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
);
581 /* Reject any messages that accumulated in backlog queue */
583 tipc_dest_list_purge(&tsk
->cong_links
);
584 tsk
->cong_link_cnt
= 0;
585 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
592 * tipc_bind - associate or disassocate TIPC name(s) with a socket
593 * @sock: socket structure
594 * @uaddr: socket address describing name(s) and desired operation
595 * @uaddr_len: size of socket address data structure
597 * Name and name sequence binding is indicated using a positive scope value;
598 * a negative scope value unbinds the specified name. Specifying no name
599 * (i.e. a socket address length of 0) unbinds all names from the socket.
601 * Returns 0 on success, errno otherwise
603 * NOTE: This routine doesn't need to take the socket lock since it doesn't
604 * access any non-constant socket information.
606 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
609 struct sock
*sk
= sock
->sk
;
610 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
611 struct tipc_sock
*tsk
= tipc_sk(sk
);
615 if (unlikely(!uaddr_len
)) {
616 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
623 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
627 if (addr
->family
!= AF_TIPC
) {
632 if (addr
->addrtype
== TIPC_ADDR_NAME
)
633 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
634 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
639 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
640 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
641 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
646 res
= (addr
->scope
> 0) ?
647 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
648 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
655 * tipc_getname - get port ID of socket or peer socket
656 * @sock: socket structure
657 * @uaddr: area for returned socket address
658 * @uaddr_len: area for returned length of socket address
659 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
661 * Returns 0 on success, errno otherwise
663 * NOTE: This routine doesn't need to take the socket lock since it only
664 * accesses socket information that is unchanging (or which changes in
665 * a completely predictable manner).
667 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
668 int *uaddr_len
, int peer
)
670 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
671 struct sock
*sk
= sock
->sk
;
672 struct tipc_sock
*tsk
= tipc_sk(sk
);
673 struct tipc_net
*tn
= net_generic(sock_net(sock
->sk
), tipc_net_id
);
675 memset(addr
, 0, sizeof(*addr
));
677 if ((!tipc_sk_connected(sk
)) &&
678 ((peer
!= 2) || (sk
->sk_state
!= TIPC_DISCONNECTING
)))
680 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
681 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
683 addr
->addr
.id
.ref
= tsk
->portid
;
684 addr
->addr
.id
.node
= tn
->own_addr
;
687 *uaddr_len
= sizeof(*addr
);
688 addr
->addrtype
= TIPC_ADDR_ID
;
689 addr
->family
= AF_TIPC
;
691 addr
->addr
.name
.domain
= 0;
697 * tipc_poll - read and possibly block on pollmask
698 * @file: file structure associated with the socket
699 * @sock: socket for which to calculate the poll bits
702 * Returns pollmask value
705 * It appears that the usual socket locking mechanisms are not useful here
706 * since the pollmask info is potentially out-of-date the moment this routine
707 * exits. TCP and other protocols seem to rely on higher level poll routines
708 * to handle any preventable race conditions, so TIPC will do the same ...
710 * IMPORTANT: The fact that a read or write operation is indicated does NOT
711 * imply that the operation will succeed, merely that it should be performed
712 * and will not block.
714 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
717 struct sock
*sk
= sock
->sk
;
718 struct tipc_sock
*tsk
= tipc_sk(sk
);
719 struct tipc_group
*grp
= tsk
->group
;
722 sock_poll_wait(file
, sk_sleep(sk
), wait
);
724 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
725 revents
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
726 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
729 switch (sk
->sk_state
) {
730 case TIPC_ESTABLISHED
:
731 case TIPC_CONNECTING
:
732 if (!tsk
->cong_link_cnt
&& !tsk_conn_cong(tsk
))
736 if (!skb_queue_empty(&sk
->sk_receive_queue
))
737 revents
|= POLLIN
| POLLRDNORM
;
740 if (!grp
|| tipc_group_size(grp
))
741 if (!tsk
->cong_link_cnt
)
743 if (!tipc_sk_type_connectionless(sk
))
745 if (skb_queue_empty(&sk
->sk_receive_queue
))
747 revents
|= POLLIN
| POLLRDNORM
;
749 case TIPC_DISCONNECTING
:
750 revents
= POLLIN
| POLLRDNORM
| POLLHUP
;
757 * tipc_sendmcast - send multicast message
758 * @sock: socket structure
759 * @seq: destination address
760 * @msg: message to send
761 * @dlen: length of data to send
762 * @timeout: timeout to wait for wakeup
764 * Called from function tipc_sendmsg(), which has done all sanity checks
765 * Returns the number of bytes sent on success, or errno
767 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
768 struct msghdr
*msg
, size_t dlen
, long timeout
)
770 struct sock
*sk
= sock
->sk
;
771 struct tipc_sock
*tsk
= tipc_sk(sk
);
772 struct tipc_msg
*hdr
= &tsk
->phdr
;
773 struct net
*net
= sock_net(sk
);
774 int mtu
= tipc_bcast_get_mtu(net
);
775 struct tipc_mc_method
*method
= &tsk
->mc_method
;
776 u32 domain
= addr_domain(net
, TIPC_CLUSTER_SCOPE
);
777 struct sk_buff_head pkts
;
778 struct tipc_nlist dsts
;
784 /* Block or return if any destination link is congested */
785 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
);
789 /* Lookup destination nodes */
790 tipc_nlist_init(&dsts
, tipc_own_addr(net
));
791 tipc_nametbl_lookup_dst_nodes(net
, seq
->type
, seq
->lower
,
792 seq
->upper
, domain
, &dsts
);
793 if (!dsts
.local
&& !dsts
.remote
)
794 return -EHOSTUNREACH
;
796 /* Build message header */
797 msg_set_type(hdr
, TIPC_MCAST_MSG
);
798 msg_set_hdr_sz(hdr
, MCAST_H_SIZE
);
799 msg_set_lookup_scope(hdr
, TIPC_CLUSTER_SCOPE
);
800 msg_set_destport(hdr
, 0);
801 msg_set_destnode(hdr
, 0);
802 msg_set_nametype(hdr
, seq
->type
);
803 msg_set_namelower(hdr
, seq
->lower
);
804 msg_set_nameupper(hdr
, seq
->upper
);
806 /* Build message as chain of buffers */
807 skb_queue_head_init(&pkts
);
808 rc
= tipc_msg_build(hdr
, msg
, 0, dlen
, mtu
, &pkts
);
810 /* Send message if build was successful */
811 if (unlikely(rc
== dlen
))
812 rc
= tipc_mcast_xmit(net
, &pkts
, method
, &dsts
,
813 &tsk
->cong_link_cnt
);
815 tipc_nlist_purge(&dsts
);
817 return rc
? rc
: dlen
;
821 * tipc_send_group_msg - send a message to a member in the group
822 * @net: network namespace
823 * @m: message to send
825 * @dnode: destination node
826 * @dport: destination port
827 * @dlen: total length of message data
829 static int tipc_send_group_msg(struct net
*net
, struct tipc_sock
*tsk
,
830 struct msghdr
*m
, struct tipc_member
*mb
,
831 u32 dnode
, u32 dport
, int dlen
)
833 u16 bc_snd_nxt
= tipc_group_bc_snd_nxt(tsk
->group
);
834 struct tipc_mc_method
*method
= &tsk
->mc_method
;
835 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
836 struct tipc_msg
*hdr
= &tsk
->phdr
;
837 struct sk_buff_head pkts
;
840 /* Complete message header */
841 msg_set_type(hdr
, TIPC_GRP_UCAST_MSG
);
842 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
843 msg_set_destport(hdr
, dport
);
844 msg_set_destnode(hdr
, dnode
);
845 msg_set_grp_bc_seqno(hdr
, bc_snd_nxt
);
847 /* Build message as chain of buffers */
848 skb_queue_head_init(&pkts
);
849 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
850 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
851 if (unlikely(rc
!= dlen
))
855 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
856 if (unlikely(rc
== -ELINKCONG
)) {
857 tipc_dest_push(&tsk
->cong_links
, dnode
, 0);
858 tsk
->cong_link_cnt
++;
861 /* Update send window */
862 tipc_group_update_member(mb
, blks
);
864 /* A broadcast sent within next EXPIRE period must follow same path */
865 method
->rcast
= true;
866 method
->mandatory
= true;
871 * tipc_send_group_unicast - send message to a member in the group
872 * @sock: socket structure
873 * @m: message to send
874 * @dlen: total length of message data
875 * @timeout: timeout to wait for wakeup
877 * Called from function tipc_sendmsg(), which has done all sanity checks
878 * Returns the number of bytes sent on success, or errno
880 static int tipc_send_group_unicast(struct socket
*sock
, struct msghdr
*m
,
881 int dlen
, long timeout
)
883 struct sock
*sk
= sock
->sk
;
884 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
885 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
886 struct tipc_sock
*tsk
= tipc_sk(sk
);
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) &&
901 !tipc_group_cong(tsk
->group
, node
, port
, blks
,
907 return -EHOSTUNREACH
;
909 rc
= tipc_send_group_msg(net
, tsk
, m
, mb
, node
, port
, dlen
);
911 return rc
? rc
: dlen
;
915 * tipc_send_group_anycast - send message to any member with given identity
916 * @sock: socket structure
917 * @m: message to send
918 * @dlen: total length of message data
919 * @timeout: timeout to wait for wakeup
921 * Called from function tipc_sendmsg(), which has done all sanity checks
922 * Returns the number of bytes sent on success, or errno
924 static int tipc_send_group_anycast(struct socket
*sock
, struct msghdr
*m
,
925 int dlen
, long timeout
)
927 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
928 struct sock
*sk
= sock
->sk
;
929 struct tipc_sock
*tsk
= tipc_sk(sk
);
930 struct list_head
*cong_links
= &tsk
->cong_links
;
931 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
932 struct tipc_member
*first
= NULL
;
933 struct tipc_member
*mbr
= NULL
;
934 struct net
*net
= sock_net(sk
);
935 u32 node
, port
, exclude
;
936 u32 type
, inst
, domain
;
937 struct list_head dsts
;
942 INIT_LIST_HEAD(&dsts
);
944 type
= dest
->addr
.name
.name
.type
;
945 inst
= dest
->addr
.name
.name
.instance
;
946 domain
= addr_domain(net
, dest
->scope
);
948 while (++lookups
< 4) {
949 exclude
= tipc_group_exclude(tsk
->group
);
953 /* Look for a non-congested destination member, if any */
955 if (!tipc_nametbl_lookup(net
, type
, inst
, domain
, &dsts
,
956 &dstcnt
, exclude
, false))
957 return -EHOSTUNREACH
;
958 tipc_dest_pop(&dsts
, &node
, &port
);
959 cong
= tipc_group_cong(tsk
->group
, node
, port
, blks
,
969 /* Start over if destination was not in member list */
973 if (likely(!cong
&& !tipc_dest_find(cong_links
, node
, 0)))
976 /* Block or return if destination link or member is congested */
977 rc
= tipc_wait_for_cond(sock
, &timeout
,
978 !tipc_dest_find(cong_links
, node
, 0) &&
980 !tipc_group_cong(tsk
->group
, node
, port
,
985 /* Send, unless destination disappeared while waiting */
990 if (unlikely(lookups
>= 4))
991 return -EHOSTUNREACH
;
993 rc
= tipc_send_group_msg(net
, tsk
, m
, mbr
, node
, port
, dlen
);
995 return rc
? rc
: dlen
;
999 * tipc_send_group_bcast - send message to all members in communication group
1000 * @sk: socket structure
1001 * @m: message to send
1002 * @dlen: total length of message data
1003 * @timeout: timeout to wait for wakeup
1005 * Called from function tipc_sendmsg(), which has done all sanity checks
1006 * Returns the number of bytes sent on success, or errno
1008 static int tipc_send_group_bcast(struct socket
*sock
, struct msghdr
*m
,
1009 int dlen
, long timeout
)
1011 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1012 struct sock
*sk
= sock
->sk
;
1013 struct net
*net
= sock_net(sk
);
1014 struct tipc_sock
*tsk
= tipc_sk(sk
);
1015 struct tipc_nlist
*dsts
;
1016 struct tipc_mc_method
*method
= &tsk
->mc_method
;
1017 bool ack
= method
->mandatory
&& method
->rcast
;
1018 int blks
= tsk_blocks(MCAST_H_SIZE
+ dlen
);
1019 struct tipc_msg
*hdr
= &tsk
->phdr
;
1020 int mtu
= tipc_bcast_get_mtu(net
);
1021 struct sk_buff_head pkts
;
1022 int rc
= -EHOSTUNREACH
;
1024 /* Block or return if any destination link or member is congested */
1025 rc
= tipc_wait_for_cond(sock
, &timeout
,
1026 !tsk
->cong_link_cnt
&& tsk
->group
&&
1027 !tipc_group_bc_cong(tsk
->group
, blks
));
1031 dsts
= tipc_group_dests(tsk
->group
);
1032 if (!dsts
->local
&& !dsts
->remote
)
1033 return -EHOSTUNREACH
;
1035 /* Complete message header */
1037 msg_set_type(hdr
, TIPC_GRP_MCAST_MSG
);
1038 msg_set_nameinst(hdr
, dest
->addr
.name
.name
.instance
);
1040 msg_set_type(hdr
, TIPC_GRP_BCAST_MSG
);
1041 msg_set_nameinst(hdr
, 0);
1043 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
1044 msg_set_destport(hdr
, 0);
1045 msg_set_destnode(hdr
, 0);
1046 msg_set_grp_bc_seqno(hdr
, tipc_group_bc_snd_nxt(tsk
->group
));
1048 /* Avoid getting stuck with repeated forced replicasts */
1049 msg_set_grp_bc_ack_req(hdr
, ack
);
1051 /* Build message as chain of buffers */
1052 skb_queue_head_init(&pkts
);
1053 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1054 if (unlikely(rc
!= dlen
))
1058 rc
= tipc_mcast_xmit(net
, &pkts
, method
, dsts
, &tsk
->cong_link_cnt
);
1062 /* Update broadcast sequence number and send windows */
1063 tipc_group_update_bc_members(tsk
->group
, blks
, ack
);
1065 /* Broadcast link is now free to choose method for next broadcast */
1066 method
->mandatory
= false;
1067 method
->expires
= jiffies
;
1073 * tipc_send_group_mcast - send message to all members with given identity
1074 * @sock: socket structure
1075 * @m: message to send
1076 * @dlen: total length of message data
1077 * @timeout: timeout to wait for wakeup
1079 * Called from function tipc_sendmsg(), which has done all sanity checks
1080 * Returns the number of bytes sent on success, or errno
1082 static int tipc_send_group_mcast(struct socket
*sock
, struct msghdr
*m
,
1083 int dlen
, long timeout
)
1085 struct sock
*sk
= sock
->sk
;
1086 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1087 struct tipc_name_seq
*seq
= &dest
->addr
.nameseq
;
1088 struct tipc_sock
*tsk
= tipc_sk(sk
);
1089 struct tipc_group
*grp
= tsk
->group
;
1090 struct net
*net
= sock_net(sk
);
1091 u32 domain
, exclude
, dstcnt
;
1092 struct list_head dsts
;
1094 INIT_LIST_HEAD(&dsts
);
1096 if (seq
->lower
!= seq
->upper
)
1099 domain
= addr_domain(net
, dest
->scope
);
1100 exclude
= tipc_group_exclude(grp
);
1101 if (!tipc_nametbl_lookup(net
, seq
->type
, seq
->lower
, domain
,
1102 &dsts
, &dstcnt
, exclude
, true))
1103 return -EHOSTUNREACH
;
1106 tipc_dest_pop(&dsts
, &dest
->addr
.id
.node
, &dest
->addr
.id
.ref
);
1107 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1110 tipc_dest_list_purge(&dsts
);
1111 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1115 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1116 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1117 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1119 * Multi-threaded: parallel calls with reference to same queues may occur
1121 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
1122 struct sk_buff_head
*inputq
)
1124 u32 scope
= TIPC_CLUSTER_SCOPE
;
1125 u32 self
= tipc_own_addr(net
);
1126 struct sk_buff
*skb
, *_skb
;
1127 u32 lower
= 0, upper
= ~0;
1128 struct sk_buff_head tmpq
;
1129 u32 portid
, oport
, onode
;
1130 struct list_head dports
;
1131 struct tipc_msg
*msg
;
1132 int user
, mtyp
, hsz
;
1134 __skb_queue_head_init(&tmpq
);
1135 INIT_LIST_HEAD(&dports
);
1137 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
1138 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
1140 user
= msg_user(msg
);
1141 mtyp
= msg_type(msg
);
1142 if (mtyp
== TIPC_GRP_UCAST_MSG
|| user
== GROUP_PROTOCOL
) {
1143 spin_lock_bh(&inputq
->lock
);
1144 if (skb_peek(arrvq
) == skb
) {
1145 __skb_dequeue(arrvq
);
1146 __skb_queue_tail(inputq
, skb
);
1149 spin_unlock_bh(&inputq
->lock
);
1152 hsz
= skb_headroom(skb
) + msg_hdr_sz(msg
);
1153 oport
= msg_origport(msg
);
1154 onode
= msg_orignode(msg
);
1156 scope
= TIPC_NODE_SCOPE
;
1158 /* Create destination port list and message clones: */
1159 if (!msg_in_group(msg
)) {
1160 lower
= msg_namelower(msg
);
1161 upper
= msg_nameupper(msg
);
1163 tipc_nametbl_mc_translate(net
, msg_nametype(msg
), lower
, upper
,
1165 while (tipc_dest_pop(&dports
, NULL
, &portid
)) {
1166 _skb
= __pskb_copy(skb
, hsz
, GFP_ATOMIC
);
1168 msg_set_destport(buf_msg(_skb
), portid
);
1169 __skb_queue_tail(&tmpq
, _skb
);
1172 pr_warn("Failed to clone mcast rcv buffer\n");
1174 /* Append to inputq if not already done by other thread */
1175 spin_lock_bh(&inputq
->lock
);
1176 if (skb_peek(arrvq
) == skb
) {
1177 skb_queue_splice_tail_init(&tmpq
, inputq
);
1178 kfree_skb(__skb_dequeue(arrvq
));
1180 spin_unlock_bh(&inputq
->lock
);
1181 __skb_queue_purge(&tmpq
);
1184 tipc_sk_rcv(net
, inputq
);
1188 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1189 * @tsk: receiving socket
1190 * @skb: pointer to message buffer.
1192 static void tipc_sk_conn_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
*skb
,
1193 struct sk_buff_head
*xmitq
)
1195 struct tipc_msg
*hdr
= buf_msg(skb
);
1196 u32 onode
= tsk_own_node(tsk
);
1197 struct sock
*sk
= &tsk
->sk
;
1198 int mtyp
= msg_type(hdr
);
1201 /* Ignore if connection cannot be validated: */
1202 if (!tsk_peer_msg(tsk
, hdr
))
1205 if (unlikely(msg_errcode(hdr
))) {
1206 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1207 tipc_node_remove_conn(sock_net(sk
), tsk_peer_node(tsk
),
1208 tsk_peer_port(tsk
));
1209 sk
->sk_state_change(sk
);
1213 tsk
->probe_unacked
= false;
1215 if (mtyp
== CONN_PROBE
) {
1216 msg_set_type(hdr
, CONN_PROBE_REPLY
);
1217 if (tipc_msg_reverse(onode
, &skb
, TIPC_OK
))
1218 __skb_queue_tail(xmitq
, skb
);
1220 } else if (mtyp
== CONN_ACK
) {
1221 conn_cong
= tsk_conn_cong(tsk
);
1222 tsk
->snt_unacked
-= msg_conn_ack(hdr
);
1223 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1224 tsk
->snd_win
= msg_adv_win(hdr
);
1226 sk
->sk_write_space(sk
);
1227 } else if (mtyp
!= CONN_PROBE_REPLY
) {
1228 pr_warn("Received unknown CONN_PROTO msg\n");
1235 * tipc_sendmsg - send message in connectionless manner
1236 * @sock: socket structure
1237 * @m: message to send
1238 * @dsz: amount of user data to be sent
1240 * Message must have an destination specified explicitly.
1241 * Used for SOCK_RDM and SOCK_DGRAM messages,
1242 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1243 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1245 * Returns the number of bytes sent on success, or errno otherwise
1247 static int tipc_sendmsg(struct socket
*sock
,
1248 struct msghdr
*m
, size_t dsz
)
1250 struct sock
*sk
= sock
->sk
;
1254 ret
= __tipc_sendmsg(sock
, m
, dsz
);
1260 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1262 struct sock
*sk
= sock
->sk
;
1263 struct net
*net
= sock_net(sk
);
1264 struct tipc_sock
*tsk
= tipc_sk(sk
);
1265 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1266 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1267 struct list_head
*clinks
= &tsk
->cong_links
;
1268 bool syn
= !tipc_sk_type_connectionless(sk
);
1269 struct tipc_group
*grp
= tsk
->group
;
1270 struct tipc_msg
*hdr
= &tsk
->phdr
;
1271 struct tipc_name_seq
*seq
;
1272 struct sk_buff_head pkts
;
1273 u32 type
, inst
, domain
;
1277 if (unlikely(dlen
> TIPC_MAX_USER_MSG_SIZE
))
1281 if (unlikely(m
->msg_namelen
< sizeof(*dest
)))
1283 if (unlikely(dest
->family
!= AF_TIPC
))
1289 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1290 if (dest
->addrtype
== TIPC_ADDR_NAME
)
1291 return tipc_send_group_anycast(sock
, m
, dlen
, timeout
);
1292 if (dest
->addrtype
== TIPC_ADDR_ID
)
1293 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1294 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1295 return tipc_send_group_mcast(sock
, m
, dlen
, timeout
);
1299 if (unlikely(!dest
)) {
1301 if (!syn
&& dest
->family
!= AF_TIPC
)
1302 return -EDESTADDRREQ
;
1305 if (unlikely(syn
)) {
1306 if (sk
->sk_state
== TIPC_LISTEN
)
1308 if (sk
->sk_state
!= TIPC_OPEN
)
1312 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1313 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
1314 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
1318 seq
= &dest
->addr
.nameseq
;
1319 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1320 return tipc_sendmcast(sock
, seq
, m
, dlen
, timeout
);
1322 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1323 type
= dest
->addr
.name
.name
.type
;
1324 inst
= dest
->addr
.name
.name
.instance
;
1325 domain
= dest
->addr
.name
.domain
;
1327 msg_set_type(hdr
, TIPC_NAMED_MSG
);
1328 msg_set_hdr_sz(hdr
, NAMED_H_SIZE
);
1329 msg_set_nametype(hdr
, type
);
1330 msg_set_nameinst(hdr
, inst
);
1331 msg_set_lookup_scope(hdr
, tipc_addr_scope(domain
));
1332 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
1333 msg_set_destnode(hdr
, dnode
);
1334 msg_set_destport(hdr
, dport
);
1335 if (unlikely(!dport
&& !dnode
))
1336 return -EHOSTUNREACH
;
1337 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
1338 dnode
= dest
->addr
.id
.node
;
1339 msg_set_type(hdr
, TIPC_DIRECT_MSG
);
1340 msg_set_lookup_scope(hdr
, 0);
1341 msg_set_destnode(hdr
, dnode
);
1342 msg_set_destport(hdr
, dest
->addr
.id
.ref
);
1343 msg_set_hdr_sz(hdr
, BASIC_H_SIZE
);
1346 /* Block or return if destination link is congested */
1347 rc
= tipc_wait_for_cond(sock
, &timeout
,
1348 !tipc_dest_find(clinks
, dnode
, 0));
1352 skb_queue_head_init(&pkts
);
1353 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
1354 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1355 if (unlikely(rc
!= dlen
))
1358 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1359 if (unlikely(rc
== -ELINKCONG
)) {
1360 tipc_dest_push(clinks
, dnode
, 0);
1361 tsk
->cong_link_cnt
++;
1365 if (unlikely(syn
&& !rc
))
1366 tipc_set_sk_state(sk
, TIPC_CONNECTING
);
1368 return rc
? rc
: dlen
;
1372 * tipc_sendstream - send stream-oriented data
1373 * @sock: socket structure
1375 * @dsz: total length of data to be transmitted
1377 * Used for SOCK_STREAM data.
1379 * Returns the number of bytes sent on success (or partial success),
1380 * or errno if no data sent
1382 static int tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1384 struct sock
*sk
= sock
->sk
;
1388 ret
= __tipc_sendstream(sock
, m
, dsz
);
1394 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1396 struct sock
*sk
= sock
->sk
;
1397 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1398 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1399 struct tipc_sock
*tsk
= tipc_sk(sk
);
1400 struct tipc_msg
*hdr
= &tsk
->phdr
;
1401 struct net
*net
= sock_net(sk
);
1402 struct sk_buff_head pkts
;
1403 u32 dnode
= tsk_peer_node(tsk
);
1407 skb_queue_head_init(&pkts
);
1409 if (unlikely(dlen
> INT_MAX
))
1412 /* Handle implicit connection setup */
1413 if (unlikely(dest
)) {
1414 rc
= __tipc_sendmsg(sock
, m
, dlen
);
1415 if (dlen
&& dlen
== rc
) {
1416 tsk
->peer_caps
= tipc_node_get_capabilities(net
, dnode
);
1417 tsk
->snt_unacked
= tsk_inc(tsk
, dlen
+ msg_hdr_sz(hdr
));
1423 rc
= tipc_wait_for_cond(sock
, &timeout
,
1424 (!tsk
->cong_link_cnt
&&
1425 !tsk_conn_cong(tsk
) &&
1426 tipc_sk_connected(sk
)));
1430 send
= min_t(size_t, dlen
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1431 rc
= tipc_msg_build(hdr
, m
, sent
, send
, tsk
->max_pkt
, &pkts
);
1432 if (unlikely(rc
!= send
))
1435 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1436 if (unlikely(rc
== -ELINKCONG
)) {
1437 tsk
->cong_link_cnt
= 1;
1441 tsk
->snt_unacked
+= tsk_inc(tsk
, send
+ MIN_H_SIZE
);
1444 } while (sent
< dlen
&& !rc
);
1446 return sent
? sent
: rc
;
1450 * tipc_send_packet - send a connection-oriented message
1451 * @sock: socket structure
1452 * @m: message to send
1453 * @dsz: length of data to be transmitted
1455 * Used for SOCK_SEQPACKET messages.
1457 * Returns the number of bytes sent on success, or errno otherwise
1459 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1461 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1464 return tipc_sendstream(sock
, m
, dsz
);
1467 /* tipc_sk_finish_conn - complete the setup of a connection
1469 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1472 struct sock
*sk
= &tsk
->sk
;
1473 struct net
*net
= sock_net(sk
);
1474 struct tipc_msg
*msg
= &tsk
->phdr
;
1476 msg_set_destnode(msg
, peer_node
);
1477 msg_set_destport(msg
, peer_port
);
1478 msg_set_type(msg
, TIPC_CONN_MSG
);
1479 msg_set_lookup_scope(msg
, 0);
1480 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1482 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
1483 tipc_set_sk_state(sk
, TIPC_ESTABLISHED
);
1484 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1485 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1486 tsk
->peer_caps
= tipc_node_get_capabilities(net
, peer_node
);
1487 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1490 /* Fall back to message based flow control */
1491 tsk
->rcv_win
= FLOWCTL_MSG_WIN
;
1492 tsk
->snd_win
= FLOWCTL_MSG_WIN
;
1496 * tipc_sk_set_orig_addr - capture sender's address for received message
1497 * @m: descriptor for message info
1498 * @hdr: received message header
1500 * Note: Address is not captured if not requested by receiver.
1502 static void tipc_sk_set_orig_addr(struct msghdr
*m
, struct sk_buff
*skb
)
1504 DECLARE_SOCKADDR(struct sockaddr_pair
*, srcaddr
, m
->msg_name
);
1505 struct tipc_msg
*hdr
= buf_msg(skb
);
1510 srcaddr
->sock
.family
= AF_TIPC
;
1511 srcaddr
->sock
.addrtype
= TIPC_ADDR_ID
;
1512 srcaddr
->sock
.scope
= 0;
1513 srcaddr
->sock
.addr
.id
.ref
= msg_origport(hdr
);
1514 srcaddr
->sock
.addr
.id
.node
= msg_orignode(hdr
);
1515 srcaddr
->sock
.addr
.name
.domain
= 0;
1516 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1518 if (!msg_in_group(hdr
))
1521 /* Group message users may also want to know sending member's id */
1522 srcaddr
->member
.family
= AF_TIPC
;
1523 srcaddr
->member
.addrtype
= TIPC_ADDR_NAME
;
1524 srcaddr
->member
.scope
= 0;
1525 srcaddr
->member
.addr
.name
.name
.type
= msg_nametype(hdr
);
1526 srcaddr
->member
.addr
.name
.name
.instance
= TIPC_SKB_CB(skb
)->orig_member
;
1527 srcaddr
->member
.addr
.name
.domain
= 0;
1528 m
->msg_namelen
= sizeof(*srcaddr
);
1532 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1533 * @m: descriptor for message info
1534 * @skb: received message buffer
1535 * @tsk: TIPC port associated with message
1537 * Note: Ancillary data is not captured if not requested by receiver.
1539 * Returns 0 if successful, otherwise errno
1541 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct sk_buff
*skb
,
1542 struct tipc_sock
*tsk
)
1544 struct tipc_msg
*msg
;
1551 if (likely(m
->msg_controllen
== 0))
1555 /* Optionally capture errored message object(s) */
1556 err
= msg
? msg_errcode(msg
) : 0;
1557 if (unlikely(err
)) {
1559 anc_data
[1] = msg_data_sz(msg
);
1560 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1564 if (skb_linearize(skb
))
1567 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1574 /* Optionally capture message destination object */
1575 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1576 switch (dest_type
) {
1577 case TIPC_NAMED_MSG
:
1579 anc_data
[0] = msg_nametype(msg
);
1580 anc_data
[1] = msg_namelower(msg
);
1581 anc_data
[2] = msg_namelower(msg
);
1583 case TIPC_MCAST_MSG
:
1585 anc_data
[0] = msg_nametype(msg
);
1586 anc_data
[1] = msg_namelower(msg
);
1587 anc_data
[2] = msg_nameupper(msg
);
1590 has_name
= (tsk
->conn_type
!= 0);
1591 anc_data
[0] = tsk
->conn_type
;
1592 anc_data
[1] = tsk
->conn_instance
;
1593 anc_data
[2] = tsk
->conn_instance
;
1599 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1607 static void tipc_sk_send_ack(struct tipc_sock
*tsk
)
1609 struct sock
*sk
= &tsk
->sk
;
1610 struct net
*net
= sock_net(sk
);
1611 struct sk_buff
*skb
= NULL
;
1612 struct tipc_msg
*msg
;
1613 u32 peer_port
= tsk_peer_port(tsk
);
1614 u32 dnode
= tsk_peer_node(tsk
);
1616 if (!tipc_sk_connected(sk
))
1618 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1619 dnode
, tsk_own_node(tsk
), peer_port
,
1620 tsk
->portid
, TIPC_OK
);
1624 msg_set_conn_ack(msg
, tsk
->rcv_unacked
);
1625 tsk
->rcv_unacked
= 0;
1627 /* Adjust to and advertize the correct window limit */
1628 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
) {
1629 tsk
->rcv_win
= tsk_adv_blocks(tsk
->sk
.sk_rcvbuf
);
1630 msg_set_adv_win(msg
, tsk
->rcv_win
);
1632 tipc_node_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1635 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1637 struct sock
*sk
= sock
->sk
;
1639 long timeo
= *timeop
;
1640 int err
= sock_error(sk
);
1646 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1647 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1648 if (sk
->sk_shutdown
& RCV_SHUTDOWN
) {
1653 timeo
= schedule_timeout(timeo
);
1657 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1662 err
= sock_intr_errno(timeo
);
1663 if (signal_pending(current
))
1666 err
= sock_error(sk
);
1670 finish_wait(sk_sleep(sk
), &wait
);
1676 * tipc_recvmsg - receive packet-oriented message
1677 * @m: descriptor for message info
1678 * @buflen: length of user buffer area
1679 * @flags: receive flags
1681 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1682 * If the complete message doesn't fit in user area, truncate it.
1684 * Returns size of returned message data, errno otherwise
1686 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
,
1687 size_t buflen
, int flags
)
1689 struct sock
*sk
= sock
->sk
;
1690 bool connected
= !tipc_sk_type_connectionless(sk
);
1691 struct tipc_sock
*tsk
= tipc_sk(sk
);
1692 int rc
, err
, hlen
, dlen
, copy
;
1693 struct sk_buff_head xmitq
;
1694 struct tipc_msg
*hdr
;
1695 struct sk_buff
*skb
;
1699 /* Catch invalid receive requests */
1700 if (unlikely(!buflen
))
1704 if (unlikely(connected
&& sk
->sk_state
== TIPC_OPEN
)) {
1708 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1710 /* Step rcv queue to first msg with data or error; wait if necessary */
1712 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1715 skb
= skb_peek(&sk
->sk_receive_queue
);
1717 dlen
= msg_data_sz(hdr
);
1718 hlen
= msg_hdr_sz(hdr
);
1719 err
= msg_errcode(hdr
);
1720 grp_evt
= msg_is_grp_evt(hdr
);
1721 if (likely(dlen
|| err
))
1723 tsk_advance_rx_queue(sk
);
1726 /* Collect msg meta data, including error code and rejected data */
1727 tipc_sk_set_orig_addr(m
, skb
);
1728 rc
= tipc_sk_anc_data_recv(m
, skb
, tsk
);
1733 /* Capture data if non-error msg, otherwise just set return value */
1735 copy
= min_t(int, dlen
, buflen
);
1736 if (unlikely(copy
!= dlen
))
1737 m
->msg_flags
|= MSG_TRUNC
;
1738 rc
= skb_copy_datagram_msg(skb
, hlen
, m
, copy
);
1742 if (err
!= TIPC_CONN_SHUTDOWN
&& connected
&& !m
->msg_control
)
1748 /* Mark message as group event if applicable */
1749 if (unlikely(grp_evt
)) {
1750 if (msg_grp_evt(hdr
) == TIPC_WITHDRAWN
)
1751 m
->msg_flags
|= MSG_EOR
;
1752 m
->msg_flags
|= MSG_OOB
;
1756 /* Caption of data or error code/rejected data was successful */
1757 if (unlikely(flags
& MSG_PEEK
))
1760 /* Send group flow control advertisement when applicable */
1761 if (tsk
->group
&& msg_in_group(hdr
) && !grp_evt
) {
1762 skb_queue_head_init(&xmitq
);
1763 tipc_group_update_rcv_win(tsk
->group
, tsk_blocks(hlen
+ dlen
),
1764 msg_orignode(hdr
), msg_origport(hdr
),
1766 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
1769 tsk_advance_rx_queue(sk
);
1771 if (likely(!connected
))
1774 /* Send connection flow control advertisement when applicable */
1775 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1776 if (tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
)
1777 tipc_sk_send_ack(tsk
);
1780 return rc
? rc
: copy
;
1784 * tipc_recvstream - receive stream-oriented data
1785 * @m: descriptor for message info
1786 * @buflen: total size of user buffer area
1787 * @flags: receive flags
1789 * Used for SOCK_STREAM messages only. If not enough data is available
1790 * will optionally wait for more; never truncates data.
1792 * Returns size of returned message data, errno otherwise
1794 static int tipc_recvstream(struct socket
*sock
, struct msghdr
*m
,
1795 size_t buflen
, int flags
)
1797 struct sock
*sk
= sock
->sk
;
1798 struct tipc_sock
*tsk
= tipc_sk(sk
);
1799 struct sk_buff
*skb
;
1800 struct tipc_msg
*hdr
;
1801 struct tipc_skb_cb
*skb_cb
;
1802 bool peek
= flags
& MSG_PEEK
;
1803 int offset
, required
, copy
, copied
= 0;
1804 int hlen
, dlen
, err
, rc
;
1807 /* Catch invalid receive attempts */
1808 if (unlikely(!buflen
))
1813 if (unlikely(sk
->sk_state
== TIPC_OPEN
)) {
1817 required
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buflen
);
1818 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1821 /* Look at first msg in receive queue; wait if necessary */
1822 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1825 skb
= skb_peek(&sk
->sk_receive_queue
);
1826 skb_cb
= TIPC_SKB_CB(skb
);
1828 dlen
= msg_data_sz(hdr
);
1829 hlen
= msg_hdr_sz(hdr
);
1830 err
= msg_errcode(hdr
);
1832 /* Discard any empty non-errored (SYN-) message */
1833 if (unlikely(!dlen
&& !err
)) {
1834 tsk_advance_rx_queue(sk
);
1838 /* Collect msg meta data, incl. error code and rejected data */
1840 tipc_sk_set_orig_addr(m
, skb
);
1841 rc
= tipc_sk_anc_data_recv(m
, skb
, tsk
);
1847 /* Copy data if msg ok, otherwise return error/partial data */
1849 offset
= skb_cb
->bytes_read
;
1850 copy
= min_t(int, dlen
- offset
, buflen
- copied
);
1851 rc
= skb_copy_datagram_msg(skb
, hlen
+ offset
, m
, copy
);
1856 if (unlikely(offset
< dlen
)) {
1858 skb_cb
->bytes_read
= offset
;
1863 if ((err
!= TIPC_CONN_SHUTDOWN
) && !m
->msg_control
)
1872 tsk_advance_rx_queue(sk
);
1874 /* Send connection flow control advertisement when applicable */
1875 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1876 if (unlikely(tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
))
1877 tipc_sk_send_ack(tsk
);
1879 /* Exit if all requested data or FIN/error received */
1880 if (copied
== buflen
|| err
)
1883 } while (!skb_queue_empty(&sk
->sk_receive_queue
) || copied
< required
);
1886 return copied
? copied
: rc
;
1890 * tipc_write_space - wake up thread if port congestion is released
1893 static void tipc_write_space(struct sock
*sk
)
1895 struct socket_wq
*wq
;
1898 wq
= rcu_dereference(sk
->sk_wq
);
1899 if (skwq_has_sleeper(wq
))
1900 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1901 POLLWRNORM
| POLLWRBAND
);
1906 * tipc_data_ready - wake up threads to indicate messages have been received
1908 * @len: the length of messages
1910 static void tipc_data_ready(struct sock
*sk
)
1912 struct socket_wq
*wq
;
1915 wq
= rcu_dereference(sk
->sk_wq
);
1916 if (skwq_has_sleeper(wq
))
1917 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1918 POLLRDNORM
| POLLRDBAND
);
1922 static void tipc_sock_destruct(struct sock
*sk
)
1924 __skb_queue_purge(&sk
->sk_receive_queue
);
1927 static void tipc_sk_proto_rcv(struct sock
*sk
,
1928 struct sk_buff_head
*inputq
,
1929 struct sk_buff_head
*xmitq
)
1931 struct sk_buff
*skb
= __skb_dequeue(inputq
);
1932 struct tipc_sock
*tsk
= tipc_sk(sk
);
1933 struct tipc_msg
*hdr
= buf_msg(skb
);
1934 struct tipc_group
*grp
= tsk
->group
;
1935 bool wakeup
= false;
1937 switch (msg_user(hdr
)) {
1939 tipc_sk_conn_proto_rcv(tsk
, skb
, xmitq
);
1942 tipc_dest_del(&tsk
->cong_links
, msg_orignode(hdr
), 0);
1943 tsk
->cong_link_cnt
--;
1946 case GROUP_PROTOCOL
:
1947 tipc_group_proto_rcv(grp
, &wakeup
, hdr
, inputq
, xmitq
);
1950 tipc_group_member_evt(tsk
->group
, &wakeup
, &sk
->sk_rcvbuf
,
1951 skb
, inputq
, xmitq
);
1959 sk
->sk_write_space(sk
);
1965 * tipc_filter_connect - Handle incoming message for a connection-based socket
1967 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1969 * Returns true if everything ok, false otherwise
1971 static bool tipc_sk_filter_connect(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
1973 struct sock
*sk
= &tsk
->sk
;
1974 struct net
*net
= sock_net(sk
);
1975 struct tipc_msg
*hdr
= buf_msg(skb
);
1976 u32 pport
= msg_origport(hdr
);
1977 u32 pnode
= msg_orignode(hdr
);
1979 if (unlikely(msg_mcast(hdr
)))
1982 switch (sk
->sk_state
) {
1983 case TIPC_CONNECTING
:
1984 /* Accept only ACK or NACK message */
1985 if (unlikely(!msg_connected(hdr
))) {
1986 if (pport
!= tsk_peer_port(tsk
) ||
1987 pnode
!= tsk_peer_node(tsk
))
1990 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1991 sk
->sk_err
= ECONNREFUSED
;
1992 sk
->sk_state_change(sk
);
1996 if (unlikely(msg_errcode(hdr
))) {
1997 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1998 sk
->sk_err
= ECONNREFUSED
;
1999 sk
->sk_state_change(sk
);
2003 if (unlikely(!msg_isdata(hdr
))) {
2004 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2005 sk
->sk_err
= EINVAL
;
2006 sk
->sk_state_change(sk
);
2010 tipc_sk_finish_conn(tsk
, msg_origport(hdr
), msg_orignode(hdr
));
2011 msg_set_importance(&tsk
->phdr
, msg_importance(hdr
));
2013 /* If 'ACK+' message, add to socket receive queue */
2014 if (msg_data_sz(hdr
))
2017 /* If empty 'ACK-' message, wake up sleeping connect() */
2018 sk
->sk_data_ready(sk
);
2020 /* 'ACK-' message is neither accepted nor rejected: */
2021 msg_set_dest_droppable(hdr
, 1);
2025 case TIPC_DISCONNECTING
:
2028 /* Accept only SYN message */
2029 if (!msg_connected(hdr
) && !(msg_errcode(hdr
)))
2032 case TIPC_ESTABLISHED
:
2033 /* Accept only connection-based messages sent by peer */
2034 if (unlikely(!tsk_peer_msg(tsk
, hdr
)))
2037 if (unlikely(msg_errcode(hdr
))) {
2038 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2039 /* Let timer expire on it's own */
2040 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
2042 sk
->sk_state_change(sk
);
2046 pr_err("Unknown sk_state %u\n", sk
->sk_state
);
2053 * rcvbuf_limit - get proper overload limit of socket receive queue
2057 * For connection oriented messages, irrespective of importance,
2058 * default queue limit is 2 MB.
2060 * For connectionless messages, queue limits are based on message
2061 * importance as follows:
2063 * TIPC_LOW_IMPORTANCE (2 MB)
2064 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2065 * TIPC_HIGH_IMPORTANCE (8 MB)
2066 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2068 * Returns overload limit according to corresponding message importance
2070 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*skb
)
2072 struct tipc_sock
*tsk
= tipc_sk(sk
);
2073 struct tipc_msg
*hdr
= buf_msg(skb
);
2075 if (unlikely(msg_in_group(hdr
)))
2076 return sk
->sk_rcvbuf
;
2078 if (unlikely(!msg_connected(hdr
)))
2079 return sk
->sk_rcvbuf
<< msg_importance(hdr
);
2081 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
2082 return sk
->sk_rcvbuf
;
2084 return FLOWCTL_MSG_LIM
;
2088 * tipc_sk_filter_rcv - validate incoming message
2090 * @skb: pointer to message.
2092 * Enqueues message on receive queue if acceptable; optionally handles
2093 * disconnect indication for a connected socket.
2095 * Called with socket lock already taken
2098 static void tipc_sk_filter_rcv(struct sock
*sk
, struct sk_buff
*skb
,
2099 struct sk_buff_head
*xmitq
)
2101 bool sk_conn
= !tipc_sk_type_connectionless(sk
);
2102 struct tipc_sock
*tsk
= tipc_sk(sk
);
2103 struct tipc_group
*grp
= tsk
->group
;
2104 struct tipc_msg
*hdr
= buf_msg(skb
);
2105 struct net
*net
= sock_net(sk
);
2106 struct sk_buff_head inputq
;
2107 int limit
, err
= TIPC_OK
;
2109 TIPC_SKB_CB(skb
)->bytes_read
= 0;
2110 __skb_queue_head_init(&inputq
);
2111 __skb_queue_tail(&inputq
, skb
);
2113 if (unlikely(!msg_isdata(hdr
)))
2114 tipc_sk_proto_rcv(sk
, &inputq
, xmitq
);
2117 tipc_group_filter_msg(grp
, &inputq
, xmitq
);
2119 /* Validate and add to receive buffer if there is space */
2120 while ((skb
= __skb_dequeue(&inputq
))) {
2122 limit
= rcvbuf_limit(sk
, skb
);
2123 if ((sk_conn
&& !tipc_sk_filter_connect(tsk
, skb
)) ||
2124 (!sk_conn
&& msg_connected(hdr
)) ||
2125 (!grp
&& msg_in_group(hdr
)))
2126 err
= TIPC_ERR_NO_PORT
;
2127 else if (sk_rmem_alloc_get(sk
) + skb
->truesize
>= limit
)
2128 err
= TIPC_ERR_OVERLOAD
;
2130 if (unlikely(err
)) {
2131 tipc_skb_reject(net
, err
, skb
, xmitq
);
2135 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
2136 skb_set_owner_r(skb
, sk
);
2137 sk
->sk_data_ready(sk
);
2142 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2146 * Caller must hold socket lock
2148 static int tipc_sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
2150 unsigned int before
= sk_rmem_alloc_get(sk
);
2151 struct sk_buff_head xmitq
;
2154 __skb_queue_head_init(&xmitq
);
2156 tipc_sk_filter_rcv(sk
, skb
, &xmitq
);
2157 added
= sk_rmem_alloc_get(sk
) - before
;
2158 atomic_add(added
, &tipc_sk(sk
)->dupl_rcvcnt
);
2160 /* Send pending response/rejected messages, if any */
2161 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2166 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2167 * inputq and try adding them to socket or backlog queue
2168 * @inputq: list of incoming buffers with potentially different destinations
2169 * @sk: socket where the buffers should be enqueued
2170 * @dport: port number for the socket
2172 * Caller must hold socket lock
2174 static void tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
2175 u32 dport
, struct sk_buff_head
*xmitq
)
2177 unsigned long time_limit
= jiffies
+ 2;
2178 struct sk_buff
*skb
;
2183 while (skb_queue_len(inputq
)) {
2184 if (unlikely(time_after_eq(jiffies
, time_limit
)))
2187 skb
= tipc_skb_dequeue(inputq
, dport
);
2191 /* Add message directly to receive queue if possible */
2192 if (!sock_owned_by_user(sk
)) {
2193 tipc_sk_filter_rcv(sk
, skb
, xmitq
);
2197 /* Try backlog, compensating for double-counted bytes */
2198 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
2199 if (!sk
->sk_backlog
.len
)
2200 atomic_set(dcnt
, 0);
2201 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
2202 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
2205 /* Overload => reject message back to sender */
2206 onode
= tipc_own_addr(sock_net(sk
));
2207 if (tipc_msg_reverse(onode
, &skb
, TIPC_ERR_OVERLOAD
))
2208 __skb_queue_tail(xmitq
, skb
);
2214 * tipc_sk_rcv - handle a chain of incoming buffers
2215 * @inputq: buffer list containing the buffers
2216 * Consumes all buffers in list until inputq is empty
2217 * Note: may be called in multiple threads referring to the same queue
2219 void tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
2221 struct sk_buff_head xmitq
;
2222 u32 dnode
, dport
= 0;
2224 struct tipc_sock
*tsk
;
2226 struct sk_buff
*skb
;
2228 __skb_queue_head_init(&xmitq
);
2229 while (skb_queue_len(inputq
)) {
2230 dport
= tipc_skb_peek_port(inputq
, dport
);
2231 tsk
= tipc_sk_lookup(net
, dport
);
2235 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
2236 tipc_sk_enqueue(inputq
, sk
, dport
, &xmitq
);
2237 spin_unlock_bh(&sk
->sk_lock
.slock
);
2239 /* Send pending response/rejected messages, if any */
2240 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2244 /* No destination socket => dequeue skb if still there */
2245 skb
= tipc_skb_dequeue(inputq
, dport
);
2249 /* Try secondary lookup if unresolved named message */
2250 err
= TIPC_ERR_NO_PORT
;
2251 if (tipc_msg_lookup_dest(net
, skb
, &err
))
2254 /* Prepare for message rejection */
2255 if (!tipc_msg_reverse(tipc_own_addr(net
), &skb
, err
))
2258 dnode
= msg_destnode(buf_msg(skb
));
2259 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
2263 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
2265 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
2266 struct sock
*sk
= sock
->sk
;
2270 int err
= sock_error(sk
);
2275 if (signal_pending(current
))
2276 return sock_intr_errno(*timeo_p
);
2278 add_wait_queue(sk_sleep(sk
), &wait
);
2279 done
= sk_wait_event(sk
, timeo_p
,
2280 sk
->sk_state
!= TIPC_CONNECTING
, &wait
);
2281 remove_wait_queue(sk_sleep(sk
), &wait
);
2287 * tipc_connect - establish a connection to another TIPC port
2288 * @sock: socket structure
2289 * @dest: socket address for destination port
2290 * @destlen: size of socket address data structure
2291 * @flags: file-related flags associated with socket
2293 * Returns 0 on success, errno otherwise
2295 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
2296 int destlen
, int flags
)
2298 struct sock
*sk
= sock
->sk
;
2299 struct tipc_sock
*tsk
= tipc_sk(sk
);
2300 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
2301 struct msghdr m
= {NULL
,};
2302 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
2306 if (destlen
!= sizeof(struct sockaddr_tipc
))
2316 if (dst
->family
== AF_UNSPEC
) {
2317 memset(&tsk
->peer
, 0, sizeof(struct sockaddr_tipc
));
2318 if (!tipc_sk_type_connectionless(sk
))
2321 } else if (dst
->family
!= AF_TIPC
) {
2324 if (dst
->addrtype
!= TIPC_ADDR_ID
&& dst
->addrtype
!= TIPC_ADDR_NAME
)
2329 /* DGRAM/RDM connect(), just save the destaddr */
2330 if (tipc_sk_type_connectionless(sk
)) {
2331 memcpy(&tsk
->peer
, dest
, destlen
);
2335 previous
= sk
->sk_state
;
2337 switch (sk
->sk_state
) {
2339 /* Send a 'SYN-' to destination */
2341 m
.msg_namelen
= destlen
;
2343 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2344 * indicate send_msg() is never blocked.
2347 m
.msg_flags
= MSG_DONTWAIT
;
2349 res
= __tipc_sendmsg(sock
, &m
, 0);
2350 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
2353 /* Just entered TIPC_CONNECTING state; the only
2354 * difference is that return value in non-blocking
2355 * case is EINPROGRESS, rather than EALREADY.
2359 case TIPC_CONNECTING
:
2361 if (previous
== TIPC_CONNECTING
)
2365 timeout
= msecs_to_jiffies(timeout
);
2366 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2367 res
= tipc_wait_for_connect(sock
, &timeout
);
2369 case TIPC_ESTABLISHED
:
2382 * tipc_listen - allow socket to listen for incoming connections
2383 * @sock: socket structure
2386 * Returns 0 on success, errno otherwise
2388 static int tipc_listen(struct socket
*sock
, int len
)
2390 struct sock
*sk
= sock
->sk
;
2394 res
= tipc_set_sk_state(sk
, TIPC_LISTEN
);
2400 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
2402 struct sock
*sk
= sock
->sk
;
2406 /* True wake-one mechanism for incoming connections: only
2407 * one process gets woken up, not the 'whole herd'.
2408 * Since we do not 'race & poll' for established sockets
2409 * anymore, the common case will execute the loop only once.
2412 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
2413 TASK_INTERRUPTIBLE
);
2414 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
2416 timeo
= schedule_timeout(timeo
);
2420 if (!skb_queue_empty(&sk
->sk_receive_queue
))
2425 err
= sock_intr_errno(timeo
);
2426 if (signal_pending(current
))
2429 finish_wait(sk_sleep(sk
), &wait
);
2434 * tipc_accept - wait for connection request
2435 * @sock: listening socket
2436 * @newsock: new socket that is to be connected
2437 * @flags: file-related flags associated with socket
2439 * Returns 0 on success, errno otherwise
2441 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
,
2444 struct sock
*new_sk
, *sk
= sock
->sk
;
2445 struct sk_buff
*buf
;
2446 struct tipc_sock
*new_tsock
;
2447 struct tipc_msg
*msg
;
2453 if (sk
->sk_state
!= TIPC_LISTEN
) {
2457 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2458 res
= tipc_wait_for_accept(sock
, timeo
);
2462 buf
= skb_peek(&sk
->sk_receive_queue
);
2464 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, kern
);
2467 security_sk_clone(sock
->sk
, new_sock
->sk
);
2469 new_sk
= new_sock
->sk
;
2470 new_tsock
= tipc_sk(new_sk
);
2473 /* we lock on new_sk; but lockdep sees the lock on sk */
2474 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2477 * Reject any stray messages received by new socket
2478 * before the socket lock was taken (very, very unlikely)
2480 tsk_rej_rx_queue(new_sk
);
2482 /* Connect new socket to it's peer */
2483 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2485 tsk_set_importance(new_tsock
, msg_importance(msg
));
2486 if (msg_named(msg
)) {
2487 new_tsock
->conn_type
= msg_nametype(msg
);
2488 new_tsock
->conn_instance
= msg_nameinst(msg
);
2492 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2493 * Respond to 'SYN+' by queuing it on new socket.
2495 if (!msg_data_sz(msg
)) {
2496 struct msghdr m
= {NULL
,};
2498 tsk_advance_rx_queue(sk
);
2499 __tipc_sendstream(new_sock
, &m
, 0);
2501 __skb_dequeue(&sk
->sk_receive_queue
);
2502 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2503 skb_set_owner_r(buf
, new_sk
);
2505 release_sock(new_sk
);
2512 * tipc_shutdown - shutdown socket connection
2513 * @sock: socket structure
2514 * @how: direction to close (must be SHUT_RDWR)
2516 * Terminates connection (if necessary), then purges socket's receive queue.
2518 * Returns 0 on success, errno otherwise
2520 static int tipc_shutdown(struct socket
*sock
, int how
)
2522 struct sock
*sk
= sock
->sk
;
2525 if (how
!= SHUT_RDWR
)
2530 __tipc_shutdown(sock
, TIPC_CONN_SHUTDOWN
);
2531 sk
->sk_shutdown
= SEND_SHUTDOWN
;
2533 if (sk
->sk_state
== TIPC_DISCONNECTING
) {
2534 /* Discard any unreceived messages */
2535 __skb_queue_purge(&sk
->sk_receive_queue
);
2537 /* Wake up anyone sleeping in poll */
2538 sk
->sk_state_change(sk
);
2548 static void tipc_sk_timeout(struct timer_list
*t
)
2550 struct sock
*sk
= from_timer(sk
, t
, sk_timer
);
2551 struct tipc_sock
*tsk
= tipc_sk(sk
);
2552 u32 peer_port
= tsk_peer_port(tsk
);
2553 u32 peer_node
= tsk_peer_node(tsk
);
2554 u32 own_node
= tsk_own_node(tsk
);
2555 u32 own_port
= tsk
->portid
;
2556 struct net
*net
= sock_net(sk
);
2557 struct sk_buff
*skb
= NULL
;
2560 if (!tipc_sk_connected(sk
))
2563 /* Try again later if socket is busy */
2564 if (sock_owned_by_user(sk
)) {
2565 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ HZ
/ 20);
2569 if (tsk
->probe_unacked
) {
2570 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2571 tipc_node_remove_conn(net
, peer_node
, peer_port
);
2572 sk
->sk_state_change(sk
);
2575 /* Send new probe */
2576 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
, 0,
2577 peer_node
, own_node
, peer_port
, own_port
,
2579 tsk
->probe_unacked
= true;
2580 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
2584 tipc_node_xmit_skb(net
, skb
, peer_node
, own_port
);
2588 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2589 struct tipc_name_seq
const *seq
)
2591 struct sock
*sk
= &tsk
->sk
;
2592 struct net
*net
= sock_net(sk
);
2593 struct publication
*publ
;
2596 if (tipc_sk_connected(sk
))
2598 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2599 if (key
== tsk
->portid
)
2602 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2603 scope
, tsk
->portid
, key
);
2604 if (unlikely(!publ
))
2607 list_add(&publ
->pport_list
, &tsk
->publications
);
2613 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2614 struct tipc_name_seq
const *seq
)
2616 struct net
*net
= sock_net(&tsk
->sk
);
2617 struct publication
*publ
;
2618 struct publication
*safe
;
2621 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2623 if (publ
->scope
!= scope
)
2625 if (publ
->type
!= seq
->type
)
2627 if (publ
->lower
!= seq
->lower
)
2629 if (publ
->upper
!= seq
->upper
)
2631 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2632 publ
->ref
, publ
->key
);
2636 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2637 publ
->ref
, publ
->key
);
2640 if (list_empty(&tsk
->publications
))
2645 /* tipc_sk_reinit: set non-zero address in all existing sockets
2646 * when we go from standalone to network mode.
2648 void tipc_sk_reinit(struct net
*net
)
2650 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2651 struct rhashtable_iter iter
;
2652 struct tipc_sock
*tsk
;
2653 struct tipc_msg
*msg
;
2655 rhashtable_walk_enter(&tn
->sk_rht
, &iter
);
2658 tsk
= ERR_PTR(rhashtable_walk_start(&iter
));
2662 while ((tsk
= rhashtable_walk_next(&iter
)) && !IS_ERR(tsk
)) {
2663 sock_hold(&tsk
->sk
);
2664 rhashtable_walk_stop(&iter
);
2665 lock_sock(&tsk
->sk
);
2667 msg_set_prevnode(msg
, tn
->own_addr
);
2668 msg_set_orignode(msg
, tn
->own_addr
);
2669 release_sock(&tsk
->sk
);
2670 rhashtable_walk_start(&iter
);
2674 rhashtable_walk_stop(&iter
);
2675 } while (tsk
== ERR_PTR(-EAGAIN
));
2677 rhashtable_walk_exit(&iter
);
2680 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2682 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2683 struct tipc_sock
*tsk
;
2686 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2688 sock_hold(&tsk
->sk
);
2694 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2696 struct sock
*sk
= &tsk
->sk
;
2697 struct net
*net
= sock_net(sk
);
2698 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2699 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2700 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2702 while (remaining
--) {
2704 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2705 portid
= TIPC_MIN_PORT
;
2706 tsk
->portid
= portid
;
2707 sock_hold(&tsk
->sk
);
2708 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2717 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2719 struct sock
*sk
= &tsk
->sk
;
2720 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2722 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2723 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
2728 static const struct rhashtable_params tsk_rht_params
= {
2730 .head_offset
= offsetof(struct tipc_sock
, node
),
2731 .key_offset
= offsetof(struct tipc_sock
, portid
),
2732 .key_len
= sizeof(u32
), /* portid */
2733 .max_size
= 1048576,
2735 .automatic_shrinking
= true,
2738 int tipc_sk_rht_init(struct net
*net
)
2740 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2742 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2745 void tipc_sk_rht_destroy(struct net
*net
)
2747 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2749 /* Wait for socket readers to complete */
2752 rhashtable_destroy(&tn
->sk_rht
);
2755 static int tipc_sk_join(struct tipc_sock
*tsk
, struct tipc_group_req
*mreq
)
2757 struct net
*net
= sock_net(&tsk
->sk
);
2758 u32 domain
= addr_domain(net
, mreq
->scope
);
2759 struct tipc_group
*grp
= tsk
->group
;
2760 struct tipc_msg
*hdr
= &tsk
->phdr
;
2761 struct tipc_name_seq seq
;
2764 if (mreq
->type
< TIPC_RESERVED_TYPES
)
2768 grp
= tipc_group_create(net
, tsk
->portid
, mreq
);
2772 msg_set_lookup_scope(hdr
, mreq
->scope
);
2773 msg_set_nametype(hdr
, mreq
->type
);
2774 msg_set_dest_droppable(hdr
, true);
2775 seq
.type
= mreq
->type
;
2776 seq
.lower
= mreq
->instance
;
2777 seq
.upper
= seq
.lower
;
2778 tipc_nametbl_build_group(net
, grp
, mreq
->type
, domain
);
2779 rc
= tipc_sk_publish(tsk
, mreq
->scope
, &seq
);
2781 tipc_group_delete(net
, grp
);
2785 /* Eliminate any risk that a broadcast overtakes the sent JOIN */
2786 tsk
->mc_method
.rcast
= true;
2787 tsk
->mc_method
.mandatory
= true;
2791 static int tipc_sk_leave(struct tipc_sock
*tsk
)
2793 struct net
*net
= sock_net(&tsk
->sk
);
2794 struct tipc_group
*grp
= tsk
->group
;
2795 struct tipc_name_seq seq
;
2800 tipc_group_self(grp
, &seq
, &scope
);
2801 tipc_group_delete(net
, grp
);
2803 tipc_sk_withdraw(tsk
, scope
, &seq
);
2808 * tipc_setsockopt - set socket option
2809 * @sock: socket structure
2810 * @lvl: option level
2811 * @opt: option identifier
2812 * @ov: pointer to new option value
2813 * @ol: length of option value
2815 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2816 * (to ease compatibility).
2818 * Returns 0 on success, errno otherwise
2820 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2821 char __user
*ov
, unsigned int ol
)
2823 struct sock
*sk
= sock
->sk
;
2824 struct tipc_sock
*tsk
= tipc_sk(sk
);
2825 struct tipc_group_req mreq
;
2829 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2831 if (lvl
!= SOL_TIPC
)
2832 return -ENOPROTOOPT
;
2835 case TIPC_IMPORTANCE
:
2836 case TIPC_SRC_DROPPABLE
:
2837 case TIPC_DEST_DROPPABLE
:
2838 case TIPC_CONN_TIMEOUT
:
2839 if (ol
< sizeof(value
))
2841 if (get_user(value
, (u32 __user
*)ov
))
2844 case TIPC_GROUP_JOIN
:
2845 if (ol
< sizeof(mreq
))
2847 if (copy_from_user(&mreq
, ov
, sizeof(mreq
)))
2858 case TIPC_IMPORTANCE
:
2859 res
= tsk_set_importance(tsk
, value
);
2861 case TIPC_SRC_DROPPABLE
:
2862 if (sock
->type
!= SOCK_STREAM
)
2863 tsk_set_unreliable(tsk
, value
);
2867 case TIPC_DEST_DROPPABLE
:
2868 tsk_set_unreturnable(tsk
, value
);
2870 case TIPC_CONN_TIMEOUT
:
2871 tipc_sk(sk
)->conn_timeout
= value
;
2873 case TIPC_MCAST_BROADCAST
:
2874 tsk
->mc_method
.rcast
= false;
2875 tsk
->mc_method
.mandatory
= true;
2877 case TIPC_MCAST_REPLICAST
:
2878 tsk
->mc_method
.rcast
= true;
2879 tsk
->mc_method
.mandatory
= true;
2881 case TIPC_GROUP_JOIN
:
2882 res
= tipc_sk_join(tsk
, &mreq
);
2884 case TIPC_GROUP_LEAVE
:
2885 res
= tipc_sk_leave(tsk
);
2897 * tipc_getsockopt - get socket option
2898 * @sock: socket structure
2899 * @lvl: option level
2900 * @opt: option identifier
2901 * @ov: receptacle for option value
2902 * @ol: receptacle for length of option value
2904 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2905 * (to ease compatibility).
2907 * Returns 0 on success, errno otherwise
2909 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2910 char __user
*ov
, int __user
*ol
)
2912 struct sock
*sk
= sock
->sk
;
2913 struct tipc_sock
*tsk
= tipc_sk(sk
);
2914 struct tipc_name_seq seq
;
2919 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2920 return put_user(0, ol
);
2921 if (lvl
!= SOL_TIPC
)
2922 return -ENOPROTOOPT
;
2923 res
= get_user(len
, ol
);
2930 case TIPC_IMPORTANCE
:
2931 value
= tsk_importance(tsk
);
2933 case TIPC_SRC_DROPPABLE
:
2934 value
= tsk_unreliable(tsk
);
2936 case TIPC_DEST_DROPPABLE
:
2937 value
= tsk_unreturnable(tsk
);
2939 case TIPC_CONN_TIMEOUT
:
2940 value
= tsk
->conn_timeout
;
2941 /* no need to set "res", since already 0 at this point */
2943 case TIPC_NODE_RECVQ_DEPTH
:
2944 value
= 0; /* was tipc_queue_size, now obsolete */
2946 case TIPC_SOCK_RECVQ_DEPTH
:
2947 value
= skb_queue_len(&sk
->sk_receive_queue
);
2949 case TIPC_GROUP_JOIN
:
2952 tipc_group_self(tsk
->group
, &seq
, &scope
);
2962 return res
; /* "get" failed */
2964 if (len
< sizeof(value
))
2967 if (copy_to_user(ov
, &value
, sizeof(value
)))
2970 return put_user(sizeof(value
), ol
);
2973 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2975 struct sock
*sk
= sock
->sk
;
2976 struct tipc_sioc_ln_req lnr
;
2977 void __user
*argp
= (void __user
*)arg
;
2980 case SIOCGETLINKNAME
:
2981 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2983 if (!tipc_node_get_linkname(sock_net(sk
),
2984 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2985 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2986 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2990 return -EADDRNOTAVAIL
;
2992 return -ENOIOCTLCMD
;
2996 static int tipc_socketpair(struct socket
*sock1
, struct socket
*sock2
)
2998 struct tipc_sock
*tsk2
= tipc_sk(sock2
->sk
);
2999 struct tipc_sock
*tsk1
= tipc_sk(sock1
->sk
);
3000 u32 onode
= tipc_own_addr(sock_net(sock1
->sk
));
3002 tsk1
->peer
.family
= AF_TIPC
;
3003 tsk1
->peer
.addrtype
= TIPC_ADDR_ID
;
3004 tsk1
->peer
.scope
= TIPC_NODE_SCOPE
;
3005 tsk1
->peer
.addr
.id
.ref
= tsk2
->portid
;
3006 tsk1
->peer
.addr
.id
.node
= onode
;
3007 tsk2
->peer
.family
= AF_TIPC
;
3008 tsk2
->peer
.addrtype
= TIPC_ADDR_ID
;
3009 tsk2
->peer
.scope
= TIPC_NODE_SCOPE
;
3010 tsk2
->peer
.addr
.id
.ref
= tsk1
->portid
;
3011 tsk2
->peer
.addr
.id
.node
= onode
;
3013 tipc_sk_finish_conn(tsk1
, tsk2
->portid
, onode
);
3014 tipc_sk_finish_conn(tsk2
, tsk1
->portid
, onode
);
3018 /* Protocol switches for the various types of TIPC sockets */
3020 static const struct proto_ops msg_ops
= {
3021 .owner
= THIS_MODULE
,
3023 .release
= tipc_release
,
3025 .connect
= tipc_connect
,
3026 .socketpair
= tipc_socketpair
,
3027 .accept
= sock_no_accept
,
3028 .getname
= tipc_getname
,
3030 .ioctl
= tipc_ioctl
,
3031 .listen
= sock_no_listen
,
3032 .shutdown
= tipc_shutdown
,
3033 .setsockopt
= tipc_setsockopt
,
3034 .getsockopt
= tipc_getsockopt
,
3035 .sendmsg
= tipc_sendmsg
,
3036 .recvmsg
= tipc_recvmsg
,
3037 .mmap
= sock_no_mmap
,
3038 .sendpage
= sock_no_sendpage
3041 static const struct proto_ops packet_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_send_packet
,
3057 .recvmsg
= tipc_recvmsg
,
3058 .mmap
= sock_no_mmap
,
3059 .sendpage
= sock_no_sendpage
3062 static const struct proto_ops stream_ops
= {
3063 .owner
= THIS_MODULE
,
3065 .release
= tipc_release
,
3067 .connect
= tipc_connect
,
3068 .socketpair
= tipc_socketpair
,
3069 .accept
= tipc_accept
,
3070 .getname
= tipc_getname
,
3072 .ioctl
= tipc_ioctl
,
3073 .listen
= tipc_listen
,
3074 .shutdown
= tipc_shutdown
,
3075 .setsockopt
= tipc_setsockopt
,
3076 .getsockopt
= tipc_getsockopt
,
3077 .sendmsg
= tipc_sendstream
,
3078 .recvmsg
= tipc_recvstream
,
3079 .mmap
= sock_no_mmap
,
3080 .sendpage
= sock_no_sendpage
3083 static const struct net_proto_family tipc_family_ops
= {
3084 .owner
= THIS_MODULE
,
3086 .create
= tipc_sk_create
3089 static struct proto tipc_proto
= {
3091 .owner
= THIS_MODULE
,
3092 .obj_size
= sizeof(struct tipc_sock
),
3093 .sysctl_rmem
= sysctl_tipc_rmem
3097 * tipc_socket_init - initialize TIPC socket interface
3099 * Returns 0 on success, errno otherwise
3101 int tipc_socket_init(void)
3105 res
= proto_register(&tipc_proto
, 1);
3107 pr_err("Failed to register TIPC protocol type\n");
3111 res
= sock_register(&tipc_family_ops
);
3113 pr_err("Failed to register TIPC socket type\n");
3114 proto_unregister(&tipc_proto
);
3122 * tipc_socket_stop - stop TIPC socket interface
3124 void tipc_socket_stop(void)
3126 sock_unregister(tipc_family_ops
.family
);
3127 proto_unregister(&tipc_proto
);
3130 /* Caller should hold socket lock for the passed tipc socket. */
3131 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
3135 struct nlattr
*nest
;
3137 peer_node
= tsk_peer_node(tsk
);
3138 peer_port
= tsk_peer_port(tsk
);
3140 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
3142 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
3144 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
3147 if (tsk
->conn_type
!= 0) {
3148 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
3150 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
3152 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
3155 nla_nest_end(skb
, nest
);
3160 nla_nest_cancel(skb
, nest
);
3165 /* Caller should hold socket lock for the passed tipc socket. */
3166 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
3167 struct tipc_sock
*tsk
)
3171 struct nlattr
*attrs
;
3172 struct net
*net
= sock_net(skb
->sk
);
3173 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
3174 struct sock
*sk
= &tsk
->sk
;
3176 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3177 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
3181 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
3183 goto genlmsg_cancel
;
3184 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
3185 goto attr_msg_cancel
;
3186 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tn
->own_addr
))
3187 goto attr_msg_cancel
;
3189 if (tipc_sk_connected(sk
)) {
3190 err
= __tipc_nl_add_sk_con(skb
, tsk
);
3192 goto attr_msg_cancel
;
3193 } else if (!list_empty(&tsk
->publications
)) {
3194 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
3195 goto attr_msg_cancel
;
3197 nla_nest_end(skb
, attrs
);
3198 genlmsg_end(skb
, hdr
);
3203 nla_nest_cancel(skb
, attrs
);
3205 genlmsg_cancel(skb
, hdr
);
3210 int tipc_dump_start(struct netlink_callback
*cb
)
3212 return __tipc_dump_start(cb
, sock_net(cb
->skb
->sk
));
3214 EXPORT_SYMBOL(tipc_dump_start
);
3216 int __tipc_dump_start(struct netlink_callback
*cb
, struct net
*net
)
3218 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3219 struct rhashtable_iter
*iter
= (void *)cb
->args
[4];
3220 struct tipc_net
*tn
= tipc_net(net
);
3223 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
3227 cb
->args
[4] = (long)iter
;
3230 rhashtable_walk_enter(&tn
->sk_rht
, iter
);
3234 int tipc_dump_done(struct netlink_callback
*cb
)
3236 struct rhashtable_iter
*hti
= (void *)cb
->args
[4];
3238 rhashtable_walk_exit(hti
);
3242 EXPORT_SYMBOL(tipc_dump_done
);
3244 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3247 struct tipc_sock
*tsk
;
3248 const struct bucket_table
*tbl
;
3249 struct rhash_head
*pos
;
3250 struct net
*net
= sock_net(skb
->sk
);
3251 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
3252 u32 tbl_id
= cb
->args
[0];
3253 u32 prev_portid
= cb
->args
[1];
3256 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
3257 for (; tbl_id
< tbl
->size
; tbl_id
++) {
3258 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
3259 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
3260 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
3261 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3265 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
3267 prev_portid
= tsk
->portid
;
3268 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3272 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3277 cb
->args
[0] = tbl_id
;
3278 cb
->args
[1] = prev_portid
;
3283 /* Caller should hold socket lock for the passed tipc socket. */
3284 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
3285 struct netlink_callback
*cb
,
3286 struct publication
*publ
)
3289 struct nlattr
*attrs
;
3291 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3292 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
3296 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
3298 goto genlmsg_cancel
;
3300 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
3301 goto attr_msg_cancel
;
3302 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
3303 goto attr_msg_cancel
;
3304 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
3305 goto attr_msg_cancel
;
3306 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
3307 goto attr_msg_cancel
;
3309 nla_nest_end(skb
, attrs
);
3310 genlmsg_end(skb
, hdr
);
3315 nla_nest_cancel(skb
, attrs
);
3317 genlmsg_cancel(skb
, hdr
);
3322 /* Caller should hold socket lock for the passed tipc socket. */
3323 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
3324 struct netlink_callback
*cb
,
3325 struct tipc_sock
*tsk
, u32
*last_publ
)
3328 struct publication
*p
;
3331 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
3332 if (p
->key
== *last_publ
)
3335 if (p
->key
!= *last_publ
) {
3336 /* We never set seq or call nl_dump_check_consistent()
3337 * this means that setting prev_seq here will cause the
3338 * consistence check to fail in the netlink callback
3339 * handler. Resulting in the last NLMSG_DONE message
3340 * having the NLM_F_DUMP_INTR flag set.
3347 p
= list_first_entry(&tsk
->publications
, struct publication
,
3351 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
3352 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
3354 *last_publ
= p
->key
;
3363 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3366 u32 tsk_portid
= cb
->args
[0];
3367 u32 last_publ
= cb
->args
[1];
3368 u32 done
= cb
->args
[2];
3369 struct net
*net
= sock_net(skb
->sk
);
3370 struct tipc_sock
*tsk
;
3373 struct nlattr
**attrs
;
3374 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
3376 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
3380 if (!attrs
[TIPC_NLA_SOCK
])
3383 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
3384 attrs
[TIPC_NLA_SOCK
],
3385 tipc_nl_sock_policy
, NULL
);
3389 if (!sock
[TIPC_NLA_SOCK_REF
])
3392 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
3398 tsk
= tipc_sk_lookup(net
, tsk_portid
);
3402 lock_sock(&tsk
->sk
);
3403 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
3406 release_sock(&tsk
->sk
);
3409 cb
->args
[0] = tsk_portid
;
3410 cb
->args
[1] = last_publ
;