2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2014, 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/jhash.h>
40 #include "name_table.h"
46 #define SS_LISTENING -1 /* socket is listening */
47 #define SS_READY -2 /* socket is connectionless */
49 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
50 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
51 #define TIPC_FWD_MSG 1
52 #define TIPC_CONN_OK 0
53 #define TIPC_CONN_PROBING 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
58 * struct tipc_sock - TIPC socket structure
59 * @sk: socket - interacts with 'port' and with user via the socket API
60 * @connected: non-zero if port is currently connected to a peer port
61 * @conn_type: TIPC type used when connection was established
62 * @conn_instance: TIPC instance used when connection was established
63 * @published: non-zero if port has one or more associated names
64 * @max_pkt: maximum packet size "hint" used when building messages sent by port
65 * @portid: unique port identity in TIPC socket hash table
66 * @phdr: preformatted message header used when sending messages
67 * @port_list: adjacent ports in TIPC's global list of ports
68 * @publications: list of publications for port
69 * @pub_count: total # of publications port has made during its lifetime
73 * @port: port - interacts with 'sk' and with the rest of the TIPC stack
74 * @peer_name: the peer of the connection, if any
75 * @conn_timeout: the time we can wait for an unresponded setup request
76 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
77 * @link_cong: non-zero if owner must sleep because of link congestion
78 * @sent_unacked: # messages sent by socket, and not yet acked by peer
79 * @rcv_unacked: # messages read by user, but not yet acked back to peer
80 * @node: hash table node
81 * @rcu: rcu struct for tipc_sock
92 struct list_head sock_list
;
93 struct list_head publications
;
96 unsigned long probing_intv
;
97 struct timer_list timer
;
103 struct rhash_head node
;
107 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
108 static void tipc_data_ready(struct sock
*sk
);
109 static void tipc_write_space(struct sock
*sk
);
110 static int tipc_release(struct socket
*sock
);
111 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
);
112 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
);
113 static void tipc_sk_timeout(unsigned long data
);
114 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
115 struct tipc_name_seq
const *seq
);
116 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
117 struct tipc_name_seq
const *seq
);
118 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
);
119 static int tipc_sk_insert(struct tipc_sock
*tsk
);
120 static void tipc_sk_remove(struct tipc_sock
*tsk
);
122 static const struct proto_ops packet_ops
;
123 static const struct proto_ops stream_ops
;
124 static const struct proto_ops msg_ops
;
126 static struct proto tipc_proto
;
127 static struct proto tipc_proto_kern
;
129 static const struct nla_policy tipc_nl_sock_policy
[TIPC_NLA_SOCK_MAX
+ 1] = {
130 [TIPC_NLA_SOCK_UNSPEC
] = { .type
= NLA_UNSPEC
},
131 [TIPC_NLA_SOCK_ADDR
] = { .type
= NLA_U32
},
132 [TIPC_NLA_SOCK_REF
] = { .type
= NLA_U32
},
133 [TIPC_NLA_SOCK_CON
] = { .type
= NLA_NESTED
},
134 [TIPC_NLA_SOCK_HAS_PUBL
] = { .type
= NLA_FLAG
}
138 * Revised TIPC socket locking policy:
140 * Most socket operations take the standard socket lock when they start
141 * and hold it until they finish (or until they need to sleep). Acquiring
142 * this lock grants the owner exclusive access to the fields of the socket
143 * data structures, with the exception of the backlog queue. A few socket
144 * operations can be done without taking the socket lock because they only
145 * read socket information that never changes during the life of the socket.
147 * Socket operations may acquire the lock for the associated TIPC port if they
148 * need to perform an operation on the port. If any routine needs to acquire
149 * both the socket lock and the port lock it must take the socket lock first
150 * to avoid the risk of deadlock.
152 * The dispatcher handling incoming messages cannot grab the socket lock in
153 * the standard fashion, since invoked it runs at the BH level and cannot block.
154 * Instead, it checks to see if the socket lock is currently owned by someone,
155 * and either handles the message itself or adds it to the socket's backlog
156 * queue; in the latter case the queued message is processed once the process
157 * owning the socket lock releases it.
159 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
160 * the problem of a blocked socket operation preventing any other operations
161 * from occurring. However, applications must be careful if they have
162 * multiple threads trying to send (or receive) on the same socket, as these
163 * operations might interfere with each other. For example, doing a connect
164 * and a receive at the same time might allow the receive to consume the
165 * ACK message meant for the connect. While additional work could be done
166 * to try and overcome this, it doesn't seem to be worthwhile at the present.
168 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
169 * that another operation that must be performed in a non-blocking manner is
170 * not delayed for very long because the lock has already been taken.
172 * NOTE: This code assumes that certain fields of a port/socket pair are
173 * constant over its lifetime; such fields can be examined without taking
174 * the socket lock and/or port lock, and do not need to be re-read even
175 * after resuming processing after waiting. These fields include:
177 * - pointer to socket sk structure (aka tipc_sock structure)
178 * - pointer to port structure
182 static u32
tsk_peer_node(struct tipc_sock
*tsk
)
184 return msg_destnode(&tsk
->phdr
);
187 static u32
tsk_peer_port(struct tipc_sock
*tsk
)
189 return msg_destport(&tsk
->phdr
);
192 static bool tsk_unreliable(struct tipc_sock
*tsk
)
194 return msg_src_droppable(&tsk
->phdr
) != 0;
197 static void tsk_set_unreliable(struct tipc_sock
*tsk
, bool unreliable
)
199 msg_set_src_droppable(&tsk
->phdr
, unreliable
? 1 : 0);
202 static bool tsk_unreturnable(struct tipc_sock
*tsk
)
204 return msg_dest_droppable(&tsk
->phdr
) != 0;
207 static void tsk_set_unreturnable(struct tipc_sock
*tsk
, bool unreturnable
)
209 msg_set_dest_droppable(&tsk
->phdr
, unreturnable
? 1 : 0);
212 static int tsk_importance(struct tipc_sock
*tsk
)
214 return msg_importance(&tsk
->phdr
);
217 static int tsk_set_importance(struct tipc_sock
*tsk
, int imp
)
219 if (imp
> TIPC_CRITICAL_IMPORTANCE
)
221 msg_set_importance(&tsk
->phdr
, (u32
)imp
);
225 static struct tipc_sock
*tipc_sk(const struct sock
*sk
)
227 return container_of(sk
, struct tipc_sock
, sk
);
230 static int tsk_conn_cong(struct tipc_sock
*tsk
)
232 return tsk
->sent_unacked
>= TIPC_FLOWCTRL_WIN
;
236 * tsk_advance_rx_queue - discard first buffer in socket receive queue
238 * Caller must hold socket lock
240 static void tsk_advance_rx_queue(struct sock
*sk
)
242 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
246 * tsk_rej_rx_queue - reject all buffers in socket receive queue
248 * Caller must hold socket lock
250 static void tsk_rej_rx_queue(struct sock
*sk
)
255 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
))) {
256 if (tipc_msg_reverse(skb
, &dnode
, TIPC_ERR_NO_PORT
))
257 tipc_link_xmit_skb(sock_net(sk
), skb
, dnode
, 0);
261 /* tsk_peer_msg - verify if message was sent by connected port's peer
263 * Handles cases where the node's network address has changed from
264 * the default of <0.0.0> to its configured setting.
266 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
268 u32 peer_port
= tsk_peer_port(tsk
);
272 if (unlikely(!tsk
->connected
))
275 if (unlikely(msg_origport(msg
) != peer_port
))
278 orig_node
= msg_orignode(msg
);
279 peer_node
= tsk_peer_node(tsk
);
281 if (likely(orig_node
== peer_node
))
284 if (!orig_node
&& (peer_node
== tipc_own_addr
))
287 if (!peer_node
&& (orig_node
== tipc_own_addr
))
294 * tipc_sk_create - create a TIPC socket
295 * @net: network namespace (must be default network)
296 * @sock: pre-allocated socket structure
297 * @protocol: protocol indicator (must be 0)
298 * @kern: caused by kernel or by userspace?
300 * This routine creates additional data structures used by the TIPC socket,
301 * initializes them, and links them together.
303 * Returns 0 on success, errno otherwise
305 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
306 int protocol
, int kern
)
308 const struct proto_ops
*ops
;
311 struct tipc_sock
*tsk
;
312 struct tipc_msg
*msg
;
314 /* Validate arguments */
315 if (unlikely(protocol
!= 0))
316 return -EPROTONOSUPPORT
;
318 switch (sock
->type
) {
321 state
= SS_UNCONNECTED
;
325 state
= SS_UNCONNECTED
;
336 /* Allocate socket's protocol area */
338 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
);
340 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto_kern
);
346 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
347 INIT_LIST_HEAD(&tsk
->publications
);
349 tipc_msg_init(msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
352 /* Finish initializing socket data structures */
355 sock_init_data(sock
, sk
);
356 if (tipc_sk_insert(tsk
)) {
357 pr_warn("Socket create failed; port numbrer exhausted\n");
360 msg_set_origport(msg
, tsk
->portid
);
361 setup_timer(&tsk
->timer
, tipc_sk_timeout
, (unsigned long)tsk
);
362 sk
->sk_backlog_rcv
= tipc_backlog_rcv
;
363 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
364 sk
->sk_data_ready
= tipc_data_ready
;
365 sk
->sk_write_space
= tipc_write_space
;
366 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
367 tsk
->sent_unacked
= 0;
368 atomic_set(&tsk
->dupl_rcvcnt
, 0);
370 if (sock
->state
== SS_READY
) {
371 tsk_set_unreturnable(tsk
, true);
372 if (sock
->type
== SOCK_DGRAM
)
373 tsk_set_unreliable(tsk
, true);
379 * tipc_sock_create_local - create TIPC socket from inside TIPC module
380 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
382 * We cannot use sock_creat_kern here because it bumps module user count.
383 * Since socket owner and creator is the same module we must make sure
384 * that module count remains zero for module local sockets, otherwise
385 * we cannot do rmmod.
387 * Returns 0 on success, errno otherwise
389 int tipc_sock_create_local(int type
, struct socket
**res
)
393 rc
= sock_create_lite(AF_TIPC
, type
, 0, res
);
395 pr_err("Failed to create kernel socket\n");
398 tipc_sk_create(&init_net
, *res
, 0, 1);
404 * tipc_sock_release_local - release socket created by tipc_sock_create_local
405 * @sock: the socket to be released.
407 * Module reference count is not incremented when such sockets are created,
408 * so we must keep it from being decremented when they are released.
410 void tipc_sock_release_local(struct socket
*sock
)
418 * tipc_sock_accept_local - accept a connection on a socket created
419 * with tipc_sock_create_local. Use this function to avoid that
420 * module reference count is inadvertently incremented.
422 * @sock: the accepting socket
423 * @newsock: reference to the new socket to be created
424 * @flags: socket flags
427 int tipc_sock_accept_local(struct socket
*sock
, struct socket
**newsock
,
430 struct sock
*sk
= sock
->sk
;
433 ret
= sock_create_lite(sk
->sk_family
, sk
->sk_type
,
434 sk
->sk_protocol
, newsock
);
438 ret
= tipc_accept(sock
, *newsock
, flags
);
440 sock_release(*newsock
);
443 (*newsock
)->ops
= sock
->ops
;
447 static void tipc_sk_callback(struct rcu_head
*head
)
449 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
455 * tipc_release - destroy a TIPC socket
456 * @sock: socket to destroy
458 * This routine cleans up any messages that are still queued on the socket.
459 * For DGRAM and RDM socket types, all queued messages are rejected.
460 * For SEQPACKET and STREAM socket types, the first message is rejected
461 * and any others are discarded. (If the first message on a STREAM socket
462 * is partially-read, it is discarded and the next one is rejected instead.)
464 * NOTE: Rejected messages are not necessarily returned to the sender! They
465 * are returned or discarded according to the "destination droppable" setting
466 * specified for the message by the sender.
468 * Returns 0 on success, errno otherwise
470 static int tipc_release(struct socket
*sock
)
472 struct sock
*sk
= sock
->sk
;
473 struct net
*net
= sock_net(sk
);
474 struct tipc_sock
*tsk
;
476 u32 dnode
, probing_state
;
479 * Exit if socket isn't fully initialized (occurs when a failed accept()
480 * releases a pre-allocated child socket that was never used)
489 * Reject all unreceived messages, except on an active connection
490 * (which disconnects locally & sends a 'FIN+' to peer)
492 dnode
= tsk_peer_node(tsk
);
493 while (sock
->state
!= SS_DISCONNECTING
) {
494 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
497 if (TIPC_SKB_CB(skb
)->handle
!= NULL
)
500 if ((sock
->state
== SS_CONNECTING
) ||
501 (sock
->state
== SS_CONNECTED
)) {
502 sock
->state
= SS_DISCONNECTING
;
504 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
506 if (tipc_msg_reverse(skb
, &dnode
, TIPC_ERR_NO_PORT
))
507 tipc_link_xmit_skb(net
, skb
, dnode
, 0);
511 tipc_sk_withdraw(tsk
, 0, NULL
);
512 probing_state
= tsk
->probing_state
;
513 if (del_timer_sync(&tsk
->timer
) && probing_state
!= TIPC_CONN_PROBING
)
516 if (tsk
->connected
) {
517 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
518 SHORT_H_SIZE
, 0, dnode
, tipc_own_addr
,
520 tsk
->portid
, TIPC_ERR_NO_PORT
);
522 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
523 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
526 /* Discard any remaining (connection-based) messages in receive queue */
527 __skb_queue_purge(&sk
->sk_receive_queue
);
529 /* Reject any messages that accumulated in backlog queue */
530 sock
->state
= SS_DISCONNECTING
;
533 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
540 * tipc_bind - associate or disassocate TIPC name(s) with a socket
541 * @sock: socket structure
542 * @uaddr: socket address describing name(s) and desired operation
543 * @uaddr_len: size of socket address data structure
545 * Name and name sequence binding is indicated using a positive scope value;
546 * a negative scope value unbinds the specified name. Specifying no name
547 * (i.e. a socket address length of 0) unbinds all names from the socket.
549 * Returns 0 on success, errno otherwise
551 * NOTE: This routine doesn't need to take the socket lock since it doesn't
552 * access any non-constant socket information.
554 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
557 struct sock
*sk
= sock
->sk
;
558 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
559 struct tipc_sock
*tsk
= tipc_sk(sk
);
563 if (unlikely(!uaddr_len
)) {
564 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
568 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
572 if (addr
->family
!= AF_TIPC
) {
577 if (addr
->addrtype
== TIPC_ADDR_NAME
)
578 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
579 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
584 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
585 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
586 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
591 res
= (addr
->scope
> 0) ?
592 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
593 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
600 * tipc_getname - get port ID of socket or peer socket
601 * @sock: socket structure
602 * @uaddr: area for returned socket address
603 * @uaddr_len: area for returned length of socket address
604 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
606 * Returns 0 on success, errno otherwise
608 * NOTE: This routine doesn't need to take the socket lock since it only
609 * accesses socket information that is unchanging (or which changes in
610 * a completely predictable manner).
612 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
613 int *uaddr_len
, int peer
)
615 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
616 struct tipc_sock
*tsk
= tipc_sk(sock
->sk
);
618 memset(addr
, 0, sizeof(*addr
));
620 if ((sock
->state
!= SS_CONNECTED
) &&
621 ((peer
!= 2) || (sock
->state
!= SS_DISCONNECTING
)))
623 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
624 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
626 addr
->addr
.id
.ref
= tsk
->portid
;
627 addr
->addr
.id
.node
= tipc_own_addr
;
630 *uaddr_len
= sizeof(*addr
);
631 addr
->addrtype
= TIPC_ADDR_ID
;
632 addr
->family
= AF_TIPC
;
634 addr
->addr
.name
.domain
= 0;
640 * tipc_poll - read and possibly block on pollmask
641 * @file: file structure associated with the socket
642 * @sock: socket for which to calculate the poll bits
645 * Returns pollmask value
648 * It appears that the usual socket locking mechanisms are not useful here
649 * since the pollmask info is potentially out-of-date the moment this routine
650 * exits. TCP and other protocols seem to rely on higher level poll routines
651 * to handle any preventable race conditions, so TIPC will do the same ...
653 * TIPC sets the returned events as follows:
655 * socket state flags set
656 * ------------ ---------
657 * unconnected no read flags
658 * POLLOUT if port is not congested
660 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
663 * connected POLLIN/POLLRDNORM if data in rx queue
664 * POLLOUT if port is not congested
666 * disconnecting POLLIN/POLLRDNORM/POLLHUP
669 * listening POLLIN if SYN in rx queue
672 * ready POLLIN/POLLRDNORM if data in rx queue
673 * [connectionless] POLLOUT (since port cannot be congested)
675 * IMPORTANT: The fact that a read or write operation is indicated does NOT
676 * imply that the operation will succeed, merely that it should be performed
677 * and will not block.
679 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
682 struct sock
*sk
= sock
->sk
;
683 struct tipc_sock
*tsk
= tipc_sk(sk
);
686 sock_poll_wait(file
, sk_sleep(sk
), wait
);
688 switch ((int)sock
->state
) {
695 if (!tsk
->link_cong
&& !tsk_conn_cong(tsk
))
700 if (!skb_queue_empty(&sk
->sk_receive_queue
))
701 mask
|= (POLLIN
| POLLRDNORM
);
703 case SS_DISCONNECTING
:
704 mask
= (POLLIN
| POLLRDNORM
| POLLHUP
);
712 * tipc_sendmcast - send multicast message
713 * @sock: socket structure
714 * @seq: destination address
715 * @msg: message to send
716 * @dsz: total length of message data
717 * @timeo: timeout to wait for wakeup
719 * Called from function tipc_sendmsg(), which has done all sanity checks
720 * Returns the number of bytes sent on success, or errno
722 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
723 struct msghdr
*msg
, size_t dsz
, long timeo
)
725 struct sock
*sk
= sock
->sk
;
726 struct net
*net
= sock_net(sk
);
727 struct tipc_msg
*mhdr
= &tipc_sk(sk
)->phdr
;
728 struct sk_buff_head head
;
732 msg_set_type(mhdr
, TIPC_MCAST_MSG
);
733 msg_set_lookup_scope(mhdr
, TIPC_CLUSTER_SCOPE
);
734 msg_set_destport(mhdr
, 0);
735 msg_set_destnode(mhdr
, 0);
736 msg_set_nametype(mhdr
, seq
->type
);
737 msg_set_namelower(mhdr
, seq
->lower
);
738 msg_set_nameupper(mhdr
, seq
->upper
);
739 msg_set_hdr_sz(mhdr
, MCAST_H_SIZE
);
742 mtu
= tipc_bclink_get_mtu();
743 __skb_queue_head_init(&head
);
744 rc
= tipc_msg_build(mhdr
, msg
, 0, dsz
, mtu
, &head
);
745 if (unlikely(rc
< 0))
749 rc
= tipc_bclink_xmit(net
, &head
);
750 if (likely(rc
>= 0)) {
756 if (rc
!= -ELINKCONG
)
758 tipc_sk(sk
)->link_cong
= 1;
759 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
761 __skb_queue_purge(&head
);
766 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
768 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff
*buf
)
770 struct tipc_msg
*msg
= buf_msg(buf
);
771 struct tipc_port_list dports
= {0, NULL
, };
772 struct tipc_port_list
*item
;
774 uint i
, last
, dst
= 0;
775 u32 scope
= TIPC_CLUSTER_SCOPE
;
777 if (in_own_node(msg_orignode(msg
)))
778 scope
= TIPC_NODE_SCOPE
;
780 /* Create destination port list: */
781 tipc_nametbl_mc_translate(net
, msg_nametype(msg
), msg_namelower(msg
),
782 msg_nameupper(msg
), scope
, &dports
);
789 for (item
= &dports
; item
; item
= item
->next
) {
790 for (i
= 0; i
< PLSIZE
&& ++dst
<= last
; i
++) {
791 b
= (dst
!= last
) ? skb_clone(buf
, GFP_ATOMIC
) : buf
;
793 pr_warn("Failed do clone mcast rcv buffer\n");
796 msg_set_destport(msg
, item
->ports
[i
]);
800 tipc_port_list_free(&dports
);
804 * tipc_sk_proto_rcv - receive a connection mng protocol message
805 * @tsk: receiving socket
806 * @dnode: node to send response message to, if any
807 * @buf: buffer containing protocol message
808 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
809 * (CONN_PROBE_REPLY) message should be forwarded.
811 static int tipc_sk_proto_rcv(struct tipc_sock
*tsk
, u32
*dnode
,
814 struct tipc_msg
*msg
= buf_msg(buf
);
817 /* Ignore if connection cannot be validated: */
818 if (!tsk_peer_msg(tsk
, msg
))
821 tsk
->probing_state
= TIPC_CONN_OK
;
823 if (msg_type(msg
) == CONN_ACK
) {
824 conn_cong
= tsk_conn_cong(tsk
);
825 tsk
->sent_unacked
-= msg_msgcnt(msg
);
827 tsk
->sk
.sk_write_space(&tsk
->sk
);
828 } else if (msg_type(msg
) == CONN_PROBE
) {
829 if (!tipc_msg_reverse(buf
, dnode
, TIPC_OK
))
831 msg_set_type(msg
, CONN_PROBE_REPLY
);
834 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
840 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
)
842 struct sock
*sk
= sock
->sk
;
843 struct tipc_sock
*tsk
= tipc_sk(sk
);
848 int err
= sock_error(sk
);
851 if (sock
->state
== SS_DISCONNECTING
)
855 if (signal_pending(current
))
856 return sock_intr_errno(*timeo_p
);
858 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
859 done
= sk_wait_event(sk
, timeo_p
, !tsk
->link_cong
);
860 finish_wait(sk_sleep(sk
), &wait
);
866 * tipc_sendmsg - send message in connectionless manner
867 * @iocb: if NULL, indicates that socket lock is already held
868 * @sock: socket structure
869 * @m: message to send
870 * @dsz: amount of user data to be sent
872 * Message must have an destination specified explicitly.
873 * Used for SOCK_RDM and SOCK_DGRAM messages,
874 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
875 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
877 * Returns the number of bytes sent on success, or errno otherwise
879 static int tipc_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
880 struct msghdr
*m
, size_t dsz
)
882 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
883 struct sock
*sk
= sock
->sk
;
884 struct tipc_sock
*tsk
= tipc_sk(sk
);
885 struct net
*net
= sock_net(sk
);
886 struct tipc_msg
*mhdr
= &tsk
->phdr
;
888 struct sk_buff_head head
;
890 struct tipc_name_seq
*seq
= &dest
->addr
.nameseq
;
896 return -EDESTADDRREQ
;
898 if (unlikely((m
->msg_namelen
< sizeof(*dest
)) ||
899 (dest
->family
!= AF_TIPC
)))
902 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
908 if (unlikely(sock
->state
!= SS_READY
)) {
909 if (sock
->state
== SS_LISTENING
) {
913 if (sock
->state
!= SS_UNCONNECTED
) {
917 if (tsk
->published
) {
921 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
922 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
923 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
927 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
929 if (dest
->addrtype
== TIPC_ADDR_MCAST
) {
930 rc
= tipc_sendmcast(sock
, seq
, m
, dsz
, timeo
);
932 } else if (dest
->addrtype
== TIPC_ADDR_NAME
) {
933 u32 type
= dest
->addr
.name
.name
.type
;
934 u32 inst
= dest
->addr
.name
.name
.instance
;
935 u32 domain
= dest
->addr
.name
.domain
;
938 msg_set_type(mhdr
, TIPC_NAMED_MSG
);
939 msg_set_hdr_sz(mhdr
, NAMED_H_SIZE
);
940 msg_set_nametype(mhdr
, type
);
941 msg_set_nameinst(mhdr
, inst
);
942 msg_set_lookup_scope(mhdr
, tipc_addr_scope(domain
));
943 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
944 msg_set_destnode(mhdr
, dnode
);
945 msg_set_destport(mhdr
, dport
);
946 if (unlikely(!dport
&& !dnode
)) {
950 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
951 dnode
= dest
->addr
.id
.node
;
952 msg_set_type(mhdr
, TIPC_DIRECT_MSG
);
953 msg_set_lookup_scope(mhdr
, 0);
954 msg_set_destnode(mhdr
, dnode
);
955 msg_set_destport(mhdr
, dest
->addr
.id
.ref
);
956 msg_set_hdr_sz(mhdr
, BASIC_H_SIZE
);
960 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
961 __skb_queue_head_init(&head
);
962 rc
= tipc_msg_build(mhdr
, m
, 0, dsz
, mtu
, &head
);
967 skb
= skb_peek(&head
);
968 TIPC_SKB_CB(skb
)->wakeup_pending
= tsk
->link_cong
;
969 rc
= tipc_link_xmit(net
, &head
, dnode
, tsk
->portid
);
970 if (likely(rc
>= 0)) {
971 if (sock
->state
!= SS_READY
)
972 sock
->state
= SS_CONNECTING
;
978 if (rc
!= -ELINKCONG
)
981 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
983 __skb_queue_purge(&head
);
992 static int tipc_wait_for_sndpkt(struct socket
*sock
, long *timeo_p
)
994 struct sock
*sk
= sock
->sk
;
995 struct tipc_sock
*tsk
= tipc_sk(sk
);
1000 int err
= sock_error(sk
);
1003 if (sock
->state
== SS_DISCONNECTING
)
1005 else if (sock
->state
!= SS_CONNECTED
)
1009 if (signal_pending(current
))
1010 return sock_intr_errno(*timeo_p
);
1012 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1013 done
= sk_wait_event(sk
, timeo_p
,
1015 !tsk_conn_cong(tsk
)) ||
1017 finish_wait(sk_sleep(sk
), &wait
);
1023 * tipc_send_stream - send stream-oriented data
1025 * @sock: socket structure
1027 * @dsz: total length of data to be transmitted
1029 * Used for SOCK_STREAM data.
1031 * Returns the number of bytes sent on success (or partial success),
1032 * or errno if no data sent
1034 static int tipc_send_stream(struct kiocb
*iocb
, struct socket
*sock
,
1035 struct msghdr
*m
, size_t dsz
)
1037 struct sock
*sk
= sock
->sk
;
1038 struct net
*net
= sock_net(sk
);
1039 struct tipc_sock
*tsk
= tipc_sk(sk
);
1040 struct tipc_msg
*mhdr
= &tsk
->phdr
;
1041 struct sk_buff_head head
;
1042 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1043 u32 portid
= tsk
->portid
;
1047 uint mtu
, send
, sent
= 0;
1049 /* Handle implied connection establishment */
1050 if (unlikely(dest
)) {
1051 rc
= tipc_sendmsg(iocb
, sock
, m
, dsz
);
1052 if (dsz
&& (dsz
== rc
))
1053 tsk
->sent_unacked
= 1;
1056 if (dsz
> (uint
)INT_MAX
)
1062 if (unlikely(sock
->state
!= SS_CONNECTED
)) {
1063 if (sock
->state
== SS_DISCONNECTING
)
1070 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1071 dnode
= tsk_peer_node(tsk
);
1075 send
= min_t(uint
, dsz
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1076 __skb_queue_head_init(&head
);
1077 rc
= tipc_msg_build(mhdr
, m
, sent
, send
, mtu
, &head
);
1078 if (unlikely(rc
< 0))
1081 if (likely(!tsk_conn_cong(tsk
))) {
1082 rc
= tipc_link_xmit(net
, &head
, dnode
, portid
);
1084 tsk
->sent_unacked
++;
1090 if (rc
== -EMSGSIZE
) {
1091 tsk
->max_pkt
= tipc_node_get_mtu(net
, dnode
,
1095 if (rc
!= -ELINKCONG
)
1099 rc
= tipc_wait_for_sndpkt(sock
, &timeo
);
1101 __skb_queue_purge(&head
);
1106 return sent
? sent
: rc
;
1110 * tipc_send_packet - send a connection-oriented message
1111 * @iocb: if NULL, indicates that socket lock is already held
1112 * @sock: socket structure
1113 * @m: message to send
1114 * @dsz: length of data to be transmitted
1116 * Used for SOCK_SEQPACKET messages.
1118 * Returns the number of bytes sent on success, or errno otherwise
1120 static int tipc_send_packet(struct kiocb
*iocb
, struct socket
*sock
,
1121 struct msghdr
*m
, size_t dsz
)
1123 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1126 return tipc_send_stream(iocb
, sock
, m
, dsz
);
1129 /* tipc_sk_finish_conn - complete the setup of a connection
1131 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1134 struct net
*net
= sock_net(&tsk
->sk
);
1135 struct tipc_msg
*msg
= &tsk
->phdr
;
1137 msg_set_destnode(msg
, peer_node
);
1138 msg_set_destport(msg
, peer_port
);
1139 msg_set_type(msg
, TIPC_CONN_MSG
);
1140 msg_set_lookup_scope(msg
, 0);
1141 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1143 tsk
->probing_intv
= CONN_PROBING_INTERVAL
;
1144 tsk
->probing_state
= TIPC_CONN_OK
;
1146 if (!mod_timer(&tsk
->timer
, jiffies
+ tsk
->probing_intv
))
1147 sock_hold(&tsk
->sk
);
1148 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1149 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1153 * set_orig_addr - capture sender's address for received message
1154 * @m: descriptor for message info
1155 * @msg: received message header
1157 * Note: Address is not captured if not requested by receiver.
1159 static void set_orig_addr(struct msghdr
*m
, struct tipc_msg
*msg
)
1161 DECLARE_SOCKADDR(struct sockaddr_tipc
*, addr
, m
->msg_name
);
1164 addr
->family
= AF_TIPC
;
1165 addr
->addrtype
= TIPC_ADDR_ID
;
1166 memset(&addr
->addr
, 0, sizeof(addr
->addr
));
1167 addr
->addr
.id
.ref
= msg_origport(msg
);
1168 addr
->addr
.id
.node
= msg_orignode(msg
);
1169 addr
->addr
.name
.domain
= 0; /* could leave uninitialized */
1170 addr
->scope
= 0; /* could leave uninitialized */
1171 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1176 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1177 * @m: descriptor for message info
1178 * @msg: received message header
1179 * @tsk: TIPC port associated with message
1181 * Note: Ancillary data is not captured if not requested by receiver.
1183 * Returns 0 if successful, otherwise errno
1185 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1186 struct tipc_sock
*tsk
)
1194 if (likely(m
->msg_controllen
== 0))
1197 /* Optionally capture errored message object(s) */
1198 err
= msg
? msg_errcode(msg
) : 0;
1199 if (unlikely(err
)) {
1201 anc_data
[1] = msg_data_sz(msg
);
1202 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1206 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1213 /* Optionally capture message destination object */
1214 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1215 switch (dest_type
) {
1216 case TIPC_NAMED_MSG
:
1218 anc_data
[0] = msg_nametype(msg
);
1219 anc_data
[1] = msg_namelower(msg
);
1220 anc_data
[2] = msg_namelower(msg
);
1222 case TIPC_MCAST_MSG
:
1224 anc_data
[0] = msg_nametype(msg
);
1225 anc_data
[1] = msg_namelower(msg
);
1226 anc_data
[2] = msg_nameupper(msg
);
1229 has_name
= (tsk
->conn_type
!= 0);
1230 anc_data
[0] = tsk
->conn_type
;
1231 anc_data
[1] = tsk
->conn_instance
;
1232 anc_data
[2] = tsk
->conn_instance
;
1238 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1246 static void tipc_sk_send_ack(struct tipc_sock
*tsk
, uint ack
)
1248 struct net
*net
= sock_net(&tsk
->sk
);
1249 struct sk_buff
*skb
= NULL
;
1250 struct tipc_msg
*msg
;
1251 u32 peer_port
= tsk_peer_port(tsk
);
1252 u32 dnode
= tsk_peer_node(tsk
);
1254 if (!tsk
->connected
)
1256 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0, dnode
,
1257 tipc_own_addr
, peer_port
, tsk
->portid
, TIPC_OK
);
1261 msg_set_msgcnt(msg
, ack
);
1262 tipc_link_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1265 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1267 struct sock
*sk
= sock
->sk
;
1269 long timeo
= *timeop
;
1273 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1274 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1275 if (sock
->state
== SS_DISCONNECTING
) {
1280 timeo
= schedule_timeout(timeo
);
1284 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1286 err
= sock_intr_errno(timeo
);
1287 if (signal_pending(current
))
1293 finish_wait(sk_sleep(sk
), &wait
);
1299 * tipc_recvmsg - receive packet-oriented message
1301 * @m: descriptor for message info
1302 * @buf_len: total size of user buffer area
1303 * @flags: receive flags
1305 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1306 * If the complete message doesn't fit in user area, truncate it.
1308 * Returns size of returned message data, errno otherwise
1310 static int tipc_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1311 struct msghdr
*m
, size_t buf_len
, int flags
)
1313 struct sock
*sk
= sock
->sk
;
1314 struct tipc_sock
*tsk
= tipc_sk(sk
);
1315 struct sk_buff
*buf
;
1316 struct tipc_msg
*msg
;
1322 /* Catch invalid receive requests */
1323 if (unlikely(!buf_len
))
1328 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1333 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1336 /* Look for a message in receive queue; wait if necessary */
1337 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1341 /* Look at first message in receive queue */
1342 buf
= skb_peek(&sk
->sk_receive_queue
);
1344 sz
= msg_data_sz(msg
);
1345 err
= msg_errcode(msg
);
1347 /* Discard an empty non-errored message & try again */
1348 if ((!sz
) && (!err
)) {
1349 tsk_advance_rx_queue(sk
);
1353 /* Capture sender's address (optional) */
1354 set_orig_addr(m
, msg
);
1356 /* Capture ancillary data (optional) */
1357 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1361 /* Capture message data (if valid) & compute return value (always) */
1363 if (unlikely(buf_len
< sz
)) {
1365 m
->msg_flags
|= MSG_TRUNC
;
1367 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
), m
, sz
);
1372 if ((sock
->state
== SS_READY
) ||
1373 ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
))
1379 /* Consume received message (optional) */
1380 if (likely(!(flags
& MSG_PEEK
))) {
1381 if ((sock
->state
!= SS_READY
) &&
1382 (++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1383 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1384 tsk
->rcv_unacked
= 0;
1386 tsk_advance_rx_queue(sk
);
1394 * tipc_recv_stream - receive stream-oriented data
1396 * @m: descriptor for message info
1397 * @buf_len: total size of user buffer area
1398 * @flags: receive flags
1400 * Used for SOCK_STREAM messages only. If not enough data is available
1401 * will optionally wait for more; never truncates data.
1403 * Returns size of returned message data, errno otherwise
1405 static int tipc_recv_stream(struct kiocb
*iocb
, struct socket
*sock
,
1406 struct msghdr
*m
, size_t buf_len
, int flags
)
1408 struct sock
*sk
= sock
->sk
;
1409 struct tipc_sock
*tsk
= tipc_sk(sk
);
1410 struct sk_buff
*buf
;
1411 struct tipc_msg
*msg
;
1414 int sz_to_copy
, target
, needed
;
1419 /* Catch invalid receive attempts */
1420 if (unlikely(!buf_len
))
1425 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1430 target
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buf_len
);
1431 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1434 /* Look for a message in receive queue; wait if necessary */
1435 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1439 /* Look at first message in receive queue */
1440 buf
= skb_peek(&sk
->sk_receive_queue
);
1442 sz
= msg_data_sz(msg
);
1443 err
= msg_errcode(msg
);
1445 /* Discard an empty non-errored message & try again */
1446 if ((!sz
) && (!err
)) {
1447 tsk_advance_rx_queue(sk
);
1451 /* Optionally capture sender's address & ancillary data of first msg */
1452 if (sz_copied
== 0) {
1453 set_orig_addr(m
, msg
);
1454 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1459 /* Capture message data (if valid) & compute return value (always) */
1461 u32 offset
= (u32
)(unsigned long)(TIPC_SKB_CB(buf
)->handle
);
1464 needed
= (buf_len
- sz_copied
);
1465 sz_to_copy
= (sz
<= needed
) ? sz
: needed
;
1467 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
) + offset
,
1472 sz_copied
+= sz_to_copy
;
1474 if (sz_to_copy
< sz
) {
1475 if (!(flags
& MSG_PEEK
))
1476 TIPC_SKB_CB(buf
)->handle
=
1477 (void *)(unsigned long)(offset
+ sz_to_copy
);
1482 goto exit
; /* can't add error msg to valid data */
1484 if ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
)
1490 /* Consume received message (optional) */
1491 if (likely(!(flags
& MSG_PEEK
))) {
1492 if (unlikely(++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1493 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1494 tsk
->rcv_unacked
= 0;
1496 tsk_advance_rx_queue(sk
);
1499 /* Loop around if more data is required */
1500 if ((sz_copied
< buf_len
) && /* didn't get all requested data */
1501 (!skb_queue_empty(&sk
->sk_receive_queue
) ||
1502 (sz_copied
< target
)) && /* and more is ready or required */
1503 (!(flags
& MSG_PEEK
)) && /* and aren't just peeking at data */
1504 (!err
)) /* and haven't reached a FIN */
1509 return sz_copied
? sz_copied
: res
;
1513 * tipc_write_space - wake up thread if port congestion is released
1516 static void tipc_write_space(struct sock
*sk
)
1518 struct socket_wq
*wq
;
1521 wq
= rcu_dereference(sk
->sk_wq
);
1522 if (wq_has_sleeper(wq
))
1523 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1524 POLLWRNORM
| POLLWRBAND
);
1529 * tipc_data_ready - wake up threads to indicate messages have been received
1531 * @len: the length of messages
1533 static void tipc_data_ready(struct sock
*sk
)
1535 struct socket_wq
*wq
;
1538 wq
= rcu_dereference(sk
->sk_wq
);
1539 if (wq_has_sleeper(wq
))
1540 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1541 POLLRDNORM
| POLLRDBAND
);
1546 * filter_connect - Handle all incoming messages for a connection-based socket
1550 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1552 static int filter_connect(struct tipc_sock
*tsk
, struct sk_buff
**buf
)
1554 struct sock
*sk
= &tsk
->sk
;
1555 struct net
*net
= sock_net(sk
);
1556 struct socket
*sock
= sk
->sk_socket
;
1557 struct tipc_msg
*msg
= buf_msg(*buf
);
1558 int retval
= -TIPC_ERR_NO_PORT
;
1563 switch ((int)sock
->state
) {
1565 /* Accept only connection-based messages sent by peer */
1566 if (tsk_peer_msg(tsk
, msg
)) {
1567 if (unlikely(msg_errcode(msg
))) {
1568 sock
->state
= SS_DISCONNECTING
;
1570 /* let timer expire on it's own */
1571 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
1578 /* Accept only ACK or NACK message */
1580 if (unlikely(!msg_connected(msg
)))
1583 if (unlikely(msg_errcode(msg
))) {
1584 sock
->state
= SS_DISCONNECTING
;
1585 sk
->sk_err
= ECONNREFUSED
;
1590 if (unlikely(msg_importance(msg
) > TIPC_CRITICAL_IMPORTANCE
)) {
1591 sock
->state
= SS_DISCONNECTING
;
1592 sk
->sk_err
= EINVAL
;
1597 tipc_sk_finish_conn(tsk
, msg_origport(msg
), msg_orignode(msg
));
1598 msg_set_importance(&tsk
->phdr
, msg_importance(msg
));
1599 sock
->state
= SS_CONNECTED
;
1601 /* If an incoming message is an 'ACK-', it should be
1602 * discarded here because it doesn't contain useful
1603 * data. In addition, we should try to wake up
1604 * connect() routine if sleeping.
1606 if (msg_data_sz(msg
) == 0) {
1609 if (waitqueue_active(sk_sleep(sk
)))
1610 wake_up_interruptible(sk_sleep(sk
));
1615 case SS_UNCONNECTED
:
1616 /* Accept only SYN message */
1617 if (!msg_connected(msg
) && !(msg_errcode(msg
)))
1620 case SS_DISCONNECTING
:
1623 pr_err("Unknown socket state %u\n", sock
->state
);
1629 * rcvbuf_limit - get proper overload limit of socket receive queue
1633 * For all connection oriented messages, irrespective of importance,
1634 * the default overload value (i.e. 67MB) is set as limit.
1636 * For all connectionless messages, by default new queue limits are
1639 * TIPC_LOW_IMPORTANCE (4 MB)
1640 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1641 * TIPC_HIGH_IMPORTANCE (16 MB)
1642 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1644 * Returns overload limit according to corresponding message importance
1646 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*buf
)
1648 struct tipc_msg
*msg
= buf_msg(buf
);
1650 if (msg_connected(msg
))
1651 return sysctl_tipc_rmem
[2];
1653 return sk
->sk_rcvbuf
>> TIPC_CRITICAL_IMPORTANCE
<<
1654 msg_importance(msg
);
1658 * filter_rcv - validate incoming message
1662 * Enqueues message on receive queue if acceptable; optionally handles
1663 * disconnect indication for a connected socket.
1665 * Called with socket lock already taken; port lock may also be taken.
1667 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1668 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1670 static int filter_rcv(struct sock
*sk
, struct sk_buff
*buf
)
1672 struct socket
*sock
= sk
->sk_socket
;
1673 struct tipc_sock
*tsk
= tipc_sk(sk
);
1674 struct tipc_msg
*msg
= buf_msg(buf
);
1675 unsigned int limit
= rcvbuf_limit(sk
, buf
);
1679 if (unlikely(msg_user(msg
) == CONN_MANAGER
))
1680 return tipc_sk_proto_rcv(tsk
, &onode
, buf
);
1682 if (unlikely(msg_user(msg
) == SOCK_WAKEUP
)) {
1685 sk
->sk_write_space(sk
);
1689 /* Reject message if it is wrong sort of message for socket */
1690 if (msg_type(msg
) > TIPC_DIRECT_MSG
)
1691 return -TIPC_ERR_NO_PORT
;
1693 if (sock
->state
== SS_READY
) {
1694 if (msg_connected(msg
))
1695 return -TIPC_ERR_NO_PORT
;
1697 rc
= filter_connect(tsk
, &buf
);
1698 if (rc
!= TIPC_OK
|| buf
== NULL
)
1702 /* Reject message if there isn't room to queue it */
1703 if (sk_rmem_alloc_get(sk
) + buf
->truesize
>= limit
)
1704 return -TIPC_ERR_OVERLOAD
;
1706 /* Enqueue message */
1707 TIPC_SKB_CB(buf
)->handle
= NULL
;
1708 __skb_queue_tail(&sk
->sk_receive_queue
, buf
);
1709 skb_set_owner_r(buf
, sk
);
1711 sk
->sk_data_ready(sk
);
1716 * tipc_backlog_rcv - handle incoming message from backlog queue
1720 * Caller must hold socket lock, but not port lock.
1724 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1728 struct tipc_sock
*tsk
= tipc_sk(sk
);
1729 uint truesize
= skb
->truesize
;
1731 rc
= filter_rcv(sk
, skb
);
1734 if (atomic_read(&tsk
->dupl_rcvcnt
) < TIPC_CONN_OVERLOAD_LIMIT
)
1735 atomic_add(truesize
, &tsk
->dupl_rcvcnt
);
1739 if ((rc
< 0) && !tipc_msg_reverse(skb
, &onode
, -rc
))
1742 tipc_link_xmit_skb(sock_net(sk
), skb
, onode
, 0);
1748 * tipc_sk_rcv - handle incoming message
1749 * @skb: buffer containing arriving message
1751 * Returns 0 if success, or errno: -EHOSTUNREACH
1753 int tipc_sk_rcv(struct net
*net
, struct sk_buff
*skb
)
1755 struct tipc_sock
*tsk
;
1757 u32 dport
= msg_destport(buf_msg(skb
));
1762 /* Validate destination and message */
1763 tsk
= tipc_sk_lookup(net
, dport
);
1764 if (unlikely(!tsk
)) {
1765 rc
= tipc_msg_eval(net
, skb
, &dnode
);
1771 spin_lock_bh(&sk
->sk_lock
.slock
);
1773 if (!sock_owned_by_user(sk
)) {
1774 rc
= filter_rcv(sk
, skb
);
1776 if (sk
->sk_backlog
.len
== 0)
1777 atomic_set(&tsk
->dupl_rcvcnt
, 0);
1778 limit
= rcvbuf_limit(sk
, skb
) + atomic_read(&tsk
->dupl_rcvcnt
);
1779 if (sk_add_backlog(sk
, skb
, limit
))
1780 rc
= -TIPC_ERR_OVERLOAD
;
1782 spin_unlock_bh(&sk
->sk_lock
.slock
);
1787 if ((rc
< 0) && !tipc_msg_reverse(skb
, &dnode
, -rc
))
1788 return -EHOSTUNREACH
;
1790 tipc_link_xmit_skb(net
, skb
, dnode
, 0);
1791 return (rc
< 0) ? -EHOSTUNREACH
: 0;
1794 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
1796 struct sock
*sk
= sock
->sk
;
1801 int err
= sock_error(sk
);
1806 if (signal_pending(current
))
1807 return sock_intr_errno(*timeo_p
);
1809 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1810 done
= sk_wait_event(sk
, timeo_p
, sock
->state
!= SS_CONNECTING
);
1811 finish_wait(sk_sleep(sk
), &wait
);
1817 * tipc_connect - establish a connection to another TIPC port
1818 * @sock: socket structure
1819 * @dest: socket address for destination port
1820 * @destlen: size of socket address data structure
1821 * @flags: file-related flags associated with socket
1823 * Returns 0 on success, errno otherwise
1825 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
1826 int destlen
, int flags
)
1828 struct sock
*sk
= sock
->sk
;
1829 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
1830 struct msghdr m
= {NULL
,};
1831 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tipc_sk(sk
)->conn_timeout
;
1832 socket_state previous
;
1837 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1838 if (sock
->state
== SS_READY
) {
1844 * Reject connection attempt using multicast address
1846 * Note: send_msg() validates the rest of the address fields,
1847 * so there's no need to do it here
1849 if (dst
->addrtype
== TIPC_ADDR_MCAST
) {
1854 previous
= sock
->state
;
1855 switch (sock
->state
) {
1856 case SS_UNCONNECTED
:
1857 /* Send a 'SYN-' to destination */
1859 m
.msg_namelen
= destlen
;
1861 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1862 * indicate send_msg() is never blocked.
1865 m
.msg_flags
= MSG_DONTWAIT
;
1867 res
= tipc_sendmsg(NULL
, sock
, &m
, 0);
1868 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
1871 /* Just entered SS_CONNECTING state; the only
1872 * difference is that return value in non-blocking
1873 * case is EINPROGRESS, rather than EALREADY.
1877 if (previous
== SS_CONNECTING
)
1881 timeout
= msecs_to_jiffies(timeout
);
1882 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1883 res
= tipc_wait_for_connect(sock
, &timeout
);
1898 * tipc_listen - allow socket to listen for incoming connections
1899 * @sock: socket structure
1902 * Returns 0 on success, errno otherwise
1904 static int tipc_listen(struct socket
*sock
, int len
)
1906 struct sock
*sk
= sock
->sk
;
1911 if (sock
->state
!= SS_UNCONNECTED
)
1914 sock
->state
= SS_LISTENING
;
1922 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
1924 struct sock
*sk
= sock
->sk
;
1928 /* True wake-one mechanism for incoming connections: only
1929 * one process gets woken up, not the 'whole herd'.
1930 * Since we do not 'race & poll' for established sockets
1931 * anymore, the common case will execute the loop only once.
1934 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
1935 TASK_INTERRUPTIBLE
);
1936 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1938 timeo
= schedule_timeout(timeo
);
1942 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1945 if (sock
->state
!= SS_LISTENING
)
1947 err
= sock_intr_errno(timeo
);
1948 if (signal_pending(current
))
1954 finish_wait(sk_sleep(sk
), &wait
);
1959 * tipc_accept - wait for connection request
1960 * @sock: listening socket
1961 * @newsock: new socket that is to be connected
1962 * @flags: file-related flags associated with socket
1964 * Returns 0 on success, errno otherwise
1966 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
)
1968 struct sock
*new_sk
, *sk
= sock
->sk
;
1969 struct sk_buff
*buf
;
1970 struct tipc_sock
*new_tsock
;
1971 struct tipc_msg
*msg
;
1977 if (sock
->state
!= SS_LISTENING
) {
1981 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
1982 res
= tipc_wait_for_accept(sock
, timeo
);
1986 buf
= skb_peek(&sk
->sk_receive_queue
);
1988 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, 1);
1992 new_sk
= new_sock
->sk
;
1993 new_tsock
= tipc_sk(new_sk
);
1996 /* we lock on new_sk; but lockdep sees the lock on sk */
1997 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2000 * Reject any stray messages received by new socket
2001 * before the socket lock was taken (very, very unlikely)
2003 tsk_rej_rx_queue(new_sk
);
2005 /* Connect new socket to it's peer */
2006 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2007 new_sock
->state
= SS_CONNECTED
;
2009 tsk_set_importance(new_tsock
, msg_importance(msg
));
2010 if (msg_named(msg
)) {
2011 new_tsock
->conn_type
= msg_nametype(msg
);
2012 new_tsock
->conn_instance
= msg_nameinst(msg
);
2016 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2017 * Respond to 'SYN+' by queuing it on new socket.
2019 if (!msg_data_sz(msg
)) {
2020 struct msghdr m
= {NULL
,};
2022 tsk_advance_rx_queue(sk
);
2023 tipc_send_packet(NULL
, new_sock
, &m
, 0);
2025 __skb_dequeue(&sk
->sk_receive_queue
);
2026 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2027 skb_set_owner_r(buf
, new_sk
);
2029 release_sock(new_sk
);
2036 * tipc_shutdown - shutdown socket connection
2037 * @sock: socket structure
2038 * @how: direction to close (must be SHUT_RDWR)
2040 * Terminates connection (if necessary), then purges socket's receive queue.
2042 * Returns 0 on success, errno otherwise
2044 static int tipc_shutdown(struct socket
*sock
, int how
)
2046 struct sock
*sk
= sock
->sk
;
2047 struct net
*net
= sock_net(sk
);
2048 struct tipc_sock
*tsk
= tipc_sk(sk
);
2049 struct sk_buff
*skb
;
2053 if (how
!= SHUT_RDWR
)
2058 switch (sock
->state
) {
2063 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2064 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
2066 if (TIPC_SKB_CB(skb
)->handle
!= NULL
) {
2070 if (tipc_msg_reverse(skb
, &dnode
, TIPC_CONN_SHUTDOWN
))
2071 tipc_link_xmit_skb(net
, skb
, dnode
,
2073 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
2075 dnode
= tsk_peer_node(tsk
);
2076 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
2077 TIPC_CONN_MSG
, SHORT_H_SIZE
,
2078 0, dnode
, tipc_own_addr
,
2080 tsk
->portid
, TIPC_CONN_SHUTDOWN
);
2081 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
2084 sock
->state
= SS_DISCONNECTING
;
2085 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
2088 case SS_DISCONNECTING
:
2090 /* Discard any unreceived messages */
2091 __skb_queue_purge(&sk
->sk_receive_queue
);
2093 /* Wake up anyone sleeping in poll */
2094 sk
->sk_state_change(sk
);
2106 static void tipc_sk_timeout(unsigned long data
)
2108 struct tipc_sock
*tsk
= (struct tipc_sock
*)data
;
2109 struct sock
*sk
= &tsk
->sk
;
2110 struct sk_buff
*skb
= NULL
;
2111 u32 peer_port
, peer_node
;
2114 if (!tsk
->connected
) {
2118 peer_port
= tsk_peer_port(tsk
);
2119 peer_node
= tsk_peer_node(tsk
);
2121 if (tsk
->probing_state
== TIPC_CONN_PROBING
) {
2122 /* Previous probe not answered -> self abort */
2123 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
2124 SHORT_H_SIZE
, 0, tipc_own_addr
,
2125 peer_node
, tsk
->portid
, peer_port
,
2128 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
,
2129 0, peer_node
, tipc_own_addr
,
2130 peer_port
, tsk
->portid
, TIPC_OK
);
2131 tsk
->probing_state
= TIPC_CONN_PROBING
;
2132 if (!mod_timer(&tsk
->timer
, jiffies
+ tsk
->probing_intv
))
2137 tipc_link_xmit_skb(sock_net(sk
), skb
, peer_node
, tsk
->portid
);
2142 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2143 struct tipc_name_seq
const *seq
)
2145 struct net
*net
= sock_net(&tsk
->sk
);
2146 struct publication
*publ
;
2151 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2152 if (key
== tsk
->portid
)
2155 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2156 scope
, tsk
->portid
, key
);
2157 if (unlikely(!publ
))
2160 list_add(&publ
->pport_list
, &tsk
->publications
);
2166 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2167 struct tipc_name_seq
const *seq
)
2169 struct net
*net
= sock_net(&tsk
->sk
);
2170 struct publication
*publ
;
2171 struct publication
*safe
;
2174 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2176 if (publ
->scope
!= scope
)
2178 if (publ
->type
!= seq
->type
)
2180 if (publ
->lower
!= seq
->lower
)
2182 if (publ
->upper
!= seq
->upper
)
2184 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2185 publ
->ref
, publ
->key
);
2189 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2190 publ
->ref
, publ
->key
);
2193 if (list_empty(&tsk
->publications
))
2198 static int tipc_sk_show(struct tipc_sock
*tsk
, char *buf
,
2199 int len
, int full_id
)
2201 struct publication
*publ
;
2205 ret
= tipc_snprintf(buf
, len
, "<%u.%u.%u:%u>:",
2206 tipc_zone(tipc_own_addr
),
2207 tipc_cluster(tipc_own_addr
),
2208 tipc_node(tipc_own_addr
), tsk
->portid
);
2210 ret
= tipc_snprintf(buf
, len
, "%-10u:", tsk
->portid
);
2212 if (tsk
->connected
) {
2213 u32 dport
= tsk_peer_port(tsk
);
2214 u32 destnode
= tsk_peer_node(tsk
);
2216 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2217 " connected to <%u.%u.%u:%u>",
2218 tipc_zone(destnode
),
2219 tipc_cluster(destnode
),
2220 tipc_node(destnode
), dport
);
2221 if (tsk
->conn_type
!= 0)
2222 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2223 " via {%u,%u}", tsk
->conn_type
,
2224 tsk
->conn_instance
);
2225 } else if (tsk
->published
) {
2226 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
, " bound to");
2227 list_for_each_entry(publ
, &tsk
->publications
, pport_list
) {
2228 if (publ
->lower
== publ
->upper
)
2229 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2230 " {%u,%u}", publ
->type
,
2233 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2234 " {%u,%u,%u}", publ
->type
,
2235 publ
->lower
, publ
->upper
);
2238 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
, "\n");
2242 struct sk_buff
*tipc_sk_socks_show(struct net
*net
)
2244 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2245 const struct bucket_table
*tbl
;
2246 struct rhash_head
*pos
;
2247 struct sk_buff
*buf
;
2248 struct tlv_desc
*rep_tlv
;
2251 struct tipc_sock
*tsk
;
2255 buf
= tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN
));
2258 rep_tlv
= (struct tlv_desc
*)buf
->data
;
2259 pb
= TLV_DATA(rep_tlv
);
2260 pb_len
= ULTRA_STRING_MAX_LEN
;
2263 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2264 for (i
= 0; i
< tbl
->size
; i
++) {
2265 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2266 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2267 str_len
+= tipc_sk_show(tsk
, pb
+ str_len
,
2268 pb_len
- str_len
, 0);
2269 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2274 str_len
+= 1; /* for "\0" */
2275 skb_put(buf
, TLV_SPACE(str_len
));
2276 TLV_SET(rep_tlv
, TIPC_TLV_ULTRA_STRING
, NULL
, str_len
);
2281 /* tipc_sk_reinit: set non-zero address in all existing sockets
2282 * when we go from standalone to network mode.
2284 void tipc_sk_reinit(struct net
*net
)
2286 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2287 const struct bucket_table
*tbl
;
2288 struct rhash_head
*pos
;
2289 struct tipc_sock
*tsk
;
2290 struct tipc_msg
*msg
;
2294 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2295 for (i
= 0; i
< tbl
->size
; i
++) {
2296 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2297 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2299 msg_set_prevnode(msg
, tipc_own_addr
);
2300 msg_set_orignode(msg
, tipc_own_addr
);
2301 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2307 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2309 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2310 struct tipc_sock
*tsk
;
2313 tsk
= rhashtable_lookup(&tn
->sk_rht
, &portid
);
2315 sock_hold(&tsk
->sk
);
2321 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2323 struct sock
*sk
= &tsk
->sk
;
2324 struct net
*net
= sock_net(sk
);
2325 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2326 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2327 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2329 while (remaining
--) {
2331 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2332 portid
= TIPC_MIN_PORT
;
2333 tsk
->portid
= portid
;
2334 sock_hold(&tsk
->sk
);
2335 if (rhashtable_lookup_insert(&tn
->sk_rht
, &tsk
->node
))
2343 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2345 struct sock
*sk
= &tsk
->sk
;
2346 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2348 if (rhashtable_remove(&tn
->sk_rht
, &tsk
->node
)) {
2349 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
2354 int tipc_sk_rht_init(struct net
*net
)
2356 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2357 struct rhashtable_params rht_params
= {
2359 .head_offset
= offsetof(struct tipc_sock
, node
),
2360 .key_offset
= offsetof(struct tipc_sock
, portid
),
2361 .key_len
= sizeof(u32
), /* portid */
2363 .max_shift
= 20, /* 1M */
2364 .min_shift
= 8, /* 256 */
2365 .grow_decision
= rht_grow_above_75
,
2366 .shrink_decision
= rht_shrink_below_30
,
2369 return rhashtable_init(&tn
->sk_rht
, &rht_params
);
2372 void tipc_sk_rht_destroy(struct net
*net
)
2374 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2376 /* Wait for socket readers to complete */
2379 rhashtable_destroy(&tn
->sk_rht
);
2383 * tipc_setsockopt - set socket option
2384 * @sock: socket structure
2385 * @lvl: option level
2386 * @opt: option identifier
2387 * @ov: pointer to new option value
2388 * @ol: length of option value
2390 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2391 * (to ease compatibility).
2393 * Returns 0 on success, errno otherwise
2395 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2396 char __user
*ov
, unsigned int ol
)
2398 struct sock
*sk
= sock
->sk
;
2399 struct tipc_sock
*tsk
= tipc_sk(sk
);
2403 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2405 if (lvl
!= SOL_TIPC
)
2406 return -ENOPROTOOPT
;
2407 if (ol
< sizeof(value
))
2409 res
= get_user(value
, (u32 __user
*)ov
);
2416 case TIPC_IMPORTANCE
:
2417 res
= tsk_set_importance(tsk
, value
);
2419 case TIPC_SRC_DROPPABLE
:
2420 if (sock
->type
!= SOCK_STREAM
)
2421 tsk_set_unreliable(tsk
, value
);
2425 case TIPC_DEST_DROPPABLE
:
2426 tsk_set_unreturnable(tsk
, value
);
2428 case TIPC_CONN_TIMEOUT
:
2429 tipc_sk(sk
)->conn_timeout
= value
;
2430 /* no need to set "res", since already 0 at this point */
2442 * tipc_getsockopt - get socket option
2443 * @sock: socket structure
2444 * @lvl: option level
2445 * @opt: option identifier
2446 * @ov: receptacle for option value
2447 * @ol: receptacle for length of option value
2449 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2450 * (to ease compatibility).
2452 * Returns 0 on success, errno otherwise
2454 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2455 char __user
*ov
, int __user
*ol
)
2457 struct sock
*sk
= sock
->sk
;
2458 struct tipc_sock
*tsk
= tipc_sk(sk
);
2463 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2464 return put_user(0, ol
);
2465 if (lvl
!= SOL_TIPC
)
2466 return -ENOPROTOOPT
;
2467 res
= get_user(len
, ol
);
2474 case TIPC_IMPORTANCE
:
2475 value
= tsk_importance(tsk
);
2477 case TIPC_SRC_DROPPABLE
:
2478 value
= tsk_unreliable(tsk
);
2480 case TIPC_DEST_DROPPABLE
:
2481 value
= tsk_unreturnable(tsk
);
2483 case TIPC_CONN_TIMEOUT
:
2484 value
= tsk
->conn_timeout
;
2485 /* no need to set "res", since already 0 at this point */
2487 case TIPC_NODE_RECVQ_DEPTH
:
2488 value
= 0; /* was tipc_queue_size, now obsolete */
2490 case TIPC_SOCK_RECVQ_DEPTH
:
2491 value
= skb_queue_len(&sk
->sk_receive_queue
);
2500 return res
; /* "get" failed */
2502 if (len
< sizeof(value
))
2505 if (copy_to_user(ov
, &value
, sizeof(value
)))
2508 return put_user(sizeof(value
), ol
);
2511 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2513 struct sock
*sk
= sock
->sk
;
2514 struct tipc_sioc_ln_req lnr
;
2515 void __user
*argp
= (void __user
*)arg
;
2518 case SIOCGETLINKNAME
:
2519 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2521 if (!tipc_node_get_linkname(sock_net(sk
),
2522 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2523 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2524 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2528 return -EADDRNOTAVAIL
;
2530 return -ENOIOCTLCMD
;
2534 /* Protocol switches for the various types of TIPC sockets */
2536 static const struct proto_ops msg_ops
= {
2537 .owner
= THIS_MODULE
,
2539 .release
= tipc_release
,
2541 .connect
= tipc_connect
,
2542 .socketpair
= sock_no_socketpair
,
2543 .accept
= sock_no_accept
,
2544 .getname
= tipc_getname
,
2546 .ioctl
= tipc_ioctl
,
2547 .listen
= sock_no_listen
,
2548 .shutdown
= tipc_shutdown
,
2549 .setsockopt
= tipc_setsockopt
,
2550 .getsockopt
= tipc_getsockopt
,
2551 .sendmsg
= tipc_sendmsg
,
2552 .recvmsg
= tipc_recvmsg
,
2553 .mmap
= sock_no_mmap
,
2554 .sendpage
= sock_no_sendpage
2557 static const struct proto_ops packet_ops
= {
2558 .owner
= THIS_MODULE
,
2560 .release
= tipc_release
,
2562 .connect
= tipc_connect
,
2563 .socketpair
= sock_no_socketpair
,
2564 .accept
= tipc_accept
,
2565 .getname
= tipc_getname
,
2567 .ioctl
= tipc_ioctl
,
2568 .listen
= tipc_listen
,
2569 .shutdown
= tipc_shutdown
,
2570 .setsockopt
= tipc_setsockopt
,
2571 .getsockopt
= tipc_getsockopt
,
2572 .sendmsg
= tipc_send_packet
,
2573 .recvmsg
= tipc_recvmsg
,
2574 .mmap
= sock_no_mmap
,
2575 .sendpage
= sock_no_sendpage
2578 static const struct proto_ops stream_ops
= {
2579 .owner
= THIS_MODULE
,
2581 .release
= tipc_release
,
2583 .connect
= tipc_connect
,
2584 .socketpair
= sock_no_socketpair
,
2585 .accept
= tipc_accept
,
2586 .getname
= tipc_getname
,
2588 .ioctl
= tipc_ioctl
,
2589 .listen
= tipc_listen
,
2590 .shutdown
= tipc_shutdown
,
2591 .setsockopt
= tipc_setsockopt
,
2592 .getsockopt
= tipc_getsockopt
,
2593 .sendmsg
= tipc_send_stream
,
2594 .recvmsg
= tipc_recv_stream
,
2595 .mmap
= sock_no_mmap
,
2596 .sendpage
= sock_no_sendpage
2599 static const struct net_proto_family tipc_family_ops
= {
2600 .owner
= THIS_MODULE
,
2602 .create
= tipc_sk_create
2605 static struct proto tipc_proto
= {
2607 .owner
= THIS_MODULE
,
2608 .obj_size
= sizeof(struct tipc_sock
),
2609 .sysctl_rmem
= sysctl_tipc_rmem
2612 static struct proto tipc_proto_kern
= {
2614 .obj_size
= sizeof(struct tipc_sock
),
2615 .sysctl_rmem
= sysctl_tipc_rmem
2619 * tipc_socket_init - initialize TIPC socket interface
2621 * Returns 0 on success, errno otherwise
2623 int tipc_socket_init(void)
2627 res
= proto_register(&tipc_proto
, 1);
2629 pr_err("Failed to register TIPC protocol type\n");
2633 res
= sock_register(&tipc_family_ops
);
2635 pr_err("Failed to register TIPC socket type\n");
2636 proto_unregister(&tipc_proto
);
2644 * tipc_socket_stop - stop TIPC socket interface
2646 void tipc_socket_stop(void)
2648 sock_unregister(tipc_family_ops
.family
);
2649 proto_unregister(&tipc_proto
);
2652 /* Caller should hold socket lock for the passed tipc socket. */
2653 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
2657 struct nlattr
*nest
;
2659 peer_node
= tsk_peer_node(tsk
);
2660 peer_port
= tsk_peer_port(tsk
);
2662 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
2664 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
2666 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
2669 if (tsk
->conn_type
!= 0) {
2670 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
2672 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
2674 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
2677 nla_nest_end(skb
, nest
);
2682 nla_nest_cancel(skb
, nest
);
2687 /* Caller should hold socket lock for the passed tipc socket. */
2688 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
2689 struct tipc_sock
*tsk
)
2693 struct nlattr
*attrs
;
2695 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2696 &tipc_genl_v2_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
2700 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
2702 goto genlmsg_cancel
;
2703 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
2704 goto attr_msg_cancel
;
2705 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tipc_own_addr
))
2706 goto attr_msg_cancel
;
2708 if (tsk
->connected
) {
2709 err
= __tipc_nl_add_sk_con(skb
, tsk
);
2711 goto attr_msg_cancel
;
2712 } else if (!list_empty(&tsk
->publications
)) {
2713 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
2714 goto attr_msg_cancel
;
2716 nla_nest_end(skb
, attrs
);
2717 genlmsg_end(skb
, hdr
);
2722 nla_nest_cancel(skb
, attrs
);
2724 genlmsg_cancel(skb
, hdr
);
2729 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2732 struct tipc_sock
*tsk
;
2733 const struct bucket_table
*tbl
;
2734 struct rhash_head
*pos
;
2735 u32 prev_portid
= cb
->args
[0];
2736 u32 portid
= prev_portid
;
2737 struct net
*net
= sock_net(skb
->sk
);
2738 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2742 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2743 for (i
= 0; i
< tbl
->size
; i
++) {
2744 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2745 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2746 portid
= tsk
->portid
;
2747 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
2748 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2752 prev_portid
= portid
;
2757 cb
->args
[0] = prev_portid
;
2762 /* Caller should hold socket lock for the passed tipc socket. */
2763 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
2764 struct netlink_callback
*cb
,
2765 struct publication
*publ
)
2768 struct nlattr
*attrs
;
2770 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2771 &tipc_genl_v2_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
2775 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
2777 goto genlmsg_cancel
;
2779 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
2780 goto attr_msg_cancel
;
2781 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
2782 goto attr_msg_cancel
;
2783 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
2784 goto attr_msg_cancel
;
2785 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
2786 goto attr_msg_cancel
;
2788 nla_nest_end(skb
, attrs
);
2789 genlmsg_end(skb
, hdr
);
2794 nla_nest_cancel(skb
, attrs
);
2796 genlmsg_cancel(skb
, hdr
);
2801 /* Caller should hold socket lock for the passed tipc socket. */
2802 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
2803 struct netlink_callback
*cb
,
2804 struct tipc_sock
*tsk
, u32
*last_publ
)
2807 struct publication
*p
;
2810 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
2811 if (p
->key
== *last_publ
)
2814 if (p
->key
!= *last_publ
) {
2815 /* We never set seq or call nl_dump_check_consistent()
2816 * this means that setting prev_seq here will cause the
2817 * consistence check to fail in the netlink callback
2818 * handler. Resulting in the last NLMSG_DONE message
2819 * having the NLM_F_DUMP_INTR flag set.
2826 p
= list_first_entry(&tsk
->publications
, struct publication
,
2830 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
2831 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
2833 *last_publ
= p
->key
;
2842 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2845 u32 tsk_portid
= cb
->args
[0];
2846 u32 last_publ
= cb
->args
[1];
2847 u32 done
= cb
->args
[2];
2848 struct net
*net
= sock_net(skb
->sk
);
2849 struct tipc_sock
*tsk
;
2852 struct nlattr
**attrs
;
2853 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
2855 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
2859 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
2860 attrs
[TIPC_NLA_SOCK
],
2861 tipc_nl_sock_policy
);
2865 if (!sock
[TIPC_NLA_SOCK_REF
])
2868 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
2874 tsk
= tipc_sk_lookup(net
, tsk_portid
);
2878 lock_sock(&tsk
->sk
);
2879 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
2882 release_sock(&tsk
->sk
);
2885 cb
->args
[0] = tsk_portid
;
2886 cb
->args
[1] = last_publ
;