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_) \
380 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
384 while ((rc_ = !(condition_))) { \
385 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
388 rc_ = tipc_sk_sock_err((sock_), timeo_); \
391 prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE); \
393 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
394 sched_annotate_sleep(); \
396 remove_wait_queue(sk_sleep(sk_), &wait_); \
402 * tipc_sk_create - create a TIPC socket
403 * @net: network namespace (must be default network)
404 * @sock: pre-allocated socket structure
405 * @protocol: protocol indicator (must be 0)
406 * @kern: caused by kernel or by userspace?
408 * This routine creates additional data structures used by the TIPC socket,
409 * initializes them, and links them together.
411 * Returns 0 on success, errno otherwise
413 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
414 int protocol
, int kern
)
417 const struct proto_ops
*ops
;
419 struct tipc_sock
*tsk
;
420 struct tipc_msg
*msg
;
422 /* Validate arguments */
423 if (unlikely(protocol
!= 0))
424 return -EPROTONOSUPPORT
;
426 switch (sock
->type
) {
441 /* Allocate socket's protocol area */
442 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
, kern
);
447 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
448 INIT_LIST_HEAD(&tsk
->publications
);
449 INIT_LIST_HEAD(&tsk
->cong_links
);
451 tn
= net_generic(sock_net(sk
), tipc_net_id
);
453 /* Finish initializing socket data structures */
455 sock_init_data(sock
, sk
);
456 tipc_set_sk_state(sk
, TIPC_OPEN
);
457 if (tipc_sk_insert(tsk
)) {
458 pr_warn("Socket create failed; port number exhausted\n");
462 /* Ensure tsk is visible before we read own_addr. */
465 tipc_msg_init(tn
->own_addr
, msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
468 msg_set_origport(msg
, tsk
->portid
);
469 timer_setup(&sk
->sk_timer
, tipc_sk_timeout
, 0);
471 sk
->sk_backlog_rcv
= tipc_sk_backlog_rcv
;
472 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
473 sk
->sk_data_ready
= tipc_data_ready
;
474 sk
->sk_write_space
= tipc_write_space
;
475 sk
->sk_destruct
= tipc_sock_destruct
;
476 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
477 atomic_set(&tsk
->dupl_rcvcnt
, 0);
479 /* Start out with safe limits until we receive an advertised window */
480 tsk
->snd_win
= tsk_adv_blocks(RCVBUF_MIN
);
481 tsk
->rcv_win
= tsk
->snd_win
;
483 if (tipc_sk_type_connectionless(sk
)) {
484 tsk_set_unreturnable(tsk
, true);
485 if (sock
->type
== SOCK_DGRAM
)
486 tsk_set_unreliable(tsk
, true);
492 static void tipc_sk_callback(struct rcu_head
*head
)
494 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
499 /* Caller should hold socket lock for the socket. */
500 static void __tipc_shutdown(struct socket
*sock
, int error
)
502 struct sock
*sk
= sock
->sk
;
503 struct tipc_sock
*tsk
= tipc_sk(sk
);
504 struct net
*net
= sock_net(sk
);
505 long timeout
= CONN_TIMEOUT_DEFAULT
;
506 u32 dnode
= tsk_peer_node(tsk
);
509 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
510 tipc_wait_for_cond(sock
, &timeout
, (!tsk
->cong_link_cnt
&&
511 !tsk_conn_cong(tsk
)));
513 /* Reject all unreceived messages, except on an active connection
514 * (which disconnects locally & sends a 'FIN+' to peer).
516 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)) != NULL
) {
517 if (TIPC_SKB_CB(skb
)->bytes_read
) {
521 if (!tipc_sk_type_connectionless(sk
) &&
522 sk
->sk_state
!= TIPC_DISCONNECTING
) {
523 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
524 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
526 tipc_sk_respond(sk
, skb
, error
);
529 if (tipc_sk_type_connectionless(sk
))
532 if (sk
->sk_state
!= TIPC_DISCONNECTING
) {
533 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
534 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0, dnode
,
535 tsk_own_node(tsk
), tsk_peer_port(tsk
),
538 tipc_node_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
539 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
540 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
545 * tipc_release - destroy a TIPC socket
546 * @sock: socket to destroy
548 * This routine cleans up any messages that are still queued on the socket.
549 * For DGRAM and RDM socket types, all queued messages are rejected.
550 * For SEQPACKET and STREAM socket types, the first message is rejected
551 * and any others are discarded. (If the first message on a STREAM socket
552 * is partially-read, it is discarded and the next one is rejected instead.)
554 * NOTE: Rejected messages are not necessarily returned to the sender! They
555 * are returned or discarded according to the "destination droppable" setting
556 * specified for the message by the sender.
558 * Returns 0 on success, errno otherwise
560 static int tipc_release(struct socket
*sock
)
562 struct sock
*sk
= sock
->sk
;
563 struct tipc_sock
*tsk
;
566 * Exit if socket isn't fully initialized (occurs when a failed accept()
567 * releases a pre-allocated child socket that was never used)
575 __tipc_shutdown(sock
, TIPC_ERR_NO_PORT
);
576 sk
->sk_shutdown
= SHUTDOWN_MASK
;
578 tipc_sk_withdraw(tsk
, 0, NULL
);
579 sk_stop_timer(sk
, &sk
->sk_timer
);
583 /* Reject any messages that accumulated in backlog queue */
585 tipc_dest_list_purge(&tsk
->cong_links
);
586 tsk
->cong_link_cnt
= 0;
587 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
594 * tipc_bind - associate or disassocate TIPC name(s) with a socket
595 * @sock: socket structure
596 * @uaddr: socket address describing name(s) and desired operation
597 * @uaddr_len: size of socket address data structure
599 * Name and name sequence binding is indicated using a positive scope value;
600 * a negative scope value unbinds the specified name. Specifying no name
601 * (i.e. a socket address length of 0) unbinds all names from the socket.
603 * Returns 0 on success, errno otherwise
605 * NOTE: This routine doesn't need to take the socket lock since it doesn't
606 * access any non-constant socket information.
608 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
611 struct sock
*sk
= sock
->sk
;
612 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
613 struct tipc_sock
*tsk
= tipc_sk(sk
);
617 if (unlikely(!uaddr_len
)) {
618 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
625 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
629 if (addr
->family
!= AF_TIPC
) {
634 if (addr
->addrtype
== TIPC_ADDR_NAME
)
635 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
636 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
641 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
642 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
643 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
648 res
= (addr
->scope
> 0) ?
649 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
650 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
657 * tipc_getname - get port ID of socket or peer socket
658 * @sock: socket structure
659 * @uaddr: area for returned socket address
660 * @uaddr_len: area for returned length of socket address
661 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
663 * Returns 0 on success, errno otherwise
665 * NOTE: This routine doesn't need to take the socket lock since it only
666 * accesses socket information that is unchanging (or which changes in
667 * a completely predictable manner).
669 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
670 int *uaddr_len
, int peer
)
672 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
673 struct sock
*sk
= sock
->sk
;
674 struct tipc_sock
*tsk
= tipc_sk(sk
);
675 struct tipc_net
*tn
= net_generic(sock_net(sock
->sk
), tipc_net_id
);
677 memset(addr
, 0, sizeof(*addr
));
679 if ((!tipc_sk_connected(sk
)) &&
680 ((peer
!= 2) || (sk
->sk_state
!= TIPC_DISCONNECTING
)))
682 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
683 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
685 addr
->addr
.id
.ref
= tsk
->portid
;
686 addr
->addr
.id
.node
= tn
->own_addr
;
689 *uaddr_len
= sizeof(*addr
);
690 addr
->addrtype
= TIPC_ADDR_ID
;
691 addr
->family
= AF_TIPC
;
693 addr
->addr
.name
.domain
= 0;
699 * tipc_poll - read and possibly block on pollmask
700 * @file: file structure associated with the socket
701 * @sock: socket for which to calculate the poll bits
704 * Returns pollmask value
707 * It appears that the usual socket locking mechanisms are not useful here
708 * since the pollmask info is potentially out-of-date the moment this routine
709 * exits. TCP and other protocols seem to rely on higher level poll routines
710 * to handle any preventable race conditions, so TIPC will do the same ...
712 * IMPORTANT: The fact that a read or write operation is indicated does NOT
713 * imply that the operation will succeed, merely that it should be performed
714 * and will not block.
716 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
719 struct sock
*sk
= sock
->sk
;
720 struct tipc_sock
*tsk
= tipc_sk(sk
);
721 struct tipc_group
*grp
= tsk
->group
;
724 sock_poll_wait(file
, sk_sleep(sk
), wait
);
726 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
727 revents
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
728 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
731 switch (sk
->sk_state
) {
732 case TIPC_ESTABLISHED
:
733 if (!tsk
->cong_link_cnt
&& !tsk_conn_cong(tsk
))
737 case TIPC_CONNECTING
:
738 if (!skb_queue_empty_lockless(&sk
->sk_receive_queue
))
739 revents
|= POLLIN
| POLLRDNORM
;
742 if (!grp
|| tipc_group_size(grp
))
743 if (!tsk
->cong_link_cnt
)
745 if (!tipc_sk_type_connectionless(sk
))
747 if (skb_queue_empty_lockless(&sk
->sk_receive_queue
))
749 revents
|= POLLIN
| POLLRDNORM
;
751 case TIPC_DISCONNECTING
:
752 revents
= POLLIN
| POLLRDNORM
| POLLHUP
;
759 * tipc_sendmcast - send multicast message
760 * @sock: socket structure
761 * @seq: destination address
762 * @msg: message to send
763 * @dlen: length of data to send
764 * @timeout: timeout to wait for wakeup
766 * Called from function tipc_sendmsg(), which has done all sanity checks
767 * Returns the number of bytes sent on success, or errno
769 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
770 struct msghdr
*msg
, size_t dlen
, long timeout
)
772 struct sock
*sk
= sock
->sk
;
773 struct tipc_sock
*tsk
= tipc_sk(sk
);
774 struct tipc_msg
*hdr
= &tsk
->phdr
;
775 struct net
*net
= sock_net(sk
);
776 int mtu
= tipc_bcast_get_mtu(net
);
777 struct tipc_mc_method
*method
= &tsk
->mc_method
;
778 u32 domain
= addr_domain(net
, TIPC_CLUSTER_SCOPE
);
779 struct sk_buff_head pkts
;
780 struct tipc_nlist dsts
;
786 /* Block or return if any destination link is congested */
787 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
);
791 /* Lookup destination nodes */
792 tipc_nlist_init(&dsts
, tipc_own_addr(net
));
793 tipc_nametbl_lookup_dst_nodes(net
, seq
->type
, seq
->lower
,
794 seq
->upper
, domain
, &dsts
);
795 if (!dsts
.local
&& !dsts
.remote
)
796 return -EHOSTUNREACH
;
798 /* Build message header */
799 msg_set_type(hdr
, TIPC_MCAST_MSG
);
800 msg_set_hdr_sz(hdr
, MCAST_H_SIZE
);
801 msg_set_lookup_scope(hdr
, TIPC_CLUSTER_SCOPE
);
802 msg_set_destport(hdr
, 0);
803 msg_set_destnode(hdr
, 0);
804 msg_set_nametype(hdr
, seq
->type
);
805 msg_set_namelower(hdr
, seq
->lower
);
806 msg_set_nameupper(hdr
, seq
->upper
);
808 /* Build message as chain of buffers */
809 skb_queue_head_init(&pkts
);
810 rc
= tipc_msg_build(hdr
, msg
, 0, dlen
, mtu
, &pkts
);
812 /* Send message if build was successful */
813 if (unlikely(rc
== dlen
))
814 rc
= tipc_mcast_xmit(net
, &pkts
, method
, &dsts
,
815 &tsk
->cong_link_cnt
);
817 tipc_nlist_purge(&dsts
);
819 return rc
? rc
: dlen
;
823 * tipc_send_group_msg - send a message to a member in the group
824 * @net: network namespace
825 * @m: message to send
827 * @dnode: destination node
828 * @dport: destination port
829 * @dlen: total length of message data
831 static int tipc_send_group_msg(struct net
*net
, struct tipc_sock
*tsk
,
832 struct msghdr
*m
, struct tipc_member
*mb
,
833 u32 dnode
, u32 dport
, int dlen
)
835 u16 bc_snd_nxt
= tipc_group_bc_snd_nxt(tsk
->group
);
836 struct tipc_mc_method
*method
= &tsk
->mc_method
;
837 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
838 struct tipc_msg
*hdr
= &tsk
->phdr
;
839 struct sk_buff_head pkts
;
842 /* Complete message header */
843 msg_set_type(hdr
, TIPC_GRP_UCAST_MSG
);
844 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
845 msg_set_destport(hdr
, dport
);
846 msg_set_destnode(hdr
, dnode
);
847 msg_set_grp_bc_seqno(hdr
, bc_snd_nxt
);
849 /* Build message as chain of buffers */
850 skb_queue_head_init(&pkts
);
851 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
852 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
853 if (unlikely(rc
!= dlen
))
857 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
858 if (unlikely(rc
== -ELINKCONG
)) {
859 tipc_dest_push(&tsk
->cong_links
, dnode
, 0);
860 tsk
->cong_link_cnt
++;
863 /* Update send window */
864 tipc_group_update_member(mb
, blks
);
866 /* A broadcast sent within next EXPIRE period must follow same path */
867 method
->rcast
= true;
868 method
->mandatory
= true;
873 * tipc_send_group_unicast - send message to a member in the group
874 * @sock: socket structure
875 * @m: message to send
876 * @dlen: total length of message data
877 * @timeout: timeout to wait for wakeup
879 * Called from function tipc_sendmsg(), which has done all sanity checks
880 * Returns the number of bytes sent on success, or errno
882 static int tipc_send_group_unicast(struct socket
*sock
, struct msghdr
*m
,
883 int dlen
, long timeout
)
885 struct sock
*sk
= sock
->sk
;
886 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
887 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
888 struct tipc_sock
*tsk
= tipc_sk(sk
);
889 struct net
*net
= sock_net(sk
);
890 struct tipc_member
*mb
= NULL
;
894 node
= dest
->addr
.id
.node
;
895 port
= dest
->addr
.id
.ref
;
897 return -EHOSTUNREACH
;
899 /* Block or return if destination link or member is congested */
900 rc
= tipc_wait_for_cond(sock
, &timeout
,
901 !tipc_dest_find(&tsk
->cong_links
, node
, 0) &&
903 !tipc_group_cong(tsk
->group
, node
, port
, blks
,
909 return -EHOSTUNREACH
;
911 rc
= tipc_send_group_msg(net
, tsk
, m
, mb
, node
, port
, dlen
);
913 return rc
? rc
: dlen
;
917 * tipc_send_group_anycast - send message to any member with given identity
918 * @sock: socket structure
919 * @m: message to send
920 * @dlen: total length of message data
921 * @timeout: timeout to wait for wakeup
923 * Called from function tipc_sendmsg(), which has done all sanity checks
924 * Returns the number of bytes sent on success, or errno
926 static int tipc_send_group_anycast(struct socket
*sock
, struct msghdr
*m
,
927 int dlen
, long timeout
)
929 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
930 struct sock
*sk
= sock
->sk
;
931 struct tipc_sock
*tsk
= tipc_sk(sk
);
932 struct list_head
*cong_links
= &tsk
->cong_links
;
933 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
934 struct tipc_member
*first
= NULL
;
935 struct tipc_member
*mbr
= NULL
;
936 struct net
*net
= sock_net(sk
);
937 u32 node
, port
, exclude
;
938 u32 type
, inst
, domain
;
939 struct list_head dsts
;
944 INIT_LIST_HEAD(&dsts
);
946 type
= dest
->addr
.name
.name
.type
;
947 inst
= dest
->addr
.name
.name
.instance
;
948 domain
= addr_domain(net
, dest
->scope
);
950 while (++lookups
< 4) {
951 exclude
= tipc_group_exclude(tsk
->group
);
955 /* Look for a non-congested destination member, if any */
957 if (!tipc_nametbl_lookup(net
, type
, inst
, domain
, &dsts
,
958 &dstcnt
, exclude
, false))
959 return -EHOSTUNREACH
;
960 tipc_dest_pop(&dsts
, &node
, &port
);
961 cong
= tipc_group_cong(tsk
->group
, node
, port
, blks
,
971 /* Start over if destination was not in member list */
975 if (likely(!cong
&& !tipc_dest_find(cong_links
, node
, 0)))
978 /* Block or return if destination link or member is congested */
979 rc
= tipc_wait_for_cond(sock
, &timeout
,
980 !tipc_dest_find(cong_links
, node
, 0) &&
982 !tipc_group_cong(tsk
->group
, node
, port
,
987 /* Send, unless destination disappeared while waiting */
992 if (unlikely(lookups
>= 4))
993 return -EHOSTUNREACH
;
995 rc
= tipc_send_group_msg(net
, tsk
, m
, mbr
, node
, port
, dlen
);
997 return rc
? rc
: dlen
;
1001 * tipc_send_group_bcast - send message to all members in communication group
1002 * @sk: socket structure
1003 * @m: message to send
1004 * @dlen: total length of message data
1005 * @timeout: timeout to wait for wakeup
1007 * Called from function tipc_sendmsg(), which has done all sanity checks
1008 * Returns the number of bytes sent on success, or errno
1010 static int tipc_send_group_bcast(struct socket
*sock
, struct msghdr
*m
,
1011 int dlen
, long timeout
)
1013 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1014 struct sock
*sk
= sock
->sk
;
1015 struct net
*net
= sock_net(sk
);
1016 struct tipc_sock
*tsk
= tipc_sk(sk
);
1017 struct tipc_nlist
*dsts
;
1018 struct tipc_mc_method
*method
= &tsk
->mc_method
;
1019 bool ack
= method
->mandatory
&& method
->rcast
;
1020 int blks
= tsk_blocks(MCAST_H_SIZE
+ dlen
);
1021 struct tipc_msg
*hdr
= &tsk
->phdr
;
1022 int mtu
= tipc_bcast_get_mtu(net
);
1023 struct sk_buff_head pkts
;
1024 int rc
= -EHOSTUNREACH
;
1026 /* Block or return if any destination link or member is congested */
1027 rc
= tipc_wait_for_cond(sock
, &timeout
,
1028 !tsk
->cong_link_cnt
&& tsk
->group
&&
1029 !tipc_group_bc_cong(tsk
->group
, blks
));
1033 dsts
= tipc_group_dests(tsk
->group
);
1034 if (!dsts
->local
&& !dsts
->remote
)
1035 return -EHOSTUNREACH
;
1037 /* Complete message header */
1039 msg_set_type(hdr
, TIPC_GRP_MCAST_MSG
);
1040 msg_set_nameinst(hdr
, dest
->addr
.name
.name
.instance
);
1042 msg_set_type(hdr
, TIPC_GRP_BCAST_MSG
);
1043 msg_set_nameinst(hdr
, 0);
1045 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
1046 msg_set_destport(hdr
, 0);
1047 msg_set_destnode(hdr
, 0);
1048 msg_set_grp_bc_seqno(hdr
, tipc_group_bc_snd_nxt(tsk
->group
));
1050 /* Avoid getting stuck with repeated forced replicasts */
1051 msg_set_grp_bc_ack_req(hdr
, ack
);
1053 /* Build message as chain of buffers */
1054 skb_queue_head_init(&pkts
);
1055 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1056 if (unlikely(rc
!= dlen
))
1060 rc
= tipc_mcast_xmit(net
, &pkts
, method
, dsts
, &tsk
->cong_link_cnt
);
1064 /* Update broadcast sequence number and send windows */
1065 tipc_group_update_bc_members(tsk
->group
, blks
, ack
);
1067 /* Broadcast link is now free to choose method for next broadcast */
1068 method
->mandatory
= false;
1069 method
->expires
= jiffies
;
1075 * tipc_send_group_mcast - send message to all members with given identity
1076 * @sock: socket structure
1077 * @m: message to send
1078 * @dlen: total length of message data
1079 * @timeout: timeout to wait for wakeup
1081 * Called from function tipc_sendmsg(), which has done all sanity checks
1082 * Returns the number of bytes sent on success, or errno
1084 static int tipc_send_group_mcast(struct socket
*sock
, struct msghdr
*m
,
1085 int dlen
, long timeout
)
1087 struct sock
*sk
= sock
->sk
;
1088 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1089 struct tipc_name_seq
*seq
= &dest
->addr
.nameseq
;
1090 struct tipc_sock
*tsk
= tipc_sk(sk
);
1091 struct tipc_group
*grp
= tsk
->group
;
1092 struct net
*net
= sock_net(sk
);
1093 u32 domain
, exclude
, dstcnt
;
1094 struct list_head dsts
;
1096 INIT_LIST_HEAD(&dsts
);
1098 if (seq
->lower
!= seq
->upper
)
1101 domain
= addr_domain(net
, dest
->scope
);
1102 exclude
= tipc_group_exclude(grp
);
1103 if (!tipc_nametbl_lookup(net
, seq
->type
, seq
->lower
, domain
,
1104 &dsts
, &dstcnt
, exclude
, true))
1105 return -EHOSTUNREACH
;
1108 tipc_dest_pop(&dsts
, &dest
->addr
.id
.node
, &dest
->addr
.id
.ref
);
1109 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1112 tipc_dest_list_purge(&dsts
);
1113 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1117 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1118 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1119 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1121 * Multi-threaded: parallel calls with reference to same queues may occur
1123 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
1124 struct sk_buff_head
*inputq
)
1126 u32 scope
= TIPC_CLUSTER_SCOPE
;
1127 u32 self
= tipc_own_addr(net
);
1128 struct sk_buff
*skb
, *_skb
;
1129 u32 lower
= 0, upper
= ~0;
1130 struct sk_buff_head tmpq
;
1131 u32 portid
, oport
, onode
;
1132 struct list_head dports
;
1133 struct tipc_msg
*msg
;
1134 int user
, mtyp
, hsz
;
1136 __skb_queue_head_init(&tmpq
);
1137 INIT_LIST_HEAD(&dports
);
1139 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
1140 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
1142 user
= msg_user(msg
);
1143 mtyp
= msg_type(msg
);
1144 if (mtyp
== TIPC_GRP_UCAST_MSG
|| user
== GROUP_PROTOCOL
) {
1145 spin_lock_bh(&inputq
->lock
);
1146 if (skb_peek(arrvq
) == skb
) {
1147 __skb_dequeue(arrvq
);
1148 __skb_queue_tail(inputq
, skb
);
1151 spin_unlock_bh(&inputq
->lock
);
1154 hsz
= skb_headroom(skb
) + msg_hdr_sz(msg
);
1155 oport
= msg_origport(msg
);
1156 onode
= msg_orignode(msg
);
1158 scope
= TIPC_NODE_SCOPE
;
1160 /* Create destination port list and message clones: */
1161 if (!msg_in_group(msg
)) {
1162 lower
= msg_namelower(msg
);
1163 upper
= msg_nameupper(msg
);
1165 tipc_nametbl_mc_translate(net
, msg_nametype(msg
), lower
, upper
,
1167 while (tipc_dest_pop(&dports
, NULL
, &portid
)) {
1168 _skb
= __pskb_copy(skb
, hsz
, GFP_ATOMIC
);
1170 msg_set_destport(buf_msg(_skb
), portid
);
1171 __skb_queue_tail(&tmpq
, _skb
);
1174 pr_warn("Failed to clone mcast rcv buffer\n");
1176 /* Append to inputq if not already done by other thread */
1177 spin_lock_bh(&inputq
->lock
);
1178 if (skb_peek(arrvq
) == skb
) {
1179 skb_queue_splice_tail_init(&tmpq
, inputq
);
1180 kfree_skb(__skb_dequeue(arrvq
));
1182 spin_unlock_bh(&inputq
->lock
);
1183 __skb_queue_purge(&tmpq
);
1186 tipc_sk_rcv(net
, inputq
);
1190 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1191 * @tsk: receiving socket
1192 * @skb: pointer to message buffer.
1194 static void tipc_sk_conn_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
*skb
,
1195 struct sk_buff_head
*xmitq
)
1197 struct tipc_msg
*hdr
= buf_msg(skb
);
1198 u32 onode
= tsk_own_node(tsk
);
1199 struct sock
*sk
= &tsk
->sk
;
1200 int mtyp
= msg_type(hdr
);
1203 /* Ignore if connection cannot be validated: */
1204 if (!tsk_peer_msg(tsk
, hdr
))
1207 if (unlikely(msg_errcode(hdr
))) {
1208 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1209 tipc_node_remove_conn(sock_net(sk
), tsk_peer_node(tsk
),
1210 tsk_peer_port(tsk
));
1211 sk
->sk_state_change(sk
);
1215 tsk
->probe_unacked
= false;
1217 if (mtyp
== CONN_PROBE
) {
1218 msg_set_type(hdr
, CONN_PROBE_REPLY
);
1219 if (tipc_msg_reverse(onode
, &skb
, TIPC_OK
))
1220 __skb_queue_tail(xmitq
, skb
);
1222 } else if (mtyp
== CONN_ACK
) {
1223 conn_cong
= tsk_conn_cong(tsk
);
1224 tsk
->snt_unacked
-= msg_conn_ack(hdr
);
1225 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1226 tsk
->snd_win
= msg_adv_win(hdr
);
1228 sk
->sk_write_space(sk
);
1229 } else if (mtyp
!= CONN_PROBE_REPLY
) {
1230 pr_warn("Received unknown CONN_PROTO msg\n");
1237 * tipc_sendmsg - send message in connectionless manner
1238 * @sock: socket structure
1239 * @m: message to send
1240 * @dsz: amount of user data to be sent
1242 * Message must have an destination specified explicitly.
1243 * Used for SOCK_RDM and SOCK_DGRAM messages,
1244 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1245 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1247 * Returns the number of bytes sent on success, or errno otherwise
1249 static int tipc_sendmsg(struct socket
*sock
,
1250 struct msghdr
*m
, size_t dsz
)
1252 struct sock
*sk
= sock
->sk
;
1256 ret
= __tipc_sendmsg(sock
, m
, dsz
);
1262 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1264 struct sock
*sk
= sock
->sk
;
1265 struct net
*net
= sock_net(sk
);
1266 struct tipc_sock
*tsk
= tipc_sk(sk
);
1267 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1268 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1269 struct list_head
*clinks
= &tsk
->cong_links
;
1270 bool syn
= !tipc_sk_type_connectionless(sk
);
1271 struct tipc_group
*grp
= tsk
->group
;
1272 struct tipc_msg
*hdr
= &tsk
->phdr
;
1273 struct tipc_name_seq
*seq
;
1274 struct sk_buff_head pkts
;
1275 u32 type
, inst
, domain
;
1279 if (unlikely(dlen
> TIPC_MAX_USER_MSG_SIZE
))
1283 if (unlikely(m
->msg_namelen
< sizeof(*dest
)))
1285 if (unlikely(dest
->family
!= AF_TIPC
))
1291 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1292 if (dest
->addrtype
== TIPC_ADDR_NAME
)
1293 return tipc_send_group_anycast(sock
, m
, dlen
, timeout
);
1294 if (dest
->addrtype
== TIPC_ADDR_ID
)
1295 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1296 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1297 return tipc_send_group_mcast(sock
, m
, dlen
, timeout
);
1301 if (unlikely(!dest
)) {
1303 if (!syn
&& dest
->family
!= AF_TIPC
)
1304 return -EDESTADDRREQ
;
1307 if (unlikely(syn
)) {
1308 if (sk
->sk_state
== TIPC_LISTEN
)
1310 if (sk
->sk_state
!= TIPC_OPEN
)
1314 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1315 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
1316 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
1320 seq
= &dest
->addr
.nameseq
;
1321 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1322 return tipc_sendmcast(sock
, seq
, m
, dlen
, timeout
);
1324 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1325 type
= dest
->addr
.name
.name
.type
;
1326 inst
= dest
->addr
.name
.name
.instance
;
1327 domain
= dest
->addr
.name
.domain
;
1329 msg_set_type(hdr
, TIPC_NAMED_MSG
);
1330 msg_set_hdr_sz(hdr
, NAMED_H_SIZE
);
1331 msg_set_nametype(hdr
, type
);
1332 msg_set_nameinst(hdr
, inst
);
1333 msg_set_lookup_scope(hdr
, tipc_addr_scope(domain
));
1334 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
1335 msg_set_destnode(hdr
, dnode
);
1336 msg_set_destport(hdr
, dport
);
1337 if (unlikely(!dport
&& !dnode
))
1338 return -EHOSTUNREACH
;
1339 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
1340 dnode
= dest
->addr
.id
.node
;
1341 msg_set_type(hdr
, TIPC_DIRECT_MSG
);
1342 msg_set_lookup_scope(hdr
, 0);
1343 msg_set_destnode(hdr
, dnode
);
1344 msg_set_destport(hdr
, dest
->addr
.id
.ref
);
1345 msg_set_hdr_sz(hdr
, BASIC_H_SIZE
);
1348 /* Block or return if destination link is congested */
1349 rc
= tipc_wait_for_cond(sock
, &timeout
,
1350 !tipc_dest_find(clinks
, dnode
, 0));
1354 skb_queue_head_init(&pkts
);
1355 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
1356 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1357 if (unlikely(rc
!= dlen
))
1360 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1361 if (unlikely(rc
== -ELINKCONG
)) {
1362 tipc_dest_push(clinks
, dnode
, 0);
1363 tsk
->cong_link_cnt
++;
1367 if (unlikely(syn
&& !rc
))
1368 tipc_set_sk_state(sk
, TIPC_CONNECTING
);
1370 return rc
? rc
: dlen
;
1374 * tipc_sendstream - send stream-oriented data
1375 * @sock: socket structure
1377 * @dsz: total length of data to be transmitted
1379 * Used for SOCK_STREAM data.
1381 * Returns the number of bytes sent on success (or partial success),
1382 * or errno if no data sent
1384 static int tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1386 struct sock
*sk
= sock
->sk
;
1390 ret
= __tipc_sendstream(sock
, m
, dsz
);
1396 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1398 struct sock
*sk
= sock
->sk
;
1399 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1400 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1401 struct tipc_sock
*tsk
= tipc_sk(sk
);
1402 struct tipc_msg
*hdr
= &tsk
->phdr
;
1403 struct net
*net
= sock_net(sk
);
1404 struct sk_buff_head pkts
;
1405 u32 dnode
= tsk_peer_node(tsk
);
1409 skb_queue_head_init(&pkts
);
1411 if (unlikely(dlen
> INT_MAX
))
1414 /* Handle implicit connection setup */
1415 if (unlikely(dest
)) {
1416 rc
= __tipc_sendmsg(sock
, m
, dlen
);
1417 if (dlen
&& dlen
== rc
) {
1418 tsk
->peer_caps
= tipc_node_get_capabilities(net
, dnode
);
1419 tsk
->snt_unacked
= tsk_inc(tsk
, dlen
+ msg_hdr_sz(hdr
));
1425 rc
= tipc_wait_for_cond(sock
, &timeout
,
1426 (!tsk
->cong_link_cnt
&&
1427 !tsk_conn_cong(tsk
) &&
1428 tipc_sk_connected(sk
)));
1432 send
= min_t(size_t, dlen
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1433 rc
= tipc_msg_build(hdr
, m
, sent
, send
, tsk
->max_pkt
, &pkts
);
1434 if (unlikely(rc
!= send
))
1437 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1438 if (unlikely(rc
== -ELINKCONG
)) {
1439 tsk
->cong_link_cnt
= 1;
1443 tsk
->snt_unacked
+= tsk_inc(tsk
, send
+ MIN_H_SIZE
);
1446 } while (sent
< dlen
&& !rc
);
1448 return sent
? sent
: rc
;
1452 * tipc_send_packet - send a connection-oriented message
1453 * @sock: socket structure
1454 * @m: message to send
1455 * @dsz: length of data to be transmitted
1457 * Used for SOCK_SEQPACKET messages.
1459 * Returns the number of bytes sent on success, or errno otherwise
1461 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1463 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1466 return tipc_sendstream(sock
, m
, dsz
);
1469 /* tipc_sk_finish_conn - complete the setup of a connection
1471 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1474 struct sock
*sk
= &tsk
->sk
;
1475 struct net
*net
= sock_net(sk
);
1476 struct tipc_msg
*msg
= &tsk
->phdr
;
1478 msg_set_destnode(msg
, peer_node
);
1479 msg_set_destport(msg
, peer_port
);
1480 msg_set_type(msg
, TIPC_CONN_MSG
);
1481 msg_set_lookup_scope(msg
, 0);
1482 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1484 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
1485 tipc_set_sk_state(sk
, TIPC_ESTABLISHED
);
1486 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1487 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1488 tsk
->peer_caps
= tipc_node_get_capabilities(net
, peer_node
);
1489 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1492 /* Fall back to message based flow control */
1493 tsk
->rcv_win
= FLOWCTL_MSG_WIN
;
1494 tsk
->snd_win
= FLOWCTL_MSG_WIN
;
1498 * tipc_sk_set_orig_addr - capture sender's address for received message
1499 * @m: descriptor for message info
1500 * @hdr: received message header
1502 * Note: Address is not captured if not requested by receiver.
1504 static void tipc_sk_set_orig_addr(struct msghdr
*m
, struct sk_buff
*skb
)
1506 DECLARE_SOCKADDR(struct sockaddr_pair
*, srcaddr
, m
->msg_name
);
1507 struct tipc_msg
*hdr
= buf_msg(skb
);
1512 srcaddr
->sock
.family
= AF_TIPC
;
1513 srcaddr
->sock
.addrtype
= TIPC_ADDR_ID
;
1514 srcaddr
->sock
.scope
= 0;
1515 srcaddr
->sock
.addr
.id
.ref
= msg_origport(hdr
);
1516 srcaddr
->sock
.addr
.id
.node
= msg_orignode(hdr
);
1517 srcaddr
->sock
.addr
.name
.domain
= 0;
1518 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1520 if (!msg_in_group(hdr
))
1523 /* Group message users may also want to know sending member's id */
1524 srcaddr
->member
.family
= AF_TIPC
;
1525 srcaddr
->member
.addrtype
= TIPC_ADDR_NAME
;
1526 srcaddr
->member
.scope
= 0;
1527 srcaddr
->member
.addr
.name
.name
.type
= msg_nametype(hdr
);
1528 srcaddr
->member
.addr
.name
.name
.instance
= TIPC_SKB_CB(skb
)->orig_member
;
1529 srcaddr
->member
.addr
.name
.domain
= 0;
1530 m
->msg_namelen
= sizeof(*srcaddr
);
1534 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1535 * @m: descriptor for message info
1536 * @skb: received message buffer
1537 * @tsk: TIPC port associated with message
1539 * Note: Ancillary data is not captured if not requested by receiver.
1541 * Returns 0 if successful, otherwise errno
1543 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct sk_buff
*skb
,
1544 struct tipc_sock
*tsk
)
1546 struct tipc_msg
*msg
;
1553 if (likely(m
->msg_controllen
== 0))
1557 /* Optionally capture errored message object(s) */
1558 err
= msg
? msg_errcode(msg
) : 0;
1559 if (unlikely(err
)) {
1561 anc_data
[1] = msg_data_sz(msg
);
1562 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1566 if (skb_linearize(skb
))
1569 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1576 /* Optionally capture message destination object */
1577 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1578 switch (dest_type
) {
1579 case TIPC_NAMED_MSG
:
1581 anc_data
[0] = msg_nametype(msg
);
1582 anc_data
[1] = msg_namelower(msg
);
1583 anc_data
[2] = msg_namelower(msg
);
1585 case TIPC_MCAST_MSG
:
1587 anc_data
[0] = msg_nametype(msg
);
1588 anc_data
[1] = msg_namelower(msg
);
1589 anc_data
[2] = msg_nameupper(msg
);
1592 has_name
= (tsk
->conn_type
!= 0);
1593 anc_data
[0] = tsk
->conn_type
;
1594 anc_data
[1] = tsk
->conn_instance
;
1595 anc_data
[2] = tsk
->conn_instance
;
1601 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1609 static void tipc_sk_send_ack(struct tipc_sock
*tsk
)
1611 struct sock
*sk
= &tsk
->sk
;
1612 struct net
*net
= sock_net(sk
);
1613 struct sk_buff
*skb
= NULL
;
1614 struct tipc_msg
*msg
;
1615 u32 peer_port
= tsk_peer_port(tsk
);
1616 u32 dnode
= tsk_peer_node(tsk
);
1618 if (!tipc_sk_connected(sk
))
1620 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1621 dnode
, tsk_own_node(tsk
), peer_port
,
1622 tsk
->portid
, TIPC_OK
);
1626 msg_set_conn_ack(msg
, tsk
->rcv_unacked
);
1627 tsk
->rcv_unacked
= 0;
1629 /* Adjust to and advertize the correct window limit */
1630 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
) {
1631 tsk
->rcv_win
= tsk_adv_blocks(tsk
->sk
.sk_rcvbuf
);
1632 msg_set_adv_win(msg
, tsk
->rcv_win
);
1634 tipc_node_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1637 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1639 struct sock
*sk
= sock
->sk
;
1641 long timeo
= *timeop
;
1642 int err
= sock_error(sk
);
1648 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1649 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1650 if (sk
->sk_shutdown
& RCV_SHUTDOWN
) {
1655 timeo
= schedule_timeout(timeo
);
1659 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1664 err
= sock_intr_errno(timeo
);
1665 if (signal_pending(current
))
1668 err
= sock_error(sk
);
1672 finish_wait(sk_sleep(sk
), &wait
);
1678 * tipc_recvmsg - receive packet-oriented message
1679 * @m: descriptor for message info
1680 * @buflen: length of user buffer area
1681 * @flags: receive flags
1683 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1684 * If the complete message doesn't fit in user area, truncate it.
1686 * Returns size of returned message data, errno otherwise
1688 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
,
1689 size_t buflen
, int flags
)
1691 struct sock
*sk
= sock
->sk
;
1692 bool connected
= !tipc_sk_type_connectionless(sk
);
1693 struct tipc_sock
*tsk
= tipc_sk(sk
);
1694 int rc
, err
, hlen
, dlen
, copy
;
1695 struct sk_buff_head xmitq
;
1696 struct tipc_msg
*hdr
;
1697 struct sk_buff
*skb
;
1701 /* Catch invalid receive requests */
1702 if (unlikely(!buflen
))
1706 if (unlikely(connected
&& sk
->sk_state
== TIPC_OPEN
)) {
1710 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1712 /* Step rcv queue to first msg with data or error; wait if necessary */
1714 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1717 skb
= skb_peek(&sk
->sk_receive_queue
);
1719 dlen
= msg_data_sz(hdr
);
1720 hlen
= msg_hdr_sz(hdr
);
1721 err
= msg_errcode(hdr
);
1722 grp_evt
= msg_is_grp_evt(hdr
);
1723 if (likely(dlen
|| err
))
1725 tsk_advance_rx_queue(sk
);
1728 /* Collect msg meta data, including error code and rejected data */
1729 tipc_sk_set_orig_addr(m
, skb
);
1730 rc
= tipc_sk_anc_data_recv(m
, skb
, tsk
);
1735 /* Capture data if non-error msg, otherwise just set return value */
1737 copy
= min_t(int, dlen
, buflen
);
1738 if (unlikely(copy
!= dlen
))
1739 m
->msg_flags
|= MSG_TRUNC
;
1740 rc
= skb_copy_datagram_msg(skb
, hlen
, m
, copy
);
1744 if (err
!= TIPC_CONN_SHUTDOWN
&& connected
&& !m
->msg_control
)
1750 /* Mark message as group event if applicable */
1751 if (unlikely(grp_evt
)) {
1752 if (msg_grp_evt(hdr
) == TIPC_WITHDRAWN
)
1753 m
->msg_flags
|= MSG_EOR
;
1754 m
->msg_flags
|= MSG_OOB
;
1758 /* Caption of data or error code/rejected data was successful */
1759 if (unlikely(flags
& MSG_PEEK
))
1762 /* Send group flow control advertisement when applicable */
1763 if (tsk
->group
&& msg_in_group(hdr
) && !grp_evt
) {
1764 skb_queue_head_init(&xmitq
);
1765 tipc_group_update_rcv_win(tsk
->group
, tsk_blocks(hlen
+ dlen
),
1766 msg_orignode(hdr
), msg_origport(hdr
),
1768 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
1771 tsk_advance_rx_queue(sk
);
1773 if (likely(!connected
))
1776 /* Send connection flow control advertisement when applicable */
1777 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1778 if (tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
)
1779 tipc_sk_send_ack(tsk
);
1782 return rc
? rc
: copy
;
1786 * tipc_recvstream - receive stream-oriented data
1787 * @m: descriptor for message info
1788 * @buflen: total size of user buffer area
1789 * @flags: receive flags
1791 * Used for SOCK_STREAM messages only. If not enough data is available
1792 * will optionally wait for more; never truncates data.
1794 * Returns size of returned message data, errno otherwise
1796 static int tipc_recvstream(struct socket
*sock
, struct msghdr
*m
,
1797 size_t buflen
, int flags
)
1799 struct sock
*sk
= sock
->sk
;
1800 struct tipc_sock
*tsk
= tipc_sk(sk
);
1801 struct sk_buff
*skb
;
1802 struct tipc_msg
*hdr
;
1803 struct tipc_skb_cb
*skb_cb
;
1804 bool peek
= flags
& MSG_PEEK
;
1805 int offset
, required
, copy
, copied
= 0;
1806 int hlen
, dlen
, err
, rc
;
1809 /* Catch invalid receive attempts */
1810 if (unlikely(!buflen
))
1815 if (unlikely(sk
->sk_state
== TIPC_OPEN
)) {
1819 required
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buflen
);
1820 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1823 /* Look at first msg in receive queue; wait if necessary */
1824 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1827 skb
= skb_peek(&sk
->sk_receive_queue
);
1828 skb_cb
= TIPC_SKB_CB(skb
);
1830 dlen
= msg_data_sz(hdr
);
1831 hlen
= msg_hdr_sz(hdr
);
1832 err
= msg_errcode(hdr
);
1834 /* Discard any empty non-errored (SYN-) message */
1835 if (unlikely(!dlen
&& !err
)) {
1836 tsk_advance_rx_queue(sk
);
1840 /* Collect msg meta data, incl. error code and rejected data */
1842 tipc_sk_set_orig_addr(m
, skb
);
1843 rc
= tipc_sk_anc_data_recv(m
, skb
, tsk
);
1849 /* Copy data if msg ok, otherwise return error/partial data */
1851 offset
= skb_cb
->bytes_read
;
1852 copy
= min_t(int, dlen
- offset
, buflen
- copied
);
1853 rc
= skb_copy_datagram_msg(skb
, hlen
+ offset
, m
, copy
);
1858 if (unlikely(offset
< dlen
)) {
1860 skb_cb
->bytes_read
= offset
;
1865 if ((err
!= TIPC_CONN_SHUTDOWN
) && !m
->msg_control
)
1874 tsk_advance_rx_queue(sk
);
1876 /* Send connection flow control advertisement when applicable */
1877 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1878 if (unlikely(tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
))
1879 tipc_sk_send_ack(tsk
);
1881 /* Exit if all requested data or FIN/error received */
1882 if (copied
== buflen
|| err
)
1885 } while (!skb_queue_empty(&sk
->sk_receive_queue
) || copied
< required
);
1888 return copied
? copied
: rc
;
1892 * tipc_write_space - wake up thread if port congestion is released
1895 static void tipc_write_space(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
, POLLOUT
|
1903 POLLWRNORM
| POLLWRBAND
);
1908 * tipc_data_ready - wake up threads to indicate messages have been received
1910 * @len: the length of messages
1912 static void tipc_data_ready(struct sock
*sk
)
1914 struct socket_wq
*wq
;
1917 wq
= rcu_dereference(sk
->sk_wq
);
1918 if (skwq_has_sleeper(wq
))
1919 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1920 POLLRDNORM
| POLLRDBAND
);
1924 static void tipc_sock_destruct(struct sock
*sk
)
1926 __skb_queue_purge(&sk
->sk_receive_queue
);
1929 static void tipc_sk_proto_rcv(struct sock
*sk
,
1930 struct sk_buff_head
*inputq
,
1931 struct sk_buff_head
*xmitq
)
1933 struct sk_buff
*skb
= __skb_dequeue(inputq
);
1934 struct tipc_sock
*tsk
= tipc_sk(sk
);
1935 struct tipc_msg
*hdr
= buf_msg(skb
);
1936 struct tipc_group
*grp
= tsk
->group
;
1937 bool wakeup
= false;
1939 switch (msg_user(hdr
)) {
1941 tipc_sk_conn_proto_rcv(tsk
, skb
, xmitq
);
1944 tipc_dest_del(&tsk
->cong_links
, msg_orignode(hdr
), 0);
1945 /* coupled with smp_rmb() in tipc_wait_for_cond() */
1947 tsk
->cong_link_cnt
--;
1950 case GROUP_PROTOCOL
:
1951 tipc_group_proto_rcv(grp
, &wakeup
, hdr
, inputq
, xmitq
);
1954 tipc_group_member_evt(tsk
->group
, &wakeup
, &sk
->sk_rcvbuf
,
1955 skb
, inputq
, xmitq
);
1963 sk
->sk_write_space(sk
);
1969 * tipc_filter_connect - Handle incoming message for a connection-based socket
1971 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1973 * Returns true if everything ok, false otherwise
1975 static bool tipc_sk_filter_connect(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
1977 struct sock
*sk
= &tsk
->sk
;
1978 struct net
*net
= sock_net(sk
);
1979 struct tipc_msg
*hdr
= buf_msg(skb
);
1980 u32 pport
= msg_origport(hdr
);
1981 u32 pnode
= msg_orignode(hdr
);
1983 if (unlikely(msg_mcast(hdr
)))
1986 switch (sk
->sk_state
) {
1987 case TIPC_CONNECTING
:
1988 /* Accept only ACK or NACK message */
1989 if (unlikely(!msg_connected(hdr
))) {
1990 if (pport
!= tsk_peer_port(tsk
) ||
1991 pnode
!= tsk_peer_node(tsk
))
1994 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1995 sk
->sk_err
= ECONNREFUSED
;
1996 sk
->sk_state_change(sk
);
2000 if (unlikely(msg_errcode(hdr
))) {
2001 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2002 sk
->sk_err
= ECONNREFUSED
;
2003 sk
->sk_state_change(sk
);
2007 if (unlikely(!msg_isdata(hdr
))) {
2008 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2009 sk
->sk_err
= EINVAL
;
2010 sk
->sk_state_change(sk
);
2014 tipc_sk_finish_conn(tsk
, msg_origport(hdr
), msg_orignode(hdr
));
2015 msg_set_importance(&tsk
->phdr
, msg_importance(hdr
));
2017 /* If 'ACK+' message, add to socket receive queue */
2018 if (msg_data_sz(hdr
))
2021 /* If empty 'ACK-' message, wake up sleeping connect() */
2022 sk
->sk_state_change(sk
);
2024 /* 'ACK-' message is neither accepted nor rejected: */
2025 msg_set_dest_droppable(hdr
, 1);
2029 case TIPC_DISCONNECTING
:
2032 /* Accept only SYN message */
2033 if (!msg_connected(hdr
) && !(msg_errcode(hdr
)))
2036 case TIPC_ESTABLISHED
:
2037 /* Accept only connection-based messages sent by peer */
2038 if (unlikely(!tsk_peer_msg(tsk
, hdr
)))
2041 if (unlikely(msg_errcode(hdr
))) {
2042 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2043 /* Let timer expire on it's own */
2044 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
2046 sk
->sk_state_change(sk
);
2050 pr_err("Unknown sk_state %u\n", sk
->sk_state
);
2057 * rcvbuf_limit - get proper overload limit of socket receive queue
2061 * For connection oriented messages, irrespective of importance,
2062 * default queue limit is 2 MB.
2064 * For connectionless messages, queue limits are based on message
2065 * importance as follows:
2067 * TIPC_LOW_IMPORTANCE (2 MB)
2068 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2069 * TIPC_HIGH_IMPORTANCE (8 MB)
2070 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2072 * Returns overload limit according to corresponding message importance
2074 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*skb
)
2076 struct tipc_sock
*tsk
= tipc_sk(sk
);
2077 struct tipc_msg
*hdr
= buf_msg(skb
);
2079 if (unlikely(msg_in_group(hdr
)))
2080 return sk
->sk_rcvbuf
;
2082 if (unlikely(!msg_connected(hdr
)))
2083 return sk
->sk_rcvbuf
<< msg_importance(hdr
);
2085 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
2086 return sk
->sk_rcvbuf
;
2088 return FLOWCTL_MSG_LIM
;
2092 * tipc_sk_filter_rcv - validate incoming message
2094 * @skb: pointer to message.
2096 * Enqueues message on receive queue if acceptable; optionally handles
2097 * disconnect indication for a connected socket.
2099 * Called with socket lock already taken
2102 static void tipc_sk_filter_rcv(struct sock
*sk
, struct sk_buff
*skb
,
2103 struct sk_buff_head
*xmitq
)
2105 bool sk_conn
= !tipc_sk_type_connectionless(sk
);
2106 struct tipc_sock
*tsk
= tipc_sk(sk
);
2107 struct tipc_group
*grp
= tsk
->group
;
2108 struct tipc_msg
*hdr
= buf_msg(skb
);
2109 struct net
*net
= sock_net(sk
);
2110 struct sk_buff_head inputq
;
2111 int limit
, err
= TIPC_OK
;
2113 TIPC_SKB_CB(skb
)->bytes_read
= 0;
2114 __skb_queue_head_init(&inputq
);
2115 __skb_queue_tail(&inputq
, skb
);
2117 if (unlikely(!msg_isdata(hdr
)))
2118 tipc_sk_proto_rcv(sk
, &inputq
, xmitq
);
2121 tipc_group_filter_msg(grp
, &inputq
, xmitq
);
2123 /* Validate and add to receive buffer if there is space */
2124 while ((skb
= __skb_dequeue(&inputq
))) {
2126 limit
= rcvbuf_limit(sk
, skb
);
2127 if ((sk_conn
&& !tipc_sk_filter_connect(tsk
, skb
)) ||
2128 (!sk_conn
&& msg_connected(hdr
)) ||
2129 (!grp
&& msg_in_group(hdr
)))
2130 err
= TIPC_ERR_NO_PORT
;
2131 else if (sk_rmem_alloc_get(sk
) + skb
->truesize
>= limit
)
2132 err
= TIPC_ERR_OVERLOAD
;
2134 if (unlikely(err
)) {
2135 tipc_skb_reject(net
, err
, skb
, xmitq
);
2139 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
2140 skb_set_owner_r(skb
, sk
);
2141 sk
->sk_data_ready(sk
);
2146 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2150 * Caller must hold socket lock
2152 static int tipc_sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
2154 unsigned int before
= sk_rmem_alloc_get(sk
);
2155 struct sk_buff_head xmitq
;
2158 __skb_queue_head_init(&xmitq
);
2160 tipc_sk_filter_rcv(sk
, skb
, &xmitq
);
2161 added
= sk_rmem_alloc_get(sk
) - before
;
2162 atomic_add(added
, &tipc_sk(sk
)->dupl_rcvcnt
);
2164 /* Send pending response/rejected messages, if any */
2165 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2170 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2171 * inputq and try adding them to socket or backlog queue
2172 * @inputq: list of incoming buffers with potentially different destinations
2173 * @sk: socket where the buffers should be enqueued
2174 * @dport: port number for the socket
2176 * Caller must hold socket lock
2178 static void tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
2179 u32 dport
, struct sk_buff_head
*xmitq
)
2181 unsigned long time_limit
= jiffies
+ 2;
2182 struct sk_buff
*skb
;
2187 while (skb_queue_len(inputq
)) {
2188 if (unlikely(time_after_eq(jiffies
, time_limit
)))
2191 skb
= tipc_skb_dequeue(inputq
, dport
);
2195 /* Add message directly to receive queue if possible */
2196 if (!sock_owned_by_user(sk
)) {
2197 tipc_sk_filter_rcv(sk
, skb
, xmitq
);
2201 /* Try backlog, compensating for double-counted bytes */
2202 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
2203 if (!sk
->sk_backlog
.len
)
2204 atomic_set(dcnt
, 0);
2205 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
2206 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
2209 /* Overload => reject message back to sender */
2210 onode
= tipc_own_addr(sock_net(sk
));
2211 if (tipc_msg_reverse(onode
, &skb
, TIPC_ERR_OVERLOAD
))
2212 __skb_queue_tail(xmitq
, skb
);
2218 * tipc_sk_rcv - handle a chain of incoming buffers
2219 * @inputq: buffer list containing the buffers
2220 * Consumes all buffers in list until inputq is empty
2221 * Note: may be called in multiple threads referring to the same queue
2223 void tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
2225 struct sk_buff_head xmitq
;
2226 u32 dnode
, dport
= 0;
2228 struct tipc_sock
*tsk
;
2230 struct sk_buff
*skb
;
2232 __skb_queue_head_init(&xmitq
);
2233 while (skb_queue_len(inputq
)) {
2234 dport
= tipc_skb_peek_port(inputq
, dport
);
2235 tsk
= tipc_sk_lookup(net
, dport
);
2239 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
2240 tipc_sk_enqueue(inputq
, sk
, dport
, &xmitq
);
2241 spin_unlock_bh(&sk
->sk_lock
.slock
);
2243 /* Send pending response/rejected messages, if any */
2244 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2248 /* No destination socket => dequeue skb if still there */
2249 skb
= tipc_skb_dequeue(inputq
, dport
);
2253 /* Try secondary lookup if unresolved named message */
2254 err
= TIPC_ERR_NO_PORT
;
2255 if (tipc_msg_lookup_dest(net
, skb
, &err
))
2258 /* Prepare for message rejection */
2259 if (!tipc_msg_reverse(tipc_own_addr(net
), &skb
, err
))
2262 dnode
= msg_destnode(buf_msg(skb
));
2263 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
2267 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
2269 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
2270 struct sock
*sk
= sock
->sk
;
2274 int err
= sock_error(sk
);
2279 if (signal_pending(current
))
2280 return sock_intr_errno(*timeo_p
);
2282 add_wait_queue(sk_sleep(sk
), &wait
);
2283 done
= sk_wait_event(sk
, timeo_p
,
2284 sk
->sk_state
!= TIPC_CONNECTING
, &wait
);
2285 remove_wait_queue(sk_sleep(sk
), &wait
);
2291 * tipc_connect - establish a connection to another TIPC port
2292 * @sock: socket structure
2293 * @dest: socket address for destination port
2294 * @destlen: size of socket address data structure
2295 * @flags: file-related flags associated with socket
2297 * Returns 0 on success, errno otherwise
2299 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
2300 int destlen
, int flags
)
2302 struct sock
*sk
= sock
->sk
;
2303 struct tipc_sock
*tsk
= tipc_sk(sk
);
2304 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
2305 struct msghdr m
= {NULL
,};
2306 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
2310 if (destlen
!= sizeof(struct sockaddr_tipc
))
2320 if (dst
->family
== AF_UNSPEC
) {
2321 memset(&tsk
->peer
, 0, sizeof(struct sockaddr_tipc
));
2322 if (!tipc_sk_type_connectionless(sk
))
2325 } else if (dst
->family
!= AF_TIPC
) {
2328 if (dst
->addrtype
!= TIPC_ADDR_ID
&& dst
->addrtype
!= TIPC_ADDR_NAME
)
2333 /* DGRAM/RDM connect(), just save the destaddr */
2334 if (tipc_sk_type_connectionless(sk
)) {
2335 memcpy(&tsk
->peer
, dest
, destlen
);
2339 previous
= sk
->sk_state
;
2341 switch (sk
->sk_state
) {
2343 /* Send a 'SYN-' to destination */
2345 m
.msg_namelen
= destlen
;
2347 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2348 * indicate send_msg() is never blocked.
2351 m
.msg_flags
= MSG_DONTWAIT
;
2353 res
= __tipc_sendmsg(sock
, &m
, 0);
2354 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
2357 /* Just entered TIPC_CONNECTING state; the only
2358 * difference is that return value in non-blocking
2359 * case is EINPROGRESS, rather than EALREADY.
2363 case TIPC_CONNECTING
:
2365 if (previous
== TIPC_CONNECTING
)
2369 timeout
= msecs_to_jiffies(timeout
);
2370 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2371 res
= tipc_wait_for_connect(sock
, &timeout
);
2373 case TIPC_ESTABLISHED
:
2386 * tipc_listen - allow socket to listen for incoming connections
2387 * @sock: socket structure
2390 * Returns 0 on success, errno otherwise
2392 static int tipc_listen(struct socket
*sock
, int len
)
2394 struct sock
*sk
= sock
->sk
;
2398 res
= tipc_set_sk_state(sk
, TIPC_LISTEN
);
2404 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
2406 struct sock
*sk
= sock
->sk
;
2410 /* True wake-one mechanism for incoming connections: only
2411 * one process gets woken up, not the 'whole herd'.
2412 * Since we do not 'race & poll' for established sockets
2413 * anymore, the common case will execute the loop only once.
2416 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
2417 TASK_INTERRUPTIBLE
);
2418 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
2420 timeo
= schedule_timeout(timeo
);
2424 if (!skb_queue_empty(&sk
->sk_receive_queue
))
2429 err
= sock_intr_errno(timeo
);
2430 if (signal_pending(current
))
2433 finish_wait(sk_sleep(sk
), &wait
);
2438 * tipc_accept - wait for connection request
2439 * @sock: listening socket
2440 * @newsock: new socket that is to be connected
2441 * @flags: file-related flags associated with socket
2443 * Returns 0 on success, errno otherwise
2445 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
,
2448 struct sock
*new_sk
, *sk
= sock
->sk
;
2449 struct sk_buff
*buf
;
2450 struct tipc_sock
*new_tsock
;
2451 struct tipc_msg
*msg
;
2457 if (sk
->sk_state
!= TIPC_LISTEN
) {
2461 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2462 res
= tipc_wait_for_accept(sock
, timeo
);
2466 buf
= skb_peek(&sk
->sk_receive_queue
);
2468 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, kern
);
2471 security_sk_clone(sock
->sk
, new_sock
->sk
);
2473 new_sk
= new_sock
->sk
;
2474 new_tsock
= tipc_sk(new_sk
);
2477 /* we lock on new_sk; but lockdep sees the lock on sk */
2478 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2481 * Reject any stray messages received by new socket
2482 * before the socket lock was taken (very, very unlikely)
2484 tsk_rej_rx_queue(new_sk
);
2486 /* Connect new socket to it's peer */
2487 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2489 tsk_set_importance(new_tsock
, msg_importance(msg
));
2490 if (msg_named(msg
)) {
2491 new_tsock
->conn_type
= msg_nametype(msg
);
2492 new_tsock
->conn_instance
= msg_nameinst(msg
);
2496 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2497 * Respond to 'SYN+' by queuing it on new socket.
2499 if (!msg_data_sz(msg
)) {
2500 struct msghdr m
= {NULL
,};
2502 tsk_advance_rx_queue(sk
);
2503 __tipc_sendstream(new_sock
, &m
, 0);
2505 __skb_dequeue(&sk
->sk_receive_queue
);
2506 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2507 skb_set_owner_r(buf
, new_sk
);
2509 release_sock(new_sk
);
2516 * tipc_shutdown - shutdown socket connection
2517 * @sock: socket structure
2518 * @how: direction to close (must be SHUT_RDWR)
2520 * Terminates connection (if necessary), then purges socket's receive queue.
2522 * Returns 0 on success, errno otherwise
2524 static int tipc_shutdown(struct socket
*sock
, int how
)
2526 struct sock
*sk
= sock
->sk
;
2529 if (how
!= SHUT_RDWR
)
2534 __tipc_shutdown(sock
, TIPC_CONN_SHUTDOWN
);
2535 sk
->sk_shutdown
= SEND_SHUTDOWN
;
2537 if (sk
->sk_state
== TIPC_DISCONNECTING
) {
2538 /* Discard any unreceived messages */
2539 __skb_queue_purge(&sk
->sk_receive_queue
);
2541 /* Wake up anyone sleeping in poll */
2542 sk
->sk_state_change(sk
);
2552 static void tipc_sk_timeout(struct timer_list
*t
)
2554 struct sock
*sk
= from_timer(sk
, t
, sk_timer
);
2555 struct tipc_sock
*tsk
= tipc_sk(sk
);
2556 u32 peer_port
= tsk_peer_port(tsk
);
2557 u32 peer_node
= tsk_peer_node(tsk
);
2558 u32 own_node
= tsk_own_node(tsk
);
2559 u32 own_port
= tsk
->portid
;
2560 struct net
*net
= sock_net(sk
);
2561 struct sk_buff
*skb
= NULL
;
2564 if (!tipc_sk_connected(sk
))
2567 /* Try again later if socket is busy */
2568 if (sock_owned_by_user(sk
)) {
2569 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ HZ
/ 20);
2573 if (tsk
->probe_unacked
) {
2574 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2575 tipc_node_remove_conn(net
, peer_node
, peer_port
);
2576 sk
->sk_state_change(sk
);
2579 /* Send new probe */
2580 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
, 0,
2581 peer_node
, own_node
, peer_port
, own_port
,
2583 tsk
->probe_unacked
= true;
2584 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
2588 tipc_node_xmit_skb(net
, skb
, peer_node
, own_port
);
2592 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2593 struct tipc_name_seq
const *seq
)
2595 struct sock
*sk
= &tsk
->sk
;
2596 struct net
*net
= sock_net(sk
);
2597 struct publication
*publ
;
2600 if (tipc_sk_connected(sk
))
2602 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2603 if (key
== tsk
->portid
)
2606 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2607 scope
, tsk
->portid
, key
);
2608 if (unlikely(!publ
))
2611 list_add(&publ
->pport_list
, &tsk
->publications
);
2617 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2618 struct tipc_name_seq
const *seq
)
2620 struct net
*net
= sock_net(&tsk
->sk
);
2621 struct publication
*publ
;
2622 struct publication
*safe
;
2625 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2627 if (publ
->scope
!= scope
)
2629 if (publ
->type
!= seq
->type
)
2631 if (publ
->lower
!= seq
->lower
)
2633 if (publ
->upper
!= seq
->upper
)
2635 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2636 publ
->ref
, publ
->key
);
2640 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2641 publ
->ref
, publ
->key
);
2644 if (list_empty(&tsk
->publications
))
2649 /* tipc_sk_reinit: set non-zero address in all existing sockets
2650 * when we go from standalone to network mode.
2652 void tipc_sk_reinit(struct net
*net
)
2654 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2655 struct rhashtable_iter iter
;
2656 struct tipc_sock
*tsk
;
2657 struct tipc_msg
*msg
;
2659 rhashtable_walk_enter(&tn
->sk_rht
, &iter
);
2662 tsk
= ERR_PTR(rhashtable_walk_start(&iter
));
2666 while ((tsk
= rhashtable_walk_next(&iter
)) && !IS_ERR(tsk
)) {
2667 sock_hold(&tsk
->sk
);
2668 rhashtable_walk_stop(&iter
);
2669 lock_sock(&tsk
->sk
);
2671 msg_set_prevnode(msg
, tn
->own_addr
);
2672 msg_set_orignode(msg
, tn
->own_addr
);
2673 release_sock(&tsk
->sk
);
2674 rhashtable_walk_start(&iter
);
2678 rhashtable_walk_stop(&iter
);
2679 } while (tsk
== ERR_PTR(-EAGAIN
));
2681 rhashtable_walk_exit(&iter
);
2684 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2686 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2687 struct tipc_sock
*tsk
;
2690 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2692 sock_hold(&tsk
->sk
);
2698 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2700 struct sock
*sk
= &tsk
->sk
;
2701 struct net
*net
= sock_net(sk
);
2702 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2703 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2704 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2706 while (remaining
--) {
2708 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2709 portid
= TIPC_MIN_PORT
;
2710 tsk
->portid
= portid
;
2711 sock_hold(&tsk
->sk
);
2712 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2721 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2723 struct sock
*sk
= &tsk
->sk
;
2724 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2726 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2727 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
2732 static const struct rhashtable_params tsk_rht_params
= {
2734 .head_offset
= offsetof(struct tipc_sock
, node
),
2735 .key_offset
= offsetof(struct tipc_sock
, portid
),
2736 .key_len
= sizeof(u32
), /* portid */
2737 .max_size
= 1048576,
2739 .automatic_shrinking
= true,
2742 int tipc_sk_rht_init(struct net
*net
)
2744 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2746 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2749 void tipc_sk_rht_destroy(struct net
*net
)
2751 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2753 /* Wait for socket readers to complete */
2756 rhashtable_destroy(&tn
->sk_rht
);
2759 static int tipc_sk_join(struct tipc_sock
*tsk
, struct tipc_group_req
*mreq
)
2761 struct net
*net
= sock_net(&tsk
->sk
);
2762 u32 domain
= addr_domain(net
, mreq
->scope
);
2763 struct tipc_group
*grp
= tsk
->group
;
2764 struct tipc_msg
*hdr
= &tsk
->phdr
;
2765 struct tipc_name_seq seq
;
2768 if (mreq
->type
< TIPC_RESERVED_TYPES
)
2772 grp
= tipc_group_create(net
, tsk
->portid
, mreq
);
2776 msg_set_lookup_scope(hdr
, mreq
->scope
);
2777 msg_set_nametype(hdr
, mreq
->type
);
2778 msg_set_dest_droppable(hdr
, true);
2779 seq
.type
= mreq
->type
;
2780 seq
.lower
= mreq
->instance
;
2781 seq
.upper
= seq
.lower
;
2782 tipc_nametbl_build_group(net
, grp
, mreq
->type
, domain
);
2783 rc
= tipc_sk_publish(tsk
, mreq
->scope
, &seq
);
2785 tipc_group_delete(net
, grp
);
2789 /* Eliminate any risk that a broadcast overtakes the sent JOIN */
2790 tsk
->mc_method
.rcast
= true;
2791 tsk
->mc_method
.mandatory
= true;
2795 static int tipc_sk_leave(struct tipc_sock
*tsk
)
2797 struct net
*net
= sock_net(&tsk
->sk
);
2798 struct tipc_group
*grp
= tsk
->group
;
2799 struct tipc_name_seq seq
;
2804 tipc_group_self(grp
, &seq
, &scope
);
2805 tipc_group_delete(net
, grp
);
2807 tipc_sk_withdraw(tsk
, scope
, &seq
);
2812 * tipc_setsockopt - set socket option
2813 * @sock: socket structure
2814 * @lvl: option level
2815 * @opt: option identifier
2816 * @ov: pointer to new option value
2817 * @ol: length of option value
2819 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2820 * (to ease compatibility).
2822 * Returns 0 on success, errno otherwise
2824 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2825 char __user
*ov
, unsigned int ol
)
2827 struct sock
*sk
= sock
->sk
;
2828 struct tipc_sock
*tsk
= tipc_sk(sk
);
2829 struct tipc_group_req mreq
;
2833 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2835 if (lvl
!= SOL_TIPC
)
2836 return -ENOPROTOOPT
;
2839 case TIPC_IMPORTANCE
:
2840 case TIPC_SRC_DROPPABLE
:
2841 case TIPC_DEST_DROPPABLE
:
2842 case TIPC_CONN_TIMEOUT
:
2843 if (ol
< sizeof(value
))
2845 if (get_user(value
, (u32 __user
*)ov
))
2848 case TIPC_GROUP_JOIN
:
2849 if (ol
< sizeof(mreq
))
2851 if (copy_from_user(&mreq
, ov
, sizeof(mreq
)))
2862 case TIPC_IMPORTANCE
:
2863 res
= tsk_set_importance(tsk
, value
);
2865 case TIPC_SRC_DROPPABLE
:
2866 if (sock
->type
!= SOCK_STREAM
)
2867 tsk_set_unreliable(tsk
, value
);
2871 case TIPC_DEST_DROPPABLE
:
2872 tsk_set_unreturnable(tsk
, value
);
2874 case TIPC_CONN_TIMEOUT
:
2875 tipc_sk(sk
)->conn_timeout
= value
;
2877 case TIPC_MCAST_BROADCAST
:
2878 tsk
->mc_method
.rcast
= false;
2879 tsk
->mc_method
.mandatory
= true;
2881 case TIPC_MCAST_REPLICAST
:
2882 tsk
->mc_method
.rcast
= true;
2883 tsk
->mc_method
.mandatory
= true;
2885 case TIPC_GROUP_JOIN
:
2886 res
= tipc_sk_join(tsk
, &mreq
);
2888 case TIPC_GROUP_LEAVE
:
2889 res
= tipc_sk_leave(tsk
);
2901 * tipc_getsockopt - get socket option
2902 * @sock: socket structure
2903 * @lvl: option level
2904 * @opt: option identifier
2905 * @ov: receptacle for option value
2906 * @ol: receptacle for length of option value
2908 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2909 * (to ease compatibility).
2911 * Returns 0 on success, errno otherwise
2913 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2914 char __user
*ov
, int __user
*ol
)
2916 struct sock
*sk
= sock
->sk
;
2917 struct tipc_sock
*tsk
= tipc_sk(sk
);
2918 struct tipc_name_seq seq
;
2923 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2924 return put_user(0, ol
);
2925 if (lvl
!= SOL_TIPC
)
2926 return -ENOPROTOOPT
;
2927 res
= get_user(len
, ol
);
2934 case TIPC_IMPORTANCE
:
2935 value
= tsk_importance(tsk
);
2937 case TIPC_SRC_DROPPABLE
:
2938 value
= tsk_unreliable(tsk
);
2940 case TIPC_DEST_DROPPABLE
:
2941 value
= tsk_unreturnable(tsk
);
2943 case TIPC_CONN_TIMEOUT
:
2944 value
= tsk
->conn_timeout
;
2945 /* no need to set "res", since already 0 at this point */
2947 case TIPC_NODE_RECVQ_DEPTH
:
2948 value
= 0; /* was tipc_queue_size, now obsolete */
2950 case TIPC_SOCK_RECVQ_DEPTH
:
2951 value
= skb_queue_len(&sk
->sk_receive_queue
);
2953 case TIPC_GROUP_JOIN
:
2956 tipc_group_self(tsk
->group
, &seq
, &scope
);
2966 return res
; /* "get" failed */
2968 if (len
< sizeof(value
))
2971 if (copy_to_user(ov
, &value
, sizeof(value
)))
2974 return put_user(sizeof(value
), ol
);
2977 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2979 struct sock
*sk
= sock
->sk
;
2980 struct tipc_sioc_ln_req lnr
;
2981 void __user
*argp
= (void __user
*)arg
;
2984 case SIOCGETLINKNAME
:
2985 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2987 if (!tipc_node_get_linkname(sock_net(sk
),
2988 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2989 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2990 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2994 return -EADDRNOTAVAIL
;
2996 return -ENOIOCTLCMD
;
3000 static int tipc_socketpair(struct socket
*sock1
, struct socket
*sock2
)
3002 struct tipc_sock
*tsk2
= tipc_sk(sock2
->sk
);
3003 struct tipc_sock
*tsk1
= tipc_sk(sock1
->sk
);
3004 u32 onode
= tipc_own_addr(sock_net(sock1
->sk
));
3006 tsk1
->peer
.family
= AF_TIPC
;
3007 tsk1
->peer
.addrtype
= TIPC_ADDR_ID
;
3008 tsk1
->peer
.scope
= TIPC_NODE_SCOPE
;
3009 tsk1
->peer
.addr
.id
.ref
= tsk2
->portid
;
3010 tsk1
->peer
.addr
.id
.node
= onode
;
3011 tsk2
->peer
.family
= AF_TIPC
;
3012 tsk2
->peer
.addrtype
= TIPC_ADDR_ID
;
3013 tsk2
->peer
.scope
= TIPC_NODE_SCOPE
;
3014 tsk2
->peer
.addr
.id
.ref
= tsk1
->portid
;
3015 tsk2
->peer
.addr
.id
.node
= onode
;
3017 tipc_sk_finish_conn(tsk1
, tsk2
->portid
, onode
);
3018 tipc_sk_finish_conn(tsk2
, tsk1
->portid
, onode
);
3022 /* Protocol switches for the various types of TIPC sockets */
3024 static const struct proto_ops msg_ops
= {
3025 .owner
= THIS_MODULE
,
3027 .release
= tipc_release
,
3029 .connect
= tipc_connect
,
3030 .socketpair
= tipc_socketpair
,
3031 .accept
= sock_no_accept
,
3032 .getname
= tipc_getname
,
3034 .ioctl
= tipc_ioctl
,
3035 .listen
= sock_no_listen
,
3036 .shutdown
= tipc_shutdown
,
3037 .setsockopt
= tipc_setsockopt
,
3038 .getsockopt
= tipc_getsockopt
,
3039 .sendmsg
= tipc_sendmsg
,
3040 .recvmsg
= tipc_recvmsg
,
3041 .mmap
= sock_no_mmap
,
3042 .sendpage
= sock_no_sendpage
3045 static const struct proto_ops packet_ops
= {
3046 .owner
= THIS_MODULE
,
3048 .release
= tipc_release
,
3050 .connect
= tipc_connect
,
3051 .socketpair
= tipc_socketpair
,
3052 .accept
= tipc_accept
,
3053 .getname
= tipc_getname
,
3055 .ioctl
= tipc_ioctl
,
3056 .listen
= tipc_listen
,
3057 .shutdown
= tipc_shutdown
,
3058 .setsockopt
= tipc_setsockopt
,
3059 .getsockopt
= tipc_getsockopt
,
3060 .sendmsg
= tipc_send_packet
,
3061 .recvmsg
= tipc_recvmsg
,
3062 .mmap
= sock_no_mmap
,
3063 .sendpage
= sock_no_sendpage
3066 static const struct proto_ops stream_ops
= {
3067 .owner
= THIS_MODULE
,
3069 .release
= tipc_release
,
3071 .connect
= tipc_connect
,
3072 .socketpair
= tipc_socketpair
,
3073 .accept
= tipc_accept
,
3074 .getname
= tipc_getname
,
3076 .ioctl
= tipc_ioctl
,
3077 .listen
= tipc_listen
,
3078 .shutdown
= tipc_shutdown
,
3079 .setsockopt
= tipc_setsockopt
,
3080 .getsockopt
= tipc_getsockopt
,
3081 .sendmsg
= tipc_sendstream
,
3082 .recvmsg
= tipc_recvstream
,
3083 .mmap
= sock_no_mmap
,
3084 .sendpage
= sock_no_sendpage
3087 static const struct net_proto_family tipc_family_ops
= {
3088 .owner
= THIS_MODULE
,
3090 .create
= tipc_sk_create
3093 static struct proto tipc_proto
= {
3095 .owner
= THIS_MODULE
,
3096 .obj_size
= sizeof(struct tipc_sock
),
3097 .sysctl_rmem
= sysctl_tipc_rmem
3101 * tipc_socket_init - initialize TIPC socket interface
3103 * Returns 0 on success, errno otherwise
3105 int tipc_socket_init(void)
3109 res
= proto_register(&tipc_proto
, 1);
3111 pr_err("Failed to register TIPC protocol type\n");
3115 res
= sock_register(&tipc_family_ops
);
3117 pr_err("Failed to register TIPC socket type\n");
3118 proto_unregister(&tipc_proto
);
3126 * tipc_socket_stop - stop TIPC socket interface
3128 void tipc_socket_stop(void)
3130 sock_unregister(tipc_family_ops
.family
);
3131 proto_unregister(&tipc_proto
);
3134 /* Caller should hold socket lock for the passed tipc socket. */
3135 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
3139 struct nlattr
*nest
;
3141 peer_node
= tsk_peer_node(tsk
);
3142 peer_port
= tsk_peer_port(tsk
);
3144 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
3146 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
3148 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
3151 if (tsk
->conn_type
!= 0) {
3152 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
3154 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
3156 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
3159 nla_nest_end(skb
, nest
);
3164 nla_nest_cancel(skb
, nest
);
3169 /* Caller should hold socket lock for the passed tipc socket. */
3170 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
3171 struct tipc_sock
*tsk
)
3175 struct nlattr
*attrs
;
3176 struct net
*net
= sock_net(skb
->sk
);
3177 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
3178 struct sock
*sk
= &tsk
->sk
;
3180 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3181 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
3185 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
3187 goto genlmsg_cancel
;
3188 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
3189 goto attr_msg_cancel
;
3190 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tn
->own_addr
))
3191 goto attr_msg_cancel
;
3193 if (tipc_sk_connected(sk
)) {
3194 err
= __tipc_nl_add_sk_con(skb
, tsk
);
3196 goto attr_msg_cancel
;
3197 } else if (!list_empty(&tsk
->publications
)) {
3198 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
3199 goto attr_msg_cancel
;
3201 nla_nest_end(skb
, attrs
);
3202 genlmsg_end(skb
, hdr
);
3207 nla_nest_cancel(skb
, attrs
);
3209 genlmsg_cancel(skb
, hdr
);
3214 int tipc_dump_start(struct netlink_callback
*cb
)
3216 return __tipc_dump_start(cb
, sock_net(cb
->skb
->sk
));
3218 EXPORT_SYMBOL(tipc_dump_start
);
3220 int __tipc_dump_start(struct netlink_callback
*cb
, struct net
*net
)
3222 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3223 struct rhashtable_iter
*iter
= (void *)cb
->args
[4];
3224 struct tipc_net
*tn
= tipc_net(net
);
3227 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
3231 cb
->args
[4] = (long)iter
;
3234 rhashtable_walk_enter(&tn
->sk_rht
, iter
);
3238 int tipc_dump_done(struct netlink_callback
*cb
)
3240 struct rhashtable_iter
*hti
= (void *)cb
->args
[4];
3242 rhashtable_walk_exit(hti
);
3246 EXPORT_SYMBOL(tipc_dump_done
);
3248 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3251 struct tipc_sock
*tsk
;
3252 const struct bucket_table
*tbl
;
3253 struct rhash_head
*pos
;
3254 struct net
*net
= sock_net(skb
->sk
);
3255 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
3256 u32 tbl_id
= cb
->args
[0];
3257 u32 prev_portid
= cb
->args
[1];
3260 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
3261 for (; tbl_id
< tbl
->size
; tbl_id
++) {
3262 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
3263 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
3264 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
3265 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3269 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
3271 prev_portid
= tsk
->portid
;
3272 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3276 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3281 cb
->args
[0] = tbl_id
;
3282 cb
->args
[1] = prev_portid
;
3287 /* Caller should hold socket lock for the passed tipc socket. */
3288 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
3289 struct netlink_callback
*cb
,
3290 struct publication
*publ
)
3293 struct nlattr
*attrs
;
3295 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3296 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
3300 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
3302 goto genlmsg_cancel
;
3304 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
3305 goto attr_msg_cancel
;
3306 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
3307 goto attr_msg_cancel
;
3308 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
3309 goto attr_msg_cancel
;
3310 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
3311 goto attr_msg_cancel
;
3313 nla_nest_end(skb
, attrs
);
3314 genlmsg_end(skb
, hdr
);
3319 nla_nest_cancel(skb
, attrs
);
3321 genlmsg_cancel(skb
, hdr
);
3326 /* Caller should hold socket lock for the passed tipc socket. */
3327 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
3328 struct netlink_callback
*cb
,
3329 struct tipc_sock
*tsk
, u32
*last_publ
)
3332 struct publication
*p
;
3335 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
3336 if (p
->key
== *last_publ
)
3339 if (p
->key
!= *last_publ
) {
3340 /* We never set seq or call nl_dump_check_consistent()
3341 * this means that setting prev_seq here will cause the
3342 * consistence check to fail in the netlink callback
3343 * handler. Resulting in the last NLMSG_DONE message
3344 * having the NLM_F_DUMP_INTR flag set.
3351 p
= list_first_entry(&tsk
->publications
, struct publication
,
3355 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
3356 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
3358 *last_publ
= p
->key
;
3367 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3370 u32 tsk_portid
= cb
->args
[0];
3371 u32 last_publ
= cb
->args
[1];
3372 u32 done
= cb
->args
[2];
3373 struct net
*net
= sock_net(skb
->sk
);
3374 struct tipc_sock
*tsk
;
3377 struct nlattr
**attrs
;
3378 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
3380 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
3384 if (!attrs
[TIPC_NLA_SOCK
])
3387 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
3388 attrs
[TIPC_NLA_SOCK
],
3389 tipc_nl_sock_policy
, NULL
);
3393 if (!sock
[TIPC_NLA_SOCK_REF
])
3396 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
3402 tsk
= tipc_sk_lookup(net
, tsk_portid
);
3406 lock_sock(&tsk
->sk
);
3407 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
3410 release_sock(&tsk
->sk
);
3413 cb
->args
[0] = tsk_portid
;
3414 cb
->args
[1] = last_publ
;