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.
38 #include "name_table.h"
41 #include <linux/export.h>
45 #define SS_LISTENING -1 /* socket is listening */
46 #define SS_READY -2 /* socket is connectionless */
48 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
49 #define CONN_PROBING_INTERVAL 3600000 /* [ms] => 1 h */
50 #define TIPC_FWD_MSG 1
51 #define TIPC_CONN_OK 0
52 #define TIPC_CONN_PROBING 1
55 * struct tipc_sock - TIPC socket structure
56 * @sk: socket - interacts with 'port' and with user via the socket API
57 * @connected: non-zero if port is currently connected to a peer port
58 * @conn_type: TIPC type used when connection was established
59 * @conn_instance: TIPC instance used when connection was established
60 * @published: non-zero if port has one or more associated names
61 * @max_pkt: maximum packet size "hint" used when building messages sent by port
62 * @ref: unique reference to port in TIPC object registry
63 * @phdr: preformatted message header used when sending messages
64 * @port_list: adjacent ports in TIPC's global list of ports
65 * @publications: list of publications for port
66 * @pub_count: total # of publications port has made during its lifetime
70 * @port: port - interacts with 'sk' and with the rest of the TIPC stack
71 * @peer_name: the peer of the connection, if any
72 * @conn_timeout: the time we can wait for an unresponded setup request
73 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
74 * @link_cong: non-zero if owner must sleep because of link congestion
75 * @sent_unacked: # messages sent by socket, and not yet acked by peer
76 * @rcv_unacked: # messages read by user, but not yet acked back to peer
87 struct list_head sock_list
;
88 struct list_head publications
;
92 struct timer_list timer
;
100 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
101 static void tipc_data_ready(struct sock
*sk
);
102 static void tipc_write_space(struct sock
*sk
);
103 static int tipc_release(struct socket
*sock
);
104 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
);
105 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
);
106 static void tipc_sk_timeout(unsigned long ref
);
107 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
108 struct tipc_name_seq
const *seq
);
109 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
110 struct tipc_name_seq
const *seq
);
111 static u32
tipc_sk_ref_acquire(struct tipc_sock
*tsk
);
112 static void tipc_sk_ref_discard(u32 ref
);
113 static struct tipc_sock
*tipc_sk_get(u32 ref
);
114 static struct tipc_sock
*tipc_sk_get_next(u32
*ref
);
115 static void tipc_sk_put(struct tipc_sock
*tsk
);
117 static const struct proto_ops packet_ops
;
118 static const struct proto_ops stream_ops
;
119 static const struct proto_ops msg_ops
;
121 static struct proto tipc_proto
;
122 static struct proto tipc_proto_kern
;
124 static const struct nla_policy tipc_nl_sock_policy
[TIPC_NLA_SOCK_MAX
+ 1] = {
125 [TIPC_NLA_SOCK_UNSPEC
] = { .type
= NLA_UNSPEC
},
126 [TIPC_NLA_SOCK_ADDR
] = { .type
= NLA_U32
},
127 [TIPC_NLA_SOCK_REF
] = { .type
= NLA_U32
},
128 [TIPC_NLA_SOCK_CON
] = { .type
= NLA_NESTED
},
129 [TIPC_NLA_SOCK_HAS_PUBL
] = { .type
= NLA_FLAG
}
133 * Revised TIPC socket locking policy:
135 * Most socket operations take the standard socket lock when they start
136 * and hold it until they finish (or until they need to sleep). Acquiring
137 * this lock grants the owner exclusive access to the fields of the socket
138 * data structures, with the exception of the backlog queue. A few socket
139 * operations can be done without taking the socket lock because they only
140 * read socket information that never changes during the life of the socket.
142 * Socket operations may acquire the lock for the associated TIPC port if they
143 * need to perform an operation on the port. If any routine needs to acquire
144 * both the socket lock and the port lock it must take the socket lock first
145 * to avoid the risk of deadlock.
147 * The dispatcher handling incoming messages cannot grab the socket lock in
148 * the standard fashion, since invoked it runs at the BH level and cannot block.
149 * Instead, it checks to see if the socket lock is currently owned by someone,
150 * and either handles the message itself or adds it to the socket's backlog
151 * queue; in the latter case the queued message is processed once the process
152 * owning the socket lock releases it.
154 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
155 * the problem of a blocked socket operation preventing any other operations
156 * from occurring. However, applications must be careful if they have
157 * multiple threads trying to send (or receive) on the same socket, as these
158 * operations might interfere with each other. For example, doing a connect
159 * and a receive at the same time might allow the receive to consume the
160 * ACK message meant for the connect. While additional work could be done
161 * to try and overcome this, it doesn't seem to be worthwhile at the present.
163 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
164 * that another operation that must be performed in a non-blocking manner is
165 * not delayed for very long because the lock has already been taken.
167 * NOTE: This code assumes that certain fields of a port/socket pair are
168 * constant over its lifetime; such fields can be examined without taking
169 * the socket lock and/or port lock, and do not need to be re-read even
170 * after resuming processing after waiting. These fields include:
172 * - pointer to socket sk structure (aka tipc_sock structure)
173 * - pointer to port structure
177 static u32
tsk_peer_node(struct tipc_sock
*tsk
)
179 return msg_destnode(&tsk
->phdr
);
182 static u32
tsk_peer_port(struct tipc_sock
*tsk
)
184 return msg_destport(&tsk
->phdr
);
187 static bool tsk_unreliable(struct tipc_sock
*tsk
)
189 return msg_src_droppable(&tsk
->phdr
) != 0;
192 static void tsk_set_unreliable(struct tipc_sock
*tsk
, bool unreliable
)
194 msg_set_src_droppable(&tsk
->phdr
, unreliable
? 1 : 0);
197 static bool tsk_unreturnable(struct tipc_sock
*tsk
)
199 return msg_dest_droppable(&tsk
->phdr
) != 0;
202 static void tsk_set_unreturnable(struct tipc_sock
*tsk
, bool unreturnable
)
204 msg_set_dest_droppable(&tsk
->phdr
, unreturnable
? 1 : 0);
207 static int tsk_importance(struct tipc_sock
*tsk
)
209 return msg_importance(&tsk
->phdr
);
212 static int tsk_set_importance(struct tipc_sock
*tsk
, int imp
)
214 if (imp
> TIPC_CRITICAL_IMPORTANCE
)
216 msg_set_importance(&tsk
->phdr
, (u32
)imp
);
220 static struct tipc_sock
*tipc_sk(const struct sock
*sk
)
222 return container_of(sk
, struct tipc_sock
, sk
);
225 static int tsk_conn_cong(struct tipc_sock
*tsk
)
227 return tsk
->sent_unacked
>= TIPC_FLOWCTRL_WIN
;
231 * tsk_advance_rx_queue - discard first buffer in socket receive queue
233 * Caller must hold socket lock
235 static void tsk_advance_rx_queue(struct sock
*sk
)
237 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
241 * tsk_rej_rx_queue - reject all buffers in socket receive queue
243 * Caller must hold socket lock
245 static void tsk_rej_rx_queue(struct sock
*sk
)
250 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
))) {
251 if (tipc_msg_reverse(skb
, &dnode
, TIPC_ERR_NO_PORT
))
252 tipc_link_xmit_skb(skb
, dnode
, 0);
256 /* tsk_peer_msg - verify if message was sent by connected port's peer
258 * Handles cases where the node's network address has changed from
259 * the default of <0.0.0> to its configured setting.
261 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
263 u32 peer_port
= tsk_peer_port(tsk
);
267 if (unlikely(!tsk
->connected
))
270 if (unlikely(msg_origport(msg
) != peer_port
))
273 orig_node
= msg_orignode(msg
);
274 peer_node
= tsk_peer_node(tsk
);
276 if (likely(orig_node
== peer_node
))
279 if (!orig_node
&& (peer_node
== tipc_own_addr
))
282 if (!peer_node
&& (orig_node
== tipc_own_addr
))
289 * tipc_sk_create - create a TIPC socket
290 * @net: network namespace (must be default network)
291 * @sock: pre-allocated socket structure
292 * @protocol: protocol indicator (must be 0)
293 * @kern: caused by kernel or by userspace?
295 * This routine creates additional data structures used by the TIPC socket,
296 * initializes them, and links them together.
298 * Returns 0 on success, errno otherwise
300 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
301 int protocol
, int kern
)
303 const struct proto_ops
*ops
;
306 struct tipc_sock
*tsk
;
307 struct tipc_msg
*msg
;
310 /* Validate arguments */
311 if (unlikely(protocol
!= 0))
312 return -EPROTONOSUPPORT
;
314 switch (sock
->type
) {
317 state
= SS_UNCONNECTED
;
321 state
= SS_UNCONNECTED
;
332 /* Allocate socket's protocol area */
334 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
);
336 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto_kern
);
342 ref
= tipc_sk_ref_acquire(tsk
);
344 pr_warn("Socket create failed; reference table exhausted\n");
347 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
349 INIT_LIST_HEAD(&tsk
->publications
);
351 tipc_msg_init(msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
353 msg_set_origport(msg
, ref
);
355 /* Finish initializing socket data structures */
358 sock_init_data(sock
, sk
);
359 k_init_timer(&tsk
->timer
, (Handler
)tipc_sk_timeout
, ref
);
360 sk
->sk_backlog_rcv
= tipc_backlog_rcv
;
361 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
362 sk
->sk_data_ready
= tipc_data_ready
;
363 sk
->sk_write_space
= tipc_write_space
;
364 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
365 tsk
->sent_unacked
= 0;
366 atomic_set(&tsk
->dupl_rcvcnt
, 0);
368 if (sock
->state
== SS_READY
) {
369 tsk_set_unreturnable(tsk
, true);
370 if (sock
->type
== SOCK_DGRAM
)
371 tsk_set_unreliable(tsk
, true);
377 * tipc_sock_create_local - create TIPC socket from inside TIPC module
378 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
380 * We cannot use sock_creat_kern here because it bumps module user count.
381 * Since socket owner and creator is the same module we must make sure
382 * that module count remains zero for module local sockets, otherwise
383 * we cannot do rmmod.
385 * Returns 0 on success, errno otherwise
387 int tipc_sock_create_local(int type
, struct socket
**res
)
391 rc
= sock_create_lite(AF_TIPC
, type
, 0, res
);
393 pr_err("Failed to create kernel socket\n");
396 tipc_sk_create(&init_net
, *res
, 0, 1);
402 * tipc_sock_release_local - release socket created by tipc_sock_create_local
403 * @sock: the socket to be released.
405 * Module reference count is not incremented when such sockets are created,
406 * so we must keep it from being decremented when they are released.
408 void tipc_sock_release_local(struct socket
*sock
)
416 * tipc_sock_accept_local - accept a connection on a socket created
417 * with tipc_sock_create_local. Use this function to avoid that
418 * module reference count is inadvertently incremented.
420 * @sock: the accepting socket
421 * @newsock: reference to the new socket to be created
422 * @flags: socket flags
425 int tipc_sock_accept_local(struct socket
*sock
, struct socket
**newsock
,
428 struct sock
*sk
= sock
->sk
;
431 ret
= sock_create_lite(sk
->sk_family
, sk
->sk_type
,
432 sk
->sk_protocol
, newsock
);
436 ret
= tipc_accept(sock
, *newsock
, flags
);
438 sock_release(*newsock
);
441 (*newsock
)->ops
= sock
->ops
;
446 * tipc_release - destroy a TIPC socket
447 * @sock: socket to destroy
449 * This routine cleans up any messages that are still queued on the socket.
450 * For DGRAM and RDM socket types, all queued messages are rejected.
451 * For SEQPACKET and STREAM socket types, the first message is rejected
452 * and any others are discarded. (If the first message on a STREAM socket
453 * is partially-read, it is discarded and the next one is rejected instead.)
455 * NOTE: Rejected messages are not necessarily returned to the sender! They
456 * are returned or discarded according to the "destination droppable" setting
457 * specified for the message by the sender.
459 * Returns 0 on success, errno otherwise
461 static int tipc_release(struct socket
*sock
)
463 struct sock
*sk
= sock
->sk
;
464 struct tipc_sock
*tsk
;
469 * Exit if socket isn't fully initialized (occurs when a failed accept()
470 * releases a pre-allocated child socket that was never used)
479 * Reject all unreceived messages, except on an active connection
480 * (which disconnects locally & sends a 'FIN+' to peer)
482 dnode
= tsk_peer_node(tsk
);
483 while (sock
->state
!= SS_DISCONNECTING
) {
484 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
487 if (TIPC_SKB_CB(skb
)->handle
!= NULL
)
490 if ((sock
->state
== SS_CONNECTING
) ||
491 (sock
->state
== SS_CONNECTED
)) {
492 sock
->state
= SS_DISCONNECTING
;
494 tipc_node_remove_conn(dnode
, tsk
->ref
);
496 if (tipc_msg_reverse(skb
, &dnode
, TIPC_ERR_NO_PORT
))
497 tipc_link_xmit_skb(skb
, dnode
, 0);
501 tipc_sk_withdraw(tsk
, 0, NULL
);
502 tipc_sk_ref_discard(tsk
->ref
);
503 k_cancel_timer(&tsk
->timer
);
504 if (tsk
->connected
) {
505 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
506 SHORT_H_SIZE
, 0, dnode
, tipc_own_addr
,
508 tsk
->ref
, TIPC_ERR_NO_PORT
);
510 tipc_link_xmit_skb(skb
, dnode
, tsk
->ref
);
511 tipc_node_remove_conn(dnode
, tsk
->ref
);
513 k_term_timer(&tsk
->timer
);
515 /* Discard any remaining (connection-based) messages in receive queue */
516 __skb_queue_purge(&sk
->sk_receive_queue
);
518 /* Reject any messages that accumulated in backlog queue */
519 sock
->state
= SS_DISCONNECTING
;
528 * tipc_bind - associate or disassocate TIPC name(s) with a socket
529 * @sock: socket structure
530 * @uaddr: socket address describing name(s) and desired operation
531 * @uaddr_len: size of socket address data structure
533 * Name and name sequence binding is indicated using a positive scope value;
534 * a negative scope value unbinds the specified name. Specifying no name
535 * (i.e. a socket address length of 0) unbinds all names from the socket.
537 * Returns 0 on success, errno otherwise
539 * NOTE: This routine doesn't need to take the socket lock since it doesn't
540 * access any non-constant socket information.
542 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
545 struct sock
*sk
= sock
->sk
;
546 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
547 struct tipc_sock
*tsk
= tipc_sk(sk
);
551 if (unlikely(!uaddr_len
)) {
552 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
556 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
560 if (addr
->family
!= AF_TIPC
) {
565 if (addr
->addrtype
== TIPC_ADDR_NAME
)
566 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
567 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
572 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
573 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
574 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
579 res
= (addr
->scope
> 0) ?
580 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
581 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
588 * tipc_getname - get port ID of socket or peer socket
589 * @sock: socket structure
590 * @uaddr: area for returned socket address
591 * @uaddr_len: area for returned length of socket address
592 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
594 * Returns 0 on success, errno otherwise
596 * NOTE: This routine doesn't need to take the socket lock since it only
597 * accesses socket information that is unchanging (or which changes in
598 * a completely predictable manner).
600 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
601 int *uaddr_len
, int peer
)
603 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
604 struct tipc_sock
*tsk
= tipc_sk(sock
->sk
);
606 memset(addr
, 0, sizeof(*addr
));
608 if ((sock
->state
!= SS_CONNECTED
) &&
609 ((peer
!= 2) || (sock
->state
!= SS_DISCONNECTING
)))
611 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
612 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
614 addr
->addr
.id
.ref
= tsk
->ref
;
615 addr
->addr
.id
.node
= tipc_own_addr
;
618 *uaddr_len
= sizeof(*addr
);
619 addr
->addrtype
= TIPC_ADDR_ID
;
620 addr
->family
= AF_TIPC
;
622 addr
->addr
.name
.domain
= 0;
628 * tipc_poll - read and possibly block on pollmask
629 * @file: file structure associated with the socket
630 * @sock: socket for which to calculate the poll bits
633 * Returns pollmask value
636 * It appears that the usual socket locking mechanisms are not useful here
637 * since the pollmask info is potentially out-of-date the moment this routine
638 * exits. TCP and other protocols seem to rely on higher level poll routines
639 * to handle any preventable race conditions, so TIPC will do the same ...
641 * TIPC sets the returned events as follows:
643 * socket state flags set
644 * ------------ ---------
645 * unconnected no read flags
646 * POLLOUT if port is not congested
648 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
651 * connected POLLIN/POLLRDNORM if data in rx queue
652 * POLLOUT if port is not congested
654 * disconnecting POLLIN/POLLRDNORM/POLLHUP
657 * listening POLLIN if SYN in rx queue
660 * ready POLLIN/POLLRDNORM if data in rx queue
661 * [connectionless] POLLOUT (since port cannot be congested)
663 * IMPORTANT: The fact that a read or write operation is indicated does NOT
664 * imply that the operation will succeed, merely that it should be performed
665 * and will not block.
667 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
670 struct sock
*sk
= sock
->sk
;
671 struct tipc_sock
*tsk
= tipc_sk(sk
);
674 sock_poll_wait(file
, sk_sleep(sk
), wait
);
676 switch ((int)sock
->state
) {
683 if (!tsk
->link_cong
&& !tsk_conn_cong(tsk
))
688 if (!skb_queue_empty(&sk
->sk_receive_queue
))
689 mask
|= (POLLIN
| POLLRDNORM
);
691 case SS_DISCONNECTING
:
692 mask
= (POLLIN
| POLLRDNORM
| POLLHUP
);
700 * tipc_sendmcast - send multicast message
701 * @sock: socket structure
702 * @seq: destination address
703 * @msg: message to send
704 * @dsz: total length of message data
705 * @timeo: timeout to wait for wakeup
707 * Called from function tipc_sendmsg(), which has done all sanity checks
708 * Returns the number of bytes sent on success, or errno
710 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
711 struct msghdr
*msg
, size_t dsz
, long timeo
)
713 struct sock
*sk
= sock
->sk
;
714 struct tipc_msg
*mhdr
= &tipc_sk(sk
)->phdr
;
715 struct sk_buff_head head
;
719 msg_set_type(mhdr
, TIPC_MCAST_MSG
);
720 msg_set_lookup_scope(mhdr
, TIPC_CLUSTER_SCOPE
);
721 msg_set_destport(mhdr
, 0);
722 msg_set_destnode(mhdr
, 0);
723 msg_set_nametype(mhdr
, seq
->type
);
724 msg_set_namelower(mhdr
, seq
->lower
);
725 msg_set_nameupper(mhdr
, seq
->upper
);
726 msg_set_hdr_sz(mhdr
, MCAST_H_SIZE
);
729 mtu
= tipc_bclink_get_mtu();
730 __skb_queue_head_init(&head
);
731 rc
= tipc_msg_build(mhdr
, msg
, 0, dsz
, mtu
, &head
);
732 if (unlikely(rc
< 0))
736 rc
= tipc_bclink_xmit(&head
);
737 if (likely(rc
>= 0)) {
743 if (rc
!= -ELINKCONG
)
745 tipc_sk(sk
)->link_cong
= 1;
746 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
748 __skb_queue_purge(&head
);
753 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
755 void tipc_sk_mcast_rcv(struct sk_buff
*buf
)
757 struct tipc_msg
*msg
= buf_msg(buf
);
758 struct tipc_port_list dports
= {0, NULL
, };
759 struct tipc_port_list
*item
;
761 uint i
, last
, dst
= 0;
762 u32 scope
= TIPC_CLUSTER_SCOPE
;
764 if (in_own_node(msg_orignode(msg
)))
765 scope
= TIPC_NODE_SCOPE
;
767 /* Create destination port list: */
768 tipc_nametbl_mc_translate(msg_nametype(msg
),
779 for (item
= &dports
; item
; item
= item
->next
) {
780 for (i
= 0; i
< PLSIZE
&& ++dst
<= last
; i
++) {
781 b
= (dst
!= last
) ? skb_clone(buf
, GFP_ATOMIC
) : buf
;
783 pr_warn("Failed do clone mcast rcv buffer\n");
786 msg_set_destport(msg
, item
->ports
[i
]);
790 tipc_port_list_free(&dports
);
794 * tipc_sk_proto_rcv - receive a connection mng protocol message
795 * @tsk: receiving socket
796 * @dnode: node to send response message to, if any
797 * @buf: buffer containing protocol message
798 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
799 * (CONN_PROBE_REPLY) message should be forwarded.
801 static int tipc_sk_proto_rcv(struct tipc_sock
*tsk
, u32
*dnode
,
804 struct tipc_msg
*msg
= buf_msg(buf
);
807 /* Ignore if connection cannot be validated: */
808 if (!tsk_peer_msg(tsk
, msg
))
811 tsk
->probing_state
= TIPC_CONN_OK
;
813 if (msg_type(msg
) == CONN_ACK
) {
814 conn_cong
= tsk_conn_cong(tsk
);
815 tsk
->sent_unacked
-= msg_msgcnt(msg
);
817 tsk
->sk
.sk_write_space(&tsk
->sk
);
818 } else if (msg_type(msg
) == CONN_PROBE
) {
819 if (!tipc_msg_reverse(buf
, dnode
, TIPC_OK
))
821 msg_set_type(msg
, CONN_PROBE_REPLY
);
824 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
830 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
)
832 struct sock
*sk
= sock
->sk
;
833 struct tipc_sock
*tsk
= tipc_sk(sk
);
838 int err
= sock_error(sk
);
841 if (sock
->state
== SS_DISCONNECTING
)
845 if (signal_pending(current
))
846 return sock_intr_errno(*timeo_p
);
848 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
849 done
= sk_wait_event(sk
, timeo_p
, !tsk
->link_cong
);
850 finish_wait(sk_sleep(sk
), &wait
);
856 * tipc_sendmsg - send message in connectionless manner
857 * @iocb: if NULL, indicates that socket lock is already held
858 * @sock: socket structure
859 * @m: message to send
860 * @dsz: amount of user data to be sent
862 * Message must have an destination specified explicitly.
863 * Used for SOCK_RDM and SOCK_DGRAM messages,
864 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
865 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
867 * Returns the number of bytes sent on success, or errno otherwise
869 static int tipc_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
870 struct msghdr
*m
, size_t dsz
)
872 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
873 struct sock
*sk
= sock
->sk
;
874 struct tipc_sock
*tsk
= tipc_sk(sk
);
875 struct tipc_msg
*mhdr
= &tsk
->phdr
;
877 struct sk_buff_head head
;
879 struct tipc_name_seq
*seq
= &dest
->addr
.nameseq
;
885 return -EDESTADDRREQ
;
887 if (unlikely((m
->msg_namelen
< sizeof(*dest
)) ||
888 (dest
->family
!= AF_TIPC
)))
891 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
897 if (unlikely(sock
->state
!= SS_READY
)) {
898 if (sock
->state
== SS_LISTENING
) {
902 if (sock
->state
!= SS_UNCONNECTED
) {
906 if (tsk
->published
) {
910 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
911 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
912 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
916 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
918 if (dest
->addrtype
== TIPC_ADDR_MCAST
) {
919 rc
= tipc_sendmcast(sock
, seq
, m
, dsz
, timeo
);
921 } else if (dest
->addrtype
== TIPC_ADDR_NAME
) {
922 u32 type
= dest
->addr
.name
.name
.type
;
923 u32 inst
= dest
->addr
.name
.name
.instance
;
924 u32 domain
= dest
->addr
.name
.domain
;
927 msg_set_type(mhdr
, TIPC_NAMED_MSG
);
928 msg_set_hdr_sz(mhdr
, NAMED_H_SIZE
);
929 msg_set_nametype(mhdr
, type
);
930 msg_set_nameinst(mhdr
, inst
);
931 msg_set_lookup_scope(mhdr
, tipc_addr_scope(domain
));
932 dport
= tipc_nametbl_translate(type
, inst
, &dnode
);
933 msg_set_destnode(mhdr
, dnode
);
934 msg_set_destport(mhdr
, dport
);
935 if (unlikely(!dport
&& !dnode
)) {
939 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
940 dnode
= dest
->addr
.id
.node
;
941 msg_set_type(mhdr
, TIPC_DIRECT_MSG
);
942 msg_set_lookup_scope(mhdr
, 0);
943 msg_set_destnode(mhdr
, dnode
);
944 msg_set_destport(mhdr
, dest
->addr
.id
.ref
);
945 msg_set_hdr_sz(mhdr
, BASIC_H_SIZE
);
949 mtu
= tipc_node_get_mtu(dnode
, tsk
->ref
);
950 __skb_queue_head_init(&head
);
951 rc
= tipc_msg_build(mhdr
, m
, 0, dsz
, mtu
, &head
);
956 skb
= skb_peek(&head
);
957 TIPC_SKB_CB(skb
)->wakeup_pending
= tsk
->link_cong
;
958 rc
= tipc_link_xmit(&head
, dnode
, tsk
->ref
);
959 if (likely(rc
>= 0)) {
960 if (sock
->state
!= SS_READY
)
961 sock
->state
= SS_CONNECTING
;
967 if (rc
!= -ELINKCONG
)
970 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
972 __skb_queue_purge(&head
);
981 static int tipc_wait_for_sndpkt(struct socket
*sock
, long *timeo_p
)
983 struct sock
*sk
= sock
->sk
;
984 struct tipc_sock
*tsk
= tipc_sk(sk
);
989 int err
= sock_error(sk
);
992 if (sock
->state
== SS_DISCONNECTING
)
994 else if (sock
->state
!= SS_CONNECTED
)
998 if (signal_pending(current
))
999 return sock_intr_errno(*timeo_p
);
1001 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1002 done
= sk_wait_event(sk
, timeo_p
,
1004 !tsk_conn_cong(tsk
)) ||
1006 finish_wait(sk_sleep(sk
), &wait
);
1012 * tipc_send_stream - send stream-oriented data
1014 * @sock: socket structure
1016 * @dsz: total length of data to be transmitted
1018 * Used for SOCK_STREAM data.
1020 * Returns the number of bytes sent on success (or partial success),
1021 * or errno if no data sent
1023 static int tipc_send_stream(struct kiocb
*iocb
, struct socket
*sock
,
1024 struct msghdr
*m
, size_t dsz
)
1026 struct sock
*sk
= sock
->sk
;
1027 struct tipc_sock
*tsk
= tipc_sk(sk
);
1028 struct tipc_msg
*mhdr
= &tsk
->phdr
;
1029 struct sk_buff_head head
;
1030 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1035 uint mtu
, send
, sent
= 0;
1037 /* Handle implied connection establishment */
1038 if (unlikely(dest
)) {
1039 rc
= tipc_sendmsg(iocb
, sock
, m
, dsz
);
1040 if (dsz
&& (dsz
== rc
))
1041 tsk
->sent_unacked
= 1;
1044 if (dsz
> (uint
)INT_MAX
)
1050 if (unlikely(sock
->state
!= SS_CONNECTED
)) {
1051 if (sock
->state
== SS_DISCONNECTING
)
1058 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1059 dnode
= tsk_peer_node(tsk
);
1063 send
= min_t(uint
, dsz
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1064 __skb_queue_head_init(&head
);
1065 rc
= tipc_msg_build(mhdr
, m
, sent
, send
, mtu
, &head
);
1066 if (unlikely(rc
< 0))
1069 if (likely(!tsk_conn_cong(tsk
))) {
1070 rc
= tipc_link_xmit(&head
, dnode
, ref
);
1072 tsk
->sent_unacked
++;
1078 if (rc
== -EMSGSIZE
) {
1079 tsk
->max_pkt
= tipc_node_get_mtu(dnode
, ref
);
1082 if (rc
!= -ELINKCONG
)
1086 rc
= tipc_wait_for_sndpkt(sock
, &timeo
);
1088 __skb_queue_purge(&head
);
1093 return sent
? sent
: rc
;
1097 * tipc_send_packet - send a connection-oriented message
1098 * @iocb: if NULL, indicates that socket lock is already held
1099 * @sock: socket structure
1100 * @m: message to send
1101 * @dsz: length of data to be transmitted
1103 * Used for SOCK_SEQPACKET messages.
1105 * Returns the number of bytes sent on success, or errno otherwise
1107 static int tipc_send_packet(struct kiocb
*iocb
, struct socket
*sock
,
1108 struct msghdr
*m
, size_t dsz
)
1110 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1113 return tipc_send_stream(iocb
, sock
, m
, dsz
);
1116 /* tipc_sk_finish_conn - complete the setup of a connection
1118 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1121 struct tipc_msg
*msg
= &tsk
->phdr
;
1123 msg_set_destnode(msg
, peer_node
);
1124 msg_set_destport(msg
, peer_port
);
1125 msg_set_type(msg
, TIPC_CONN_MSG
);
1126 msg_set_lookup_scope(msg
, 0);
1127 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1129 tsk
->probing_interval
= CONN_PROBING_INTERVAL
;
1130 tsk
->probing_state
= TIPC_CONN_OK
;
1132 k_start_timer(&tsk
->timer
, tsk
->probing_interval
);
1133 tipc_node_add_conn(peer_node
, tsk
->ref
, peer_port
);
1134 tsk
->max_pkt
= tipc_node_get_mtu(peer_node
, tsk
->ref
);
1138 * set_orig_addr - capture sender's address for received message
1139 * @m: descriptor for message info
1140 * @msg: received message header
1142 * Note: Address is not captured if not requested by receiver.
1144 static void set_orig_addr(struct msghdr
*m
, struct tipc_msg
*msg
)
1146 DECLARE_SOCKADDR(struct sockaddr_tipc
*, addr
, m
->msg_name
);
1149 addr
->family
= AF_TIPC
;
1150 addr
->addrtype
= TIPC_ADDR_ID
;
1151 memset(&addr
->addr
, 0, sizeof(addr
->addr
));
1152 addr
->addr
.id
.ref
= msg_origport(msg
);
1153 addr
->addr
.id
.node
= msg_orignode(msg
);
1154 addr
->addr
.name
.domain
= 0; /* could leave uninitialized */
1155 addr
->scope
= 0; /* could leave uninitialized */
1156 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1161 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1162 * @m: descriptor for message info
1163 * @msg: received message header
1164 * @tsk: TIPC port associated with message
1166 * Note: Ancillary data is not captured if not requested by receiver.
1168 * Returns 0 if successful, otherwise errno
1170 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1171 struct tipc_sock
*tsk
)
1179 if (likely(m
->msg_controllen
== 0))
1182 /* Optionally capture errored message object(s) */
1183 err
= msg
? msg_errcode(msg
) : 0;
1184 if (unlikely(err
)) {
1186 anc_data
[1] = msg_data_sz(msg
);
1187 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1191 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1198 /* Optionally capture message destination object */
1199 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1200 switch (dest_type
) {
1201 case TIPC_NAMED_MSG
:
1203 anc_data
[0] = msg_nametype(msg
);
1204 anc_data
[1] = msg_namelower(msg
);
1205 anc_data
[2] = msg_namelower(msg
);
1207 case TIPC_MCAST_MSG
:
1209 anc_data
[0] = msg_nametype(msg
);
1210 anc_data
[1] = msg_namelower(msg
);
1211 anc_data
[2] = msg_nameupper(msg
);
1214 has_name
= (tsk
->conn_type
!= 0);
1215 anc_data
[0] = tsk
->conn_type
;
1216 anc_data
[1] = tsk
->conn_instance
;
1217 anc_data
[2] = tsk
->conn_instance
;
1223 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1231 static void tipc_sk_send_ack(struct tipc_sock
*tsk
, uint ack
)
1233 struct sk_buff
*skb
= NULL
;
1234 struct tipc_msg
*msg
;
1235 u32 peer_port
= tsk_peer_port(tsk
);
1236 u32 dnode
= tsk_peer_node(tsk
);
1238 if (!tsk
->connected
)
1240 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0, dnode
,
1241 tipc_own_addr
, peer_port
, tsk
->ref
, TIPC_OK
);
1245 msg_set_msgcnt(msg
, ack
);
1246 tipc_link_xmit_skb(skb
, dnode
, msg_link_selector(msg
));
1249 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1251 struct sock
*sk
= sock
->sk
;
1253 long timeo
= *timeop
;
1257 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1258 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1259 if (sock
->state
== SS_DISCONNECTING
) {
1264 timeo
= schedule_timeout(timeo
);
1268 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1270 err
= sock_intr_errno(timeo
);
1271 if (signal_pending(current
))
1277 finish_wait(sk_sleep(sk
), &wait
);
1283 * tipc_recvmsg - receive packet-oriented message
1285 * @m: descriptor for message info
1286 * @buf_len: total size of user buffer area
1287 * @flags: receive flags
1289 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1290 * If the complete message doesn't fit in user area, truncate it.
1292 * Returns size of returned message data, errno otherwise
1294 static int tipc_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1295 struct msghdr
*m
, size_t buf_len
, int flags
)
1297 struct sock
*sk
= sock
->sk
;
1298 struct tipc_sock
*tsk
= tipc_sk(sk
);
1299 struct sk_buff
*buf
;
1300 struct tipc_msg
*msg
;
1306 /* Catch invalid receive requests */
1307 if (unlikely(!buf_len
))
1312 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1317 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1320 /* Look for a message in receive queue; wait if necessary */
1321 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1325 /* Look at first message in receive queue */
1326 buf
= skb_peek(&sk
->sk_receive_queue
);
1328 sz
= msg_data_sz(msg
);
1329 err
= msg_errcode(msg
);
1331 /* Discard an empty non-errored message & try again */
1332 if ((!sz
) && (!err
)) {
1333 tsk_advance_rx_queue(sk
);
1337 /* Capture sender's address (optional) */
1338 set_orig_addr(m
, msg
);
1340 /* Capture ancillary data (optional) */
1341 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1345 /* Capture message data (if valid) & compute return value (always) */
1347 if (unlikely(buf_len
< sz
)) {
1349 m
->msg_flags
|= MSG_TRUNC
;
1351 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
), m
, sz
);
1356 if ((sock
->state
== SS_READY
) ||
1357 ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
))
1363 /* Consume received message (optional) */
1364 if (likely(!(flags
& MSG_PEEK
))) {
1365 if ((sock
->state
!= SS_READY
) &&
1366 (++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1367 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1368 tsk
->rcv_unacked
= 0;
1370 tsk_advance_rx_queue(sk
);
1378 * tipc_recv_stream - receive stream-oriented data
1380 * @m: descriptor for message info
1381 * @buf_len: total size of user buffer area
1382 * @flags: receive flags
1384 * Used for SOCK_STREAM messages only. If not enough data is available
1385 * will optionally wait for more; never truncates data.
1387 * Returns size of returned message data, errno otherwise
1389 static int tipc_recv_stream(struct kiocb
*iocb
, struct socket
*sock
,
1390 struct msghdr
*m
, size_t buf_len
, int flags
)
1392 struct sock
*sk
= sock
->sk
;
1393 struct tipc_sock
*tsk
= tipc_sk(sk
);
1394 struct sk_buff
*buf
;
1395 struct tipc_msg
*msg
;
1398 int sz_to_copy
, target
, needed
;
1403 /* Catch invalid receive attempts */
1404 if (unlikely(!buf_len
))
1409 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1414 target
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buf_len
);
1415 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1418 /* Look for a message in receive queue; wait if necessary */
1419 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1423 /* Look at first message in receive queue */
1424 buf
= skb_peek(&sk
->sk_receive_queue
);
1426 sz
= msg_data_sz(msg
);
1427 err
= msg_errcode(msg
);
1429 /* Discard an empty non-errored message & try again */
1430 if ((!sz
) && (!err
)) {
1431 tsk_advance_rx_queue(sk
);
1435 /* Optionally capture sender's address & ancillary data of first msg */
1436 if (sz_copied
== 0) {
1437 set_orig_addr(m
, msg
);
1438 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1443 /* Capture message data (if valid) & compute return value (always) */
1445 u32 offset
= (u32
)(unsigned long)(TIPC_SKB_CB(buf
)->handle
);
1448 needed
= (buf_len
- sz_copied
);
1449 sz_to_copy
= (sz
<= needed
) ? sz
: needed
;
1451 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
) + offset
,
1456 sz_copied
+= sz_to_copy
;
1458 if (sz_to_copy
< sz
) {
1459 if (!(flags
& MSG_PEEK
))
1460 TIPC_SKB_CB(buf
)->handle
=
1461 (void *)(unsigned long)(offset
+ sz_to_copy
);
1466 goto exit
; /* can't add error msg to valid data */
1468 if ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
)
1474 /* Consume received message (optional) */
1475 if (likely(!(flags
& MSG_PEEK
))) {
1476 if (unlikely(++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1477 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1478 tsk
->rcv_unacked
= 0;
1480 tsk_advance_rx_queue(sk
);
1483 /* Loop around if more data is required */
1484 if ((sz_copied
< buf_len
) && /* didn't get all requested data */
1485 (!skb_queue_empty(&sk
->sk_receive_queue
) ||
1486 (sz_copied
< target
)) && /* and more is ready or required */
1487 (!(flags
& MSG_PEEK
)) && /* and aren't just peeking at data */
1488 (!err
)) /* and haven't reached a FIN */
1493 return sz_copied
? sz_copied
: res
;
1497 * tipc_write_space - wake up thread if port congestion is released
1500 static void tipc_write_space(struct sock
*sk
)
1502 struct socket_wq
*wq
;
1505 wq
= rcu_dereference(sk
->sk_wq
);
1506 if (wq_has_sleeper(wq
))
1507 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1508 POLLWRNORM
| POLLWRBAND
);
1513 * tipc_data_ready - wake up threads to indicate messages have been received
1515 * @len: the length of messages
1517 static void tipc_data_ready(struct sock
*sk
)
1519 struct socket_wq
*wq
;
1522 wq
= rcu_dereference(sk
->sk_wq
);
1523 if (wq_has_sleeper(wq
))
1524 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1525 POLLRDNORM
| POLLRDBAND
);
1530 * filter_connect - Handle all incoming messages for a connection-based socket
1534 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1536 static int filter_connect(struct tipc_sock
*tsk
, struct sk_buff
**buf
)
1538 struct sock
*sk
= &tsk
->sk
;
1539 struct socket
*sock
= sk
->sk_socket
;
1540 struct tipc_msg
*msg
= buf_msg(*buf
);
1541 int retval
= -TIPC_ERR_NO_PORT
;
1546 switch ((int)sock
->state
) {
1548 /* Accept only connection-based messages sent by peer */
1549 if (tsk_peer_msg(tsk
, msg
)) {
1550 if (unlikely(msg_errcode(msg
))) {
1551 sock
->state
= SS_DISCONNECTING
;
1553 /* let timer expire on it's own */
1554 tipc_node_remove_conn(tsk_peer_node(tsk
),
1561 /* Accept only ACK or NACK message */
1563 if (unlikely(!msg_connected(msg
)))
1566 if (unlikely(msg_errcode(msg
))) {
1567 sock
->state
= SS_DISCONNECTING
;
1568 sk
->sk_err
= ECONNREFUSED
;
1573 if (unlikely(msg_importance(msg
) > TIPC_CRITICAL_IMPORTANCE
)) {
1574 sock
->state
= SS_DISCONNECTING
;
1575 sk
->sk_err
= EINVAL
;
1580 tipc_sk_finish_conn(tsk
, msg_origport(msg
), msg_orignode(msg
));
1581 msg_set_importance(&tsk
->phdr
, msg_importance(msg
));
1582 sock
->state
= SS_CONNECTED
;
1584 /* If an incoming message is an 'ACK-', it should be
1585 * discarded here because it doesn't contain useful
1586 * data. In addition, we should try to wake up
1587 * connect() routine if sleeping.
1589 if (msg_data_sz(msg
) == 0) {
1592 if (waitqueue_active(sk_sleep(sk
)))
1593 wake_up_interruptible(sk_sleep(sk
));
1598 case SS_UNCONNECTED
:
1599 /* Accept only SYN message */
1600 if (!msg_connected(msg
) && !(msg_errcode(msg
)))
1603 case SS_DISCONNECTING
:
1606 pr_err("Unknown socket state %u\n", sock
->state
);
1612 * rcvbuf_limit - get proper overload limit of socket receive queue
1616 * For all connection oriented messages, irrespective of importance,
1617 * the default overload value (i.e. 67MB) is set as limit.
1619 * For all connectionless messages, by default new queue limits are
1622 * TIPC_LOW_IMPORTANCE (4 MB)
1623 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1624 * TIPC_HIGH_IMPORTANCE (16 MB)
1625 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1627 * Returns overload limit according to corresponding message importance
1629 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*buf
)
1631 struct tipc_msg
*msg
= buf_msg(buf
);
1633 if (msg_connected(msg
))
1634 return sysctl_tipc_rmem
[2];
1636 return sk
->sk_rcvbuf
>> TIPC_CRITICAL_IMPORTANCE
<<
1637 msg_importance(msg
);
1641 * filter_rcv - validate incoming message
1645 * Enqueues message on receive queue if acceptable; optionally handles
1646 * disconnect indication for a connected socket.
1648 * Called with socket lock already taken; port lock may also be taken.
1650 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1651 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1653 static int filter_rcv(struct sock
*sk
, struct sk_buff
*buf
)
1655 struct socket
*sock
= sk
->sk_socket
;
1656 struct tipc_sock
*tsk
= tipc_sk(sk
);
1657 struct tipc_msg
*msg
= buf_msg(buf
);
1658 unsigned int limit
= rcvbuf_limit(sk
, buf
);
1662 if (unlikely(msg_user(msg
) == CONN_MANAGER
))
1663 return tipc_sk_proto_rcv(tsk
, &onode
, buf
);
1665 if (unlikely(msg_user(msg
) == SOCK_WAKEUP
)) {
1668 sk
->sk_write_space(sk
);
1672 /* Reject message if it is wrong sort of message for socket */
1673 if (msg_type(msg
) > TIPC_DIRECT_MSG
)
1674 return -TIPC_ERR_NO_PORT
;
1676 if (sock
->state
== SS_READY
) {
1677 if (msg_connected(msg
))
1678 return -TIPC_ERR_NO_PORT
;
1680 rc
= filter_connect(tsk
, &buf
);
1681 if (rc
!= TIPC_OK
|| buf
== NULL
)
1685 /* Reject message if there isn't room to queue it */
1686 if (sk_rmem_alloc_get(sk
) + buf
->truesize
>= limit
)
1687 return -TIPC_ERR_OVERLOAD
;
1689 /* Enqueue message */
1690 TIPC_SKB_CB(buf
)->handle
= NULL
;
1691 __skb_queue_tail(&sk
->sk_receive_queue
, buf
);
1692 skb_set_owner_r(buf
, sk
);
1694 sk
->sk_data_ready(sk
);
1699 * tipc_backlog_rcv - handle incoming message from backlog queue
1703 * Caller must hold socket lock, but not port lock.
1707 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1711 struct tipc_sock
*tsk
= tipc_sk(sk
);
1712 uint truesize
= skb
->truesize
;
1714 rc
= filter_rcv(sk
, skb
);
1717 if (atomic_read(&tsk
->dupl_rcvcnt
) < TIPC_CONN_OVERLOAD_LIMIT
)
1718 atomic_add(truesize
, &tsk
->dupl_rcvcnt
);
1722 if ((rc
< 0) && !tipc_msg_reverse(skb
, &onode
, -rc
))
1725 tipc_link_xmit_skb(skb
, onode
, 0);
1731 * tipc_sk_rcv - handle incoming message
1732 * @skb: buffer containing arriving message
1734 * Returns 0 if success, or errno: -EHOSTUNREACH
1736 int tipc_sk_rcv(struct sk_buff
*skb
)
1738 struct tipc_sock
*tsk
;
1740 u32 dport
= msg_destport(buf_msg(skb
));
1745 /* Validate destination and message */
1746 tsk
= tipc_sk_get(dport
);
1747 if (unlikely(!tsk
)) {
1748 rc
= tipc_msg_eval(skb
, &dnode
);
1754 spin_lock_bh(&sk
->sk_lock
.slock
);
1756 if (!sock_owned_by_user(sk
)) {
1757 rc
= filter_rcv(sk
, skb
);
1759 if (sk
->sk_backlog
.len
== 0)
1760 atomic_set(&tsk
->dupl_rcvcnt
, 0);
1761 limit
= rcvbuf_limit(sk
, skb
) + atomic_read(&tsk
->dupl_rcvcnt
);
1762 if (sk_add_backlog(sk
, skb
, limit
))
1763 rc
= -TIPC_ERR_OVERLOAD
;
1765 spin_unlock_bh(&sk
->sk_lock
.slock
);
1770 if ((rc
< 0) && !tipc_msg_reverse(skb
, &dnode
, -rc
))
1771 return -EHOSTUNREACH
;
1773 tipc_link_xmit_skb(skb
, dnode
, 0);
1774 return (rc
< 0) ? -EHOSTUNREACH
: 0;
1777 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
1779 struct sock
*sk
= sock
->sk
;
1784 int err
= sock_error(sk
);
1789 if (signal_pending(current
))
1790 return sock_intr_errno(*timeo_p
);
1792 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1793 done
= sk_wait_event(sk
, timeo_p
, sock
->state
!= SS_CONNECTING
);
1794 finish_wait(sk_sleep(sk
), &wait
);
1800 * tipc_connect - establish a connection to another TIPC port
1801 * @sock: socket structure
1802 * @dest: socket address for destination port
1803 * @destlen: size of socket address data structure
1804 * @flags: file-related flags associated with socket
1806 * Returns 0 on success, errno otherwise
1808 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
1809 int destlen
, int flags
)
1811 struct sock
*sk
= sock
->sk
;
1812 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
1813 struct msghdr m
= {NULL
,};
1814 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tipc_sk(sk
)->conn_timeout
;
1815 socket_state previous
;
1820 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1821 if (sock
->state
== SS_READY
) {
1827 * Reject connection attempt using multicast address
1829 * Note: send_msg() validates the rest of the address fields,
1830 * so there's no need to do it here
1832 if (dst
->addrtype
== TIPC_ADDR_MCAST
) {
1837 previous
= sock
->state
;
1838 switch (sock
->state
) {
1839 case SS_UNCONNECTED
:
1840 /* Send a 'SYN-' to destination */
1842 m
.msg_namelen
= destlen
;
1844 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1845 * indicate send_msg() is never blocked.
1848 m
.msg_flags
= MSG_DONTWAIT
;
1850 res
= tipc_sendmsg(NULL
, sock
, &m
, 0);
1851 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
1854 /* Just entered SS_CONNECTING state; the only
1855 * difference is that return value in non-blocking
1856 * case is EINPROGRESS, rather than EALREADY.
1860 if (previous
== SS_CONNECTING
)
1864 timeout
= msecs_to_jiffies(timeout
);
1865 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1866 res
= tipc_wait_for_connect(sock
, &timeout
);
1881 * tipc_listen - allow socket to listen for incoming connections
1882 * @sock: socket structure
1885 * Returns 0 on success, errno otherwise
1887 static int tipc_listen(struct socket
*sock
, int len
)
1889 struct sock
*sk
= sock
->sk
;
1894 if (sock
->state
!= SS_UNCONNECTED
)
1897 sock
->state
= SS_LISTENING
;
1905 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
1907 struct sock
*sk
= sock
->sk
;
1911 /* True wake-one mechanism for incoming connections: only
1912 * one process gets woken up, not the 'whole herd'.
1913 * Since we do not 'race & poll' for established sockets
1914 * anymore, the common case will execute the loop only once.
1917 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
1918 TASK_INTERRUPTIBLE
);
1919 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1921 timeo
= schedule_timeout(timeo
);
1925 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1928 if (sock
->state
!= SS_LISTENING
)
1930 err
= sock_intr_errno(timeo
);
1931 if (signal_pending(current
))
1937 finish_wait(sk_sleep(sk
), &wait
);
1942 * tipc_accept - wait for connection request
1943 * @sock: listening socket
1944 * @newsock: new socket that is to be connected
1945 * @flags: file-related flags associated with socket
1947 * Returns 0 on success, errno otherwise
1949 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
)
1951 struct sock
*new_sk
, *sk
= sock
->sk
;
1952 struct sk_buff
*buf
;
1953 struct tipc_sock
*new_tsock
;
1954 struct tipc_msg
*msg
;
1960 if (sock
->state
!= SS_LISTENING
) {
1964 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
1965 res
= tipc_wait_for_accept(sock
, timeo
);
1969 buf
= skb_peek(&sk
->sk_receive_queue
);
1971 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, 1);
1975 new_sk
= new_sock
->sk
;
1976 new_tsock
= tipc_sk(new_sk
);
1979 /* we lock on new_sk; but lockdep sees the lock on sk */
1980 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
1983 * Reject any stray messages received by new socket
1984 * before the socket lock was taken (very, very unlikely)
1986 tsk_rej_rx_queue(new_sk
);
1988 /* Connect new socket to it's peer */
1989 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
1990 new_sock
->state
= SS_CONNECTED
;
1992 tsk_set_importance(new_tsock
, msg_importance(msg
));
1993 if (msg_named(msg
)) {
1994 new_tsock
->conn_type
= msg_nametype(msg
);
1995 new_tsock
->conn_instance
= msg_nameinst(msg
);
1999 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2000 * Respond to 'SYN+' by queuing it on new socket.
2002 if (!msg_data_sz(msg
)) {
2003 struct msghdr m
= {NULL
,};
2005 tsk_advance_rx_queue(sk
);
2006 tipc_send_packet(NULL
, new_sock
, &m
, 0);
2008 __skb_dequeue(&sk
->sk_receive_queue
);
2009 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2010 skb_set_owner_r(buf
, new_sk
);
2012 release_sock(new_sk
);
2019 * tipc_shutdown - shutdown socket connection
2020 * @sock: socket structure
2021 * @how: direction to close (must be SHUT_RDWR)
2023 * Terminates connection (if necessary), then purges socket's receive queue.
2025 * Returns 0 on success, errno otherwise
2027 static int tipc_shutdown(struct socket
*sock
, int how
)
2029 struct sock
*sk
= sock
->sk
;
2030 struct tipc_sock
*tsk
= tipc_sk(sk
);
2031 struct sk_buff
*skb
;
2035 if (how
!= SHUT_RDWR
)
2040 switch (sock
->state
) {
2045 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2046 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
2048 if (TIPC_SKB_CB(skb
)->handle
!= NULL
) {
2052 if (tipc_msg_reverse(skb
, &dnode
, TIPC_CONN_SHUTDOWN
))
2053 tipc_link_xmit_skb(skb
, dnode
, tsk
->ref
);
2054 tipc_node_remove_conn(dnode
, tsk
->ref
);
2056 dnode
= tsk_peer_node(tsk
);
2057 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
2058 TIPC_CONN_MSG
, SHORT_H_SIZE
,
2059 0, dnode
, tipc_own_addr
,
2061 tsk
->ref
, TIPC_CONN_SHUTDOWN
);
2062 tipc_link_xmit_skb(skb
, dnode
, tsk
->ref
);
2065 sock
->state
= SS_DISCONNECTING
;
2066 tipc_node_remove_conn(dnode
, tsk
->ref
);
2069 case SS_DISCONNECTING
:
2071 /* Discard any unreceived messages */
2072 __skb_queue_purge(&sk
->sk_receive_queue
);
2074 /* Wake up anyone sleeping in poll */
2075 sk
->sk_state_change(sk
);
2087 static void tipc_sk_timeout(unsigned long ref
)
2089 struct tipc_sock
*tsk
;
2091 struct sk_buff
*skb
= NULL
;
2092 u32 peer_port
, peer_node
;
2094 tsk
= tipc_sk_get(ref
);
2100 if (!tsk
->connected
) {
2104 peer_port
= tsk_peer_port(tsk
);
2105 peer_node
= tsk_peer_node(tsk
);
2107 if (tsk
->probing_state
== TIPC_CONN_PROBING
) {
2108 /* Previous probe not answered -> self abort */
2109 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
2110 SHORT_H_SIZE
, 0, tipc_own_addr
,
2111 peer_node
, ref
, peer_port
,
2114 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
,
2115 0, peer_node
, tipc_own_addr
,
2116 peer_port
, ref
, TIPC_OK
);
2117 tsk
->probing_state
= TIPC_CONN_PROBING
;
2118 k_start_timer(&tsk
->timer
, tsk
->probing_interval
);
2122 tipc_link_xmit_skb(skb
, peer_node
, ref
);
2127 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2128 struct tipc_name_seq
const *seq
)
2130 struct publication
*publ
;
2135 key
= tsk
->ref
+ tsk
->pub_count
+ 1;
2136 if (key
== tsk
->ref
)
2139 publ
= tipc_nametbl_publish(seq
->type
, seq
->lower
, seq
->upper
,
2140 scope
, tsk
->ref
, key
);
2141 if (unlikely(!publ
))
2144 list_add(&publ
->pport_list
, &tsk
->publications
);
2150 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2151 struct tipc_name_seq
const *seq
)
2153 struct publication
*publ
;
2154 struct publication
*safe
;
2157 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2159 if (publ
->scope
!= scope
)
2161 if (publ
->type
!= seq
->type
)
2163 if (publ
->lower
!= seq
->lower
)
2165 if (publ
->upper
!= seq
->upper
)
2167 tipc_nametbl_withdraw(publ
->type
, publ
->lower
,
2168 publ
->ref
, publ
->key
);
2172 tipc_nametbl_withdraw(publ
->type
, publ
->lower
,
2173 publ
->ref
, publ
->key
);
2176 if (list_empty(&tsk
->publications
))
2181 static int tipc_sk_show(struct tipc_sock
*tsk
, char *buf
,
2182 int len
, int full_id
)
2184 struct publication
*publ
;
2188 ret
= tipc_snprintf(buf
, len
, "<%u.%u.%u:%u>:",
2189 tipc_zone(tipc_own_addr
),
2190 tipc_cluster(tipc_own_addr
),
2191 tipc_node(tipc_own_addr
), tsk
->ref
);
2193 ret
= tipc_snprintf(buf
, len
, "%-10u:", tsk
->ref
);
2195 if (tsk
->connected
) {
2196 u32 dport
= tsk_peer_port(tsk
);
2197 u32 destnode
= tsk_peer_node(tsk
);
2199 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2200 " connected to <%u.%u.%u:%u>",
2201 tipc_zone(destnode
),
2202 tipc_cluster(destnode
),
2203 tipc_node(destnode
), dport
);
2204 if (tsk
->conn_type
!= 0)
2205 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2206 " via {%u,%u}", tsk
->conn_type
,
2207 tsk
->conn_instance
);
2208 } else if (tsk
->published
) {
2209 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
, " bound to");
2210 list_for_each_entry(publ
, &tsk
->publications
, pport_list
) {
2211 if (publ
->lower
== publ
->upper
)
2212 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2213 " {%u,%u}", publ
->type
,
2216 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2217 " {%u,%u,%u}", publ
->type
,
2218 publ
->lower
, publ
->upper
);
2221 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
, "\n");
2225 struct sk_buff
*tipc_sk_socks_show(void)
2227 struct sk_buff
*buf
;
2228 struct tlv_desc
*rep_tlv
;
2231 struct tipc_sock
*tsk
;
2235 buf
= tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN
));
2238 rep_tlv
= (struct tlv_desc
*)buf
->data
;
2239 pb
= TLV_DATA(rep_tlv
);
2240 pb_len
= ULTRA_STRING_MAX_LEN
;
2242 tsk
= tipc_sk_get_next(&ref
);
2243 for (; tsk
; tsk
= tipc_sk_get_next(&ref
)) {
2244 lock_sock(&tsk
->sk
);
2245 str_len
+= tipc_sk_show(tsk
, pb
+ str_len
,
2246 pb_len
- str_len
, 0);
2247 release_sock(&tsk
->sk
);
2250 str_len
+= 1; /* for "\0" */
2251 skb_put(buf
, TLV_SPACE(str_len
));
2252 TLV_SET(rep_tlv
, TIPC_TLV_ULTRA_STRING
, NULL
, str_len
);
2257 /* tipc_sk_reinit: set non-zero address in all existing sockets
2258 * when we go from standalone to network mode.
2260 void tipc_sk_reinit(void)
2262 struct tipc_msg
*msg
;
2264 struct tipc_sock
*tsk
= tipc_sk_get_next(&ref
);
2266 for (; tsk
; tsk
= tipc_sk_get_next(&ref
)) {
2267 lock_sock(&tsk
->sk
);
2269 msg_set_prevnode(msg
, tipc_own_addr
);
2270 msg_set_orignode(msg
, tipc_own_addr
);
2271 release_sock(&tsk
->sk
);
2277 * struct reference - TIPC socket reference entry
2278 * @tsk: pointer to socket associated with reference entry
2279 * @ref: reference value for socket (combines instance & array index info)
2282 struct tipc_sock
*tsk
;
2287 * struct tipc_ref_table - table of TIPC socket reference entries
2288 * @entries: pointer to array of reference entries
2289 * @capacity: array index of first unusable entry
2290 * @init_point: array index of first uninitialized entry
2291 * @first_free: array index of first unused socket reference entry
2292 * @last_free: array index of last unused socket reference entry
2293 * @index_mask: bitmask for array index portion of reference values
2294 * @start_mask: initial value for instance value portion of reference values
2297 struct reference
*entries
;
2306 /* Socket reference table consists of 2**N entries.
2308 * State Socket ptr Reference
2309 * ----- ---------- ---------
2310 * In use non-NULL XXXX|own index
2311 * (XXXX changes each time entry is acquired)
2312 * Free NULL YYYY|next free index
2313 * (YYYY is one more than last used XXXX)
2314 * Uninitialized NULL 0
2316 * Entry 0 is not used; this allows index 0 to denote the end of the free list.
2318 * Note that a reference value of 0 does not necessarily indicate that an
2319 * entry is uninitialized, since the last entry in the free list could also
2320 * have a reference value of 0 (although this is unlikely).
2323 static struct ref_table tipc_ref_table
;
2325 static DEFINE_RWLOCK(ref_table_lock
);
2328 * tipc_ref_table_init - create reference table for sockets
2330 int tipc_sk_ref_table_init(u32 req_sz
, u32 start
)
2332 struct reference
*table
;
2335 /* account for unused entry, then round up size to a power of 2 */
2338 for (actual_sz
= 16; actual_sz
< req_sz
; actual_sz
<<= 1) {
2342 /* allocate table & mark all entries as uninitialized */
2343 table
= vzalloc(actual_sz
* sizeof(struct reference
));
2347 tipc_ref_table
.entries
= table
;
2348 tipc_ref_table
.capacity
= req_sz
;
2349 tipc_ref_table
.init_point
= 1;
2350 tipc_ref_table
.first_free
= 0;
2351 tipc_ref_table
.last_free
= 0;
2352 tipc_ref_table
.index_mask
= actual_sz
- 1;
2353 tipc_ref_table
.start_mask
= start
& ~tipc_ref_table
.index_mask
;
2359 * tipc_ref_table_stop - destroy reference table for sockets
2361 void tipc_sk_ref_table_stop(void)
2363 if (!tipc_ref_table
.entries
)
2365 vfree(tipc_ref_table
.entries
);
2366 tipc_ref_table
.entries
= NULL
;
2369 /* tipc_ref_acquire - create reference to a socket
2371 * Register an socket pointer in the reference table.
2372 * Returns a unique reference value that is used from then on to retrieve the
2373 * socket pointer, or to determine if the socket has been deregistered.
2375 u32
tipc_sk_ref_acquire(struct tipc_sock
*tsk
)
2379 u32 next_plus_upper
;
2381 struct reference
*entry
;
2383 if (unlikely(!tsk
)) {
2384 pr_err("Attempt to acquire ref. to non-existent obj\n");
2387 if (unlikely(!tipc_ref_table
.entries
)) {
2388 pr_err("Ref. table not found in acquisition attempt\n");
2392 /* Take a free entry, if available; otherwise initialize a new one */
2393 write_lock_bh(&ref_table_lock
);
2394 index
= tipc_ref_table
.first_free
;
2395 entry
= &tipc_ref_table
.entries
[index
];
2397 if (likely(index
)) {
2398 index
= tipc_ref_table
.first_free
;
2399 entry
= &tipc_ref_table
.entries
[index
];
2400 index_mask
= tipc_ref_table
.index_mask
;
2401 next_plus_upper
= entry
->ref
;
2402 tipc_ref_table
.first_free
= next_plus_upper
& index_mask
;
2403 ref
= (next_plus_upper
& ~index_mask
) + index
;
2405 } else if (tipc_ref_table
.init_point
< tipc_ref_table
.capacity
) {
2406 index
= tipc_ref_table
.init_point
++;
2407 entry
= &tipc_ref_table
.entries
[index
];
2408 ref
= tipc_ref_table
.start_mask
+ index
;
2415 write_unlock_bh(&ref_table_lock
);
2419 /* tipc_sk_ref_discard - invalidate reference to an socket
2421 * Disallow future references to an socket and free up the entry for re-use.
2423 void tipc_sk_ref_discard(u32 ref
)
2425 struct reference
*entry
;
2429 if (unlikely(!tipc_ref_table
.entries
)) {
2430 pr_err("Ref. table not found during discard attempt\n");
2434 index_mask
= tipc_ref_table
.index_mask
;
2435 index
= ref
& index_mask
;
2436 entry
= &tipc_ref_table
.entries
[index
];
2438 write_lock_bh(&ref_table_lock
);
2440 if (unlikely(!entry
->tsk
)) {
2441 pr_err("Attempt to discard ref. to non-existent socket\n");
2444 if (unlikely(entry
->ref
!= ref
)) {
2445 pr_err("Attempt to discard non-existent reference\n");
2449 /* Mark entry as unused; increment instance part of entry's
2450 * reference to invalidate any subsequent references
2454 entry
->ref
= (ref
& ~index_mask
) + (index_mask
+ 1);
2456 /* Append entry to free entry list */
2457 if (unlikely(tipc_ref_table
.first_free
== 0))
2458 tipc_ref_table
.first_free
= index
;
2460 tipc_ref_table
.entries
[tipc_ref_table
.last_free
].ref
|= index
;
2461 tipc_ref_table
.last_free
= index
;
2463 write_unlock_bh(&ref_table_lock
);
2466 /* tipc_sk_get - find referenced socket and return pointer to it
2468 struct tipc_sock
*tipc_sk_get(u32 ref
)
2470 struct reference
*entry
;
2471 struct tipc_sock
*tsk
;
2473 if (unlikely(!tipc_ref_table
.entries
))
2475 read_lock_bh(&ref_table_lock
);
2476 entry
= &tipc_ref_table
.entries
[ref
& tipc_ref_table
.index_mask
];
2478 if (likely(tsk
&& (entry
->ref
== ref
)))
2479 sock_hold(&tsk
->sk
);
2482 read_unlock_bh(&ref_table_lock
);
2486 /* tipc_sk_get_next - lock & return next socket after referenced one
2488 struct tipc_sock
*tipc_sk_get_next(u32
*ref
)
2490 struct reference
*entry
;
2491 struct tipc_sock
*tsk
= NULL
;
2492 uint index
= *ref
& tipc_ref_table
.index_mask
;
2494 read_lock_bh(&ref_table_lock
);
2495 while (++index
< tipc_ref_table
.capacity
) {
2496 entry
= &tipc_ref_table
.entries
[index
];
2500 sock_hold(&tsk
->sk
);
2504 read_unlock_bh(&ref_table_lock
);
2508 static void tipc_sk_put(struct tipc_sock
*tsk
)
2514 * tipc_setsockopt - set socket option
2515 * @sock: socket structure
2516 * @lvl: option level
2517 * @opt: option identifier
2518 * @ov: pointer to new option value
2519 * @ol: length of option value
2521 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2522 * (to ease compatibility).
2524 * Returns 0 on success, errno otherwise
2526 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2527 char __user
*ov
, unsigned int ol
)
2529 struct sock
*sk
= sock
->sk
;
2530 struct tipc_sock
*tsk
= tipc_sk(sk
);
2534 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2536 if (lvl
!= SOL_TIPC
)
2537 return -ENOPROTOOPT
;
2538 if (ol
< sizeof(value
))
2540 res
= get_user(value
, (u32 __user
*)ov
);
2547 case TIPC_IMPORTANCE
:
2548 res
= tsk_set_importance(tsk
, value
);
2550 case TIPC_SRC_DROPPABLE
:
2551 if (sock
->type
!= SOCK_STREAM
)
2552 tsk_set_unreliable(tsk
, value
);
2556 case TIPC_DEST_DROPPABLE
:
2557 tsk_set_unreturnable(tsk
, value
);
2559 case TIPC_CONN_TIMEOUT
:
2560 tipc_sk(sk
)->conn_timeout
= value
;
2561 /* no need to set "res", since already 0 at this point */
2573 * tipc_getsockopt - get socket option
2574 * @sock: socket structure
2575 * @lvl: option level
2576 * @opt: option identifier
2577 * @ov: receptacle for option value
2578 * @ol: receptacle for length of option value
2580 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2581 * (to ease compatibility).
2583 * Returns 0 on success, errno otherwise
2585 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2586 char __user
*ov
, int __user
*ol
)
2588 struct sock
*sk
= sock
->sk
;
2589 struct tipc_sock
*tsk
= tipc_sk(sk
);
2594 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2595 return put_user(0, ol
);
2596 if (lvl
!= SOL_TIPC
)
2597 return -ENOPROTOOPT
;
2598 res
= get_user(len
, ol
);
2605 case TIPC_IMPORTANCE
:
2606 value
= tsk_importance(tsk
);
2608 case TIPC_SRC_DROPPABLE
:
2609 value
= tsk_unreliable(tsk
);
2611 case TIPC_DEST_DROPPABLE
:
2612 value
= tsk_unreturnable(tsk
);
2614 case TIPC_CONN_TIMEOUT
:
2615 value
= tsk
->conn_timeout
;
2616 /* no need to set "res", since already 0 at this point */
2618 case TIPC_NODE_RECVQ_DEPTH
:
2619 value
= 0; /* was tipc_queue_size, now obsolete */
2621 case TIPC_SOCK_RECVQ_DEPTH
:
2622 value
= skb_queue_len(&sk
->sk_receive_queue
);
2631 return res
; /* "get" failed */
2633 if (len
< sizeof(value
))
2636 if (copy_to_user(ov
, &value
, sizeof(value
)))
2639 return put_user(sizeof(value
), ol
);
2642 static int tipc_ioctl(struct socket
*sk
, unsigned int cmd
, unsigned long arg
)
2644 struct tipc_sioc_ln_req lnr
;
2645 void __user
*argp
= (void __user
*)arg
;
2648 case SIOCGETLINKNAME
:
2649 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2651 if (!tipc_node_get_linkname(lnr
.bearer_id
& 0xffff, lnr
.peer
,
2652 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2653 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2657 return -EADDRNOTAVAIL
;
2659 return -ENOIOCTLCMD
;
2663 /* Protocol switches for the various types of TIPC sockets */
2665 static const struct proto_ops msg_ops
= {
2666 .owner
= THIS_MODULE
,
2668 .release
= tipc_release
,
2670 .connect
= tipc_connect
,
2671 .socketpair
= sock_no_socketpair
,
2672 .accept
= sock_no_accept
,
2673 .getname
= tipc_getname
,
2675 .ioctl
= tipc_ioctl
,
2676 .listen
= sock_no_listen
,
2677 .shutdown
= tipc_shutdown
,
2678 .setsockopt
= tipc_setsockopt
,
2679 .getsockopt
= tipc_getsockopt
,
2680 .sendmsg
= tipc_sendmsg
,
2681 .recvmsg
= tipc_recvmsg
,
2682 .mmap
= sock_no_mmap
,
2683 .sendpage
= sock_no_sendpage
2686 static const struct proto_ops packet_ops
= {
2687 .owner
= THIS_MODULE
,
2689 .release
= tipc_release
,
2691 .connect
= tipc_connect
,
2692 .socketpair
= sock_no_socketpair
,
2693 .accept
= tipc_accept
,
2694 .getname
= tipc_getname
,
2696 .ioctl
= tipc_ioctl
,
2697 .listen
= tipc_listen
,
2698 .shutdown
= tipc_shutdown
,
2699 .setsockopt
= tipc_setsockopt
,
2700 .getsockopt
= tipc_getsockopt
,
2701 .sendmsg
= tipc_send_packet
,
2702 .recvmsg
= tipc_recvmsg
,
2703 .mmap
= sock_no_mmap
,
2704 .sendpage
= sock_no_sendpage
2707 static const struct proto_ops stream_ops
= {
2708 .owner
= THIS_MODULE
,
2710 .release
= tipc_release
,
2712 .connect
= tipc_connect
,
2713 .socketpair
= sock_no_socketpair
,
2714 .accept
= tipc_accept
,
2715 .getname
= tipc_getname
,
2717 .ioctl
= tipc_ioctl
,
2718 .listen
= tipc_listen
,
2719 .shutdown
= tipc_shutdown
,
2720 .setsockopt
= tipc_setsockopt
,
2721 .getsockopt
= tipc_getsockopt
,
2722 .sendmsg
= tipc_send_stream
,
2723 .recvmsg
= tipc_recv_stream
,
2724 .mmap
= sock_no_mmap
,
2725 .sendpage
= sock_no_sendpage
2728 static const struct net_proto_family tipc_family_ops
= {
2729 .owner
= THIS_MODULE
,
2731 .create
= tipc_sk_create
2734 static struct proto tipc_proto
= {
2736 .owner
= THIS_MODULE
,
2737 .obj_size
= sizeof(struct tipc_sock
),
2738 .sysctl_rmem
= sysctl_tipc_rmem
2741 static struct proto tipc_proto_kern
= {
2743 .obj_size
= sizeof(struct tipc_sock
),
2744 .sysctl_rmem
= sysctl_tipc_rmem
2748 * tipc_socket_init - initialize TIPC socket interface
2750 * Returns 0 on success, errno otherwise
2752 int tipc_socket_init(void)
2756 res
= proto_register(&tipc_proto
, 1);
2758 pr_err("Failed to register TIPC protocol type\n");
2762 res
= sock_register(&tipc_family_ops
);
2764 pr_err("Failed to register TIPC socket type\n");
2765 proto_unregister(&tipc_proto
);
2773 * tipc_socket_stop - stop TIPC socket interface
2775 void tipc_socket_stop(void)
2777 sock_unregister(tipc_family_ops
.family
);
2778 proto_unregister(&tipc_proto
);
2781 /* Caller should hold socket lock for the passed tipc socket. */
2782 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
2786 struct nlattr
*nest
;
2788 peer_node
= tsk_peer_node(tsk
);
2789 peer_port
= tsk_peer_port(tsk
);
2791 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
2793 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
2795 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
2798 if (tsk
->conn_type
!= 0) {
2799 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
2801 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
2803 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
2806 nla_nest_end(skb
, nest
);
2811 nla_nest_cancel(skb
, nest
);
2816 /* Caller should hold socket lock for the passed tipc socket. */
2817 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
2818 struct tipc_sock
*tsk
)
2822 struct nlattr
*attrs
;
2824 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2825 &tipc_genl_v2_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
2829 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
2831 goto genlmsg_cancel
;
2832 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->ref
))
2833 goto attr_msg_cancel
;
2834 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tipc_own_addr
))
2835 goto attr_msg_cancel
;
2837 if (tsk
->connected
) {
2838 err
= __tipc_nl_add_sk_con(skb
, tsk
);
2840 goto attr_msg_cancel
;
2841 } else if (!list_empty(&tsk
->publications
)) {
2842 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
2843 goto attr_msg_cancel
;
2845 nla_nest_end(skb
, attrs
);
2846 genlmsg_end(skb
, hdr
);
2851 nla_nest_cancel(skb
, attrs
);
2853 genlmsg_cancel(skb
, hdr
);
2858 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2861 struct tipc_sock
*tsk
;
2862 u32 prev_ref
= cb
->args
[0];
2865 tsk
= tipc_sk_get_next(&ref
);
2866 for (; tsk
; tsk
= tipc_sk_get_next(&ref
)) {
2867 lock_sock(&tsk
->sk
);
2868 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
2869 release_sock(&tsk
->sk
);
2877 cb
->args
[0] = prev_ref
;
2882 /* Caller should hold socket lock for the passed tipc socket. */
2883 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
2884 struct netlink_callback
*cb
,
2885 struct publication
*publ
)
2888 struct nlattr
*attrs
;
2890 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2891 &tipc_genl_v2_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
2895 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
2897 goto genlmsg_cancel
;
2899 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
2900 goto attr_msg_cancel
;
2901 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
2902 goto attr_msg_cancel
;
2903 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
2904 goto attr_msg_cancel
;
2905 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
2906 goto attr_msg_cancel
;
2908 nla_nest_end(skb
, attrs
);
2909 genlmsg_end(skb
, hdr
);
2914 nla_nest_cancel(skb
, attrs
);
2916 genlmsg_cancel(skb
, hdr
);
2921 /* Caller should hold socket lock for the passed tipc socket. */
2922 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
2923 struct netlink_callback
*cb
,
2924 struct tipc_sock
*tsk
, u32
*last_publ
)
2927 struct publication
*p
;
2930 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
2931 if (p
->key
== *last_publ
)
2934 if (p
->key
!= *last_publ
) {
2935 /* We never set seq or call nl_dump_check_consistent()
2936 * this means that setting prev_seq here will cause the
2937 * consistence check to fail in the netlink callback
2938 * handler. Resulting in the last NLMSG_DONE message
2939 * having the NLM_F_DUMP_INTR flag set.
2946 p
= list_first_entry(&tsk
->publications
, struct publication
,
2950 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
2951 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
2953 *last_publ
= p
->key
;
2962 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2965 u32 tsk_ref
= cb
->args
[0];
2966 u32 last_publ
= cb
->args
[1];
2967 u32 done
= cb
->args
[2];
2968 struct tipc_sock
*tsk
;
2971 struct nlattr
**attrs
;
2972 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
2974 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
2978 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
2979 attrs
[TIPC_NLA_SOCK
],
2980 tipc_nl_sock_policy
);
2984 if (!sock
[TIPC_NLA_SOCK_REF
])
2987 tsk_ref
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
2993 tsk
= tipc_sk_get(tsk_ref
);
2997 lock_sock(&tsk
->sk
);
2998 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
3001 release_sock(&tsk
->sk
);
3004 cb
->args
[0] = tsk_ref
;
3005 cb
->args
[1] = last_publ
;