler)(struct sk_buff *skb,
3480 struct netlink_callback *cb,
3481 struct tipc_sock *tsk))
3483 struct rhashtable_iter *iter = (void *)cb->args[4];
3484 struct tipc_sock *tsk;
3485 int err;
3487 rhashtable_walk_start(iter);
3488 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3489 if (IS_ERR(tsk)) {
3490 err = PTR_ERR(tsk);
3491 if (err == -EAGAIN) {
3492 err = 0;
3493 continue;
3495 break;
3498 sock_hold(&tsk->sk);
3499 rhashtable_walk_stop(iter);
3500 lock_sock(&tsk->sk);
3501 err = skb_handler(skb, cb, tsk);
3502 if (err) {
3503 release_sock(&tsk->sk);
3504 sock_put(&tsk->sk);
3505 goto out;
3507 release_sock(&tsk->sk);
3508 rhashtable_walk_start(iter);
3509 sock_put(&tsk->sk);
3511 rhashtable_walk_stop(iter);
3512 out:
3513 return skb->len;
3515 EXPORT_SYMBOL(tipc_nl_sk_walk);
3517 int tipc_dump_start(struct netlink_callback *cb)
3519 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3521 EXPORT_SYMBOL(tipc_dump_start);
3523 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3525 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3526 struct rhashtable_iter *iter = (void *)cb->args[4];
3527 struct tipc_net *tn = tipc_net(net);
3529 if (!iter) {
3530 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3531 if (!iter)
3532 return -ENOMEM;
3534 cb->args[4] = (long)iter;
3537 rhashtable_walk_enter(&tn->sk_rht, iter);
3538 return 0;
3541 int tipc_dump_done(struct netlink_callback *cb)
3543 struct rhashtable_iter *hti = (void *)cb->args[4];
3545 rhashtable_walk_exit(hti);
3546 kfree(hti);
3547 return 0;
3549 EXPORT_SYMBOL(tipc_dump_done);
3551 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3552 struct tipc_sock *tsk, u32 sk_filter_state,
3553 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3555 struct sock *sk = &tsk->sk;
3556 struct nlattr *attrs;
3557 struct nlattr *stat;
3559 /*filter response w.r.t sk_state*/
3560 if (!(sk_filter_state & (1 << sk->sk_state)))
3561 return 0;
3563 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3564 if (!attrs)
3565 goto msg_cancel;
3567 if (__tipc_nl_add_sk_info(skb, tsk))
3568 goto attr_msg_cancel;
3570 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3571 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3572 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3573 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3574 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3575 sock_i_uid(sk))) ||
3576 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3577 tipc_diag_gen_cookie(sk),
3578 TIPC_NLA_SOCK_PAD))
3579 goto attr_msg_cancel;
3581 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3582 if (!stat)
3583 goto attr_msg_cancel;
3585 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3586 skb_queue_len(&sk->sk_receive_queue)) ||
3587 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3588 skb_queue_len(&sk->sk_write_queue)) ||
3589 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3590 atomic_read(&sk->sk_drops)))
3591 goto stat_msg_cancel;
3593 if (tsk->cong_link_cnt &&
3594 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3595 goto stat_msg_cancel;
3597 if (tsk_conn_cong(tsk) &&
3598 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3599 goto stat_msg_cancel;
3601 nla_nest_end(skb, stat);
3603 if (tsk->group)
3604 if (tipc_group_fill_sock_diag(tsk->group, skb))
3605 goto stat_msg_cancel;
3607 nla_nest_end(skb, attrs);
3609 return 0;
3611 stat_msg_cancel:
3612 nla_nest_cancel(skb, stat);
3613 attr_msg_cancel:
3614 nla_nest_cancel(skb, attrs);
3615 msg_cancel:
3616 return -EMSGSIZE;
3618 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3620 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3622 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3625 /* Caller should hold socket lock for the passed tipc socket. */
3626 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3627 struct netlink_callback *cb,
3628 struct publication *publ)
3630 void *hdr;
3631 struct nlattr *attrs;
3633 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3634 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3635 if (!hdr)
3636 goto msg_cancel;
3638 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3639 if (!attrs)
3640 goto genlmsg_cancel;
3642 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3643 goto attr_msg_cancel;
3644 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3645 goto attr_msg_cancel;
3646 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3647 goto attr_msg_cancel;
3648 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3649 goto attr_msg_cancel;
3651 nla_nest_end(skb, attrs);
3652 genlmsg_end(skb, hdr);
3654 return 0;
3656 attr_msg_cancel:
3657 nla_nest_cancel(skb, attrs);
3658 genlmsg_cancel:
3659 genlmsg_cancel(skb, hdr);
3660 msg_cancel:
3661 return -EMSGSIZE;
3664 /* Caller should hold socket lock for the passed tipc socket. */
3665 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3666 struct netlink_callback *cb,
3667 struct tipc_sock *tsk, u32 *last_publ)
3669 int err;
3670 struct publication *p;
3672 if (*last_publ) {
3673 list_for_each_entry(p, &tsk->publications, binding_sock) {
3674 if (p->key == *last_publ)
3675 break;
3677 if (p->key != *last_publ) {
3678 /* We never set seq or call nl_dump_check_consistent()
3679 * this means that setting prev_seq here will cause the
3680 * consistence check to fail in the netlink callback
3681 * handler. Resulting in the last NLMSG_DONE message
3682 * having the NLM_F_DUMP_INTR flag set.
3684 cb->prev_seq = 1;
3685 *last_publ = 0;
3686 return -EPIPE;
3688 } else {
3689 p = list_first_entry(&tsk->publications, struct publication,
3690 binding_sock);
3693 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3694 err = __tipc_nl_add_sk_publ(skb, cb, p);
3695 if (err) {
3696 *last_publ = p->key;
3697 return err;
3700 *last_publ = 0;
3702 return 0;
3705 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3707 int err;
3708 u32 tsk_portid = cb->args[0];
3709 u32 last_publ = cb->args[1];
3710 u32 done = cb->args[2];
3711 struct net *net = sock_net(skb->sk);
3712 struct tipc_sock *tsk;
3714 if (!tsk_portid) {
3715 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3716 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3718 if (!attrs[TIPC_NLA_SOCK])
3719 return -EINVAL;
3721 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3722 attrs[TIPC_NLA_SOCK],
3723 tipc_nl_sock_policy, NULL);
3724 if (err)
3725 return err;
3727 if (!sock[TIPC_NLA_SOCK_REF])
3728 return -EINVAL;
3730 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3733 if (done)
3734 return 0;
3736 tsk = tipc_sk_lookup(net, tsk_portid);
3737 if (!tsk)
3738 return -EINVAL;
3740 lock_sock(&tsk->sk);
3741 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3742 if (!err)
3743 done = 1;
3744 release_sock(&tsk->sk);
3745 sock_put(&tsk->sk);
3747 cb->args[0] = tsk_portid;
3748 cb->args[1] = last_publ;
3749 cb->args[2] = done;
3751 return skb->len;
3755 * tipc_sk_filtering - check if a socket should be traced
3756 * @sk: the socket to be examined
3757 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3758 * (portid, sock type, name type, name lower, name upper)
3760 * Returns true if the socket meets the socket tuple data
3761 * (value 0 = 'any') or when there is no tuple set (all = 0),
3762 * otherwise false
3764 bool tipc_sk_filtering(struct sock *sk)
3766 struct tipc_sock *tsk;
3767 struct publication *p;
3768 u32 _port, _sktype, _type, _lower, _upper;
3769 u32 type = 0, lower = 0, upper = 0;
3771 if (!sk)
3772 return true;
3774 tsk = tipc_sk(sk);
3776 _port = sysctl_tipc_sk_filter[0];
3777 _sktype = sysctl_tipc_sk_filter[1];
3778 _type = sysctl_tipc_sk_filter[2];
3779 _lower = sysctl_tipc_sk_filter[3];
3780 _upper = sysctl_tipc_sk_filter[4];
3782 if (!_port && !_sktype && !_type && !_lower && !_upper)
3783 return true;
3785 if (_port)
3786 return (_port == tsk->portid);
3788 if (_sktype && _sktype != sk->sk_type)
3789 return false;
3791 if (tsk->published) {
3792 p = list_first_entry_or_null(&tsk->publications,
3793 struct publication, binding_sock);
3794 if (p) {
3795 type = p->type;
3796 lower = p->lower;
3797 upper = p->upper;
3801 if (!tipc_sk_type_connectionless(sk)) {
3802 type = tsk->conn_type;
3803 lower = tsk->conn_instance;
3804 upper = tsk->conn_instance;
3807 if ((_type && _type != type) || (_lower && _lower != lower) ||
3808 (_upper && _upper != upper))
3809 return false;
3811 return true;
3814 u32 tipc_sock_get_portid(struct sock *sk)
3816 return (sk) ? (tipc_sk(sk))->portid : 0;
3820 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3821 * both the rcv and backlog queues are considered
3822 * @sk: tipc sk to be checked
3823 * @skb: tipc msg to be checked
3825 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3828 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3830 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3831 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3832 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3834 return (qsize > lim * 90 / 100);
3838 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3839 * only the rcv queue is considered
3840 * @sk: tipc sk to be checked
3841 * @skb: tipc msg to be checked
3843 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3846 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3848 unsigned int lim = rcvbuf_limit(sk, skb);
3849 unsigned int qsize = sk_rmem_alloc_get(sk);
3851 return (qsize > lim * 90 / 100);
3855 * tipc_sk_dump - dump TIPC socket
3856 * @sk: tipc sk to be dumped
3857 * @dqueues: bitmask to decide if any socket queue to be dumped?
3858 * - TIPC_DUMP_NONE: don't dump socket queues
3859 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3860 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3861 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3862 * - TIPC_DUMP_ALL: dump all the socket queues above
3863 * @buf: returned buffer of dump data in format
3865 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3867 int i = 0;
3868 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3869 struct tipc_sock *tsk;
3870 struct publication *p;
3871 bool tsk_connected;
3873 if (!sk) {
3874 i += scnprintf(buf, sz, "sk data: (null)\n");
3875 return i;
3878 tsk = tipc_sk(sk);
3879 tsk_connected = !tipc_sk_type_connectionless(sk);
3881 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3882 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3883 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3884 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3885 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3886 if (tsk_connected) {
3887 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3888 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3889 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3890 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3892 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3893 if (tsk->published) {
3894 p = list_first_entry_or_null(&tsk->publications,
3895 struct publication, binding_sock);
3896 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3897 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3898 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3900 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3901 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3902 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3903 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3904 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3905 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3906 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3907 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3908 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3909 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3910 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3911 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3912 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3913 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3915 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3916 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3917 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3920 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3921 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3922 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3925 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3926 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3927 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3928 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3929 i += scnprintf(buf + i, sz - i, " tail ");
3930 i += tipc_skb_dump(sk->sk_backlog.tail, false,
3931 buf + i);
3935 return i;