bus: mhi: core: Fix some error return code
[linux/fpc-iii.git] / net / tipc / socket.c
blob87466607097f18f5d6ea658e7123cf7ae949d775
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 void tipc_sk_send_ack(struct tipc_sock *tsk)
1744 struct sock *sk = &tsk->sk;
1745 struct net *net = sock_net(sk);
1746 struct sk_buff *skb = NULL;
1747 struct tipc_msg *msg;
1748 u32 peer_port = tsk_peer_port(tsk);
1749 u32 dnode = tsk_peer_node(tsk);
1751 if (!tipc_sk_connected(sk))
1752 return;
1753 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1754 dnode, tsk_own_node(tsk), peer_port,
1755 tsk->portid, TIPC_OK);
1756 if (!skb)
1757 return;
1758 msg = buf_msg(skb);
1759 msg_set_conn_ack(msg, tsk->rcv_unacked);
1760 tsk->rcv_unacked = 0;
1762 /* Adjust to and advertize the correct window limit */
1763 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1764 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1765 msg_set_adv_win(msg, tsk->rcv_win);
1767 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1770 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1772 struct sock *sk = sock->sk;
1773 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1774 long timeo = *timeop;
1775 int err = sock_error(sk);
1777 if (err)
1778 return err;
1780 for (;;) {
1781 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1782 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1783 err = -ENOTCONN;
1784 break;
1786 add_wait_queue(sk_sleep(sk), &wait);
1787 release_sock(sk);
1788 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1789 sched_annotate_sleep();
1790 lock_sock(sk);
1791 remove_wait_queue(sk_sleep(sk), &wait);
1793 err = 0;
1794 if (!skb_queue_empty(&sk->sk_receive_queue))
1795 break;
1796 err = -EAGAIN;
1797 if (!timeo)
1798 break;
1799 err = sock_intr_errno(timeo);
1800 if (signal_pending(current))
1801 break;
1803 err = sock_error(sk);
1804 if (err)
1805 break;
1807 *timeop = timeo;
1808 return err;
1812 * tipc_recvmsg - receive packet-oriented message
1813 * @m: descriptor for message info
1814 * @buflen: length of user buffer area
1815 * @flags: receive flags
1817 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1818 * If the complete message doesn't fit in user area, truncate it.
1820 * Returns size of returned message data, errno otherwise
1822 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1823 size_t buflen, int flags)
1825 struct sock *sk = sock->sk;
1826 bool connected = !tipc_sk_type_connectionless(sk);
1827 struct tipc_sock *tsk = tipc_sk(sk);
1828 int rc, err, hlen, dlen, copy;
1829 struct sk_buff_head xmitq;
1830 struct tipc_msg *hdr;
1831 struct sk_buff *skb;
1832 bool grp_evt;
1833 long timeout;
1835 /* Catch invalid receive requests */
1836 if (unlikely(!buflen))
1837 return -EINVAL;
1839 lock_sock(sk);
1840 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1841 rc = -ENOTCONN;
1842 goto exit;
1844 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1846 /* Step rcv queue to first msg with data or error; wait if necessary */
1847 do {
1848 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1849 if (unlikely(rc))
1850 goto exit;
1851 skb = skb_peek(&sk->sk_receive_queue);
1852 hdr = buf_msg(skb);
1853 dlen = msg_data_sz(hdr);
1854 hlen = msg_hdr_sz(hdr);
1855 err = msg_errcode(hdr);
1856 grp_evt = msg_is_grp_evt(hdr);
1857 if (likely(dlen || err))
1858 break;
1859 tsk_advance_rx_queue(sk);
1860 } while (1);
1862 /* Collect msg meta data, including error code and rejected data */
1863 tipc_sk_set_orig_addr(m, skb);
1864 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1865 if (unlikely(rc))
1866 goto exit;
1867 hdr = buf_msg(skb);
1869 /* Capture data if non-error msg, otherwise just set return value */
1870 if (likely(!err)) {
1871 copy = min_t(int, dlen, buflen);
1872 if (unlikely(copy != dlen))
1873 m->msg_flags |= MSG_TRUNC;
1874 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1875 } else {
1876 copy = 0;
1877 rc = 0;
1878 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1879 rc = -ECONNRESET;
1881 if (unlikely(rc))
1882 goto exit;
1884 /* Mark message as group event if applicable */
1885 if (unlikely(grp_evt)) {
1886 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1887 m->msg_flags |= MSG_EOR;
1888 m->msg_flags |= MSG_OOB;
1889 copy = 0;
1892 /* Caption of data or error code/rejected data was successful */
1893 if (unlikely(flags & MSG_PEEK))
1894 goto exit;
1896 /* Send group flow control advertisement when applicable */
1897 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1898 __skb_queue_head_init(&xmitq);
1899 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1900 msg_orignode(hdr), msg_origport(hdr),
1901 &xmitq);
1902 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1905 tsk_advance_rx_queue(sk);
1907 if (likely(!connected))
1908 goto exit;
1910 /* Send connection flow control advertisement when applicable */
1911 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1912 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1913 tipc_sk_send_ack(tsk);
1914 exit:
1915 release_sock(sk);
1916 return rc ? rc : copy;
1920 * tipc_recvstream - receive stream-oriented data
1921 * @m: descriptor for message info
1922 * @buflen: total size of user buffer area
1923 * @flags: receive flags
1925 * Used for SOCK_STREAM messages only. If not enough data is available
1926 * will optionally wait for more; never truncates data.
1928 * Returns size of returned message data, errno otherwise
1930 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1931 size_t buflen, int flags)
1933 struct sock *sk = sock->sk;
1934 struct tipc_sock *tsk = tipc_sk(sk);
1935 struct sk_buff *skb;
1936 struct tipc_msg *hdr;
1937 struct tipc_skb_cb *skb_cb;
1938 bool peek = flags & MSG_PEEK;
1939 int offset, required, copy, copied = 0;
1940 int hlen, dlen, err, rc;
1941 bool ack = false;
1942 long timeout;
1944 /* Catch invalid receive attempts */
1945 if (unlikely(!buflen))
1946 return -EINVAL;
1948 lock_sock(sk);
1950 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1951 rc = -ENOTCONN;
1952 goto exit;
1954 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1955 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1957 do {
1958 /* Look at first msg in receive queue; wait if necessary */
1959 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1960 if (unlikely(rc))
1961 break;
1962 skb = skb_peek(&sk->sk_receive_queue);
1963 skb_cb = TIPC_SKB_CB(skb);
1964 hdr = buf_msg(skb);
1965 dlen = msg_data_sz(hdr);
1966 hlen = msg_hdr_sz(hdr);
1967 err = msg_errcode(hdr);
1969 /* Discard any empty non-errored (SYN-) message */
1970 if (unlikely(!dlen && !err)) {
1971 tsk_advance_rx_queue(sk);
1972 continue;
1975 /* Collect msg meta data, incl. error code and rejected data */
1976 if (!copied) {
1977 tipc_sk_set_orig_addr(m, skb);
1978 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1979 if (rc)
1980 break;
1981 hdr = buf_msg(skb);
1984 /* Copy data if msg ok, otherwise return error/partial data */
1985 if (likely(!err)) {
1986 ack = msg_ack_required(hdr);
1987 offset = skb_cb->bytes_read;
1988 copy = min_t(int, dlen - offset, buflen - copied);
1989 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1990 if (unlikely(rc))
1991 break;
1992 copied += copy;
1993 offset += copy;
1994 if (unlikely(offset < dlen)) {
1995 if (!peek)
1996 skb_cb->bytes_read = offset;
1997 break;
1999 } else {
2000 rc = 0;
2001 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2002 rc = -ECONNRESET;
2003 if (copied || rc)
2004 break;
2007 if (unlikely(peek))
2008 break;
2010 tsk_advance_rx_queue(sk);
2012 /* Send connection flow control advertisement when applicable */
2013 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2014 if (ack || tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2015 tipc_sk_send_ack(tsk);
2017 /* Exit if all requested data or FIN/error received */
2018 if (copied == buflen || err)
2019 break;
2021 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2022 exit:
2023 release_sock(sk);
2024 return copied ? copied : rc;
2028 * tipc_write_space - wake up thread if port congestion is released
2029 * @sk: socket
2031 static void tipc_write_space(struct sock *sk)
2033 struct socket_wq *wq;
2035 rcu_read_lock();
2036 wq = rcu_dereference(sk->sk_wq);
2037 if (skwq_has_sleeper(wq))
2038 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2039 EPOLLWRNORM | EPOLLWRBAND);
2040 rcu_read_unlock();
2044 * tipc_data_ready - wake up threads to indicate messages have been received
2045 * @sk: socket
2046 * @len: the length of messages
2048 static void tipc_data_ready(struct sock *sk)
2050 struct socket_wq *wq;
2052 rcu_read_lock();
2053 wq = rcu_dereference(sk->sk_wq);
2054 if (skwq_has_sleeper(wq))
2055 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2056 EPOLLRDNORM | EPOLLRDBAND);
2057 rcu_read_unlock();
2060 static void tipc_sock_destruct(struct sock *sk)
2062 __skb_queue_purge(&sk->sk_receive_queue);
2065 static void tipc_sk_proto_rcv(struct sock *sk,
2066 struct sk_buff_head *inputq,
2067 struct sk_buff_head *xmitq)
2069 struct sk_buff *skb = __skb_dequeue(inputq);
2070 struct tipc_sock *tsk = tipc_sk(sk);
2071 struct tipc_msg *hdr = buf_msg(skb);
2072 struct tipc_group *grp = tsk->group;
2073 bool wakeup = false;
2075 switch (msg_user(hdr)) {
2076 case CONN_MANAGER:
2077 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2078 return;
2079 case SOCK_WAKEUP:
2080 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2081 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2082 smp_wmb();
2083 tsk->cong_link_cnt--;
2084 wakeup = true;
2085 tipc_sk_push_backlog(tsk);
2086 break;
2087 case GROUP_PROTOCOL:
2088 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2089 break;
2090 case TOP_SRV:
2091 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2092 hdr, inputq, xmitq);
2093 break;
2094 default:
2095 break;
2098 if (wakeup)
2099 sk->sk_write_space(sk);
2101 kfree_skb(skb);
2105 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2106 * @tsk: TIPC socket
2107 * @skb: pointer to message buffer.
2108 * Returns true if message should be added to receive queue, false otherwise
2110 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2112 struct sock *sk = &tsk->sk;
2113 struct net *net = sock_net(sk);
2114 struct tipc_msg *hdr = buf_msg(skb);
2115 bool con_msg = msg_connected(hdr);
2116 u32 pport = tsk_peer_port(tsk);
2117 u32 pnode = tsk_peer_node(tsk);
2118 u32 oport = msg_origport(hdr);
2119 u32 onode = msg_orignode(hdr);
2120 int err = msg_errcode(hdr);
2121 unsigned long delay;
2123 if (unlikely(msg_mcast(hdr)))
2124 return false;
2125 tsk->oneway = 0;
2127 switch (sk->sk_state) {
2128 case TIPC_CONNECTING:
2129 /* Setup ACK */
2130 if (likely(con_msg)) {
2131 if (err)
2132 break;
2133 tipc_sk_finish_conn(tsk, oport, onode);
2134 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2135 /* ACK+ message with data is added to receive queue */
2136 if (msg_data_sz(hdr))
2137 return true;
2138 /* Empty ACK-, - wake up sleeping connect() and drop */
2139 sk->sk_state_change(sk);
2140 msg_set_dest_droppable(hdr, 1);
2141 return false;
2143 /* Ignore connectionless message if not from listening socket */
2144 if (oport != pport || onode != pnode)
2145 return false;
2147 /* Rejected SYN */
2148 if (err != TIPC_ERR_OVERLOAD)
2149 break;
2151 /* Prepare for new setup attempt if we have a SYN clone */
2152 if (skb_queue_empty(&sk->sk_write_queue))
2153 break;
2154 get_random_bytes(&delay, 2);
2155 delay %= (tsk->conn_timeout / 4);
2156 delay = msecs_to_jiffies(delay + 100);
2157 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2158 return false;
2159 case TIPC_OPEN:
2160 case TIPC_DISCONNECTING:
2161 return false;
2162 case TIPC_LISTEN:
2163 /* Accept only SYN message */
2164 if (!msg_is_syn(hdr) &&
2165 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2166 return false;
2167 if (!con_msg && !err)
2168 return true;
2169 return false;
2170 case TIPC_ESTABLISHED:
2171 if (!skb_queue_empty(&sk->sk_write_queue))
2172 tipc_sk_push_backlog(tsk);
2173 /* Accept only connection-based messages sent by peer */
2174 if (likely(con_msg && !err && pport == oport && pnode == onode))
2175 return true;
2176 if (!tsk_peer_msg(tsk, hdr))
2177 return false;
2178 if (!err)
2179 return true;
2180 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2181 tipc_node_remove_conn(net, pnode, tsk->portid);
2182 sk->sk_state_change(sk);
2183 return true;
2184 default:
2185 pr_err("Unknown sk_state %u\n", sk->sk_state);
2187 /* Abort connection setup attempt */
2188 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2189 sk->sk_err = ECONNREFUSED;
2190 sk->sk_state_change(sk);
2191 return true;
2195 * rcvbuf_limit - get proper overload limit of socket receive queue
2196 * @sk: socket
2197 * @skb: message
2199 * For connection oriented messages, irrespective of importance,
2200 * default queue limit is 2 MB.
2202 * For connectionless messages, queue limits are based on message
2203 * importance as follows:
2205 * TIPC_LOW_IMPORTANCE (2 MB)
2206 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2207 * TIPC_HIGH_IMPORTANCE (8 MB)
2208 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2210 * Returns overload limit according to corresponding message importance
2212 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2214 struct tipc_sock *tsk = tipc_sk(sk);
2215 struct tipc_msg *hdr = buf_msg(skb);
2217 if (unlikely(msg_in_group(hdr)))
2218 return READ_ONCE(sk->sk_rcvbuf);
2220 if (unlikely(!msg_connected(hdr)))
2221 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2223 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2224 return READ_ONCE(sk->sk_rcvbuf);
2226 return FLOWCTL_MSG_LIM;
2230 * tipc_sk_filter_rcv - validate incoming message
2231 * @sk: socket
2232 * @skb: pointer to message.
2234 * Enqueues message on receive queue if acceptable; optionally handles
2235 * disconnect indication for a connected socket.
2237 * Called with socket lock already taken
2240 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2241 struct sk_buff_head *xmitq)
2243 bool sk_conn = !tipc_sk_type_connectionless(sk);
2244 struct tipc_sock *tsk = tipc_sk(sk);
2245 struct tipc_group *grp = tsk->group;
2246 struct tipc_msg *hdr = buf_msg(skb);
2247 struct net *net = sock_net(sk);
2248 struct sk_buff_head inputq;
2249 int mtyp = msg_type(hdr);
2250 int limit, err = TIPC_OK;
2252 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2253 TIPC_SKB_CB(skb)->bytes_read = 0;
2254 __skb_queue_head_init(&inputq);
2255 __skb_queue_tail(&inputq, skb);
2257 if (unlikely(!msg_isdata(hdr)))
2258 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2260 if (unlikely(grp))
2261 tipc_group_filter_msg(grp, &inputq, xmitq);
2263 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2264 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2266 /* Validate and add to receive buffer if there is space */
2267 while ((skb = __skb_dequeue(&inputq))) {
2268 hdr = buf_msg(skb);
2269 limit = rcvbuf_limit(sk, skb);
2270 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2271 (!sk_conn && msg_connected(hdr)) ||
2272 (!grp && msg_in_group(hdr)))
2273 err = TIPC_ERR_NO_PORT;
2274 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2275 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2276 "err_overload2!");
2277 atomic_inc(&sk->sk_drops);
2278 err = TIPC_ERR_OVERLOAD;
2281 if (unlikely(err)) {
2282 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2283 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2284 "@filter_rcv!");
2285 __skb_queue_tail(xmitq, skb);
2287 err = TIPC_OK;
2288 continue;
2290 __skb_queue_tail(&sk->sk_receive_queue, skb);
2291 skb_set_owner_r(skb, sk);
2292 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2293 "rcvq >90% allocated!");
2294 sk->sk_data_ready(sk);
2299 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2300 * @sk: socket
2301 * @skb: message
2303 * Caller must hold socket lock
2305 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2307 unsigned int before = sk_rmem_alloc_get(sk);
2308 struct sk_buff_head xmitq;
2309 unsigned int added;
2311 __skb_queue_head_init(&xmitq);
2313 tipc_sk_filter_rcv(sk, skb, &xmitq);
2314 added = sk_rmem_alloc_get(sk) - before;
2315 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2317 /* Send pending response/rejected messages, if any */
2318 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2319 return 0;
2323 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2324 * inputq and try adding them to socket or backlog queue
2325 * @inputq: list of incoming buffers with potentially different destinations
2326 * @sk: socket where the buffers should be enqueued
2327 * @dport: port number for the socket
2329 * Caller must hold socket lock
2331 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2332 u32 dport, struct sk_buff_head *xmitq)
2334 unsigned long time_limit = jiffies + 2;
2335 struct sk_buff *skb;
2336 unsigned int lim;
2337 atomic_t *dcnt;
2338 u32 onode;
2340 while (skb_queue_len(inputq)) {
2341 if (unlikely(time_after_eq(jiffies, time_limit)))
2342 return;
2344 skb = tipc_skb_dequeue(inputq, dport);
2345 if (unlikely(!skb))
2346 return;
2348 /* Add message directly to receive queue if possible */
2349 if (!sock_owned_by_user(sk)) {
2350 tipc_sk_filter_rcv(sk, skb, xmitq);
2351 continue;
2354 /* Try backlog, compensating for double-counted bytes */
2355 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2356 if (!sk->sk_backlog.len)
2357 atomic_set(dcnt, 0);
2358 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2359 if (likely(!sk_add_backlog(sk, skb, lim))) {
2360 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2361 "bklg & rcvq >90% allocated!");
2362 continue;
2365 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2366 /* Overload => reject message back to sender */
2367 onode = tipc_own_addr(sock_net(sk));
2368 atomic_inc(&sk->sk_drops);
2369 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2370 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2371 "@sk_enqueue!");
2372 __skb_queue_tail(xmitq, skb);
2374 break;
2379 * tipc_sk_rcv - handle a chain of incoming buffers
2380 * @inputq: buffer list containing the buffers
2381 * Consumes all buffers in list until inputq is empty
2382 * Note: may be called in multiple threads referring to the same queue
2384 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2386 struct sk_buff_head xmitq;
2387 u32 dnode, dport = 0;
2388 int err;
2389 struct tipc_sock *tsk;
2390 struct sock *sk;
2391 struct sk_buff *skb;
2393 __skb_queue_head_init(&xmitq);
2394 while (skb_queue_len(inputq)) {
2395 dport = tipc_skb_peek_port(inputq, dport);
2396 tsk = tipc_sk_lookup(net, dport);
2398 if (likely(tsk)) {
2399 sk = &tsk->sk;
2400 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2401 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2402 spin_unlock_bh(&sk->sk_lock.slock);
2404 /* Send pending response/rejected messages, if any */
2405 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2406 sock_put(sk);
2407 continue;
2409 /* No destination socket => dequeue skb if still there */
2410 skb = tipc_skb_dequeue(inputq, dport);
2411 if (!skb)
2412 return;
2414 /* Try secondary lookup if unresolved named message */
2415 err = TIPC_ERR_NO_PORT;
2416 if (tipc_msg_lookup_dest(net, skb, &err))
2417 goto xmit;
2419 /* Prepare for message rejection */
2420 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2421 continue;
2423 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2424 xmit:
2425 dnode = msg_destnode(buf_msg(skb));
2426 tipc_node_xmit_skb(net, skb, dnode, dport);
2430 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2432 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2433 struct sock *sk = sock->sk;
2434 int done;
2436 do {
2437 int err = sock_error(sk);
2438 if (err)
2439 return err;
2440 if (!*timeo_p)
2441 return -ETIMEDOUT;
2442 if (signal_pending(current))
2443 return sock_intr_errno(*timeo_p);
2444 if (sk->sk_state == TIPC_DISCONNECTING)
2445 break;
2447 add_wait_queue(sk_sleep(sk), &wait);
2448 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2449 &wait);
2450 remove_wait_queue(sk_sleep(sk), &wait);
2451 } while (!done);
2452 return 0;
2455 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2457 if (addr->family != AF_TIPC)
2458 return false;
2459 if (addr->addrtype == TIPC_SERVICE_RANGE)
2460 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2461 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2462 addr->addrtype == TIPC_SOCKET_ADDR);
2466 * tipc_connect - establish a connection to another TIPC port
2467 * @sock: socket structure
2468 * @dest: socket address for destination port
2469 * @destlen: size of socket address data structure
2470 * @flags: file-related flags associated with socket
2472 * Returns 0 on success, errno otherwise
2474 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2475 int destlen, int flags)
2477 struct sock *sk = sock->sk;
2478 struct tipc_sock *tsk = tipc_sk(sk);
2479 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2480 struct msghdr m = {NULL,};
2481 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2482 int previous;
2483 int res = 0;
2485 if (destlen != sizeof(struct sockaddr_tipc))
2486 return -EINVAL;
2488 lock_sock(sk);
2490 if (tsk->group) {
2491 res = -EINVAL;
2492 goto exit;
2495 if (dst->family == AF_UNSPEC) {
2496 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2497 if (!tipc_sk_type_connectionless(sk))
2498 res = -EINVAL;
2499 goto exit;
2501 if (!tipc_sockaddr_is_sane(dst)) {
2502 res = -EINVAL;
2503 goto exit;
2505 /* DGRAM/RDM connect(), just save the destaddr */
2506 if (tipc_sk_type_connectionless(sk)) {
2507 memcpy(&tsk->peer, dest, destlen);
2508 goto exit;
2509 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2510 res = -EINVAL;
2511 goto exit;
2514 previous = sk->sk_state;
2516 switch (sk->sk_state) {
2517 case TIPC_OPEN:
2518 /* Send a 'SYN-' to destination */
2519 m.msg_name = dest;
2520 m.msg_namelen = destlen;
2522 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2523 * indicate send_msg() is never blocked.
2525 if (!timeout)
2526 m.msg_flags = MSG_DONTWAIT;
2528 res = __tipc_sendmsg(sock, &m, 0);
2529 if ((res < 0) && (res != -EWOULDBLOCK))
2530 goto exit;
2532 /* Just entered TIPC_CONNECTING state; the only
2533 * difference is that return value in non-blocking
2534 * case is EINPROGRESS, rather than EALREADY.
2536 res = -EINPROGRESS;
2537 /* fall through */
2538 case TIPC_CONNECTING:
2539 if (!timeout) {
2540 if (previous == TIPC_CONNECTING)
2541 res = -EALREADY;
2542 goto exit;
2544 timeout = msecs_to_jiffies(timeout);
2545 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2546 res = tipc_wait_for_connect(sock, &timeout);
2547 break;
2548 case TIPC_ESTABLISHED:
2549 res = -EISCONN;
2550 break;
2551 default:
2552 res = -EINVAL;
2555 exit:
2556 release_sock(sk);
2557 return res;
2561 * tipc_listen - allow socket to listen for incoming connections
2562 * @sock: socket structure
2563 * @len: (unused)
2565 * Returns 0 on success, errno otherwise
2567 static int tipc_listen(struct socket *sock, int len)
2569 struct sock *sk = sock->sk;
2570 int res;
2572 lock_sock(sk);
2573 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2574 release_sock(sk);
2576 return res;
2579 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2581 struct sock *sk = sock->sk;
2582 DEFINE_WAIT(wait);
2583 int err;
2585 /* True wake-one mechanism for incoming connections: only
2586 * one process gets woken up, not the 'whole herd'.
2587 * Since we do not 'race & poll' for established sockets
2588 * anymore, the common case will execute the loop only once.
2590 for (;;) {
2591 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2592 TASK_INTERRUPTIBLE);
2593 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2594 release_sock(sk);
2595 timeo = schedule_timeout(timeo);
2596 lock_sock(sk);
2598 err = 0;
2599 if (!skb_queue_empty(&sk->sk_receive_queue))
2600 break;
2601 err = -EAGAIN;
2602 if (!timeo)
2603 break;
2604 err = sock_intr_errno(timeo);
2605 if (signal_pending(current))
2606 break;
2608 finish_wait(sk_sleep(sk), &wait);
2609 return err;
2613 * tipc_accept - wait for connection request
2614 * @sock: listening socket
2615 * @newsock: new socket that is to be connected
2616 * @flags: file-related flags associated with socket
2618 * Returns 0 on success, errno otherwise
2620 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2621 bool kern)
2623 struct sock *new_sk, *sk = sock->sk;
2624 struct sk_buff *buf;
2625 struct tipc_sock *new_tsock;
2626 struct tipc_msg *msg;
2627 long timeo;
2628 int res;
2630 lock_sock(sk);
2632 if (sk->sk_state != TIPC_LISTEN) {
2633 res = -EINVAL;
2634 goto exit;
2636 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2637 res = tipc_wait_for_accept(sock, timeo);
2638 if (res)
2639 goto exit;
2641 buf = skb_peek(&sk->sk_receive_queue);
2643 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2644 if (res)
2645 goto exit;
2646 security_sk_clone(sock->sk, new_sock->sk);
2648 new_sk = new_sock->sk;
2649 new_tsock = tipc_sk(new_sk);
2650 msg = buf_msg(buf);
2652 /* we lock on new_sk; but lockdep sees the lock on sk */
2653 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2656 * Reject any stray messages received by new socket
2657 * before the socket lock was taken (very, very unlikely)
2659 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2661 /* Connect new socket to it's peer */
2662 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2664 tsk_set_importance(new_tsock, msg_importance(msg));
2665 if (msg_named(msg)) {
2666 new_tsock->conn_type = msg_nametype(msg);
2667 new_tsock->conn_instance = msg_nameinst(msg);
2671 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2672 * Respond to 'SYN+' by queuing it on new socket.
2674 if (!msg_data_sz(msg)) {
2675 struct msghdr m = {NULL,};
2677 tsk_advance_rx_queue(sk);
2678 __tipc_sendstream(new_sock, &m, 0);
2679 } else {
2680 __skb_dequeue(&sk->sk_receive_queue);
2681 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2682 skb_set_owner_r(buf, new_sk);
2684 release_sock(new_sk);
2685 exit:
2686 release_sock(sk);
2687 return res;
2691 * tipc_shutdown - shutdown socket connection
2692 * @sock: socket structure
2693 * @how: direction to close (must be SHUT_RDWR)
2695 * Terminates connection (if necessary), then purges socket's receive queue.
2697 * Returns 0 on success, errno otherwise
2699 static int tipc_shutdown(struct socket *sock, int how)
2701 struct sock *sk = sock->sk;
2702 int res;
2704 if (how != SHUT_RDWR)
2705 return -EINVAL;
2707 lock_sock(sk);
2709 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2710 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2711 sk->sk_shutdown = SEND_SHUTDOWN;
2713 if (sk->sk_state == TIPC_DISCONNECTING) {
2714 /* Discard any unreceived messages */
2715 __skb_queue_purge(&sk->sk_receive_queue);
2717 /* Wake up anyone sleeping in poll */
2718 sk->sk_state_change(sk);
2719 res = 0;
2720 } else {
2721 res = -ENOTCONN;
2724 release_sock(sk);
2725 return res;
2728 static void tipc_sk_check_probing_state(struct sock *sk,
2729 struct sk_buff_head *list)
2731 struct tipc_sock *tsk = tipc_sk(sk);
2732 u32 pnode = tsk_peer_node(tsk);
2733 u32 pport = tsk_peer_port(tsk);
2734 u32 self = tsk_own_node(tsk);
2735 u32 oport = tsk->portid;
2736 struct sk_buff *skb;
2738 if (tsk->probe_unacked) {
2739 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2740 sk->sk_err = ECONNABORTED;
2741 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2742 sk->sk_state_change(sk);
2743 return;
2745 /* Prepare new probe */
2746 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2747 pnode, self, pport, oport, TIPC_OK);
2748 if (skb)
2749 __skb_queue_tail(list, skb);
2750 tsk->probe_unacked = true;
2751 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2754 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2756 struct tipc_sock *tsk = tipc_sk(sk);
2758 /* Try again later if dest link is congested */
2759 if (tsk->cong_link_cnt) {
2760 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2761 return;
2763 /* Prepare SYN for retransmit */
2764 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2767 static void tipc_sk_timeout(struct timer_list *t)
2769 struct sock *sk = from_timer(sk, t, sk_timer);
2770 struct tipc_sock *tsk = tipc_sk(sk);
2771 u32 pnode = tsk_peer_node(tsk);
2772 struct sk_buff_head list;
2773 int rc = 0;
2775 __skb_queue_head_init(&list);
2776 bh_lock_sock(sk);
2778 /* Try again later if socket is busy */
2779 if (sock_owned_by_user(sk)) {
2780 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2781 bh_unlock_sock(sk);
2782 sock_put(sk);
2783 return;
2786 if (sk->sk_state == TIPC_ESTABLISHED)
2787 tipc_sk_check_probing_state(sk, &list);
2788 else if (sk->sk_state == TIPC_CONNECTING)
2789 tipc_sk_retry_connect(sk, &list);
2791 bh_unlock_sock(sk);
2793 if (!skb_queue_empty(&list))
2794 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2796 /* SYN messages may cause link congestion */
2797 if (rc == -ELINKCONG) {
2798 tipc_dest_push(&tsk->cong_links, pnode, 0);
2799 tsk->cong_link_cnt = 1;
2801 sock_put(sk);
2804 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2805 struct tipc_name_seq const *seq)
2807 struct sock *sk = &tsk->sk;
2808 struct net *net = sock_net(sk);
2809 struct publication *publ;
2810 u32 key;
2812 if (scope != TIPC_NODE_SCOPE)
2813 scope = TIPC_CLUSTER_SCOPE;
2815 if (tipc_sk_connected(sk))
2816 return -EINVAL;
2817 key = tsk->portid + tsk->pub_count + 1;
2818 if (key == tsk->portid)
2819 return -EADDRINUSE;
2821 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2822 scope, tsk->portid, key);
2823 if (unlikely(!publ))
2824 return -EINVAL;
2826 list_add(&publ->binding_sock, &tsk->publications);
2827 tsk->pub_count++;
2828 tsk->published = 1;
2829 return 0;
2832 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2833 struct tipc_name_seq const *seq)
2835 struct net *net = sock_net(&tsk->sk);
2836 struct publication *publ;
2837 struct publication *safe;
2838 int rc = -EINVAL;
2840 if (scope != TIPC_NODE_SCOPE)
2841 scope = TIPC_CLUSTER_SCOPE;
2843 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2844 if (seq) {
2845 if (publ->scope != scope)
2846 continue;
2847 if (publ->type != seq->type)
2848 continue;
2849 if (publ->lower != seq->lower)
2850 continue;
2851 if (publ->upper != seq->upper)
2852 break;
2853 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2854 publ->upper, publ->key);
2855 rc = 0;
2856 break;
2858 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2859 publ->upper, publ->key);
2860 rc = 0;
2862 if (list_empty(&tsk->publications))
2863 tsk->published = 0;
2864 return rc;
2867 /* tipc_sk_reinit: set non-zero address in all existing sockets
2868 * when we go from standalone to network mode.
2870 void tipc_sk_reinit(struct net *net)
2872 struct tipc_net *tn = net_generic(net, tipc_net_id);
2873 struct rhashtable_iter iter;
2874 struct tipc_sock *tsk;
2875 struct tipc_msg *msg;
2877 rhashtable_walk_enter(&tn->sk_rht, &iter);
2879 do {
2880 rhashtable_walk_start(&iter);
2882 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2883 sock_hold(&tsk->sk);
2884 rhashtable_walk_stop(&iter);
2885 lock_sock(&tsk->sk);
2886 msg = &tsk->phdr;
2887 msg_set_prevnode(msg, tipc_own_addr(net));
2888 msg_set_orignode(msg, tipc_own_addr(net));
2889 release_sock(&tsk->sk);
2890 rhashtable_walk_start(&iter);
2891 sock_put(&tsk->sk);
2894 rhashtable_walk_stop(&iter);
2895 } while (tsk == ERR_PTR(-EAGAIN));
2897 rhashtable_walk_exit(&iter);
2900 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2902 struct tipc_net *tn = net_generic(net, tipc_net_id);
2903 struct tipc_sock *tsk;
2905 rcu_read_lock();
2906 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2907 if (tsk)
2908 sock_hold(&tsk->sk);
2909 rcu_read_unlock();
2911 return tsk;
2914 static int tipc_sk_insert(struct tipc_sock *tsk)
2916 struct sock *sk = &tsk->sk;
2917 struct net *net = sock_net(sk);
2918 struct tipc_net *tn = net_generic(net, tipc_net_id);
2919 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2920 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2922 while (remaining--) {
2923 portid++;
2924 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2925 portid = TIPC_MIN_PORT;
2926 tsk->portid = portid;
2927 sock_hold(&tsk->sk);
2928 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2929 tsk_rht_params))
2930 return 0;
2931 sock_put(&tsk->sk);
2934 return -1;
2937 static void tipc_sk_remove(struct tipc_sock *tsk)
2939 struct sock *sk = &tsk->sk;
2940 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2942 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2943 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2944 __sock_put(sk);
2948 static const struct rhashtable_params tsk_rht_params = {
2949 .nelem_hint = 192,
2950 .head_offset = offsetof(struct tipc_sock, node),
2951 .key_offset = offsetof(struct tipc_sock, portid),
2952 .key_len = sizeof(u32), /* portid */
2953 .max_size = 1048576,
2954 .min_size = 256,
2955 .automatic_shrinking = true,
2958 int tipc_sk_rht_init(struct net *net)
2960 struct tipc_net *tn = net_generic(net, tipc_net_id);
2962 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2965 void tipc_sk_rht_destroy(struct net *net)
2967 struct tipc_net *tn = net_generic(net, tipc_net_id);
2969 /* Wait for socket readers to complete */
2970 synchronize_net();
2972 rhashtable_destroy(&tn->sk_rht);
2975 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2977 struct net *net = sock_net(&tsk->sk);
2978 struct tipc_group *grp = tsk->group;
2979 struct tipc_msg *hdr = &tsk->phdr;
2980 struct tipc_name_seq seq;
2981 int rc;
2983 if (mreq->type < TIPC_RESERVED_TYPES)
2984 return -EACCES;
2985 if (mreq->scope > TIPC_NODE_SCOPE)
2986 return -EINVAL;
2987 if (grp)
2988 return -EACCES;
2989 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2990 if (!grp)
2991 return -ENOMEM;
2992 tsk->group = grp;
2993 msg_set_lookup_scope(hdr, mreq->scope);
2994 msg_set_nametype(hdr, mreq->type);
2995 msg_set_dest_droppable(hdr, true);
2996 seq.type = mreq->type;
2997 seq.lower = mreq->instance;
2998 seq.upper = seq.lower;
2999 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3000 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3001 if (rc) {
3002 tipc_group_delete(net, grp);
3003 tsk->group = NULL;
3004 return rc;
3006 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3007 tsk->mc_method.rcast = true;
3008 tsk->mc_method.mandatory = true;
3009 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3010 return rc;
3013 static int tipc_sk_leave(struct tipc_sock *tsk)
3015 struct net *net = sock_net(&tsk->sk);
3016 struct tipc_group *grp = tsk->group;
3017 struct tipc_name_seq seq;
3018 int scope;
3020 if (!grp)
3021 return -EINVAL;
3022 tipc_group_self(grp, &seq, &scope);
3023 tipc_group_delete(net, grp);
3024 tsk->group = NULL;
3025 tipc_sk_withdraw(tsk, scope, &seq);
3026 return 0;
3030 * tipc_setsockopt - set socket option
3031 * @sock: socket structure
3032 * @lvl: option level
3033 * @opt: option identifier
3034 * @ov: pointer to new option value
3035 * @ol: length of option value
3037 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3038 * (to ease compatibility).
3040 * Returns 0 on success, errno otherwise
3042 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3043 char __user *ov, unsigned int ol)
3045 struct sock *sk = sock->sk;
3046 struct tipc_sock *tsk = tipc_sk(sk);
3047 struct tipc_group_req mreq;
3048 u32 value = 0;
3049 int res = 0;
3051 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3052 return 0;
3053 if (lvl != SOL_TIPC)
3054 return -ENOPROTOOPT;
3056 switch (opt) {
3057 case TIPC_IMPORTANCE:
3058 case TIPC_SRC_DROPPABLE:
3059 case TIPC_DEST_DROPPABLE:
3060 case TIPC_CONN_TIMEOUT:
3061 case TIPC_NODELAY:
3062 if (ol < sizeof(value))
3063 return -EINVAL;
3064 if (get_user(value, (u32 __user *)ov))
3065 return -EFAULT;
3066 break;
3067 case TIPC_GROUP_JOIN:
3068 if (ol < sizeof(mreq))
3069 return -EINVAL;
3070 if (copy_from_user(&mreq, ov, sizeof(mreq)))
3071 return -EFAULT;
3072 break;
3073 default:
3074 if (ov || ol)
3075 return -EINVAL;
3078 lock_sock(sk);
3080 switch (opt) {
3081 case TIPC_IMPORTANCE:
3082 res = tsk_set_importance(tsk, value);
3083 break;
3084 case TIPC_SRC_DROPPABLE:
3085 if (sock->type != SOCK_STREAM)
3086 tsk_set_unreliable(tsk, value);
3087 else
3088 res = -ENOPROTOOPT;
3089 break;
3090 case TIPC_DEST_DROPPABLE:
3091 tsk_set_unreturnable(tsk, value);
3092 break;
3093 case TIPC_CONN_TIMEOUT:
3094 tipc_sk(sk)->conn_timeout = value;
3095 break;
3096 case TIPC_MCAST_BROADCAST:
3097 tsk->mc_method.rcast = false;
3098 tsk->mc_method.mandatory = true;
3099 break;
3100 case TIPC_MCAST_REPLICAST:
3101 tsk->mc_method.rcast = true;
3102 tsk->mc_method.mandatory = true;
3103 break;
3104 case TIPC_GROUP_JOIN:
3105 res = tipc_sk_join(tsk, &mreq);
3106 break;
3107 case TIPC_GROUP_LEAVE:
3108 res = tipc_sk_leave(tsk);
3109 break;
3110 case TIPC_NODELAY:
3111 tsk->nodelay = !!value;
3112 tsk_set_nagle(tsk);
3113 break;
3114 default:
3115 res = -EINVAL;
3118 release_sock(sk);
3120 return res;
3124 * tipc_getsockopt - get socket option
3125 * @sock: socket structure
3126 * @lvl: option level
3127 * @opt: option identifier
3128 * @ov: receptacle for option value
3129 * @ol: receptacle for length of option value
3131 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3132 * (to ease compatibility).
3134 * Returns 0 on success, errno otherwise
3136 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3137 char __user *ov, int __user *ol)
3139 struct sock *sk = sock->sk;
3140 struct tipc_sock *tsk = tipc_sk(sk);
3141 struct tipc_name_seq seq;
3142 int len, scope;
3143 u32 value;
3144 int res;
3146 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3147 return put_user(0, ol);
3148 if (lvl != SOL_TIPC)
3149 return -ENOPROTOOPT;
3150 res = get_user(len, ol);
3151 if (res)
3152 return res;
3154 lock_sock(sk);
3156 switch (opt) {
3157 case TIPC_IMPORTANCE:
3158 value = tsk_importance(tsk);
3159 break;
3160 case TIPC_SRC_DROPPABLE:
3161 value = tsk_unreliable(tsk);
3162 break;
3163 case TIPC_DEST_DROPPABLE:
3164 value = tsk_unreturnable(tsk);
3165 break;
3166 case TIPC_CONN_TIMEOUT:
3167 value = tsk->conn_timeout;
3168 /* no need to set "res", since already 0 at this point */
3169 break;
3170 case TIPC_NODE_RECVQ_DEPTH:
3171 value = 0; /* was tipc_queue_size, now obsolete */
3172 break;
3173 case TIPC_SOCK_RECVQ_DEPTH:
3174 value = skb_queue_len(&sk->sk_receive_queue);
3175 break;
3176 case TIPC_SOCK_RECVQ_USED:
3177 value = sk_rmem_alloc_get(sk);
3178 break;
3179 case TIPC_GROUP_JOIN:
3180 seq.type = 0;
3181 if (tsk->group)
3182 tipc_group_self(tsk->group, &seq, &scope);
3183 value = seq.type;
3184 break;
3185 default:
3186 res = -EINVAL;
3189 release_sock(sk);
3191 if (res)
3192 return res; /* "get" failed */
3194 if (len < sizeof(value))
3195 return -EINVAL;
3197 if (copy_to_user(ov, &value, sizeof(value)))
3198 return -EFAULT;
3200 return put_user(sizeof(value), ol);
3203 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3205 struct net *net = sock_net(sock->sk);
3206 struct tipc_sioc_nodeid_req nr = {0};
3207 struct tipc_sioc_ln_req lnr;
3208 void __user *argp = (void __user *)arg;
3210 switch (cmd) {
3211 case SIOCGETLINKNAME:
3212 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3213 return -EFAULT;
3214 if (!tipc_node_get_linkname(net,
3215 lnr.bearer_id & 0xffff, lnr.peer,
3216 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3217 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3218 return -EFAULT;
3219 return 0;
3221 return -EADDRNOTAVAIL;
3222 case SIOCGETNODEID:
3223 if (copy_from_user(&nr, argp, sizeof(nr)))
3224 return -EFAULT;
3225 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3226 return -EADDRNOTAVAIL;
3227 if (copy_to_user(argp, &nr, sizeof(nr)))
3228 return -EFAULT;
3229 return 0;
3230 default:
3231 return -ENOIOCTLCMD;
3235 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3237 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3238 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3239 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3241 tsk1->peer.family = AF_TIPC;
3242 tsk1->peer.addrtype = TIPC_ADDR_ID;
3243 tsk1->peer.scope = TIPC_NODE_SCOPE;
3244 tsk1->peer.addr.id.ref = tsk2->portid;
3245 tsk1->peer.addr.id.node = onode;
3246 tsk2->peer.family = AF_TIPC;
3247 tsk2->peer.addrtype = TIPC_ADDR_ID;
3248 tsk2->peer.scope = TIPC_NODE_SCOPE;
3249 tsk2->peer.addr.id.ref = tsk1->portid;
3250 tsk2->peer.addr.id.node = onode;
3252 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3253 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3254 return 0;
3257 /* Protocol switches for the various types of TIPC sockets */
3259 static const struct proto_ops msg_ops = {
3260 .owner = THIS_MODULE,
3261 .family = AF_TIPC,
3262 .release = tipc_release,
3263 .bind = tipc_bind,
3264 .connect = tipc_connect,
3265 .socketpair = tipc_socketpair,
3266 .accept = sock_no_accept,
3267 .getname = tipc_getname,
3268 .poll = tipc_poll,
3269 .ioctl = tipc_ioctl,
3270 .listen = sock_no_listen,
3271 .shutdown = tipc_shutdown,
3272 .setsockopt = tipc_setsockopt,
3273 .getsockopt = tipc_getsockopt,
3274 .sendmsg = tipc_sendmsg,
3275 .recvmsg = tipc_recvmsg,
3276 .mmap = sock_no_mmap,
3277 .sendpage = sock_no_sendpage
3280 static const struct proto_ops packet_ops = {
3281 .owner = THIS_MODULE,
3282 .family = AF_TIPC,
3283 .release = tipc_release,
3284 .bind = tipc_bind,
3285 .connect = tipc_connect,
3286 .socketpair = tipc_socketpair,
3287 .accept = tipc_accept,
3288 .getname = tipc_getname,
3289 .poll = tipc_poll,
3290 .ioctl = tipc_ioctl,
3291 .listen = tipc_listen,
3292 .shutdown = tipc_shutdown,
3293 .setsockopt = tipc_setsockopt,
3294 .getsockopt = tipc_getsockopt,
3295 .sendmsg = tipc_send_packet,
3296 .recvmsg = tipc_recvmsg,
3297 .mmap = sock_no_mmap,
3298 .sendpage = sock_no_sendpage
3301 static const struct proto_ops stream_ops = {
3302 .owner = THIS_MODULE,
3303 .family = AF_TIPC,
3304 .release = tipc_release,
3305 .bind = tipc_bind,
3306 .connect = tipc_connect,
3307 .socketpair = tipc_socketpair,
3308 .accept = tipc_accept,
3309 .getname = tipc_getname,
3310 .poll = tipc_poll,
3311 .ioctl = tipc_ioctl,
3312 .listen = tipc_listen,
3313 .shutdown = tipc_shutdown,
3314 .setsockopt = tipc_setsockopt,
3315 .getsockopt = tipc_getsockopt,
3316 .sendmsg = tipc_sendstream,
3317 .recvmsg = tipc_recvstream,
3318 .mmap = sock_no_mmap,
3319 .sendpage = sock_no_sendpage
3322 static const struct net_proto_family tipc_family_ops = {
3323 .owner = THIS_MODULE,
3324 .family = AF_TIPC,
3325 .create = tipc_sk_create
3328 static struct proto tipc_proto = {
3329 .name = "TIPC",
3330 .owner = THIS_MODULE,
3331 .obj_size = sizeof(struct tipc_sock),
3332 .sysctl_rmem = sysctl_tipc_rmem
3336 * tipc_socket_init - initialize TIPC socket interface
3338 * Returns 0 on success, errno otherwise
3340 int tipc_socket_init(void)
3342 int res;
3344 res = proto_register(&tipc_proto, 1);
3345 if (res) {
3346 pr_err("Failed to register TIPC protocol type\n");
3347 goto out;
3350 res = sock_register(&tipc_family_ops);
3351 if (res) {
3352 pr_err("Failed to register TIPC socket type\n");
3353 proto_unregister(&tipc_proto);
3354 goto out;
3356 out:
3357 return res;
3361 * tipc_socket_stop - stop TIPC socket interface
3363 void tipc_socket_stop(void)
3365 sock_unregister(tipc_family_ops.family);
3366 proto_unregister(&tipc_proto);
3369 /* Caller should hold socket lock for the passed tipc socket. */
3370 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3372 u32 peer_node;
3373 u32 peer_port;
3374 struct nlattr *nest;
3376 peer_node = tsk_peer_node(tsk);
3377 peer_port = tsk_peer_port(tsk);
3379 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3380 if (!nest)
3381 return -EMSGSIZE;
3383 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3384 goto msg_full;
3385 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3386 goto msg_full;
3388 if (tsk->conn_type != 0) {
3389 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3390 goto msg_full;
3391 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3392 goto msg_full;
3393 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3394 goto msg_full;
3396 nla_nest_end(skb, nest);
3398 return 0;
3400 msg_full:
3401 nla_nest_cancel(skb, nest);
3403 return -EMSGSIZE;
3406 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3407 *tsk)
3409 struct net *net = sock_net(skb->sk);
3410 struct sock *sk = &tsk->sk;
3412 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3413 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3414 return -EMSGSIZE;
3416 if (tipc_sk_connected(sk)) {
3417 if (__tipc_nl_add_sk_con(skb, tsk))
3418 return -EMSGSIZE;
3419 } else if (!list_empty(&tsk->publications)) {
3420 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3421 return -EMSGSIZE;
3423 return 0;
3426 /* Caller should hold socket lock for the passed tipc socket. */
3427 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3428 struct tipc_sock *tsk)
3430 struct nlattr *attrs;
3431 void *hdr;
3433 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3434 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3435 if (!hdr)
3436 goto msg_cancel;
3438 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3439 if (!attrs)
3440 goto genlmsg_cancel;
3442 if (__tipc_nl_add_sk_info(skb, tsk))
3443 goto attr_msg_cancel;
3445 nla_nest_end(skb, attrs);
3446 genlmsg_end(skb, hdr);
3448 return 0;
3450 attr_msg_cancel:
3451 nla_nest_cancel(skb, attrs);
3452 genlmsg_cancel:
3453 genlmsg_cancel(skb, hdr);
3454 msg_cancel:
3455 return -EMSGSIZE;
3458 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3459 int (*skb_handler)(struct sk_buff *skb,
3460 struct netlink_callback *cb,
3461 struct tipc_sock *tsk))
3463 struct rhashtable_iter *iter = (void *)cb->args[4];
3464 struct tipc_sock *tsk;
3465 int err;
3467 rhashtable_walk_start(iter);
3468 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3469 if (IS_ERR(tsk)) {
3470 err = PTR_ERR(tsk);
3471 if (err == -EAGAIN) {
3472 err = 0;
3473 continue;
3475 break;
3478 sock_hold(&tsk->sk);
3479 rhashtable_walk_stop(iter);
3480 lock_sock(&tsk->sk);
3481 err = skb_handler(skb, cb, tsk);
3482 if (err) {
3483 release_sock(&tsk->sk);
3484 sock_put(&tsk->sk);
3485 goto out;
3487 release_sock(&tsk->sk);
3488 rhashtable_walk_start(iter);
3489 sock_put(&tsk->sk);
3491 rhashtable_walk_stop(iter);
3492 out:
3493 return skb->len;
3495 EXPORT_SYMBOL(tipc_nl_sk_walk);
3497 int tipc_dump_start(struct netlink_callback *cb)
3499 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3501 EXPORT_SYMBOL(tipc_dump_start);
3503 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3505 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3506 struct rhashtable_iter *iter = (void *)cb->args[4];
3507 struct tipc_net *tn = tipc_net(net);
3509 if (!iter) {
3510 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3511 if (!iter)
3512 return -ENOMEM;
3514 cb->args[4] = (long)iter;
3517 rhashtable_walk_enter(&tn->sk_rht, iter);
3518 return 0;
3521 int tipc_dump_done(struct netlink_callback *cb)
3523 struct rhashtable_iter *hti = (void *)cb->args[4];
3525 rhashtable_walk_exit(hti);
3526 kfree(hti);
3527 return 0;
3529 EXPORT_SYMBOL(tipc_dump_done);
3531 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3532 struct tipc_sock *tsk, u32 sk_filter_state,
3533 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3535 struct sock *sk = &tsk->sk;
3536 struct nlattr *attrs;
3537 struct nlattr *stat;
3539 /*filter response w.r.t sk_state*/
3540 if (!(sk_filter_state & (1 << sk->sk_state)))
3541 return 0;
3543 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3544 if (!attrs)
3545 goto msg_cancel;
3547 if (__tipc_nl_add_sk_info(skb, tsk))
3548 goto attr_msg_cancel;
3550 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3551 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3552 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3553 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3554 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3555 sock_i_uid(sk))) ||
3556 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3557 tipc_diag_gen_cookie(sk),
3558 TIPC_NLA_SOCK_PAD))
3559 goto attr_msg_cancel;
3561 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3562 if (!stat)
3563 goto attr_msg_cancel;
3565 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3566 skb_queue_len(&sk->sk_receive_queue)) ||
3567 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3568 skb_queue_len(&sk->sk_write_queue)) ||
3569 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3570 atomic_read(&sk->sk_drops)))
3571 goto stat_msg_cancel;
3573 if (tsk->cong_link_cnt &&
3574 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3575 goto stat_msg_cancel;
3577 if (tsk_conn_cong(tsk) &&
3578 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3579 goto stat_msg_cancel;
3581 nla_nest_end(skb, stat);
3583 if (tsk->group)
3584 if (tipc_group_fill_sock_diag(tsk->group, skb))
3585 goto stat_msg_cancel;
3587 nla_nest_end(skb, attrs);
3589 return 0;
3591 stat_msg_cancel:
3592 nla_nest_cancel(skb, stat);
3593 attr_msg_cancel:
3594 nla_nest_cancel(skb, attrs);
3595 msg_cancel:
3596 return -EMSGSIZE;
3598 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3600 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3602 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3605 /* Caller should hold socket lock for the passed tipc socket. */
3606 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3607 struct netlink_callback *cb,
3608 struct publication *publ)
3610 void *hdr;
3611 struct nlattr *attrs;
3613 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3614 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3615 if (!hdr)
3616 goto msg_cancel;
3618 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3619 if (!attrs)
3620 goto genlmsg_cancel;
3622 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3623 goto attr_msg_cancel;
3624 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3625 goto attr_msg_cancel;
3626 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3627 goto attr_msg_cancel;
3628 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3629 goto attr_msg_cancel;
3631 nla_nest_end(skb, attrs);
3632 genlmsg_end(skb, hdr);
3634 return 0;
3636 attr_msg_cancel:
3637 nla_nest_cancel(skb, attrs);
3638 genlmsg_cancel:
3639 genlmsg_cancel(skb, hdr);
3640 msg_cancel:
3641 return -EMSGSIZE;
3644 /* Caller should hold socket lock for the passed tipc socket. */
3645 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3646 struct netlink_callback *cb,
3647 struct tipc_sock *tsk, u32 *last_publ)
3649 int err;
3650 struct publication *p;
3652 if (*last_publ) {
3653 list_for_each_entry(p, &tsk->publications, binding_sock) {
3654 if (p->key == *last_publ)
3655 break;
3657 if (p->key != *last_publ) {
3658 /* We never set seq or call nl_dump_check_consistent()
3659 * this means that setting prev_seq here will cause the
3660 * consistence check to fail in the netlink callback
3661 * handler. Resulting in the last NLMSG_DONE message
3662 * having the NLM_F_DUMP_INTR flag set.
3664 cb->prev_seq = 1;
3665 *last_publ = 0;
3666 return -EPIPE;
3668 } else {
3669 p = list_first_entry(&tsk->publications, struct publication,
3670 binding_sock);
3673 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3674 err = __tipc_nl_add_sk_publ(skb, cb, p);
3675 if (err) {
3676 *last_publ = p->key;
3677 return err;
3680 *last_publ = 0;
3682 return 0;
3685 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3687 int err;
3688 u32 tsk_portid = cb->args[0];
3689 u32 last_publ = cb->args[1];
3690 u32 done = cb->args[2];
3691 struct net *net = sock_net(skb->sk);
3692 struct tipc_sock *tsk;
3694 if (!tsk_portid) {
3695 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3696 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3698 if (!attrs[TIPC_NLA_SOCK])
3699 return -EINVAL;
3701 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3702 attrs[TIPC_NLA_SOCK],
3703 tipc_nl_sock_policy, NULL);
3704 if (err)
3705 return err;
3707 if (!sock[TIPC_NLA_SOCK_REF])
3708 return -EINVAL;
3710 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3713 if (done)
3714 return 0;
3716 tsk = tipc_sk_lookup(net, tsk_portid);
3717 if (!tsk)
3718 return -EINVAL;
3720 lock_sock(&tsk->sk);
3721 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3722 if (!err)
3723 done = 1;
3724 release_sock(&tsk->sk);
3725 sock_put(&tsk->sk);
3727 cb->args[0] = tsk_portid;
3728 cb->args[1] = last_publ;
3729 cb->args[2] = done;
3731 return skb->len;
3735 * tipc_sk_filtering - check if a socket should be traced
3736 * @sk: the socket to be examined
3737 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3738 * (portid, sock type, name type, name lower, name upper)
3740 * Returns true if the socket meets the socket tuple data
3741 * (value 0 = 'any') or when there is no tuple set (all = 0),
3742 * otherwise false
3744 bool tipc_sk_filtering(struct sock *sk)
3746 struct tipc_sock *tsk;
3747 struct publication *p;
3748 u32 _port, _sktype, _type, _lower, _upper;
3749 u32 type = 0, lower = 0, upper = 0;
3751 if (!sk)
3752 return true;
3754 tsk = tipc_sk(sk);
3756 _port = sysctl_tipc_sk_filter[0];
3757 _sktype = sysctl_tipc_sk_filter[1];
3758 _type = sysctl_tipc_sk_filter[2];
3759 _lower = sysctl_tipc_sk_filter[3];
3760 _upper = sysctl_tipc_sk_filter[4];
3762 if (!_port && !_sktype && !_type && !_lower && !_upper)
3763 return true;
3765 if (_port)
3766 return (_port == tsk->portid);
3768 if (_sktype && _sktype != sk->sk_type)
3769 return false;
3771 if (tsk->published) {
3772 p = list_first_entry_or_null(&tsk->publications,
3773 struct publication, binding_sock);
3774 if (p) {
3775 type = p->type;
3776 lower = p->lower;
3777 upper = p->upper;
3781 if (!tipc_sk_type_connectionless(sk)) {
3782 type = tsk->conn_type;
3783 lower = tsk->conn_instance;
3784 upper = tsk->conn_instance;
3787 if ((_type && _type != type) || (_lower && _lower != lower) ||
3788 (_upper && _upper != upper))
3789 return false;
3791 return true;
3794 u32 tipc_sock_get_portid(struct sock *sk)
3796 return (sk) ? (tipc_sk(sk))->portid : 0;
3800 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3801 * both the rcv and backlog queues are considered
3802 * @sk: tipc sk to be checked
3803 * @skb: tipc msg to be checked
3805 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3808 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3810 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3811 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3812 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3814 return (qsize > lim * 90 / 100);
3818 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3819 * only the rcv queue is considered
3820 * @sk: tipc sk to be checked
3821 * @skb: tipc msg to be checked
3823 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3826 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3828 unsigned int lim = rcvbuf_limit(sk, skb);
3829 unsigned int qsize = sk_rmem_alloc_get(sk);
3831 return (qsize > lim * 90 / 100);
3835 * tipc_sk_dump - dump TIPC socket
3836 * @sk: tipc sk to be dumped
3837 * @dqueues: bitmask to decide if any socket queue to be dumped?
3838 * - TIPC_DUMP_NONE: don't dump socket queues
3839 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3840 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3841 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3842 * - TIPC_DUMP_ALL: dump all the socket queues above
3843 * @buf: returned buffer of dump data in format
3845 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3847 int i = 0;
3848 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3849 struct tipc_sock *tsk;
3850 struct publication *p;
3851 bool tsk_connected;
3853 if (!sk) {
3854 i += scnprintf(buf, sz, "sk data: (null)\n");
3855 return i;
3858 tsk = tipc_sk(sk);
3859 tsk_connected = !tipc_sk_type_connectionless(sk);
3861 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3862 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3863 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3864 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3865 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3866 if (tsk_connected) {
3867 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3868 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3869 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3870 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3872 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3873 if (tsk->published) {
3874 p = list_first_entry_or_null(&tsk->publications,
3875 struct publication, binding_sock);
3876 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3877 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3878 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3880 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3881 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3882 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3883 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3884 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3885 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3886 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3887 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3888 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3889 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3890 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3891 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3892 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3893 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3895 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3896 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3897 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3900 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3901 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3902 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3905 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3906 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3907 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3908 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3909 i += scnprintf(buf + i, sz - i, " tail ");
3910 i += tipc_skb_dump(sk->sk_backlog.tail, false,
3911 buf + i);
3915 return i;