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
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>
41 #include "name_table.h"
44 #include "name_distr.h"
50 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
52 #define TIPC_FWD_MSG 1
53 #define TIPC_MAX_PORT 0xffffffff
54 #define TIPC_MIN_PORT 1
55 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
58 TIPC_LISTEN
= TCP_LISTEN
,
59 TIPC_ESTABLISHED
= TCP_ESTABLISHED
,
60 TIPC_OPEN
= TCP_CLOSE
,
61 TIPC_DISCONNECTING
= TCP_CLOSE_WAIT
,
62 TIPC_CONNECTING
= TCP_SYN_SENT
,
65 struct sockaddr_pair
{
66 struct sockaddr_tipc sock
;
67 struct sockaddr_tipc member
;
71 * struct tipc_sock - TIPC socket structure
72 * @sk: socket - interacts with 'port' and with user via the socket API
73 * @conn_type: TIPC type used when connection was established
74 * @conn_instance: TIPC instance used when connection was established
75 * @published: non-zero if port has one or more associated names
76 * @max_pkt: maximum packet size "hint" used when building messages sent by port
77 * @portid: unique port identity in TIPC socket hash table
78 * @phdr: preformatted message header used when sending messages
79 * #cong_links: list of congested links
80 * @publications: list of publications for port
81 * @blocking_link: address of the congested link we are currently sleeping on
82 * @pub_count: total # of publications port has made during its lifetime
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86 * @cong_link_cnt: number of congested links
87 * @snt_unacked: # messages sent by socket, and not yet acked by peer
88 * @rcv_unacked: # messages read by user, but not yet acked back to peer
89 * @peer: 'connected' peer for dgram/rdm
90 * @node: hash table node
91 * @mc_method: cookie for use between socket and broadcast layer
92 * @rcu: rcu struct for tipc_sock
101 struct tipc_msg phdr
;
102 struct list_head cong_links
;
103 struct list_head publications
;
106 atomic_t dupl_rcvcnt
;
114 struct sockaddr_tipc peer
;
115 struct rhash_head node
;
116 struct tipc_mc_method mc_method
;
118 struct tipc_group
*group
;
121 static int tipc_sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
122 static void tipc_data_ready(struct sock
*sk
);
123 static void tipc_write_space(struct sock
*sk
);
124 static void tipc_sock_destruct(struct sock
*sk
);
125 static int tipc_release(struct socket
*sock
);
126 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
,
128 static void tipc_sk_timeout(struct timer_list
*t
);
129 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
130 struct tipc_name_seq
const *seq
);
131 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
132 struct tipc_name_seq
const *seq
);
133 static int tipc_sk_leave(struct tipc_sock
*tsk
);
134 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
);
135 static int tipc_sk_insert(struct tipc_sock
*tsk
);
136 static void tipc_sk_remove(struct tipc_sock
*tsk
);
137 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
);
138 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dsz
);
140 static const struct proto_ops packet_ops
;
141 static const struct proto_ops stream_ops
;
142 static const struct proto_ops msg_ops
;
143 static struct proto tipc_proto
;
144 static const struct rhashtable_params tsk_rht_params
;
146 static u32
tsk_own_node(struct tipc_sock
*tsk
)
148 return msg_prevnode(&tsk
->phdr
);
151 static u32
tsk_peer_node(struct tipc_sock
*tsk
)
153 return msg_destnode(&tsk
->phdr
);
156 static u32
tsk_peer_port(struct tipc_sock
*tsk
)
158 return msg_destport(&tsk
->phdr
);
161 static bool tsk_unreliable(struct tipc_sock
*tsk
)
163 return msg_src_droppable(&tsk
->phdr
) != 0;
166 static void tsk_set_unreliable(struct tipc_sock
*tsk
, bool unreliable
)
168 msg_set_src_droppable(&tsk
->phdr
, unreliable
? 1 : 0);
171 static bool tsk_unreturnable(struct tipc_sock
*tsk
)
173 return msg_dest_droppable(&tsk
->phdr
) != 0;
176 static void tsk_set_unreturnable(struct tipc_sock
*tsk
, bool unreturnable
)
178 msg_set_dest_droppable(&tsk
->phdr
, unreturnable
? 1 : 0);
181 static int tsk_importance(struct tipc_sock
*tsk
)
183 return msg_importance(&tsk
->phdr
);
186 static int tsk_set_importance(struct tipc_sock
*tsk
, int imp
)
188 if (imp
> TIPC_CRITICAL_IMPORTANCE
)
190 msg_set_importance(&tsk
->phdr
, (u32
)imp
);
194 static struct tipc_sock
*tipc_sk(const struct sock
*sk
)
196 return container_of(sk
, struct tipc_sock
, sk
);
199 static bool tsk_conn_cong(struct tipc_sock
*tsk
)
201 return tsk
->snt_unacked
> tsk
->snd_win
;
204 static u16
tsk_blocks(int len
)
206 return ((len
/ FLOWCTL_BLK_SZ
) + 1);
209 /* tsk_blocks(): translate a buffer size in bytes to number of
210 * advertisable blocks, taking into account the ratio truesize(len)/len
211 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
213 static u16
tsk_adv_blocks(int len
)
215 return len
/ FLOWCTL_BLK_SZ
/ 4;
218 /* tsk_inc(): increment counter for sent or received data
219 * - If block based flow control is not supported by peer we
220 * fall back to message based ditto, incrementing the counter
222 static u16
tsk_inc(struct tipc_sock
*tsk
, int msglen
)
224 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
225 return ((msglen
/ FLOWCTL_BLK_SZ
) + 1);
230 * tsk_advance_rx_queue - discard first buffer in socket receive queue
232 * Caller must hold socket lock
234 static void tsk_advance_rx_queue(struct sock
*sk
)
236 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
239 /* tipc_sk_respond() : send response message back to sender
241 static void tipc_sk_respond(struct sock
*sk
, struct sk_buff
*skb
, int err
)
245 u32 onode
= tipc_own_addr(sock_net(sk
));
247 if (!tipc_msg_reverse(onode
, &skb
, err
))
250 dnode
= msg_destnode(buf_msg(skb
));
251 selector
= msg_origport(buf_msg(skb
));
252 tipc_node_xmit_skb(sock_net(sk
), skb
, dnode
, selector
);
256 * tsk_rej_rx_queue - reject all buffers in socket receive queue
258 * Caller must hold socket lock
260 static void tsk_rej_rx_queue(struct sock
*sk
)
264 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)))
265 tipc_sk_respond(sk
, skb
, TIPC_ERR_NO_PORT
);
268 static bool tipc_sk_connected(struct sock
*sk
)
270 return sk
->sk_state
== TIPC_ESTABLISHED
;
273 /* tipc_sk_type_connectionless - check if the socket is datagram socket
276 * Returns true if connection less, false otherwise
278 static bool tipc_sk_type_connectionless(struct sock
*sk
)
280 return sk
->sk_type
== SOCK_RDM
|| sk
->sk_type
== SOCK_DGRAM
;
283 /* tsk_peer_msg - verify if message was sent by connected port's peer
285 * Handles cases where the node's network address has changed from
286 * the default of <0.0.0> to its configured setting.
288 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
290 struct sock
*sk
= &tsk
->sk
;
291 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
292 u32 peer_port
= tsk_peer_port(tsk
);
296 if (unlikely(!tipc_sk_connected(sk
)))
299 if (unlikely(msg_origport(msg
) != peer_port
))
302 orig_node
= msg_orignode(msg
);
303 peer_node
= tsk_peer_node(tsk
);
305 if (likely(orig_node
== peer_node
))
308 if (!orig_node
&& (peer_node
== tn
->own_addr
))
311 if (!peer_node
&& (orig_node
== tn
->own_addr
))
317 /* tipc_set_sk_state - set the sk_state of the socket
320 * Caller must hold socket lock
322 * Returns 0 on success, errno otherwise
324 static int tipc_set_sk_state(struct sock
*sk
, int state
)
326 int oldsk_state
= sk
->sk_state
;
334 case TIPC_CONNECTING
:
335 if (oldsk_state
== TIPC_OPEN
)
338 case TIPC_ESTABLISHED
:
339 if (oldsk_state
== TIPC_CONNECTING
||
340 oldsk_state
== TIPC_OPEN
)
343 case TIPC_DISCONNECTING
:
344 if (oldsk_state
== TIPC_CONNECTING
||
345 oldsk_state
== TIPC_ESTABLISHED
)
351 sk
->sk_state
= state
;
356 static int tipc_sk_sock_err(struct socket
*sock
, long *timeout
)
358 struct sock
*sk
= sock
->sk
;
359 int err
= sock_error(sk
);
360 int typ
= sock
->type
;
364 if (typ
== SOCK_STREAM
|| typ
== SOCK_SEQPACKET
) {
365 if (sk
->sk_state
== TIPC_DISCONNECTING
)
367 else if (!tipc_sk_connected(sk
))
372 if (signal_pending(current
))
373 return sock_intr_errno(*timeout
);
378 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
383 while ((rc_ = !(condition_))) { \
384 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
386 rc_ = tipc_sk_sock_err((sock_), timeo_); \
389 prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE); \
391 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
392 sched_annotate_sleep(); \
394 remove_wait_queue(sk_sleep(sk_), &wait_); \
400 * tipc_sk_create - create a TIPC socket
401 * @net: network namespace (must be default network)
402 * @sock: pre-allocated socket structure
403 * @protocol: protocol indicator (must be 0)
404 * @kern: caused by kernel or by userspace?
406 * This routine creates additional data structures used by the TIPC socket,
407 * initializes them, and links them together.
409 * Returns 0 on success, errno otherwise
411 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
412 int protocol
, int kern
)
415 const struct proto_ops
*ops
;
417 struct tipc_sock
*tsk
;
418 struct tipc_msg
*msg
;
420 /* Validate arguments */
421 if (unlikely(protocol
!= 0))
422 return -EPROTONOSUPPORT
;
424 switch (sock
->type
) {
439 /* Allocate socket's protocol area */
440 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
, kern
);
445 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
446 INIT_LIST_HEAD(&tsk
->publications
);
447 INIT_LIST_HEAD(&tsk
->cong_links
);
449 tn
= net_generic(sock_net(sk
), tipc_net_id
);
451 /* Finish initializing socket data structures */
453 sock_init_data(sock
, sk
);
454 tipc_set_sk_state(sk
, TIPC_OPEN
);
455 if (tipc_sk_insert(tsk
)) {
456 pr_warn("Socket create failed; port number exhausted\n");
460 /* Ensure tsk is visible before we read own_addr. */
463 tipc_msg_init(tn
->own_addr
, msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
466 msg_set_origport(msg
, tsk
->portid
);
467 timer_setup(&sk
->sk_timer
, tipc_sk_timeout
, 0);
469 sk
->sk_backlog_rcv
= tipc_sk_backlog_rcv
;
470 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
471 sk
->sk_data_ready
= tipc_data_ready
;
472 sk
->sk_write_space
= tipc_write_space
;
473 sk
->sk_destruct
= tipc_sock_destruct
;
474 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
475 atomic_set(&tsk
->dupl_rcvcnt
, 0);
477 /* Start out with safe limits until we receive an advertised window */
478 tsk
->snd_win
= tsk_adv_blocks(RCVBUF_MIN
);
479 tsk
->rcv_win
= tsk
->snd_win
;
481 if (tipc_sk_type_connectionless(sk
)) {
482 tsk_set_unreturnable(tsk
, true);
483 if (sock
->type
== SOCK_DGRAM
)
484 tsk_set_unreliable(tsk
, true);
490 static void tipc_sk_callback(struct rcu_head
*head
)
492 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
497 /* Caller should hold socket lock for the socket. */
498 static void __tipc_shutdown(struct socket
*sock
, int error
)
500 struct sock
*sk
= sock
->sk
;
501 struct tipc_sock
*tsk
= tipc_sk(sk
);
502 struct net
*net
= sock_net(sk
);
503 long timeout
= CONN_TIMEOUT_DEFAULT
;
504 u32 dnode
= tsk_peer_node(tsk
);
507 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
508 tipc_wait_for_cond(sock
, &timeout
, (!tsk
->cong_link_cnt
&&
509 !tsk_conn_cong(tsk
)));
511 /* Reject all unreceived messages, except on an active connection
512 * (which disconnects locally & sends a 'FIN+' to peer).
514 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)) != NULL
) {
515 if (TIPC_SKB_CB(skb
)->bytes_read
) {
519 if (!tipc_sk_type_connectionless(sk
) &&
520 sk
->sk_state
!= TIPC_DISCONNECTING
) {
521 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
522 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
524 tipc_sk_respond(sk
, skb
, error
);
527 if (tipc_sk_type_connectionless(sk
))
530 if (sk
->sk_state
!= TIPC_DISCONNECTING
) {
531 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
532 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0, dnode
,
533 tsk_own_node(tsk
), tsk_peer_port(tsk
),
536 tipc_node_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
537 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
538 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
543 * tipc_release - destroy a TIPC socket
544 * @sock: socket to destroy
546 * This routine cleans up any messages that are still queued on the socket.
547 * For DGRAM and RDM socket types, all queued messages are rejected.
548 * For SEQPACKET and STREAM socket types, the first message is rejected
549 * and any others are discarded. (If the first message on a STREAM socket
550 * is partially-read, it is discarded and the next one is rejected instead.)
552 * NOTE: Rejected messages are not necessarily returned to the sender! They
553 * are returned or discarded according to the "destination droppable" setting
554 * specified for the message by the sender.
556 * Returns 0 on success, errno otherwise
558 static int tipc_release(struct socket
*sock
)
560 struct sock
*sk
= sock
->sk
;
561 struct tipc_sock
*tsk
;
564 * Exit if socket isn't fully initialized (occurs when a failed accept()
565 * releases a pre-allocated child socket that was never used)
573 __tipc_shutdown(sock
, TIPC_ERR_NO_PORT
);
574 sk
->sk_shutdown
= SHUTDOWN_MASK
;
576 tipc_sk_withdraw(tsk
, 0, NULL
);
577 sk_stop_timer(sk
, &sk
->sk_timer
);
580 /* Reject any messages that accumulated in backlog queue */
582 tipc_dest_list_purge(&tsk
->cong_links
);
583 tsk
->cong_link_cnt
= 0;
584 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
591 * tipc_bind - associate or disassocate TIPC name(s) with a socket
592 * @sock: socket structure
593 * @uaddr: socket address describing name(s) and desired operation
594 * @uaddr_len: size of socket address data structure
596 * Name and name sequence binding is indicated using a positive scope value;
597 * a negative scope value unbinds the specified name. Specifying no name
598 * (i.e. a socket address length of 0) unbinds all names from the socket.
600 * Returns 0 on success, errno otherwise
602 * NOTE: This routine doesn't need to take the socket lock since it doesn't
603 * access any non-constant socket information.
605 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
608 struct sock
*sk
= sock
->sk
;
609 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
610 struct tipc_sock
*tsk
= tipc_sk(sk
);
614 if (unlikely(!uaddr_len
)) {
615 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
622 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
626 if (addr
->family
!= AF_TIPC
) {
631 if (addr
->addrtype
== TIPC_ADDR_NAME
)
632 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
633 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
638 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
639 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
640 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
645 res
= (addr
->scope
> 0) ?
646 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
647 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
654 * tipc_getname - get port ID of socket or peer socket
655 * @sock: socket structure
656 * @uaddr: area for returned socket address
657 * @uaddr_len: area for returned length of socket address
658 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
660 * Returns 0 on success, errno otherwise
662 * NOTE: This routine doesn't need to take the socket lock since it only
663 * accesses socket information that is unchanging (or which changes in
664 * a completely predictable manner).
666 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
667 int *uaddr_len
, int peer
)
669 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
670 struct sock
*sk
= sock
->sk
;
671 struct tipc_sock
*tsk
= tipc_sk(sk
);
672 struct tipc_net
*tn
= net_generic(sock_net(sock
->sk
), tipc_net_id
);
674 memset(addr
, 0, sizeof(*addr
));
676 if ((!tipc_sk_connected(sk
)) &&
677 ((peer
!= 2) || (sk
->sk_state
!= TIPC_DISCONNECTING
)))
679 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
680 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
682 addr
->addr
.id
.ref
= tsk
->portid
;
683 addr
->addr
.id
.node
= tn
->own_addr
;
686 *uaddr_len
= sizeof(*addr
);
687 addr
->addrtype
= TIPC_ADDR_ID
;
688 addr
->family
= AF_TIPC
;
690 addr
->addr
.name
.domain
= 0;
696 * tipc_poll - read and possibly block on pollmask
697 * @file: file structure associated with the socket
698 * @sock: socket for which to calculate the poll bits
701 * Returns pollmask value
704 * It appears that the usual socket locking mechanisms are not useful here
705 * since the pollmask info is potentially out-of-date the moment this routine
706 * exits. TCP and other protocols seem to rely on higher level poll routines
707 * to handle any preventable race conditions, so TIPC will do the same ...
709 * IMPORTANT: The fact that a read or write operation is indicated does NOT
710 * imply that the operation will succeed, merely that it should be performed
711 * and will not block.
713 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
716 struct sock
*sk
= sock
->sk
;
717 struct tipc_sock
*tsk
= tipc_sk(sk
);
718 struct tipc_group
*grp
= tsk
->group
;
721 sock_poll_wait(file
, sk_sleep(sk
), wait
);
723 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
724 revents
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
725 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
728 switch (sk
->sk_state
) {
729 case TIPC_ESTABLISHED
:
730 if (!tsk
->cong_link_cnt
&& !tsk_conn_cong(tsk
))
734 case TIPC_CONNECTING
:
735 if (!skb_queue_empty(&sk
->sk_receive_queue
))
736 revents
|= POLLIN
| POLLRDNORM
;
739 if (!grp
|| tipc_group_size(grp
))
740 if (!tsk
->cong_link_cnt
)
742 if (!tipc_sk_type_connectionless(sk
))
744 if (skb_queue_empty(&sk
->sk_receive_queue
))
746 revents
|= POLLIN
| POLLRDNORM
;
748 case TIPC_DISCONNECTING
:
749 revents
= POLLIN
| POLLRDNORM
| POLLHUP
;
756 * tipc_sendmcast - send multicast message
757 * @sock: socket structure
758 * @seq: destination address
759 * @msg: message to send
760 * @dlen: length of data to send
761 * @timeout: timeout to wait for wakeup
763 * Called from function tipc_sendmsg(), which has done all sanity checks
764 * Returns the number of bytes sent on success, or errno
766 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
767 struct msghdr
*msg
, size_t dlen
, long timeout
)
769 struct sock
*sk
= sock
->sk
;
770 struct tipc_sock
*tsk
= tipc_sk(sk
);
771 struct tipc_msg
*hdr
= &tsk
->phdr
;
772 struct net
*net
= sock_net(sk
);
773 int mtu
= tipc_bcast_get_mtu(net
);
774 struct tipc_mc_method
*method
= &tsk
->mc_method
;
775 u32 domain
= addr_domain(net
, TIPC_CLUSTER_SCOPE
);
776 struct sk_buff_head pkts
;
777 struct tipc_nlist dsts
;
783 /* Block or return if any destination link is congested */
784 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
);
788 /* Lookup destination nodes */
789 tipc_nlist_init(&dsts
, tipc_own_addr(net
));
790 tipc_nametbl_lookup_dst_nodes(net
, seq
->type
, seq
->lower
,
791 seq
->upper
, domain
, &dsts
);
792 if (!dsts
.local
&& !dsts
.remote
)
793 return -EHOSTUNREACH
;
795 /* Build message header */
796 msg_set_type(hdr
, TIPC_MCAST_MSG
);
797 msg_set_hdr_sz(hdr
, MCAST_H_SIZE
);
798 msg_set_lookup_scope(hdr
, TIPC_CLUSTER_SCOPE
);
799 msg_set_destport(hdr
, 0);
800 msg_set_destnode(hdr
, 0);
801 msg_set_nametype(hdr
, seq
->type
);
802 msg_set_namelower(hdr
, seq
->lower
);
803 msg_set_nameupper(hdr
, seq
->upper
);
805 /* Build message as chain of buffers */
806 skb_queue_head_init(&pkts
);
807 rc
= tipc_msg_build(hdr
, msg
, 0, dlen
, mtu
, &pkts
);
809 /* Send message if build was successful */
810 if (unlikely(rc
== dlen
))
811 rc
= tipc_mcast_xmit(net
, &pkts
, method
, &dsts
,
812 &tsk
->cong_link_cnt
);
814 tipc_nlist_purge(&dsts
);
816 return rc
? rc
: dlen
;
820 * tipc_send_group_msg - send a message to a member in the group
821 * @net: network namespace
822 * @m: message to send
824 * @dnode: destination node
825 * @dport: destination port
826 * @dlen: total length of message data
828 static int tipc_send_group_msg(struct net
*net
, struct tipc_sock
*tsk
,
829 struct msghdr
*m
, struct tipc_member
*mb
,
830 u32 dnode
, u32 dport
, int dlen
)
832 u16 bc_snd_nxt
= tipc_group_bc_snd_nxt(tsk
->group
);
833 struct tipc_mc_method
*method
= &tsk
->mc_method
;
834 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
835 struct tipc_msg
*hdr
= &tsk
->phdr
;
836 struct sk_buff_head pkts
;
839 /* Complete message header */
840 msg_set_type(hdr
, TIPC_GRP_UCAST_MSG
);
841 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
842 msg_set_destport(hdr
, dport
);
843 msg_set_destnode(hdr
, dnode
);
844 msg_set_grp_bc_seqno(hdr
, bc_snd_nxt
);
846 /* Build message as chain of buffers */
847 skb_queue_head_init(&pkts
);
848 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
849 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
850 if (unlikely(rc
!= dlen
))
854 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
855 if (unlikely(rc
== -ELINKCONG
)) {
856 tipc_dest_push(&tsk
->cong_links
, dnode
, 0);
857 tsk
->cong_link_cnt
++;
860 /* Update send window */
861 tipc_group_update_member(mb
, blks
);
863 /* A broadcast sent within next EXPIRE period must follow same path */
864 method
->rcast
= true;
865 method
->mandatory
= true;
870 * tipc_send_group_unicast - send message to a member in the group
871 * @sock: socket structure
872 * @m: message to send
873 * @dlen: total length of message data
874 * @timeout: timeout to wait for wakeup
876 * Called from function tipc_sendmsg(), which has done all sanity checks
877 * Returns the number of bytes sent on success, or errno
879 static int tipc_send_group_unicast(struct socket
*sock
, struct msghdr
*m
,
880 int dlen
, long timeout
)
882 struct sock
*sk
= sock
->sk
;
883 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
884 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
885 struct tipc_sock
*tsk
= tipc_sk(sk
);
886 struct tipc_group
*grp
= tsk
->group
;
887 struct net
*net
= sock_net(sk
);
888 struct tipc_member
*mb
= NULL
;
892 node
= dest
->addr
.id
.node
;
893 port
= dest
->addr
.id
.ref
;
895 return -EHOSTUNREACH
;
897 /* Block or return if destination link or member is congested */
898 rc
= tipc_wait_for_cond(sock
, &timeout
,
899 !tipc_dest_find(&tsk
->cong_links
, node
, 0) &&
900 !tipc_group_cong(grp
, node
, port
, blks
, &mb
));
905 return -EHOSTUNREACH
;
907 rc
= tipc_send_group_msg(net
, tsk
, m
, mb
, node
, port
, dlen
);
909 return rc
? rc
: dlen
;
913 * tipc_send_group_anycast - send message to any member with given identity
914 * @sock: socket structure
915 * @m: message to send
916 * @dlen: total length of message data
917 * @timeout: timeout to wait for wakeup
919 * Called from function tipc_sendmsg(), which has done all sanity checks
920 * Returns the number of bytes sent on success, or errno
922 static int tipc_send_group_anycast(struct socket
*sock
, struct msghdr
*m
,
923 int dlen
, long timeout
)
925 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
926 struct sock
*sk
= sock
->sk
;
927 struct tipc_sock
*tsk
= tipc_sk(sk
);
928 struct list_head
*cong_links
= &tsk
->cong_links
;
929 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
930 struct tipc_group
*grp
= tsk
->group
;
931 struct tipc_member
*first
= NULL
;
932 struct tipc_member
*mbr
= NULL
;
933 struct net
*net
= sock_net(sk
);
934 u32 node
, port
, exclude
;
935 u32 type
, inst
, domain
;
936 struct list_head dsts
;
941 INIT_LIST_HEAD(&dsts
);
943 type
= dest
->addr
.name
.name
.type
;
944 inst
= dest
->addr
.name
.name
.instance
;
945 domain
= addr_domain(net
, dest
->scope
);
946 exclude
= tipc_group_exclude(grp
);
948 while (++lookups
< 4) {
951 /* Look for a non-congested destination member, if any */
953 if (!tipc_nametbl_lookup(net
, type
, inst
, domain
, &dsts
,
954 &dstcnt
, exclude
, false))
955 return -EHOSTUNREACH
;
956 tipc_dest_pop(&dsts
, &node
, &port
);
957 cong
= tipc_group_cong(grp
, node
, port
, blks
, &mbr
);
966 /* Start over if destination was not in member list */
970 if (likely(!cong
&& !tipc_dest_find(cong_links
, node
, 0)))
973 /* Block or return if destination link or member is congested */
974 rc
= tipc_wait_for_cond(sock
, &timeout
,
975 !tipc_dest_find(cong_links
, node
, 0) &&
976 !tipc_group_cong(grp
, node
, port
,
981 /* Send, unless destination disappeared while waiting */
986 if (unlikely(lookups
>= 4))
987 return -EHOSTUNREACH
;
989 rc
= tipc_send_group_msg(net
, tsk
, m
, mbr
, node
, port
, dlen
);
991 return rc
? rc
: dlen
;
995 * tipc_send_group_bcast - send message to all members in communication group
996 * @sk: socket structure
997 * @m: message to send
998 * @dlen: total length of message data
999 * @timeout: timeout to wait for wakeup
1001 * Called from function tipc_sendmsg(), which has done all sanity checks
1002 * Returns the number of bytes sent on success, or errno
1004 static int tipc_send_group_bcast(struct socket
*sock
, struct msghdr
*m
,
1005 int dlen
, long timeout
)
1007 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1008 struct sock
*sk
= sock
->sk
;
1009 struct net
*net
= sock_net(sk
);
1010 struct tipc_sock
*tsk
= tipc_sk(sk
);
1011 struct tipc_group
*grp
= tsk
->group
;
1012 struct tipc_nlist
*dsts
= tipc_group_dests(grp
);
1013 struct tipc_mc_method
*method
= &tsk
->mc_method
;
1014 bool ack
= method
->mandatory
&& method
->rcast
;
1015 int blks
= tsk_blocks(MCAST_H_SIZE
+ dlen
);
1016 struct tipc_msg
*hdr
= &tsk
->phdr
;
1017 int mtu
= tipc_bcast_get_mtu(net
);
1018 struct sk_buff_head pkts
;
1019 int rc
= -EHOSTUNREACH
;
1021 if (!dsts
->local
&& !dsts
->remote
)
1022 return -EHOSTUNREACH
;
1024 /* Block or return if any destination link or member is congested */
1025 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
&&
1026 !tipc_group_bc_cong(grp
, blks
));
1030 /* Complete message header */
1032 msg_set_type(hdr
, TIPC_GRP_MCAST_MSG
);
1033 msg_set_nameinst(hdr
, dest
->addr
.name
.name
.instance
);
1035 msg_set_type(hdr
, TIPC_GRP_BCAST_MSG
);
1036 msg_set_nameinst(hdr
, 0);
1038 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
1039 msg_set_destport(hdr
, 0);
1040 msg_set_destnode(hdr
, 0);
1041 msg_set_grp_bc_seqno(hdr
, tipc_group_bc_snd_nxt(grp
));
1043 /* Avoid getting stuck with repeated forced replicasts */
1044 msg_set_grp_bc_ack_req(hdr
, ack
);
1046 /* Build message as chain of buffers */
1047 skb_queue_head_init(&pkts
);
1048 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1049 if (unlikely(rc
!= dlen
))
1053 rc
= tipc_mcast_xmit(net
, &pkts
, method
, dsts
, &tsk
->cong_link_cnt
);
1057 /* Update broadcast sequence number and send windows */
1058 tipc_group_update_bc_members(tsk
->group
, blks
, ack
);
1060 /* Broadcast link is now free to choose method for next broadcast */
1061 method
->mandatory
= false;
1062 method
->expires
= jiffies
;
1068 * tipc_send_group_mcast - send message to all members with given identity
1069 * @sock: socket structure
1070 * @m: message to send
1071 * @dlen: total length of message data
1072 * @timeout: timeout to wait for wakeup
1074 * Called from function tipc_sendmsg(), which has done all sanity checks
1075 * Returns the number of bytes sent on success, or errno
1077 static int tipc_send_group_mcast(struct socket
*sock
, struct msghdr
*m
,
1078 int dlen
, long timeout
)
1080 struct sock
*sk
= sock
->sk
;
1081 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1082 struct tipc_name_seq
*seq
= &dest
->addr
.nameseq
;
1083 struct tipc_sock
*tsk
= tipc_sk(sk
);
1084 struct tipc_group
*grp
= tsk
->group
;
1085 struct net
*net
= sock_net(sk
);
1086 u32 domain
, exclude
, dstcnt
;
1087 struct list_head dsts
;
1089 INIT_LIST_HEAD(&dsts
);
1091 if (seq
->lower
!= seq
->upper
)
1094 domain
= addr_domain(net
, dest
->scope
);
1095 exclude
= tipc_group_exclude(grp
);
1096 if (!tipc_nametbl_lookup(net
, seq
->type
, seq
->lower
, domain
,
1097 &dsts
, &dstcnt
, exclude
, true))
1098 return -EHOSTUNREACH
;
1101 tipc_dest_pop(&dsts
, &dest
->addr
.id
.node
, &dest
->addr
.id
.ref
);
1102 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1105 tipc_dest_list_purge(&dsts
);
1106 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1110 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1111 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1112 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1114 * Multi-threaded: parallel calls with reference to same queues may occur
1116 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
1117 struct sk_buff_head
*inputq
)
1119 u32 scope
= TIPC_CLUSTER_SCOPE
;
1120 u32 self
= tipc_own_addr(net
);
1121 struct sk_buff
*skb
, *_skb
;
1122 u32 lower
= 0, upper
= ~0;
1123 struct sk_buff_head tmpq
;
1124 u32 portid
, oport
, onode
;
1125 struct list_head dports
;
1126 struct tipc_msg
*msg
;
1127 int user
, mtyp
, hsz
;
1129 __skb_queue_head_init(&tmpq
);
1130 INIT_LIST_HEAD(&dports
);
1132 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
1133 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
1135 user
= msg_user(msg
);
1136 mtyp
= msg_type(msg
);
1137 if (mtyp
== TIPC_GRP_UCAST_MSG
|| user
== GROUP_PROTOCOL
) {
1138 spin_lock_bh(&inputq
->lock
);
1139 if (skb_peek(arrvq
) == skb
) {
1140 __skb_dequeue(arrvq
);
1141 __skb_queue_tail(inputq
, skb
);
1143 refcount_dec(&skb
->users
);
1144 spin_unlock_bh(&inputq
->lock
);
1147 hsz
= skb_headroom(skb
) + msg_hdr_sz(msg
);
1148 oport
= msg_origport(msg
);
1149 onode
= msg_orignode(msg
);
1151 scope
= TIPC_NODE_SCOPE
;
1153 /* Create destination port list and message clones: */
1154 if (!msg_in_group(msg
)) {
1155 lower
= msg_namelower(msg
);
1156 upper
= msg_nameupper(msg
);
1158 tipc_nametbl_mc_translate(net
, msg_nametype(msg
), lower
, upper
,
1160 while (tipc_dest_pop(&dports
, NULL
, &portid
)) {
1161 _skb
= __pskb_copy(skb
, hsz
, GFP_ATOMIC
);
1163 msg_set_destport(buf_msg(_skb
), portid
);
1164 __skb_queue_tail(&tmpq
, _skb
);
1167 pr_warn("Failed to clone mcast rcv buffer\n");
1169 /* Append to inputq if not already done by other thread */
1170 spin_lock_bh(&inputq
->lock
);
1171 if (skb_peek(arrvq
) == skb
) {
1172 skb_queue_splice_tail_init(&tmpq
, inputq
);
1173 kfree_skb(__skb_dequeue(arrvq
));
1175 spin_unlock_bh(&inputq
->lock
);
1176 __skb_queue_purge(&tmpq
);
1179 tipc_sk_rcv(net
, inputq
);
1183 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1184 * @tsk: receiving socket
1185 * @skb: pointer to message buffer.
1187 static void tipc_sk_conn_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
*skb
,
1188 struct sk_buff_head
*xmitq
)
1190 struct tipc_msg
*hdr
= buf_msg(skb
);
1191 u32 onode
= tsk_own_node(tsk
);
1192 struct sock
*sk
= &tsk
->sk
;
1193 int mtyp
= msg_type(hdr
);
1196 /* Ignore if connection cannot be validated: */
1197 if (!tsk_peer_msg(tsk
, hdr
))
1200 if (unlikely(msg_errcode(hdr
))) {
1201 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1202 tipc_node_remove_conn(sock_net(sk
), tsk_peer_node(tsk
),
1203 tsk_peer_port(tsk
));
1204 sk
->sk_state_change(sk
);
1208 tsk
->probe_unacked
= false;
1210 if (mtyp
== CONN_PROBE
) {
1211 msg_set_type(hdr
, CONN_PROBE_REPLY
);
1212 if (tipc_msg_reverse(onode
, &skb
, TIPC_OK
))
1213 __skb_queue_tail(xmitq
, skb
);
1215 } else if (mtyp
== CONN_ACK
) {
1216 conn_cong
= tsk_conn_cong(tsk
);
1217 tsk
->snt_unacked
-= msg_conn_ack(hdr
);
1218 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1219 tsk
->snd_win
= msg_adv_win(hdr
);
1221 sk
->sk_write_space(sk
);
1222 } else if (mtyp
!= CONN_PROBE_REPLY
) {
1223 pr_warn("Received unknown CONN_PROTO msg\n");
1230 * tipc_sendmsg - send message in connectionless manner
1231 * @sock: socket structure
1232 * @m: message to send
1233 * @dsz: amount of user data to be sent
1235 * Message must have an destination specified explicitly.
1236 * Used for SOCK_RDM and SOCK_DGRAM messages,
1237 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1238 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1240 * Returns the number of bytes sent on success, or errno otherwise
1242 static int tipc_sendmsg(struct socket
*sock
,
1243 struct msghdr
*m
, size_t dsz
)
1245 struct sock
*sk
= sock
->sk
;
1249 ret
= __tipc_sendmsg(sock
, m
, dsz
);
1255 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1257 struct sock
*sk
= sock
->sk
;
1258 struct net
*net
= sock_net(sk
);
1259 struct tipc_sock
*tsk
= tipc_sk(sk
);
1260 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1261 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1262 struct list_head
*clinks
= &tsk
->cong_links
;
1263 bool syn
= !tipc_sk_type_connectionless(sk
);
1264 struct tipc_group
*grp
= tsk
->group
;
1265 struct tipc_msg
*hdr
= &tsk
->phdr
;
1266 struct tipc_name_seq
*seq
;
1267 struct sk_buff_head pkts
;
1268 u32 type
, inst
, domain
;
1272 if (unlikely(dlen
> TIPC_MAX_USER_MSG_SIZE
))
1276 if (unlikely(m
->msg_namelen
< sizeof(*dest
)))
1278 if (unlikely(dest
->family
!= AF_TIPC
))
1284 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1285 if (dest
->addrtype
== TIPC_ADDR_NAME
)
1286 return tipc_send_group_anycast(sock
, m
, dlen
, timeout
);
1287 if (dest
->addrtype
== TIPC_ADDR_ID
)
1288 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1289 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1290 return tipc_send_group_mcast(sock
, m
, dlen
, timeout
);
1294 if (unlikely(!dest
)) {
1296 if (!syn
|| dest
->family
!= AF_TIPC
)
1297 return -EDESTADDRREQ
;
1300 if (unlikely(syn
)) {
1301 if (sk
->sk_state
== TIPC_LISTEN
)
1303 if (sk
->sk_state
!= TIPC_OPEN
)
1307 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1308 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
1309 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
1313 seq
= &dest
->addr
.nameseq
;
1314 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1315 return tipc_sendmcast(sock
, seq
, m
, dlen
, timeout
);
1317 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1318 type
= dest
->addr
.name
.name
.type
;
1319 inst
= dest
->addr
.name
.name
.instance
;
1320 domain
= dest
->addr
.name
.domain
;
1322 msg_set_type(hdr
, TIPC_NAMED_MSG
);
1323 msg_set_hdr_sz(hdr
, NAMED_H_SIZE
);
1324 msg_set_nametype(hdr
, type
);
1325 msg_set_nameinst(hdr
, inst
);
1326 msg_set_lookup_scope(hdr
, tipc_addr_scope(domain
));
1327 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
1328 msg_set_destnode(hdr
, dnode
);
1329 msg_set_destport(hdr
, dport
);
1330 if (unlikely(!dport
&& !dnode
))
1331 return -EHOSTUNREACH
;
1332 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
1333 dnode
= dest
->addr
.id
.node
;
1334 msg_set_type(hdr
, TIPC_DIRECT_MSG
);
1335 msg_set_lookup_scope(hdr
, 0);
1336 msg_set_destnode(hdr
, dnode
);
1337 msg_set_destport(hdr
, dest
->addr
.id
.ref
);
1338 msg_set_hdr_sz(hdr
, BASIC_H_SIZE
);
1341 /* Block or return if destination link is congested */
1342 rc
= tipc_wait_for_cond(sock
, &timeout
,
1343 !tipc_dest_find(clinks
, dnode
, 0));
1347 skb_queue_head_init(&pkts
);
1348 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
1349 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1350 if (unlikely(rc
!= dlen
))
1353 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1354 if (unlikely(rc
== -ELINKCONG
)) {
1355 tipc_dest_push(clinks
, dnode
, 0);
1356 tsk
->cong_link_cnt
++;
1360 if (unlikely(syn
&& !rc
))
1361 tipc_set_sk_state(sk
, TIPC_CONNECTING
);
1363 return rc
? rc
: dlen
;
1367 * tipc_sendstream - send stream-oriented data
1368 * @sock: socket structure
1370 * @dsz: total length of data to be transmitted
1372 * Used for SOCK_STREAM data.
1374 * Returns the number of bytes sent on success (or partial success),
1375 * or errno if no data sent
1377 static int tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1379 struct sock
*sk
= sock
->sk
;
1383 ret
= __tipc_sendstream(sock
, m
, dsz
);
1389 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1391 struct sock
*sk
= sock
->sk
;
1392 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1393 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1394 struct tipc_sock
*tsk
= tipc_sk(sk
);
1395 struct tipc_msg
*hdr
= &tsk
->phdr
;
1396 struct net
*net
= sock_net(sk
);
1397 struct sk_buff_head pkts
;
1398 u32 dnode
= tsk_peer_node(tsk
);
1402 skb_queue_head_init(&pkts
);
1404 if (unlikely(dlen
> INT_MAX
))
1407 /* Handle implicit connection setup */
1408 if (unlikely(dest
)) {
1409 rc
= __tipc_sendmsg(sock
, m
, dlen
);
1410 if (dlen
&& (dlen
== rc
))
1411 tsk
->snt_unacked
= tsk_inc(tsk
, dlen
+ msg_hdr_sz(hdr
));
1416 rc
= tipc_wait_for_cond(sock
, &timeout
,
1417 (!tsk
->cong_link_cnt
&&
1418 !tsk_conn_cong(tsk
) &&
1419 tipc_sk_connected(sk
)));
1423 send
= min_t(size_t, dlen
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1424 rc
= tipc_msg_build(hdr
, m
, sent
, send
, tsk
->max_pkt
, &pkts
);
1425 if (unlikely(rc
!= send
))
1428 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1429 if (unlikely(rc
== -ELINKCONG
)) {
1430 tsk
->cong_link_cnt
= 1;
1434 tsk
->snt_unacked
+= tsk_inc(tsk
, send
+ MIN_H_SIZE
);
1437 } while (sent
< dlen
&& !rc
);
1439 return sent
? sent
: rc
;
1443 * tipc_send_packet - send a connection-oriented message
1444 * @sock: socket structure
1445 * @m: message to send
1446 * @dsz: length of data to be transmitted
1448 * Used for SOCK_SEQPACKET messages.
1450 * Returns the number of bytes sent on success, or errno otherwise
1452 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1454 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1457 return tipc_sendstream(sock
, m
, dsz
);
1460 /* tipc_sk_finish_conn - complete the setup of a connection
1462 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1465 struct sock
*sk
= &tsk
->sk
;
1466 struct net
*net
= sock_net(sk
);
1467 struct tipc_msg
*msg
= &tsk
->phdr
;
1469 msg_set_destnode(msg
, peer_node
);
1470 msg_set_destport(msg
, peer_port
);
1471 msg_set_type(msg
, TIPC_CONN_MSG
);
1472 msg_set_lookup_scope(msg
, 0);
1473 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1475 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
1476 tipc_set_sk_state(sk
, TIPC_ESTABLISHED
);
1477 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1478 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1479 tsk
->peer_caps
= tipc_node_get_capabilities(net
, peer_node
);
1480 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1483 /* Fall back to message based flow control */
1484 tsk
->rcv_win
= FLOWCTL_MSG_WIN
;
1485 tsk
->snd_win
= FLOWCTL_MSG_WIN
;
1489 * tipc_sk_set_orig_addr - capture sender's address for received message
1490 * @m: descriptor for message info
1491 * @hdr: received message header
1493 * Note: Address is not captured if not requested by receiver.
1495 static void tipc_sk_set_orig_addr(struct msghdr
*m
, struct sk_buff
*skb
)
1497 DECLARE_SOCKADDR(struct sockaddr_pair
*, srcaddr
, m
->msg_name
);
1498 struct tipc_msg
*hdr
= buf_msg(skb
);
1503 srcaddr
->sock
.family
= AF_TIPC
;
1504 srcaddr
->sock
.addrtype
= TIPC_ADDR_ID
;
1505 srcaddr
->sock
.addr
.id
.ref
= msg_origport(hdr
);
1506 srcaddr
->sock
.addr
.id
.node
= msg_orignode(hdr
);
1507 srcaddr
->sock
.addr
.name
.domain
= 0;
1508 srcaddr
->sock
.scope
= 0;
1509 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1511 if (!msg_in_group(hdr
))
1514 /* Group message users may also want to know sending member's id */
1515 srcaddr
->member
.family
= AF_TIPC
;
1516 srcaddr
->member
.addrtype
= TIPC_ADDR_NAME
;
1517 srcaddr
->member
.addr
.name
.name
.type
= msg_nametype(hdr
);
1518 srcaddr
->member
.addr
.name
.name
.instance
= TIPC_SKB_CB(skb
)->orig_member
;
1519 srcaddr
->member
.addr
.name
.domain
= 0;
1520 m
->msg_namelen
= sizeof(*srcaddr
);
1524 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1525 * @m: descriptor for message info
1526 * @msg: received message header
1527 * @tsk: TIPC port associated with message
1529 * Note: Ancillary data is not captured if not requested by receiver.
1531 * Returns 0 if successful, otherwise errno
1533 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1534 struct tipc_sock
*tsk
)
1542 if (likely(m
->msg_controllen
== 0))
1545 /* Optionally capture errored message object(s) */
1546 err
= msg
? msg_errcode(msg
) : 0;
1547 if (unlikely(err
)) {
1549 anc_data
[1] = msg_data_sz(msg
);
1550 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1554 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1561 /* Optionally capture message destination object */
1562 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1563 switch (dest_type
) {
1564 case TIPC_NAMED_MSG
:
1566 anc_data
[0] = msg_nametype(msg
);
1567 anc_data
[1] = msg_namelower(msg
);
1568 anc_data
[2] = msg_namelower(msg
);
1570 case TIPC_MCAST_MSG
:
1572 anc_data
[0] = msg_nametype(msg
);
1573 anc_data
[1] = msg_namelower(msg
);
1574 anc_data
[2] = msg_nameupper(msg
);
1577 has_name
= (tsk
->conn_type
!= 0);
1578 anc_data
[0] = tsk
->conn_type
;
1579 anc_data
[1] = tsk
->conn_instance
;
1580 anc_data
[2] = tsk
->conn_instance
;
1586 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1594 static void tipc_sk_send_ack(struct tipc_sock
*tsk
)
1596 struct sock
*sk
= &tsk
->sk
;
1597 struct net
*net
= sock_net(sk
);
1598 struct sk_buff
*skb
= NULL
;
1599 struct tipc_msg
*msg
;
1600 u32 peer_port
= tsk_peer_port(tsk
);
1601 u32 dnode
= tsk_peer_node(tsk
);
1603 if (!tipc_sk_connected(sk
))
1605 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1606 dnode
, tsk_own_node(tsk
), peer_port
,
1607 tsk
->portid
, TIPC_OK
);
1611 msg_set_conn_ack(msg
, tsk
->rcv_unacked
);
1612 tsk
->rcv_unacked
= 0;
1614 /* Adjust to and advertize the correct window limit */
1615 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
) {
1616 tsk
->rcv_win
= tsk_adv_blocks(tsk
->sk
.sk_rcvbuf
);
1617 msg_set_adv_win(msg
, tsk
->rcv_win
);
1619 tipc_node_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1622 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1624 struct sock
*sk
= sock
->sk
;
1626 long timeo
= *timeop
;
1627 int err
= sock_error(sk
);
1633 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1634 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1635 if (sk
->sk_shutdown
& RCV_SHUTDOWN
) {
1640 timeo
= schedule_timeout(timeo
);
1644 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1649 err
= sock_intr_errno(timeo
);
1650 if (signal_pending(current
))
1653 err
= sock_error(sk
);
1657 finish_wait(sk_sleep(sk
), &wait
);
1663 * tipc_recvmsg - receive packet-oriented message
1664 * @m: descriptor for message info
1665 * @buflen: length of user buffer area
1666 * @flags: receive flags
1668 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1669 * If the complete message doesn't fit in user area, truncate it.
1671 * Returns size of returned message data, errno otherwise
1673 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
,
1674 size_t buflen
, int flags
)
1676 struct sock
*sk
= sock
->sk
;
1677 bool connected
= !tipc_sk_type_connectionless(sk
);
1678 struct tipc_sock
*tsk
= tipc_sk(sk
);
1679 int rc
, err
, hlen
, dlen
, copy
;
1680 struct sk_buff_head xmitq
;
1681 struct tipc_msg
*hdr
;
1682 struct sk_buff
*skb
;
1686 /* Catch invalid receive requests */
1687 if (unlikely(!buflen
))
1691 if (unlikely(connected
&& sk
->sk_state
== TIPC_OPEN
)) {
1695 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1697 /* Step rcv queue to first msg with data or error; wait if necessary */
1699 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1702 skb
= skb_peek(&sk
->sk_receive_queue
);
1704 dlen
= msg_data_sz(hdr
);
1705 hlen
= msg_hdr_sz(hdr
);
1706 err
= msg_errcode(hdr
);
1707 grp_evt
= msg_is_grp_evt(hdr
);
1708 if (likely(dlen
|| err
))
1710 tsk_advance_rx_queue(sk
);
1713 /* Collect msg meta data, including error code and rejected data */
1714 tipc_sk_set_orig_addr(m
, skb
);
1715 rc
= tipc_sk_anc_data_recv(m
, hdr
, tsk
);
1719 /* Capture data if non-error msg, otherwise just set return value */
1721 copy
= min_t(int, dlen
, buflen
);
1722 if (unlikely(copy
!= dlen
))
1723 m
->msg_flags
|= MSG_TRUNC
;
1724 rc
= skb_copy_datagram_msg(skb
, hlen
, m
, copy
);
1728 if (err
!= TIPC_CONN_SHUTDOWN
&& connected
&& !m
->msg_control
)
1734 /* Mark message as group event if applicable */
1735 if (unlikely(grp_evt
)) {
1736 if (msg_grp_evt(hdr
) == TIPC_WITHDRAWN
)
1737 m
->msg_flags
|= MSG_EOR
;
1738 m
->msg_flags
|= MSG_OOB
;
1742 /* Caption of data or error code/rejected data was successful */
1743 if (unlikely(flags
& MSG_PEEK
))
1746 /* Send group flow control advertisement when applicable */
1747 if (tsk
->group
&& msg_in_group(hdr
) && !grp_evt
) {
1748 skb_queue_head_init(&xmitq
);
1749 tipc_group_update_rcv_win(tsk
->group
, tsk_blocks(hlen
+ dlen
),
1750 msg_orignode(hdr
), msg_origport(hdr
),
1752 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
1755 tsk_advance_rx_queue(sk
);
1757 if (likely(!connected
))
1760 /* Send connection flow control advertisement when applicable */
1761 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1762 if (tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
)
1763 tipc_sk_send_ack(tsk
);
1766 return rc
? rc
: copy
;
1770 * tipc_recvstream - receive stream-oriented data
1771 * @m: descriptor for message info
1772 * @buflen: total size of user buffer area
1773 * @flags: receive flags
1775 * Used for SOCK_STREAM messages only. If not enough data is available
1776 * will optionally wait for more; never truncates data.
1778 * Returns size of returned message data, errno otherwise
1780 static int tipc_recvstream(struct socket
*sock
, struct msghdr
*m
,
1781 size_t buflen
, int flags
)
1783 struct sock
*sk
= sock
->sk
;
1784 struct tipc_sock
*tsk
= tipc_sk(sk
);
1785 struct sk_buff
*skb
;
1786 struct tipc_msg
*hdr
;
1787 struct tipc_skb_cb
*skb_cb
;
1788 bool peek
= flags
& MSG_PEEK
;
1789 int offset
, required
, copy
, copied
= 0;
1790 int hlen
, dlen
, err
, rc
;
1793 /* Catch invalid receive attempts */
1794 if (unlikely(!buflen
))
1799 if (unlikely(sk
->sk_state
== TIPC_OPEN
)) {
1803 required
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buflen
);
1804 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1807 /* Look at first msg in receive queue; wait if necessary */
1808 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1811 skb
= skb_peek(&sk
->sk_receive_queue
);
1812 skb_cb
= TIPC_SKB_CB(skb
);
1814 dlen
= msg_data_sz(hdr
);
1815 hlen
= msg_hdr_sz(hdr
);
1816 err
= msg_errcode(hdr
);
1818 /* Discard any empty non-errored (SYN-) message */
1819 if (unlikely(!dlen
&& !err
)) {
1820 tsk_advance_rx_queue(sk
);
1824 /* Collect msg meta data, incl. error code and rejected data */
1826 tipc_sk_set_orig_addr(m
, skb
);
1827 rc
= tipc_sk_anc_data_recv(m
, hdr
, tsk
);
1832 /* Copy data if msg ok, otherwise return error/partial data */
1834 offset
= skb_cb
->bytes_read
;
1835 copy
= min_t(int, dlen
- offset
, buflen
- copied
);
1836 rc
= skb_copy_datagram_msg(skb
, hlen
+ offset
, m
, copy
);
1841 if (unlikely(offset
< dlen
)) {
1843 skb_cb
->bytes_read
= offset
;
1848 if ((err
!= TIPC_CONN_SHUTDOWN
) && !m
->msg_control
)
1857 tsk_advance_rx_queue(sk
);
1859 /* Send connection flow control advertisement when applicable */
1860 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1861 if (unlikely(tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
))
1862 tipc_sk_send_ack(tsk
);
1864 /* Exit if all requested data or FIN/error received */
1865 if (copied
== buflen
|| err
)
1868 } while (!skb_queue_empty(&sk
->sk_receive_queue
) || copied
< required
);
1871 return copied
? copied
: rc
;
1875 * tipc_write_space - wake up thread if port congestion is released
1878 static void tipc_write_space(struct sock
*sk
)
1880 struct socket_wq
*wq
;
1883 wq
= rcu_dereference(sk
->sk_wq
);
1884 if (skwq_has_sleeper(wq
))
1885 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1886 POLLWRNORM
| POLLWRBAND
);
1891 * tipc_data_ready - wake up threads to indicate messages have been received
1893 * @len: the length of messages
1895 static void tipc_data_ready(struct sock
*sk
)
1897 struct socket_wq
*wq
;
1900 wq
= rcu_dereference(sk
->sk_wq
);
1901 if (skwq_has_sleeper(wq
))
1902 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1903 POLLRDNORM
| POLLRDBAND
);
1907 static void tipc_sock_destruct(struct sock
*sk
)
1909 __skb_queue_purge(&sk
->sk_receive_queue
);
1912 static void tipc_sk_proto_rcv(struct sock
*sk
,
1913 struct sk_buff_head
*inputq
,
1914 struct sk_buff_head
*xmitq
)
1916 struct sk_buff
*skb
= __skb_dequeue(inputq
);
1917 struct tipc_sock
*tsk
= tipc_sk(sk
);
1918 struct tipc_msg
*hdr
= buf_msg(skb
);
1919 struct tipc_group
*grp
= tsk
->group
;
1920 bool wakeup
= false;
1922 switch (msg_user(hdr
)) {
1924 tipc_sk_conn_proto_rcv(tsk
, skb
, xmitq
);
1927 tipc_dest_del(&tsk
->cong_links
, msg_orignode(hdr
), 0);
1928 tsk
->cong_link_cnt
--;
1931 case GROUP_PROTOCOL
:
1932 tipc_group_proto_rcv(grp
, &wakeup
, hdr
, inputq
, xmitq
);
1935 tipc_group_member_evt(tsk
->group
, &wakeup
, &sk
->sk_rcvbuf
,
1936 skb
, inputq
, xmitq
);
1944 sk
->sk_write_space(sk
);
1950 * tipc_filter_connect - Handle incoming message for a connection-based socket
1952 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1954 * Returns true if everything ok, false otherwise
1956 static bool tipc_sk_filter_connect(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
1958 struct sock
*sk
= &tsk
->sk
;
1959 struct net
*net
= sock_net(sk
);
1960 struct tipc_msg
*hdr
= buf_msg(skb
);
1961 u32 pport
= msg_origport(hdr
);
1962 u32 pnode
= msg_orignode(hdr
);
1964 if (unlikely(msg_mcast(hdr
)))
1967 switch (sk
->sk_state
) {
1968 case TIPC_CONNECTING
:
1969 /* Accept only ACK or NACK message */
1970 if (unlikely(!msg_connected(hdr
))) {
1971 if (pport
!= tsk_peer_port(tsk
) ||
1972 pnode
!= tsk_peer_node(tsk
))
1975 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1976 sk
->sk_err
= ECONNREFUSED
;
1977 sk
->sk_state_change(sk
);
1981 if (unlikely(msg_errcode(hdr
))) {
1982 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1983 sk
->sk_err
= ECONNREFUSED
;
1984 sk
->sk_state_change(sk
);
1988 if (unlikely(!msg_isdata(hdr
))) {
1989 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1990 sk
->sk_err
= EINVAL
;
1991 sk
->sk_state_change(sk
);
1995 tipc_sk_finish_conn(tsk
, msg_origport(hdr
), msg_orignode(hdr
));
1996 msg_set_importance(&tsk
->phdr
, msg_importance(hdr
));
1998 /* If 'ACK+' message, add to socket receive queue */
1999 if (msg_data_sz(hdr
))
2002 /* If empty 'ACK-' message, wake up sleeping connect() */
2003 sk
->sk_data_ready(sk
);
2005 /* 'ACK-' message is neither accepted nor rejected: */
2006 msg_set_dest_droppable(hdr
, 1);
2010 case TIPC_DISCONNECTING
:
2013 /* Accept only SYN message */
2014 if (!msg_connected(hdr
) && !(msg_errcode(hdr
)))
2017 case TIPC_ESTABLISHED
:
2018 /* Accept only connection-based messages sent by peer */
2019 if (unlikely(!tsk_peer_msg(tsk
, hdr
)))
2022 if (unlikely(msg_errcode(hdr
))) {
2023 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2024 /* Let timer expire on it's own */
2025 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
2027 sk
->sk_state_change(sk
);
2031 pr_err("Unknown sk_state %u\n", sk
->sk_state
);
2038 * rcvbuf_limit - get proper overload limit of socket receive queue
2042 * For connection oriented messages, irrespective of importance,
2043 * default queue limit is 2 MB.
2045 * For connectionless messages, queue limits are based on message
2046 * importance as follows:
2048 * TIPC_LOW_IMPORTANCE (2 MB)
2049 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2050 * TIPC_HIGH_IMPORTANCE (8 MB)
2051 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2053 * Returns overload limit according to corresponding message importance
2055 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*skb
)
2057 struct tipc_sock
*tsk
= tipc_sk(sk
);
2058 struct tipc_msg
*hdr
= buf_msg(skb
);
2060 if (unlikely(msg_in_group(hdr
)))
2061 return sk
->sk_rcvbuf
;
2063 if (unlikely(!msg_connected(hdr
)))
2064 return sk
->sk_rcvbuf
<< msg_importance(hdr
);
2066 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
2067 return sk
->sk_rcvbuf
;
2069 return FLOWCTL_MSG_LIM
;
2073 * tipc_sk_filter_rcv - validate incoming message
2075 * @skb: pointer to message.
2077 * Enqueues message on receive queue if acceptable; optionally handles
2078 * disconnect indication for a connected socket.
2080 * Called with socket lock already taken
2083 static void tipc_sk_filter_rcv(struct sock
*sk
, struct sk_buff
*skb
,
2084 struct sk_buff_head
*xmitq
)
2086 bool sk_conn
= !tipc_sk_type_connectionless(sk
);
2087 struct tipc_sock
*tsk
= tipc_sk(sk
);
2088 struct tipc_group
*grp
= tsk
->group
;
2089 struct tipc_msg
*hdr
= buf_msg(skb
);
2090 struct net
*net
= sock_net(sk
);
2091 struct sk_buff_head inputq
;
2092 int limit
, err
= TIPC_OK
;
2094 TIPC_SKB_CB(skb
)->bytes_read
= 0;
2095 __skb_queue_head_init(&inputq
);
2096 __skb_queue_tail(&inputq
, skb
);
2098 if (unlikely(!msg_isdata(hdr
)))
2099 tipc_sk_proto_rcv(sk
, &inputq
, xmitq
);
2102 tipc_group_filter_msg(grp
, &inputq
, xmitq
);
2104 /* Validate and add to receive buffer if there is space */
2105 while ((skb
= __skb_dequeue(&inputq
))) {
2107 limit
= rcvbuf_limit(sk
, skb
);
2108 if ((sk_conn
&& !tipc_sk_filter_connect(tsk
, skb
)) ||
2109 (!sk_conn
&& msg_connected(hdr
)) ||
2110 (!grp
&& msg_in_group(hdr
)))
2111 err
= TIPC_ERR_NO_PORT
;
2112 else if (sk_rmem_alloc_get(sk
) + skb
->truesize
>= limit
)
2113 err
= TIPC_ERR_OVERLOAD
;
2115 if (unlikely(err
)) {
2116 tipc_skb_reject(net
, err
, skb
, xmitq
);
2120 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
2121 skb_set_owner_r(skb
, sk
);
2122 sk
->sk_data_ready(sk
);
2127 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2131 * Caller must hold socket lock
2133 static int tipc_sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
2135 unsigned int before
= sk_rmem_alloc_get(sk
);
2136 struct sk_buff_head xmitq
;
2139 __skb_queue_head_init(&xmitq
);
2141 tipc_sk_filter_rcv(sk
, skb
, &xmitq
);
2142 added
= sk_rmem_alloc_get(sk
) - before
;
2143 atomic_add(added
, &tipc_sk(sk
)->dupl_rcvcnt
);
2145 /* Send pending response/rejected messages, if any */
2146 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2151 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2152 * inputq and try adding them to socket or backlog queue
2153 * @inputq: list of incoming buffers with potentially different destinations
2154 * @sk: socket where the buffers should be enqueued
2155 * @dport: port number for the socket
2157 * Caller must hold socket lock
2159 static void tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
2160 u32 dport
, struct sk_buff_head
*xmitq
)
2162 unsigned long time_limit
= jiffies
+ 2;
2163 struct sk_buff
*skb
;
2168 while (skb_queue_len(inputq
)) {
2169 if (unlikely(time_after_eq(jiffies
, time_limit
)))
2172 skb
= tipc_skb_dequeue(inputq
, dport
);
2176 /* Add message directly to receive queue if possible */
2177 if (!sock_owned_by_user(sk
)) {
2178 tipc_sk_filter_rcv(sk
, skb
, xmitq
);
2182 /* Try backlog, compensating for double-counted bytes */
2183 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
2184 if (!sk
->sk_backlog
.len
)
2185 atomic_set(dcnt
, 0);
2186 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
2187 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
2190 /* Overload => reject message back to sender */
2191 onode
= tipc_own_addr(sock_net(sk
));
2192 if (tipc_msg_reverse(onode
, &skb
, TIPC_ERR_OVERLOAD
))
2193 __skb_queue_tail(xmitq
, skb
);
2199 * tipc_sk_rcv - handle a chain of incoming buffers
2200 * @inputq: buffer list containing the buffers
2201 * Consumes all buffers in list until inputq is empty
2202 * Note: may be called in multiple threads referring to the same queue
2204 void tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
2206 struct sk_buff_head xmitq
;
2207 u32 dnode
, dport
= 0;
2209 struct tipc_sock
*tsk
;
2211 struct sk_buff
*skb
;
2213 __skb_queue_head_init(&xmitq
);
2214 while (skb_queue_len(inputq
)) {
2215 dport
= tipc_skb_peek_port(inputq
, dport
);
2216 tsk
= tipc_sk_lookup(net
, dport
);
2220 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
2221 tipc_sk_enqueue(inputq
, sk
, dport
, &xmitq
);
2222 spin_unlock_bh(&sk
->sk_lock
.slock
);
2224 /* Send pending response/rejected messages, if any */
2225 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2229 /* No destination socket => dequeue skb if still there */
2230 skb
= tipc_skb_dequeue(inputq
, dport
);
2234 /* Try secondary lookup if unresolved named message */
2235 err
= TIPC_ERR_NO_PORT
;
2236 if (tipc_msg_lookup_dest(net
, skb
, &err
))
2239 /* Prepare for message rejection */
2240 if (!tipc_msg_reverse(tipc_own_addr(net
), &skb
, err
))
2243 dnode
= msg_destnode(buf_msg(skb
));
2244 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
2248 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
2250 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
2251 struct sock
*sk
= sock
->sk
;
2255 int err
= sock_error(sk
);
2260 if (signal_pending(current
))
2261 return sock_intr_errno(*timeo_p
);
2263 add_wait_queue(sk_sleep(sk
), &wait
);
2264 done
= sk_wait_event(sk
, timeo_p
,
2265 sk
->sk_state
!= TIPC_CONNECTING
, &wait
);
2266 remove_wait_queue(sk_sleep(sk
), &wait
);
2272 * tipc_connect - establish a connection to another TIPC port
2273 * @sock: socket structure
2274 * @dest: socket address for destination port
2275 * @destlen: size of socket address data structure
2276 * @flags: file-related flags associated with socket
2278 * Returns 0 on success, errno otherwise
2280 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
2281 int destlen
, int flags
)
2283 struct sock
*sk
= sock
->sk
;
2284 struct tipc_sock
*tsk
= tipc_sk(sk
);
2285 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
2286 struct msghdr m
= {NULL
,};
2287 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
2291 if (destlen
!= sizeof(struct sockaddr_tipc
))
2301 if (dst
->family
== AF_UNSPEC
) {
2302 memset(&tsk
->peer
, 0, sizeof(struct sockaddr_tipc
));
2303 if (!tipc_sk_type_connectionless(sk
))
2306 } else if (dst
->family
!= AF_TIPC
) {
2309 if (dst
->addrtype
!= TIPC_ADDR_ID
&& dst
->addrtype
!= TIPC_ADDR_NAME
)
2314 /* DGRAM/RDM connect(), just save the destaddr */
2315 if (tipc_sk_type_connectionless(sk
)) {
2316 memcpy(&tsk
->peer
, dest
, destlen
);
2320 previous
= sk
->sk_state
;
2322 switch (sk
->sk_state
) {
2324 /* Send a 'SYN-' to destination */
2326 m
.msg_namelen
= destlen
;
2328 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2329 * indicate send_msg() is never blocked.
2332 m
.msg_flags
= MSG_DONTWAIT
;
2334 res
= __tipc_sendmsg(sock
, &m
, 0);
2335 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
2338 /* Just entered TIPC_CONNECTING state; the only
2339 * difference is that return value in non-blocking
2340 * case is EINPROGRESS, rather than EALREADY.
2344 case TIPC_CONNECTING
:
2346 if (previous
== TIPC_CONNECTING
)
2350 timeout
= msecs_to_jiffies(timeout
);
2351 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2352 res
= tipc_wait_for_connect(sock
, &timeout
);
2354 case TIPC_ESTABLISHED
:
2367 * tipc_listen - allow socket to listen for incoming connections
2368 * @sock: socket structure
2371 * Returns 0 on success, errno otherwise
2373 static int tipc_listen(struct socket
*sock
, int len
)
2375 struct sock
*sk
= sock
->sk
;
2379 res
= tipc_set_sk_state(sk
, TIPC_LISTEN
);
2385 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
2387 struct sock
*sk
= sock
->sk
;
2391 /* True wake-one mechanism for incoming connections: only
2392 * one process gets woken up, not the 'whole herd'.
2393 * Since we do not 'race & poll' for established sockets
2394 * anymore, the common case will execute the loop only once.
2397 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
2398 TASK_INTERRUPTIBLE
);
2399 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
2401 timeo
= schedule_timeout(timeo
);
2405 if (!skb_queue_empty(&sk
->sk_receive_queue
))
2410 err
= sock_intr_errno(timeo
);
2411 if (signal_pending(current
))
2414 finish_wait(sk_sleep(sk
), &wait
);
2419 * tipc_accept - wait for connection request
2420 * @sock: listening socket
2421 * @newsock: new socket that is to be connected
2422 * @flags: file-related flags associated with socket
2424 * Returns 0 on success, errno otherwise
2426 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
,
2429 struct sock
*new_sk
, *sk
= sock
->sk
;
2430 struct sk_buff
*buf
;
2431 struct tipc_sock
*new_tsock
;
2432 struct tipc_msg
*msg
;
2438 if (sk
->sk_state
!= TIPC_LISTEN
) {
2442 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2443 res
= tipc_wait_for_accept(sock
, timeo
);
2447 buf
= skb_peek(&sk
->sk_receive_queue
);
2449 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, kern
);
2452 security_sk_clone(sock
->sk
, new_sock
->sk
);
2454 new_sk
= new_sock
->sk
;
2455 new_tsock
= tipc_sk(new_sk
);
2458 /* we lock on new_sk; but lockdep sees the lock on sk */
2459 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2462 * Reject any stray messages received by new socket
2463 * before the socket lock was taken (very, very unlikely)
2465 tsk_rej_rx_queue(new_sk
);
2467 /* Connect new socket to it's peer */
2468 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2470 tsk_set_importance(new_tsock
, msg_importance(msg
));
2471 if (msg_named(msg
)) {
2472 new_tsock
->conn_type
= msg_nametype(msg
);
2473 new_tsock
->conn_instance
= msg_nameinst(msg
);
2477 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2478 * Respond to 'SYN+' by queuing it on new socket.
2480 if (!msg_data_sz(msg
)) {
2481 struct msghdr m
= {NULL
,};
2483 tsk_advance_rx_queue(sk
);
2484 __tipc_sendstream(new_sock
, &m
, 0);
2486 __skb_dequeue(&sk
->sk_receive_queue
);
2487 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2488 skb_set_owner_r(buf
, new_sk
);
2490 release_sock(new_sk
);
2497 * tipc_shutdown - shutdown socket connection
2498 * @sock: socket structure
2499 * @how: direction to close (must be SHUT_RDWR)
2501 * Terminates connection (if necessary), then purges socket's receive queue.
2503 * Returns 0 on success, errno otherwise
2505 static int tipc_shutdown(struct socket
*sock
, int how
)
2507 struct sock
*sk
= sock
->sk
;
2510 if (how
!= SHUT_RDWR
)
2515 __tipc_shutdown(sock
, TIPC_CONN_SHUTDOWN
);
2516 sk
->sk_shutdown
= SEND_SHUTDOWN
;
2518 if (sk
->sk_state
== TIPC_DISCONNECTING
) {
2519 /* Discard any unreceived messages */
2520 __skb_queue_purge(&sk
->sk_receive_queue
);
2522 /* Wake up anyone sleeping in poll */
2523 sk
->sk_state_change(sk
);
2533 static void tipc_sk_timeout(struct timer_list
*t
)
2535 struct sock
*sk
= from_timer(sk
, t
, sk_timer
);
2536 struct tipc_sock
*tsk
= tipc_sk(sk
);
2537 u32 peer_port
= tsk_peer_port(tsk
);
2538 u32 peer_node
= tsk_peer_node(tsk
);
2539 u32 own_node
= tsk_own_node(tsk
);
2540 u32 own_port
= tsk
->portid
;
2541 struct net
*net
= sock_net(sk
);
2542 struct sk_buff
*skb
= NULL
;
2545 if (!tipc_sk_connected(sk
))
2548 /* Try again later if socket is busy */
2549 if (sock_owned_by_user(sk
)) {
2550 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ HZ
/ 20);
2554 if (tsk
->probe_unacked
) {
2555 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2556 tipc_node_remove_conn(net
, peer_node
, peer_port
);
2557 sk
->sk_state_change(sk
);
2560 /* Send new probe */
2561 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
, 0,
2562 peer_node
, own_node
, peer_port
, own_port
,
2564 tsk
->probe_unacked
= true;
2565 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
2569 tipc_node_xmit_skb(net
, skb
, peer_node
, own_port
);
2573 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2574 struct tipc_name_seq
const *seq
)
2576 struct sock
*sk
= &tsk
->sk
;
2577 struct net
*net
= sock_net(sk
);
2578 struct publication
*publ
;
2581 if (tipc_sk_connected(sk
))
2583 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2584 if (key
== tsk
->portid
)
2587 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2588 scope
, tsk
->portid
, key
);
2589 if (unlikely(!publ
))
2592 list_add(&publ
->pport_list
, &tsk
->publications
);
2598 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2599 struct tipc_name_seq
const *seq
)
2601 struct net
*net
= sock_net(&tsk
->sk
);
2602 struct publication
*publ
;
2603 struct publication
*safe
;
2606 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2608 if (publ
->scope
!= scope
)
2610 if (publ
->type
!= seq
->type
)
2612 if (publ
->lower
!= seq
->lower
)
2614 if (publ
->upper
!= seq
->upper
)
2616 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2617 publ
->ref
, publ
->key
);
2621 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2622 publ
->ref
, publ
->key
);
2625 if (list_empty(&tsk
->publications
))
2630 /* tipc_sk_reinit: set non-zero address in all existing sockets
2631 * when we go from standalone to network mode.
2633 void tipc_sk_reinit(struct net
*net
)
2635 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2636 struct rhashtable_iter iter
;
2637 struct tipc_sock
*tsk
;
2638 struct tipc_msg
*msg
;
2640 rhashtable_walk_enter(&tn
->sk_rht
, &iter
);
2643 tsk
= ERR_PTR(rhashtable_walk_start(&iter
));
2647 while ((tsk
= rhashtable_walk_next(&iter
)) && !IS_ERR(tsk
)) {
2648 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2650 msg_set_prevnode(msg
, tn
->own_addr
);
2651 msg_set_orignode(msg
, tn
->own_addr
);
2652 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2655 rhashtable_walk_stop(&iter
);
2656 } while (tsk
== ERR_PTR(-EAGAIN
));
2659 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2661 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2662 struct tipc_sock
*tsk
;
2665 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2667 sock_hold(&tsk
->sk
);
2673 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2675 struct sock
*sk
= &tsk
->sk
;
2676 struct net
*net
= sock_net(sk
);
2677 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2678 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2679 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2681 while (remaining
--) {
2683 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2684 portid
= TIPC_MIN_PORT
;
2685 tsk
->portid
= portid
;
2686 sock_hold(&tsk
->sk
);
2687 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2696 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2698 struct sock
*sk
= &tsk
->sk
;
2699 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2701 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2702 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
2707 static const struct rhashtable_params tsk_rht_params
= {
2709 .head_offset
= offsetof(struct tipc_sock
, node
),
2710 .key_offset
= offsetof(struct tipc_sock
, portid
),
2711 .key_len
= sizeof(u32
), /* portid */
2712 .max_size
= 1048576,
2714 .automatic_shrinking
= true,
2717 int tipc_sk_rht_init(struct net
*net
)
2719 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2721 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2724 void tipc_sk_rht_destroy(struct net
*net
)
2726 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2728 /* Wait for socket readers to complete */
2731 rhashtable_destroy(&tn
->sk_rht
);
2734 static int tipc_sk_join(struct tipc_sock
*tsk
, struct tipc_group_req
*mreq
)
2736 struct net
*net
= sock_net(&tsk
->sk
);
2737 u32 domain
= addr_domain(net
, mreq
->scope
);
2738 struct tipc_group
*grp
= tsk
->group
;
2739 struct tipc_msg
*hdr
= &tsk
->phdr
;
2740 struct tipc_name_seq seq
;
2743 if (mreq
->type
< TIPC_RESERVED_TYPES
)
2747 grp
= tipc_group_create(net
, tsk
->portid
, mreq
);
2751 msg_set_lookup_scope(hdr
, mreq
->scope
);
2752 msg_set_nametype(hdr
, mreq
->type
);
2753 msg_set_dest_droppable(hdr
, true);
2754 seq
.type
= mreq
->type
;
2755 seq
.lower
= mreq
->instance
;
2756 seq
.upper
= seq
.lower
;
2757 tipc_nametbl_build_group(net
, grp
, mreq
->type
, domain
);
2758 rc
= tipc_sk_publish(tsk
, mreq
->scope
, &seq
);
2760 tipc_group_delete(net
, grp
);
2764 /* Eliminate any risk that a broadcast overtakes the sent JOIN */
2765 tsk
->mc_method
.rcast
= true;
2766 tsk
->mc_method
.mandatory
= true;
2770 static int tipc_sk_leave(struct tipc_sock
*tsk
)
2772 struct net
*net
= sock_net(&tsk
->sk
);
2773 struct tipc_group
*grp
= tsk
->group
;
2774 struct tipc_name_seq seq
;
2779 tipc_group_self(grp
, &seq
, &scope
);
2780 tipc_group_delete(net
, grp
);
2782 tipc_sk_withdraw(tsk
, scope
, &seq
);
2787 * tipc_setsockopt - set socket option
2788 * @sock: socket structure
2789 * @lvl: option level
2790 * @opt: option identifier
2791 * @ov: pointer to new option value
2792 * @ol: length of option value
2794 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2795 * (to ease compatibility).
2797 * Returns 0 on success, errno otherwise
2799 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2800 char __user
*ov
, unsigned int ol
)
2802 struct sock
*sk
= sock
->sk
;
2803 struct tipc_sock
*tsk
= tipc_sk(sk
);
2804 struct tipc_group_req mreq
;
2808 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2810 if (lvl
!= SOL_TIPC
)
2811 return -ENOPROTOOPT
;
2814 case TIPC_IMPORTANCE
:
2815 case TIPC_SRC_DROPPABLE
:
2816 case TIPC_DEST_DROPPABLE
:
2817 case TIPC_CONN_TIMEOUT
:
2818 if (ol
< sizeof(value
))
2820 if (get_user(value
, (u32 __user
*)ov
))
2823 case TIPC_GROUP_JOIN
:
2824 if (ol
< sizeof(mreq
))
2826 if (copy_from_user(&mreq
, ov
, sizeof(mreq
)))
2837 case TIPC_IMPORTANCE
:
2838 res
= tsk_set_importance(tsk
, value
);
2840 case TIPC_SRC_DROPPABLE
:
2841 if (sock
->type
!= SOCK_STREAM
)
2842 tsk_set_unreliable(tsk
, value
);
2846 case TIPC_DEST_DROPPABLE
:
2847 tsk_set_unreturnable(tsk
, value
);
2849 case TIPC_CONN_TIMEOUT
:
2850 tipc_sk(sk
)->conn_timeout
= value
;
2852 case TIPC_MCAST_BROADCAST
:
2853 tsk
->mc_method
.rcast
= false;
2854 tsk
->mc_method
.mandatory
= true;
2856 case TIPC_MCAST_REPLICAST
:
2857 tsk
->mc_method
.rcast
= true;
2858 tsk
->mc_method
.mandatory
= true;
2860 case TIPC_GROUP_JOIN
:
2861 res
= tipc_sk_join(tsk
, &mreq
);
2863 case TIPC_GROUP_LEAVE
:
2864 res
= tipc_sk_leave(tsk
);
2876 * tipc_getsockopt - get socket option
2877 * @sock: socket structure
2878 * @lvl: option level
2879 * @opt: option identifier
2880 * @ov: receptacle for option value
2881 * @ol: receptacle for length of option value
2883 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2884 * (to ease compatibility).
2886 * Returns 0 on success, errno otherwise
2888 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2889 char __user
*ov
, int __user
*ol
)
2891 struct sock
*sk
= sock
->sk
;
2892 struct tipc_sock
*tsk
= tipc_sk(sk
);
2893 struct tipc_name_seq seq
;
2898 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2899 return put_user(0, ol
);
2900 if (lvl
!= SOL_TIPC
)
2901 return -ENOPROTOOPT
;
2902 res
= get_user(len
, ol
);
2909 case TIPC_IMPORTANCE
:
2910 value
= tsk_importance(tsk
);
2912 case TIPC_SRC_DROPPABLE
:
2913 value
= tsk_unreliable(tsk
);
2915 case TIPC_DEST_DROPPABLE
:
2916 value
= tsk_unreturnable(tsk
);
2918 case TIPC_CONN_TIMEOUT
:
2919 value
= tsk
->conn_timeout
;
2920 /* no need to set "res", since already 0 at this point */
2922 case TIPC_NODE_RECVQ_DEPTH
:
2923 value
= 0; /* was tipc_queue_size, now obsolete */
2925 case TIPC_SOCK_RECVQ_DEPTH
:
2926 value
= skb_queue_len(&sk
->sk_receive_queue
);
2928 case TIPC_GROUP_JOIN
:
2931 tipc_group_self(tsk
->group
, &seq
, &scope
);
2941 return res
; /* "get" failed */
2943 if (len
< sizeof(value
))
2946 if (copy_to_user(ov
, &value
, sizeof(value
)))
2949 return put_user(sizeof(value
), ol
);
2952 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2954 struct sock
*sk
= sock
->sk
;
2955 struct tipc_sioc_ln_req lnr
;
2956 void __user
*argp
= (void __user
*)arg
;
2959 case SIOCGETLINKNAME
:
2960 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2962 if (!tipc_node_get_linkname(sock_net(sk
),
2963 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2964 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2965 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2969 return -EADDRNOTAVAIL
;
2971 return -ENOIOCTLCMD
;
2975 static int tipc_socketpair(struct socket
*sock1
, struct socket
*sock2
)
2977 struct tipc_sock
*tsk2
= tipc_sk(sock2
->sk
);
2978 struct tipc_sock
*tsk1
= tipc_sk(sock1
->sk
);
2979 u32 onode
= tipc_own_addr(sock_net(sock1
->sk
));
2981 tsk1
->peer
.family
= AF_TIPC
;
2982 tsk1
->peer
.addrtype
= TIPC_ADDR_ID
;
2983 tsk1
->peer
.scope
= TIPC_NODE_SCOPE
;
2984 tsk1
->peer
.addr
.id
.ref
= tsk2
->portid
;
2985 tsk1
->peer
.addr
.id
.node
= onode
;
2986 tsk2
->peer
.family
= AF_TIPC
;
2987 tsk2
->peer
.addrtype
= TIPC_ADDR_ID
;
2988 tsk2
->peer
.scope
= TIPC_NODE_SCOPE
;
2989 tsk2
->peer
.addr
.id
.ref
= tsk1
->portid
;
2990 tsk2
->peer
.addr
.id
.node
= onode
;
2992 tipc_sk_finish_conn(tsk1
, tsk2
->portid
, onode
);
2993 tipc_sk_finish_conn(tsk2
, tsk1
->portid
, onode
);
2997 /* Protocol switches for the various types of TIPC sockets */
2999 static const struct proto_ops msg_ops
= {
3000 .owner
= THIS_MODULE
,
3002 .release
= tipc_release
,
3004 .connect
= tipc_connect
,
3005 .socketpair
= tipc_socketpair
,
3006 .accept
= sock_no_accept
,
3007 .getname
= tipc_getname
,
3009 .ioctl
= tipc_ioctl
,
3010 .listen
= sock_no_listen
,
3011 .shutdown
= tipc_shutdown
,
3012 .setsockopt
= tipc_setsockopt
,
3013 .getsockopt
= tipc_getsockopt
,
3014 .sendmsg
= tipc_sendmsg
,
3015 .recvmsg
= tipc_recvmsg
,
3016 .mmap
= sock_no_mmap
,
3017 .sendpage
= sock_no_sendpage
3020 static const struct proto_ops packet_ops
= {
3021 .owner
= THIS_MODULE
,
3023 .release
= tipc_release
,
3025 .connect
= tipc_connect
,
3026 .socketpair
= tipc_socketpair
,
3027 .accept
= tipc_accept
,
3028 .getname
= tipc_getname
,
3030 .ioctl
= tipc_ioctl
,
3031 .listen
= tipc_listen
,
3032 .shutdown
= tipc_shutdown
,
3033 .setsockopt
= tipc_setsockopt
,
3034 .getsockopt
= tipc_getsockopt
,
3035 .sendmsg
= tipc_send_packet
,
3036 .recvmsg
= tipc_recvmsg
,
3037 .mmap
= sock_no_mmap
,
3038 .sendpage
= sock_no_sendpage
3041 static const struct proto_ops stream_ops
= {
3042 .owner
= THIS_MODULE
,
3044 .release
= tipc_release
,
3046 .connect
= tipc_connect
,
3047 .socketpair
= tipc_socketpair
,
3048 .accept
= tipc_accept
,
3049 .getname
= tipc_getname
,
3051 .ioctl
= tipc_ioctl
,
3052 .listen
= tipc_listen
,
3053 .shutdown
= tipc_shutdown
,
3054 .setsockopt
= tipc_setsockopt
,
3055 .getsockopt
= tipc_getsockopt
,
3056 .sendmsg
= tipc_sendstream
,
3057 .recvmsg
= tipc_recvstream
,
3058 .mmap
= sock_no_mmap
,
3059 .sendpage
= sock_no_sendpage
3062 static const struct net_proto_family tipc_family_ops
= {
3063 .owner
= THIS_MODULE
,
3065 .create
= tipc_sk_create
3068 static struct proto tipc_proto
= {
3070 .owner
= THIS_MODULE
,
3071 .obj_size
= sizeof(struct tipc_sock
),
3072 .sysctl_rmem
= sysctl_tipc_rmem
3076 * tipc_socket_init - initialize TIPC socket interface
3078 * Returns 0 on success, errno otherwise
3080 int tipc_socket_init(void)
3084 res
= proto_register(&tipc_proto
, 1);
3086 pr_err("Failed to register TIPC protocol type\n");
3090 res
= sock_register(&tipc_family_ops
);
3092 pr_err("Failed to register TIPC socket type\n");
3093 proto_unregister(&tipc_proto
);
3101 * tipc_socket_stop - stop TIPC socket interface
3103 void tipc_socket_stop(void)
3105 sock_unregister(tipc_family_ops
.family
);
3106 proto_unregister(&tipc_proto
);
3109 /* Caller should hold socket lock for the passed tipc socket. */
3110 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
3114 struct nlattr
*nest
;
3116 peer_node
= tsk_peer_node(tsk
);
3117 peer_port
= tsk_peer_port(tsk
);
3119 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
3121 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
3123 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
3126 if (tsk
->conn_type
!= 0) {
3127 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
3129 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
3131 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
3134 nla_nest_end(skb
, nest
);
3139 nla_nest_cancel(skb
, nest
);
3144 /* Caller should hold socket lock for the passed tipc socket. */
3145 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
3146 struct tipc_sock
*tsk
)
3150 struct nlattr
*attrs
;
3151 struct net
*net
= sock_net(skb
->sk
);
3152 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
3153 struct sock
*sk
= &tsk
->sk
;
3155 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3156 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
3160 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
3162 goto genlmsg_cancel
;
3163 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
3164 goto attr_msg_cancel
;
3165 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tn
->own_addr
))
3166 goto attr_msg_cancel
;
3168 if (tipc_sk_connected(sk
)) {
3169 err
= __tipc_nl_add_sk_con(skb
, tsk
);
3171 goto attr_msg_cancel
;
3172 } else if (!list_empty(&tsk
->publications
)) {
3173 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
3174 goto attr_msg_cancel
;
3176 nla_nest_end(skb
, attrs
);
3177 genlmsg_end(skb
, hdr
);
3182 nla_nest_cancel(skb
, attrs
);
3184 genlmsg_cancel(skb
, hdr
);
3189 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3192 struct tipc_sock
*tsk
;
3193 const struct bucket_table
*tbl
;
3194 struct rhash_head
*pos
;
3195 struct net
*net
= sock_net(skb
->sk
);
3196 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
3197 u32 tbl_id
= cb
->args
[0];
3198 u32 prev_portid
= cb
->args
[1];
3201 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
3202 for (; tbl_id
< tbl
->size
; tbl_id
++) {
3203 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
3204 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
3205 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
3206 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3210 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
3212 prev_portid
= tsk
->portid
;
3213 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3217 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3222 cb
->args
[0] = tbl_id
;
3223 cb
->args
[1] = prev_portid
;
3228 /* Caller should hold socket lock for the passed tipc socket. */
3229 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
3230 struct netlink_callback
*cb
,
3231 struct publication
*publ
)
3234 struct nlattr
*attrs
;
3236 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3237 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
3241 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
3243 goto genlmsg_cancel
;
3245 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
3246 goto attr_msg_cancel
;
3247 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
3248 goto attr_msg_cancel
;
3249 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
3250 goto attr_msg_cancel
;
3251 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
3252 goto attr_msg_cancel
;
3254 nla_nest_end(skb
, attrs
);
3255 genlmsg_end(skb
, hdr
);
3260 nla_nest_cancel(skb
, attrs
);
3262 genlmsg_cancel(skb
, hdr
);
3267 /* Caller should hold socket lock for the passed tipc socket. */
3268 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
3269 struct netlink_callback
*cb
,
3270 struct tipc_sock
*tsk
, u32
*last_publ
)
3273 struct publication
*p
;
3276 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
3277 if (p
->key
== *last_publ
)
3280 if (p
->key
!= *last_publ
) {
3281 /* We never set seq or call nl_dump_check_consistent()
3282 * this means that setting prev_seq here will cause the
3283 * consistence check to fail in the netlink callback
3284 * handler. Resulting in the last NLMSG_DONE message
3285 * having the NLM_F_DUMP_INTR flag set.
3292 p
= list_first_entry(&tsk
->publications
, struct publication
,
3296 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
3297 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
3299 *last_publ
= p
->key
;
3308 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3311 u32 tsk_portid
= cb
->args
[0];
3312 u32 last_publ
= cb
->args
[1];
3313 u32 done
= cb
->args
[2];
3314 struct net
*net
= sock_net(skb
->sk
);
3315 struct tipc_sock
*tsk
;
3318 struct nlattr
**attrs
;
3319 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
3321 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
3325 if (!attrs
[TIPC_NLA_SOCK
])
3328 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
3329 attrs
[TIPC_NLA_SOCK
],
3330 tipc_nl_sock_policy
, NULL
);
3334 if (!sock
[TIPC_NLA_SOCK_REF
])
3337 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
3343 tsk
= tipc_sk_lookup(net
, tsk_portid
);
3347 lock_sock(&tsk
->sk
);
3348 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
3351 release_sock(&tsk
->sk
);
3354 cb
->args
[0] = tsk_portid
;
3355 cb
->args
[1] = last_publ
;