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(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 /* Protects tipc socket hash table mutations */
183 static struct rhashtable tipc_sk_rht
;
185 static u32
tsk_peer_node(struct tipc_sock
*tsk
)
187 return msg_destnode(&tsk
->phdr
);
190 static u32
tsk_peer_port(struct tipc_sock
*tsk
)
192 return msg_destport(&tsk
->phdr
);
195 static bool tsk_unreliable(struct tipc_sock
*tsk
)
197 return msg_src_droppable(&tsk
->phdr
) != 0;
200 static void tsk_set_unreliable(struct tipc_sock
*tsk
, bool unreliable
)
202 msg_set_src_droppable(&tsk
->phdr
, unreliable
? 1 : 0);
205 static bool tsk_unreturnable(struct tipc_sock
*tsk
)
207 return msg_dest_droppable(&tsk
->phdr
) != 0;
210 static void tsk_set_unreturnable(struct tipc_sock
*tsk
, bool unreturnable
)
212 msg_set_dest_droppable(&tsk
->phdr
, unreturnable
? 1 : 0);
215 static int tsk_importance(struct tipc_sock
*tsk
)
217 return msg_importance(&tsk
->phdr
);
220 static int tsk_set_importance(struct tipc_sock
*tsk
, int imp
)
222 if (imp
> TIPC_CRITICAL_IMPORTANCE
)
224 msg_set_importance(&tsk
->phdr
, (u32
)imp
);
228 static struct tipc_sock
*tipc_sk(const struct sock
*sk
)
230 return container_of(sk
, struct tipc_sock
, sk
);
233 static int tsk_conn_cong(struct tipc_sock
*tsk
)
235 return tsk
->sent_unacked
>= TIPC_FLOWCTRL_WIN
;
239 * tsk_advance_rx_queue - discard first buffer in socket receive queue
241 * Caller must hold socket lock
243 static void tsk_advance_rx_queue(struct sock
*sk
)
245 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
249 * tsk_rej_rx_queue - reject all buffers in socket receive queue
251 * Caller must hold socket lock
253 static void tsk_rej_rx_queue(struct sock
*sk
)
258 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
))) {
259 if (tipc_msg_reverse(skb
, &dnode
, TIPC_ERR_NO_PORT
))
260 tipc_link_xmit_skb(sock_net(sk
), skb
, dnode
, 0);
264 /* tsk_peer_msg - verify if message was sent by connected port's peer
266 * Handles cases where the node's network address has changed from
267 * the default of <0.0.0> to its configured setting.
269 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
271 u32 peer_port
= tsk_peer_port(tsk
);
275 if (unlikely(!tsk
->connected
))
278 if (unlikely(msg_origport(msg
) != peer_port
))
281 orig_node
= msg_orignode(msg
);
282 peer_node
= tsk_peer_node(tsk
);
284 if (likely(orig_node
== peer_node
))
287 if (!orig_node
&& (peer_node
== tipc_own_addr
))
290 if (!peer_node
&& (orig_node
== tipc_own_addr
))
297 * tipc_sk_create - create a TIPC socket
298 * @net: network namespace (must be default network)
299 * @sock: pre-allocated socket structure
300 * @protocol: protocol indicator (must be 0)
301 * @kern: caused by kernel or by userspace?
303 * This routine creates additional data structures used by the TIPC socket,
304 * initializes them, and links them together.
306 * Returns 0 on success, errno otherwise
308 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
309 int protocol
, int kern
)
311 const struct proto_ops
*ops
;
314 struct tipc_sock
*tsk
;
315 struct tipc_msg
*msg
;
317 /* Validate arguments */
318 if (unlikely(protocol
!= 0))
319 return -EPROTONOSUPPORT
;
321 switch (sock
->type
) {
324 state
= SS_UNCONNECTED
;
328 state
= SS_UNCONNECTED
;
339 /* Allocate socket's protocol area */
341 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
);
343 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto_kern
);
349 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
350 INIT_LIST_HEAD(&tsk
->publications
);
352 tipc_msg_init(msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
355 /* Finish initializing socket data structures */
358 sock_init_data(sock
, sk
);
359 if (tipc_sk_insert(tsk
)) {
360 pr_warn("Socket create failed; port numbrer exhausted\n");
363 msg_set_origport(msg
, tsk
->portid
);
364 setup_timer(&tsk
->timer
, tipc_sk_timeout
, (unsigned long)tsk
);
365 sk
->sk_backlog_rcv
= tipc_backlog_rcv
;
366 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
367 sk
->sk_data_ready
= tipc_data_ready
;
368 sk
->sk_write_space
= tipc_write_space
;
369 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
370 tsk
->sent_unacked
= 0;
371 atomic_set(&tsk
->dupl_rcvcnt
, 0);
373 if (sock
->state
== SS_READY
) {
374 tsk_set_unreturnable(tsk
, true);
375 if (sock
->type
== SOCK_DGRAM
)
376 tsk_set_unreliable(tsk
, true);
382 * tipc_sock_create_local - create TIPC socket from inside TIPC module
383 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
385 * We cannot use sock_creat_kern here because it bumps module user count.
386 * Since socket owner and creator is the same module we must make sure
387 * that module count remains zero for module local sockets, otherwise
388 * we cannot do rmmod.
390 * Returns 0 on success, errno otherwise
392 int tipc_sock_create_local(int type
, struct socket
**res
)
396 rc
= sock_create_lite(AF_TIPC
, type
, 0, res
);
398 pr_err("Failed to create kernel socket\n");
401 tipc_sk_create(&init_net
, *res
, 0, 1);
407 * tipc_sock_release_local - release socket created by tipc_sock_create_local
408 * @sock: the socket to be released.
410 * Module reference count is not incremented when such sockets are created,
411 * so we must keep it from being decremented when they are released.
413 void tipc_sock_release_local(struct socket
*sock
)
421 * tipc_sock_accept_local - accept a connection on a socket created
422 * with tipc_sock_create_local. Use this function to avoid that
423 * module reference count is inadvertently incremented.
425 * @sock: the accepting socket
426 * @newsock: reference to the new socket to be created
427 * @flags: socket flags
430 int tipc_sock_accept_local(struct socket
*sock
, struct socket
**newsock
,
433 struct sock
*sk
= sock
->sk
;
436 ret
= sock_create_lite(sk
->sk_family
, sk
->sk_type
,
437 sk
->sk_protocol
, newsock
);
441 ret
= tipc_accept(sock
, *newsock
, flags
);
443 sock_release(*newsock
);
446 (*newsock
)->ops
= sock
->ops
;
450 static void tipc_sk_callback(struct rcu_head
*head
)
452 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
458 * tipc_release - destroy a TIPC socket
459 * @sock: socket to destroy
461 * This routine cleans up any messages that are still queued on the socket.
462 * For DGRAM and RDM socket types, all queued messages are rejected.
463 * For SEQPACKET and STREAM socket types, the first message is rejected
464 * and any others are discarded. (If the first message on a STREAM socket
465 * is partially-read, it is discarded and the next one is rejected instead.)
467 * NOTE: Rejected messages are not necessarily returned to the sender! They
468 * are returned or discarded according to the "destination droppable" setting
469 * specified for the message by the sender.
471 * Returns 0 on success, errno otherwise
473 static int tipc_release(struct socket
*sock
)
475 struct sock
*sk
= sock
->sk
;
476 struct net
*net
= sock_net(sk
);
477 struct tipc_sock
*tsk
;
479 u32 dnode
, probing_state
;
482 * Exit if socket isn't fully initialized (occurs when a failed accept()
483 * releases a pre-allocated child socket that was never used)
492 * Reject all unreceived messages, except on an active connection
493 * (which disconnects locally & sends a 'FIN+' to peer)
495 dnode
= tsk_peer_node(tsk
);
496 while (sock
->state
!= SS_DISCONNECTING
) {
497 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
500 if (TIPC_SKB_CB(skb
)->handle
!= NULL
)
503 if ((sock
->state
== SS_CONNECTING
) ||
504 (sock
->state
== SS_CONNECTED
)) {
505 sock
->state
= SS_DISCONNECTING
;
507 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
509 if (tipc_msg_reverse(skb
, &dnode
, TIPC_ERR_NO_PORT
))
510 tipc_link_xmit_skb(net
, skb
, dnode
, 0);
514 tipc_sk_withdraw(tsk
, 0, NULL
);
515 probing_state
= tsk
->probing_state
;
516 if (del_timer_sync(&tsk
->timer
) && probing_state
!= TIPC_CONN_PROBING
)
519 if (tsk
->connected
) {
520 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
521 SHORT_H_SIZE
, 0, dnode
, tipc_own_addr
,
523 tsk
->portid
, TIPC_ERR_NO_PORT
);
525 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
526 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
529 /* Discard any remaining (connection-based) messages in receive queue */
530 __skb_queue_purge(&sk
->sk_receive_queue
);
532 /* Reject any messages that accumulated in backlog queue */
533 sock
->state
= SS_DISCONNECTING
;
536 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
543 * tipc_bind - associate or disassocate TIPC name(s) with a socket
544 * @sock: socket structure
545 * @uaddr: socket address describing name(s) and desired operation
546 * @uaddr_len: size of socket address data structure
548 * Name and name sequence binding is indicated using a positive scope value;
549 * a negative scope value unbinds the specified name. Specifying no name
550 * (i.e. a socket address length of 0) unbinds all names from the socket.
552 * Returns 0 on success, errno otherwise
554 * NOTE: This routine doesn't need to take the socket lock since it doesn't
555 * access any non-constant socket information.
557 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
560 struct sock
*sk
= sock
->sk
;
561 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
562 struct tipc_sock
*tsk
= tipc_sk(sk
);
566 if (unlikely(!uaddr_len
)) {
567 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
571 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
575 if (addr
->family
!= AF_TIPC
) {
580 if (addr
->addrtype
== TIPC_ADDR_NAME
)
581 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
582 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
587 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
588 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
589 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
594 res
= (addr
->scope
> 0) ?
595 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
596 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
603 * tipc_getname - get port ID of socket or peer socket
604 * @sock: socket structure
605 * @uaddr: area for returned socket address
606 * @uaddr_len: area for returned length of socket address
607 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
609 * Returns 0 on success, errno otherwise
611 * NOTE: This routine doesn't need to take the socket lock since it only
612 * accesses socket information that is unchanging (or which changes in
613 * a completely predictable manner).
615 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
616 int *uaddr_len
, int peer
)
618 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
619 struct tipc_sock
*tsk
= tipc_sk(sock
->sk
);
621 memset(addr
, 0, sizeof(*addr
));
623 if ((sock
->state
!= SS_CONNECTED
) &&
624 ((peer
!= 2) || (sock
->state
!= SS_DISCONNECTING
)))
626 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
627 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
629 addr
->addr
.id
.ref
= tsk
->portid
;
630 addr
->addr
.id
.node
= tipc_own_addr
;
633 *uaddr_len
= sizeof(*addr
);
634 addr
->addrtype
= TIPC_ADDR_ID
;
635 addr
->family
= AF_TIPC
;
637 addr
->addr
.name
.domain
= 0;
643 * tipc_poll - read and possibly block on pollmask
644 * @file: file structure associated with the socket
645 * @sock: socket for which to calculate the poll bits
648 * Returns pollmask value
651 * It appears that the usual socket locking mechanisms are not useful here
652 * since the pollmask info is potentially out-of-date the moment this routine
653 * exits. TCP and other protocols seem to rely on higher level poll routines
654 * to handle any preventable race conditions, so TIPC will do the same ...
656 * TIPC sets the returned events as follows:
658 * socket state flags set
659 * ------------ ---------
660 * unconnected no read flags
661 * POLLOUT if port is not congested
663 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
666 * connected POLLIN/POLLRDNORM if data in rx queue
667 * POLLOUT if port is not congested
669 * disconnecting POLLIN/POLLRDNORM/POLLHUP
672 * listening POLLIN if SYN in rx queue
675 * ready POLLIN/POLLRDNORM if data in rx queue
676 * [connectionless] POLLOUT (since port cannot be congested)
678 * IMPORTANT: The fact that a read or write operation is indicated does NOT
679 * imply that the operation will succeed, merely that it should be performed
680 * and will not block.
682 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
685 struct sock
*sk
= sock
->sk
;
686 struct tipc_sock
*tsk
= tipc_sk(sk
);
689 sock_poll_wait(file
, sk_sleep(sk
), wait
);
691 switch ((int)sock
->state
) {
698 if (!tsk
->link_cong
&& !tsk_conn_cong(tsk
))
703 if (!skb_queue_empty(&sk
->sk_receive_queue
))
704 mask
|= (POLLIN
| POLLRDNORM
);
706 case SS_DISCONNECTING
:
707 mask
= (POLLIN
| POLLRDNORM
| POLLHUP
);
715 * tipc_sendmcast - send multicast message
716 * @sock: socket structure
717 * @seq: destination address
718 * @msg: message to send
719 * @dsz: total length of message data
720 * @timeo: timeout to wait for wakeup
722 * Called from function tipc_sendmsg(), which has done all sanity checks
723 * Returns the number of bytes sent on success, or errno
725 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
726 struct msghdr
*msg
, size_t dsz
, long timeo
)
728 struct sock
*sk
= sock
->sk
;
729 struct net
*net
= sock_net(sk
);
730 struct tipc_msg
*mhdr
= &tipc_sk(sk
)->phdr
;
731 struct sk_buff_head head
;
735 msg_set_type(mhdr
, TIPC_MCAST_MSG
);
736 msg_set_lookup_scope(mhdr
, TIPC_CLUSTER_SCOPE
);
737 msg_set_destport(mhdr
, 0);
738 msg_set_destnode(mhdr
, 0);
739 msg_set_nametype(mhdr
, seq
->type
);
740 msg_set_namelower(mhdr
, seq
->lower
);
741 msg_set_nameupper(mhdr
, seq
->upper
);
742 msg_set_hdr_sz(mhdr
, MCAST_H_SIZE
);
745 mtu
= tipc_bclink_get_mtu();
746 __skb_queue_head_init(&head
);
747 rc
= tipc_msg_build(mhdr
, msg
, 0, dsz
, mtu
, &head
);
748 if (unlikely(rc
< 0))
752 rc
= tipc_bclink_xmit(net
, &head
);
753 if (likely(rc
>= 0)) {
759 if (rc
!= -ELINKCONG
)
761 tipc_sk(sk
)->link_cong
= 1;
762 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
764 __skb_queue_purge(&head
);
769 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
771 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff
*buf
)
773 struct tipc_msg
*msg
= buf_msg(buf
);
774 struct tipc_port_list dports
= {0, NULL
, };
775 struct tipc_port_list
*item
;
777 uint i
, last
, dst
= 0;
778 u32 scope
= TIPC_CLUSTER_SCOPE
;
780 if (in_own_node(msg_orignode(msg
)))
781 scope
= TIPC_NODE_SCOPE
;
783 /* Create destination port list: */
784 tipc_nametbl_mc_translate(msg_nametype(msg
),
795 for (item
= &dports
; item
; item
= item
->next
) {
796 for (i
= 0; i
< PLSIZE
&& ++dst
<= last
; i
++) {
797 b
= (dst
!= last
) ? skb_clone(buf
, GFP_ATOMIC
) : buf
;
799 pr_warn("Failed do clone mcast rcv buffer\n");
802 msg_set_destport(msg
, item
->ports
[i
]);
806 tipc_port_list_free(&dports
);
810 * tipc_sk_proto_rcv - receive a connection mng protocol message
811 * @tsk: receiving socket
812 * @dnode: node to send response message to, if any
813 * @buf: buffer containing protocol message
814 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
815 * (CONN_PROBE_REPLY) message should be forwarded.
817 static int tipc_sk_proto_rcv(struct tipc_sock
*tsk
, u32
*dnode
,
820 struct tipc_msg
*msg
= buf_msg(buf
);
823 /* Ignore if connection cannot be validated: */
824 if (!tsk_peer_msg(tsk
, msg
))
827 tsk
->probing_state
= TIPC_CONN_OK
;
829 if (msg_type(msg
) == CONN_ACK
) {
830 conn_cong
= tsk_conn_cong(tsk
);
831 tsk
->sent_unacked
-= msg_msgcnt(msg
);
833 tsk
->sk
.sk_write_space(&tsk
->sk
);
834 } else if (msg_type(msg
) == CONN_PROBE
) {
835 if (!tipc_msg_reverse(buf
, dnode
, TIPC_OK
))
837 msg_set_type(msg
, CONN_PROBE_REPLY
);
840 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
846 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
)
848 struct sock
*sk
= sock
->sk
;
849 struct tipc_sock
*tsk
= tipc_sk(sk
);
854 int err
= sock_error(sk
);
857 if (sock
->state
== SS_DISCONNECTING
)
861 if (signal_pending(current
))
862 return sock_intr_errno(*timeo_p
);
864 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
865 done
= sk_wait_event(sk
, timeo_p
, !tsk
->link_cong
);
866 finish_wait(sk_sleep(sk
), &wait
);
872 * tipc_sendmsg - send message in connectionless manner
873 * @iocb: if NULL, indicates that socket lock is already held
874 * @sock: socket structure
875 * @m: message to send
876 * @dsz: amount of user data to be sent
878 * Message must have an destination specified explicitly.
879 * Used for SOCK_RDM and SOCK_DGRAM messages,
880 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
881 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
883 * Returns the number of bytes sent on success, or errno otherwise
885 static int tipc_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
886 struct msghdr
*m
, size_t dsz
)
888 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
889 struct sock
*sk
= sock
->sk
;
890 struct tipc_sock
*tsk
= tipc_sk(sk
);
891 struct net
*net
= sock_net(sk
);
892 struct tipc_msg
*mhdr
= &tsk
->phdr
;
894 struct sk_buff_head head
;
896 struct tipc_name_seq
*seq
= &dest
->addr
.nameseq
;
902 return -EDESTADDRREQ
;
904 if (unlikely((m
->msg_namelen
< sizeof(*dest
)) ||
905 (dest
->family
!= AF_TIPC
)))
908 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
914 if (unlikely(sock
->state
!= SS_READY
)) {
915 if (sock
->state
== SS_LISTENING
) {
919 if (sock
->state
!= SS_UNCONNECTED
) {
923 if (tsk
->published
) {
927 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
928 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
929 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
933 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
935 if (dest
->addrtype
== TIPC_ADDR_MCAST
) {
936 rc
= tipc_sendmcast(sock
, seq
, m
, dsz
, timeo
);
938 } else if (dest
->addrtype
== TIPC_ADDR_NAME
) {
939 u32 type
= dest
->addr
.name
.name
.type
;
940 u32 inst
= dest
->addr
.name
.name
.instance
;
941 u32 domain
= dest
->addr
.name
.domain
;
944 msg_set_type(mhdr
, TIPC_NAMED_MSG
);
945 msg_set_hdr_sz(mhdr
, NAMED_H_SIZE
);
946 msg_set_nametype(mhdr
, type
);
947 msg_set_nameinst(mhdr
, inst
);
948 msg_set_lookup_scope(mhdr
, tipc_addr_scope(domain
));
949 dport
= tipc_nametbl_translate(type
, inst
, &dnode
);
950 msg_set_destnode(mhdr
, dnode
);
951 msg_set_destport(mhdr
, dport
);
952 if (unlikely(!dport
&& !dnode
)) {
956 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
957 dnode
= dest
->addr
.id
.node
;
958 msg_set_type(mhdr
, TIPC_DIRECT_MSG
);
959 msg_set_lookup_scope(mhdr
, 0);
960 msg_set_destnode(mhdr
, dnode
);
961 msg_set_destport(mhdr
, dest
->addr
.id
.ref
);
962 msg_set_hdr_sz(mhdr
, BASIC_H_SIZE
);
966 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
967 __skb_queue_head_init(&head
);
968 rc
= tipc_msg_build(mhdr
, m
, 0, dsz
, mtu
, &head
);
973 skb
= skb_peek(&head
);
974 TIPC_SKB_CB(skb
)->wakeup_pending
= tsk
->link_cong
;
975 rc
= tipc_link_xmit(net
, &head
, dnode
, tsk
->portid
);
976 if (likely(rc
>= 0)) {
977 if (sock
->state
!= SS_READY
)
978 sock
->state
= SS_CONNECTING
;
984 if (rc
!= -ELINKCONG
)
987 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
989 __skb_queue_purge(&head
);
998 static int tipc_wait_for_sndpkt(struct socket
*sock
, long *timeo_p
)
1000 struct sock
*sk
= sock
->sk
;
1001 struct tipc_sock
*tsk
= tipc_sk(sk
);
1006 int err
= sock_error(sk
);
1009 if (sock
->state
== SS_DISCONNECTING
)
1011 else if (sock
->state
!= SS_CONNECTED
)
1015 if (signal_pending(current
))
1016 return sock_intr_errno(*timeo_p
);
1018 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1019 done
= sk_wait_event(sk
, timeo_p
,
1021 !tsk_conn_cong(tsk
)) ||
1023 finish_wait(sk_sleep(sk
), &wait
);
1029 * tipc_send_stream - send stream-oriented data
1031 * @sock: socket structure
1033 * @dsz: total length of data to be transmitted
1035 * Used for SOCK_STREAM data.
1037 * Returns the number of bytes sent on success (or partial success),
1038 * or errno if no data sent
1040 static int tipc_send_stream(struct kiocb
*iocb
, struct socket
*sock
,
1041 struct msghdr
*m
, size_t dsz
)
1043 struct sock
*sk
= sock
->sk
;
1044 struct net
*net
= sock_net(sk
);
1045 struct tipc_sock
*tsk
= tipc_sk(sk
);
1046 struct tipc_msg
*mhdr
= &tsk
->phdr
;
1047 struct sk_buff_head head
;
1048 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1049 u32 portid
= tsk
->portid
;
1053 uint mtu
, send
, sent
= 0;
1055 /* Handle implied connection establishment */
1056 if (unlikely(dest
)) {
1057 rc
= tipc_sendmsg(iocb
, sock
, m
, dsz
);
1058 if (dsz
&& (dsz
== rc
))
1059 tsk
->sent_unacked
= 1;
1062 if (dsz
> (uint
)INT_MAX
)
1068 if (unlikely(sock
->state
!= SS_CONNECTED
)) {
1069 if (sock
->state
== SS_DISCONNECTING
)
1076 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1077 dnode
= tsk_peer_node(tsk
);
1081 send
= min_t(uint
, dsz
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1082 __skb_queue_head_init(&head
);
1083 rc
= tipc_msg_build(mhdr
, m
, sent
, send
, mtu
, &head
);
1084 if (unlikely(rc
< 0))
1087 if (likely(!tsk_conn_cong(tsk
))) {
1088 rc
= tipc_link_xmit(net
, &head
, dnode
, portid
);
1090 tsk
->sent_unacked
++;
1096 if (rc
== -EMSGSIZE
) {
1097 tsk
->max_pkt
= tipc_node_get_mtu(net
, dnode
,
1101 if (rc
!= -ELINKCONG
)
1105 rc
= tipc_wait_for_sndpkt(sock
, &timeo
);
1107 __skb_queue_purge(&head
);
1112 return sent
? sent
: rc
;
1116 * tipc_send_packet - send a connection-oriented message
1117 * @iocb: if NULL, indicates that socket lock is already held
1118 * @sock: socket structure
1119 * @m: message to send
1120 * @dsz: length of data to be transmitted
1122 * Used for SOCK_SEQPACKET messages.
1124 * Returns the number of bytes sent on success, or errno otherwise
1126 static int tipc_send_packet(struct kiocb
*iocb
, struct socket
*sock
,
1127 struct msghdr
*m
, size_t dsz
)
1129 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1132 return tipc_send_stream(iocb
, sock
, m
, dsz
);
1135 /* tipc_sk_finish_conn - complete the setup of a connection
1137 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1140 struct net
*net
= sock_net(&tsk
->sk
);
1141 struct tipc_msg
*msg
= &tsk
->phdr
;
1143 msg_set_destnode(msg
, peer_node
);
1144 msg_set_destport(msg
, peer_port
);
1145 msg_set_type(msg
, TIPC_CONN_MSG
);
1146 msg_set_lookup_scope(msg
, 0);
1147 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1149 tsk
->probing_intv
= CONN_PROBING_INTERVAL
;
1150 tsk
->probing_state
= TIPC_CONN_OK
;
1152 if (!mod_timer(&tsk
->timer
, jiffies
+ tsk
->probing_intv
))
1153 sock_hold(&tsk
->sk
);
1154 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1155 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1159 * set_orig_addr - capture sender's address for received message
1160 * @m: descriptor for message info
1161 * @msg: received message header
1163 * Note: Address is not captured if not requested by receiver.
1165 static void set_orig_addr(struct msghdr
*m
, struct tipc_msg
*msg
)
1167 DECLARE_SOCKADDR(struct sockaddr_tipc
*, addr
, m
->msg_name
);
1170 addr
->family
= AF_TIPC
;
1171 addr
->addrtype
= TIPC_ADDR_ID
;
1172 memset(&addr
->addr
, 0, sizeof(addr
->addr
));
1173 addr
->addr
.id
.ref
= msg_origport(msg
);
1174 addr
->addr
.id
.node
= msg_orignode(msg
);
1175 addr
->addr
.name
.domain
= 0; /* could leave uninitialized */
1176 addr
->scope
= 0; /* could leave uninitialized */
1177 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1182 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1183 * @m: descriptor for message info
1184 * @msg: received message header
1185 * @tsk: TIPC port associated with message
1187 * Note: Ancillary data is not captured if not requested by receiver.
1189 * Returns 0 if successful, otherwise errno
1191 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1192 struct tipc_sock
*tsk
)
1200 if (likely(m
->msg_controllen
== 0))
1203 /* Optionally capture errored message object(s) */
1204 err
= msg
? msg_errcode(msg
) : 0;
1205 if (unlikely(err
)) {
1207 anc_data
[1] = msg_data_sz(msg
);
1208 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1212 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1219 /* Optionally capture message destination object */
1220 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1221 switch (dest_type
) {
1222 case TIPC_NAMED_MSG
:
1224 anc_data
[0] = msg_nametype(msg
);
1225 anc_data
[1] = msg_namelower(msg
);
1226 anc_data
[2] = msg_namelower(msg
);
1228 case TIPC_MCAST_MSG
:
1230 anc_data
[0] = msg_nametype(msg
);
1231 anc_data
[1] = msg_namelower(msg
);
1232 anc_data
[2] = msg_nameupper(msg
);
1235 has_name
= (tsk
->conn_type
!= 0);
1236 anc_data
[0] = tsk
->conn_type
;
1237 anc_data
[1] = tsk
->conn_instance
;
1238 anc_data
[2] = tsk
->conn_instance
;
1244 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1252 static void tipc_sk_send_ack(struct tipc_sock
*tsk
, uint ack
)
1254 struct net
*net
= sock_net(&tsk
->sk
);
1255 struct sk_buff
*skb
= NULL
;
1256 struct tipc_msg
*msg
;
1257 u32 peer_port
= tsk_peer_port(tsk
);
1258 u32 dnode
= tsk_peer_node(tsk
);
1260 if (!tsk
->connected
)
1262 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0, dnode
,
1263 tipc_own_addr
, peer_port
, tsk
->portid
, TIPC_OK
);
1267 msg_set_msgcnt(msg
, ack
);
1268 tipc_link_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1271 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1273 struct sock
*sk
= sock
->sk
;
1275 long timeo
= *timeop
;
1279 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1280 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1281 if (sock
->state
== SS_DISCONNECTING
) {
1286 timeo
= schedule_timeout(timeo
);
1290 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1292 err
= sock_intr_errno(timeo
);
1293 if (signal_pending(current
))
1299 finish_wait(sk_sleep(sk
), &wait
);
1305 * tipc_recvmsg - receive packet-oriented message
1307 * @m: descriptor for message info
1308 * @buf_len: total size of user buffer area
1309 * @flags: receive flags
1311 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1312 * If the complete message doesn't fit in user area, truncate it.
1314 * Returns size of returned message data, errno otherwise
1316 static int tipc_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1317 struct msghdr
*m
, size_t buf_len
, int flags
)
1319 struct sock
*sk
= sock
->sk
;
1320 struct tipc_sock
*tsk
= tipc_sk(sk
);
1321 struct sk_buff
*buf
;
1322 struct tipc_msg
*msg
;
1328 /* Catch invalid receive requests */
1329 if (unlikely(!buf_len
))
1334 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1339 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1342 /* Look for a message in receive queue; wait if necessary */
1343 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1347 /* Look at first message in receive queue */
1348 buf
= skb_peek(&sk
->sk_receive_queue
);
1350 sz
= msg_data_sz(msg
);
1351 err
= msg_errcode(msg
);
1353 /* Discard an empty non-errored message & try again */
1354 if ((!sz
) && (!err
)) {
1355 tsk_advance_rx_queue(sk
);
1359 /* Capture sender's address (optional) */
1360 set_orig_addr(m
, msg
);
1362 /* Capture ancillary data (optional) */
1363 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1367 /* Capture message data (if valid) & compute return value (always) */
1369 if (unlikely(buf_len
< sz
)) {
1371 m
->msg_flags
|= MSG_TRUNC
;
1373 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
), m
, sz
);
1378 if ((sock
->state
== SS_READY
) ||
1379 ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
))
1385 /* Consume received message (optional) */
1386 if (likely(!(flags
& MSG_PEEK
))) {
1387 if ((sock
->state
!= SS_READY
) &&
1388 (++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1389 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1390 tsk
->rcv_unacked
= 0;
1392 tsk_advance_rx_queue(sk
);
1400 * tipc_recv_stream - receive stream-oriented data
1402 * @m: descriptor for message info
1403 * @buf_len: total size of user buffer area
1404 * @flags: receive flags
1406 * Used for SOCK_STREAM messages only. If not enough data is available
1407 * will optionally wait for more; never truncates data.
1409 * Returns size of returned message data, errno otherwise
1411 static int tipc_recv_stream(struct kiocb
*iocb
, struct socket
*sock
,
1412 struct msghdr
*m
, size_t buf_len
, int flags
)
1414 struct sock
*sk
= sock
->sk
;
1415 struct tipc_sock
*tsk
= tipc_sk(sk
);
1416 struct sk_buff
*buf
;
1417 struct tipc_msg
*msg
;
1420 int sz_to_copy
, target
, needed
;
1425 /* Catch invalid receive attempts */
1426 if (unlikely(!buf_len
))
1431 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1436 target
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buf_len
);
1437 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1440 /* Look for a message in receive queue; wait if necessary */
1441 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1445 /* Look at first message in receive queue */
1446 buf
= skb_peek(&sk
->sk_receive_queue
);
1448 sz
= msg_data_sz(msg
);
1449 err
= msg_errcode(msg
);
1451 /* Discard an empty non-errored message & try again */
1452 if ((!sz
) && (!err
)) {
1453 tsk_advance_rx_queue(sk
);
1457 /* Optionally capture sender's address & ancillary data of first msg */
1458 if (sz_copied
== 0) {
1459 set_orig_addr(m
, msg
);
1460 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1465 /* Capture message data (if valid) & compute return value (always) */
1467 u32 offset
= (u32
)(unsigned long)(TIPC_SKB_CB(buf
)->handle
);
1470 needed
= (buf_len
- sz_copied
);
1471 sz_to_copy
= (sz
<= needed
) ? sz
: needed
;
1473 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
) + offset
,
1478 sz_copied
+= sz_to_copy
;
1480 if (sz_to_copy
< sz
) {
1481 if (!(flags
& MSG_PEEK
))
1482 TIPC_SKB_CB(buf
)->handle
=
1483 (void *)(unsigned long)(offset
+ sz_to_copy
);
1488 goto exit
; /* can't add error msg to valid data */
1490 if ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
)
1496 /* Consume received message (optional) */
1497 if (likely(!(flags
& MSG_PEEK
))) {
1498 if (unlikely(++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1499 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1500 tsk
->rcv_unacked
= 0;
1502 tsk_advance_rx_queue(sk
);
1505 /* Loop around if more data is required */
1506 if ((sz_copied
< buf_len
) && /* didn't get all requested data */
1507 (!skb_queue_empty(&sk
->sk_receive_queue
) ||
1508 (sz_copied
< target
)) && /* and more is ready or required */
1509 (!(flags
& MSG_PEEK
)) && /* and aren't just peeking at data */
1510 (!err
)) /* and haven't reached a FIN */
1515 return sz_copied
? sz_copied
: res
;
1519 * tipc_write_space - wake up thread if port congestion is released
1522 static void tipc_write_space(struct sock
*sk
)
1524 struct socket_wq
*wq
;
1527 wq
= rcu_dereference(sk
->sk_wq
);
1528 if (wq_has_sleeper(wq
))
1529 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1530 POLLWRNORM
| POLLWRBAND
);
1535 * tipc_data_ready - wake up threads to indicate messages have been received
1537 * @len: the length of messages
1539 static void tipc_data_ready(struct sock
*sk
)
1541 struct socket_wq
*wq
;
1544 wq
= rcu_dereference(sk
->sk_wq
);
1545 if (wq_has_sleeper(wq
))
1546 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1547 POLLRDNORM
| POLLRDBAND
);
1552 * filter_connect - Handle all incoming messages for a connection-based socket
1556 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1558 static int filter_connect(struct tipc_sock
*tsk
, struct sk_buff
**buf
)
1560 struct sock
*sk
= &tsk
->sk
;
1561 struct net
*net
= sock_net(sk
);
1562 struct socket
*sock
= sk
->sk_socket
;
1563 struct tipc_msg
*msg
= buf_msg(*buf
);
1564 int retval
= -TIPC_ERR_NO_PORT
;
1569 switch ((int)sock
->state
) {
1571 /* Accept only connection-based messages sent by peer */
1572 if (tsk_peer_msg(tsk
, msg
)) {
1573 if (unlikely(msg_errcode(msg
))) {
1574 sock
->state
= SS_DISCONNECTING
;
1576 /* let timer expire on it's own */
1577 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
1584 /* Accept only ACK or NACK message */
1586 if (unlikely(!msg_connected(msg
)))
1589 if (unlikely(msg_errcode(msg
))) {
1590 sock
->state
= SS_DISCONNECTING
;
1591 sk
->sk_err
= ECONNREFUSED
;
1596 if (unlikely(msg_importance(msg
) > TIPC_CRITICAL_IMPORTANCE
)) {
1597 sock
->state
= SS_DISCONNECTING
;
1598 sk
->sk_err
= EINVAL
;
1603 tipc_sk_finish_conn(tsk
, msg_origport(msg
), msg_orignode(msg
));
1604 msg_set_importance(&tsk
->phdr
, msg_importance(msg
));
1605 sock
->state
= SS_CONNECTED
;
1607 /* If an incoming message is an 'ACK-', it should be
1608 * discarded here because it doesn't contain useful
1609 * data. In addition, we should try to wake up
1610 * connect() routine if sleeping.
1612 if (msg_data_sz(msg
) == 0) {
1615 if (waitqueue_active(sk_sleep(sk
)))
1616 wake_up_interruptible(sk_sleep(sk
));
1621 case SS_UNCONNECTED
:
1622 /* Accept only SYN message */
1623 if (!msg_connected(msg
) && !(msg_errcode(msg
)))
1626 case SS_DISCONNECTING
:
1629 pr_err("Unknown socket state %u\n", sock
->state
);
1635 * rcvbuf_limit - get proper overload limit of socket receive queue
1639 * For all connection oriented messages, irrespective of importance,
1640 * the default overload value (i.e. 67MB) is set as limit.
1642 * For all connectionless messages, by default new queue limits are
1645 * TIPC_LOW_IMPORTANCE (4 MB)
1646 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1647 * TIPC_HIGH_IMPORTANCE (16 MB)
1648 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1650 * Returns overload limit according to corresponding message importance
1652 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*buf
)
1654 struct tipc_msg
*msg
= buf_msg(buf
);
1656 if (msg_connected(msg
))
1657 return sysctl_tipc_rmem
[2];
1659 return sk
->sk_rcvbuf
>> TIPC_CRITICAL_IMPORTANCE
<<
1660 msg_importance(msg
);
1664 * filter_rcv - validate incoming message
1668 * Enqueues message on receive queue if acceptable; optionally handles
1669 * disconnect indication for a connected socket.
1671 * Called with socket lock already taken; port lock may also be taken.
1673 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1674 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1676 static int filter_rcv(struct sock
*sk
, struct sk_buff
*buf
)
1678 struct socket
*sock
= sk
->sk_socket
;
1679 struct tipc_sock
*tsk
= tipc_sk(sk
);
1680 struct tipc_msg
*msg
= buf_msg(buf
);
1681 unsigned int limit
= rcvbuf_limit(sk
, buf
);
1685 if (unlikely(msg_user(msg
) == CONN_MANAGER
))
1686 return tipc_sk_proto_rcv(tsk
, &onode
, buf
);
1688 if (unlikely(msg_user(msg
) == SOCK_WAKEUP
)) {
1691 sk
->sk_write_space(sk
);
1695 /* Reject message if it is wrong sort of message for socket */
1696 if (msg_type(msg
) > TIPC_DIRECT_MSG
)
1697 return -TIPC_ERR_NO_PORT
;
1699 if (sock
->state
== SS_READY
) {
1700 if (msg_connected(msg
))
1701 return -TIPC_ERR_NO_PORT
;
1703 rc
= filter_connect(tsk
, &buf
);
1704 if (rc
!= TIPC_OK
|| buf
== NULL
)
1708 /* Reject message if there isn't room to queue it */
1709 if (sk_rmem_alloc_get(sk
) + buf
->truesize
>= limit
)
1710 return -TIPC_ERR_OVERLOAD
;
1712 /* Enqueue message */
1713 TIPC_SKB_CB(buf
)->handle
= NULL
;
1714 __skb_queue_tail(&sk
->sk_receive_queue
, buf
);
1715 skb_set_owner_r(buf
, sk
);
1717 sk
->sk_data_ready(sk
);
1722 * tipc_backlog_rcv - handle incoming message from backlog queue
1726 * Caller must hold socket lock, but not port lock.
1730 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1734 struct tipc_sock
*tsk
= tipc_sk(sk
);
1735 uint truesize
= skb
->truesize
;
1737 rc
= filter_rcv(sk
, skb
);
1740 if (atomic_read(&tsk
->dupl_rcvcnt
) < TIPC_CONN_OVERLOAD_LIMIT
)
1741 atomic_add(truesize
, &tsk
->dupl_rcvcnt
);
1745 if ((rc
< 0) && !tipc_msg_reverse(skb
, &onode
, -rc
))
1748 tipc_link_xmit_skb(sock_net(sk
), skb
, onode
, 0);
1754 * tipc_sk_rcv - handle incoming message
1755 * @skb: buffer containing arriving message
1757 * Returns 0 if success, or errno: -EHOSTUNREACH
1759 int tipc_sk_rcv(struct net
*net
, struct sk_buff
*skb
)
1761 struct tipc_sock
*tsk
;
1763 u32 dport
= msg_destport(buf_msg(skb
));
1768 /* Validate destination and message */
1769 tsk
= tipc_sk_lookup(dport
);
1770 if (unlikely(!tsk
)) {
1771 rc
= tipc_msg_eval(skb
, &dnode
);
1777 spin_lock_bh(&sk
->sk_lock
.slock
);
1779 if (!sock_owned_by_user(sk
)) {
1780 rc
= filter_rcv(sk
, skb
);
1782 if (sk
->sk_backlog
.len
== 0)
1783 atomic_set(&tsk
->dupl_rcvcnt
, 0);
1784 limit
= rcvbuf_limit(sk
, skb
) + atomic_read(&tsk
->dupl_rcvcnt
);
1785 if (sk_add_backlog(sk
, skb
, limit
))
1786 rc
= -TIPC_ERR_OVERLOAD
;
1788 spin_unlock_bh(&sk
->sk_lock
.slock
);
1793 if ((rc
< 0) && !tipc_msg_reverse(skb
, &dnode
, -rc
))
1794 return -EHOSTUNREACH
;
1796 tipc_link_xmit_skb(net
, skb
, dnode
, 0);
1797 return (rc
< 0) ? -EHOSTUNREACH
: 0;
1800 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
1802 struct sock
*sk
= sock
->sk
;
1807 int err
= sock_error(sk
);
1812 if (signal_pending(current
))
1813 return sock_intr_errno(*timeo_p
);
1815 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1816 done
= sk_wait_event(sk
, timeo_p
, sock
->state
!= SS_CONNECTING
);
1817 finish_wait(sk_sleep(sk
), &wait
);
1823 * tipc_connect - establish a connection to another TIPC port
1824 * @sock: socket structure
1825 * @dest: socket address for destination port
1826 * @destlen: size of socket address data structure
1827 * @flags: file-related flags associated with socket
1829 * Returns 0 on success, errno otherwise
1831 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
1832 int destlen
, int flags
)
1834 struct sock
*sk
= sock
->sk
;
1835 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
1836 struct msghdr m
= {NULL
,};
1837 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tipc_sk(sk
)->conn_timeout
;
1838 socket_state previous
;
1843 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1844 if (sock
->state
== SS_READY
) {
1850 * Reject connection attempt using multicast address
1852 * Note: send_msg() validates the rest of the address fields,
1853 * so there's no need to do it here
1855 if (dst
->addrtype
== TIPC_ADDR_MCAST
) {
1860 previous
= sock
->state
;
1861 switch (sock
->state
) {
1862 case SS_UNCONNECTED
:
1863 /* Send a 'SYN-' to destination */
1865 m
.msg_namelen
= destlen
;
1867 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1868 * indicate send_msg() is never blocked.
1871 m
.msg_flags
= MSG_DONTWAIT
;
1873 res
= tipc_sendmsg(NULL
, sock
, &m
, 0);
1874 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
1877 /* Just entered SS_CONNECTING state; the only
1878 * difference is that return value in non-blocking
1879 * case is EINPROGRESS, rather than EALREADY.
1883 if (previous
== SS_CONNECTING
)
1887 timeout
= msecs_to_jiffies(timeout
);
1888 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1889 res
= tipc_wait_for_connect(sock
, &timeout
);
1904 * tipc_listen - allow socket to listen for incoming connections
1905 * @sock: socket structure
1908 * Returns 0 on success, errno otherwise
1910 static int tipc_listen(struct socket
*sock
, int len
)
1912 struct sock
*sk
= sock
->sk
;
1917 if (sock
->state
!= SS_UNCONNECTED
)
1920 sock
->state
= SS_LISTENING
;
1928 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
1930 struct sock
*sk
= sock
->sk
;
1934 /* True wake-one mechanism for incoming connections: only
1935 * one process gets woken up, not the 'whole herd'.
1936 * Since we do not 'race & poll' for established sockets
1937 * anymore, the common case will execute the loop only once.
1940 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
1941 TASK_INTERRUPTIBLE
);
1942 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1944 timeo
= schedule_timeout(timeo
);
1948 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1951 if (sock
->state
!= SS_LISTENING
)
1953 err
= sock_intr_errno(timeo
);
1954 if (signal_pending(current
))
1960 finish_wait(sk_sleep(sk
), &wait
);
1965 * tipc_accept - wait for connection request
1966 * @sock: listening socket
1967 * @newsock: new socket that is to be connected
1968 * @flags: file-related flags associated with socket
1970 * Returns 0 on success, errno otherwise
1972 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
)
1974 struct sock
*new_sk
, *sk
= sock
->sk
;
1975 struct sk_buff
*buf
;
1976 struct tipc_sock
*new_tsock
;
1977 struct tipc_msg
*msg
;
1983 if (sock
->state
!= SS_LISTENING
) {
1987 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
1988 res
= tipc_wait_for_accept(sock
, timeo
);
1992 buf
= skb_peek(&sk
->sk_receive_queue
);
1994 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, 1);
1998 new_sk
= new_sock
->sk
;
1999 new_tsock
= tipc_sk(new_sk
);
2002 /* we lock on new_sk; but lockdep sees the lock on sk */
2003 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2006 * Reject any stray messages received by new socket
2007 * before the socket lock was taken (very, very unlikely)
2009 tsk_rej_rx_queue(new_sk
);
2011 /* Connect new socket to it's peer */
2012 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2013 new_sock
->state
= SS_CONNECTED
;
2015 tsk_set_importance(new_tsock
, msg_importance(msg
));
2016 if (msg_named(msg
)) {
2017 new_tsock
->conn_type
= msg_nametype(msg
);
2018 new_tsock
->conn_instance
= msg_nameinst(msg
);
2022 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2023 * Respond to 'SYN+' by queuing it on new socket.
2025 if (!msg_data_sz(msg
)) {
2026 struct msghdr m
= {NULL
,};
2028 tsk_advance_rx_queue(sk
);
2029 tipc_send_packet(NULL
, new_sock
, &m
, 0);
2031 __skb_dequeue(&sk
->sk_receive_queue
);
2032 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2033 skb_set_owner_r(buf
, new_sk
);
2035 release_sock(new_sk
);
2042 * tipc_shutdown - shutdown socket connection
2043 * @sock: socket structure
2044 * @how: direction to close (must be SHUT_RDWR)
2046 * Terminates connection (if necessary), then purges socket's receive queue.
2048 * Returns 0 on success, errno otherwise
2050 static int tipc_shutdown(struct socket
*sock
, int how
)
2052 struct sock
*sk
= sock
->sk
;
2053 struct net
*net
= sock_net(sk
);
2054 struct tipc_sock
*tsk
= tipc_sk(sk
);
2055 struct sk_buff
*skb
;
2059 if (how
!= SHUT_RDWR
)
2064 switch (sock
->state
) {
2069 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2070 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
2072 if (TIPC_SKB_CB(skb
)->handle
!= NULL
) {
2076 if (tipc_msg_reverse(skb
, &dnode
, TIPC_CONN_SHUTDOWN
))
2077 tipc_link_xmit_skb(net
, skb
, dnode
,
2079 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
2081 dnode
= tsk_peer_node(tsk
);
2082 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
2083 TIPC_CONN_MSG
, SHORT_H_SIZE
,
2084 0, dnode
, tipc_own_addr
,
2086 tsk
->portid
, TIPC_CONN_SHUTDOWN
);
2087 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
2090 sock
->state
= SS_DISCONNECTING
;
2091 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
2094 case SS_DISCONNECTING
:
2096 /* Discard any unreceived messages */
2097 __skb_queue_purge(&sk
->sk_receive_queue
);
2099 /* Wake up anyone sleeping in poll */
2100 sk
->sk_state_change(sk
);
2112 static void tipc_sk_timeout(unsigned long data
)
2114 struct tipc_sock
*tsk
= (struct tipc_sock
*)data
;
2115 struct sock
*sk
= &tsk
->sk
;
2116 struct sk_buff
*skb
= NULL
;
2117 u32 peer_port
, peer_node
;
2120 if (!tsk
->connected
) {
2124 peer_port
= tsk_peer_port(tsk
);
2125 peer_node
= tsk_peer_node(tsk
);
2127 if (tsk
->probing_state
== TIPC_CONN_PROBING
) {
2128 /* Previous probe not answered -> self abort */
2129 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
, TIPC_CONN_MSG
,
2130 SHORT_H_SIZE
, 0, tipc_own_addr
,
2131 peer_node
, tsk
->portid
, peer_port
,
2134 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
,
2135 0, peer_node
, tipc_own_addr
,
2136 peer_port
, tsk
->portid
, TIPC_OK
);
2137 tsk
->probing_state
= TIPC_CONN_PROBING
;
2138 if (!mod_timer(&tsk
->timer
, jiffies
+ tsk
->probing_intv
))
2143 tipc_link_xmit_skb(sock_net(sk
), skb
, peer_node
, tsk
->portid
);
2148 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2149 struct tipc_name_seq
const *seq
)
2151 struct net
*net
= sock_net(&tsk
->sk
);
2152 struct publication
*publ
;
2157 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2158 if (key
== tsk
->portid
)
2161 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2162 scope
, tsk
->portid
, key
);
2163 if (unlikely(!publ
))
2166 list_add(&publ
->pport_list
, &tsk
->publications
);
2172 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2173 struct tipc_name_seq
const *seq
)
2175 struct net
*net
= sock_net(&tsk
->sk
);
2176 struct publication
*publ
;
2177 struct publication
*safe
;
2180 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2182 if (publ
->scope
!= scope
)
2184 if (publ
->type
!= seq
->type
)
2186 if (publ
->lower
!= seq
->lower
)
2188 if (publ
->upper
!= seq
->upper
)
2190 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2191 publ
->ref
, publ
->key
);
2195 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2196 publ
->ref
, publ
->key
);
2199 if (list_empty(&tsk
->publications
))
2204 static int tipc_sk_show(struct tipc_sock
*tsk
, char *buf
,
2205 int len
, int full_id
)
2207 struct publication
*publ
;
2211 ret
= tipc_snprintf(buf
, len
, "<%u.%u.%u:%u>:",
2212 tipc_zone(tipc_own_addr
),
2213 tipc_cluster(tipc_own_addr
),
2214 tipc_node(tipc_own_addr
), tsk
->portid
);
2216 ret
= tipc_snprintf(buf
, len
, "%-10u:", tsk
->portid
);
2218 if (tsk
->connected
) {
2219 u32 dport
= tsk_peer_port(tsk
);
2220 u32 destnode
= tsk_peer_node(tsk
);
2222 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2223 " connected to <%u.%u.%u:%u>",
2224 tipc_zone(destnode
),
2225 tipc_cluster(destnode
),
2226 tipc_node(destnode
), dport
);
2227 if (tsk
->conn_type
!= 0)
2228 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2229 " via {%u,%u}", tsk
->conn_type
,
2230 tsk
->conn_instance
);
2231 } else if (tsk
->published
) {
2232 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
, " bound to");
2233 list_for_each_entry(publ
, &tsk
->publications
, pport_list
) {
2234 if (publ
->lower
== publ
->upper
)
2235 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2236 " {%u,%u}", publ
->type
,
2239 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
,
2240 " {%u,%u,%u}", publ
->type
,
2241 publ
->lower
, publ
->upper
);
2244 ret
+= tipc_snprintf(buf
+ ret
, len
- ret
, "\n");
2248 struct sk_buff
*tipc_sk_socks_show(void)
2250 const struct bucket_table
*tbl
;
2251 struct rhash_head
*pos
;
2252 struct sk_buff
*buf
;
2253 struct tlv_desc
*rep_tlv
;
2256 struct tipc_sock
*tsk
;
2260 buf
= tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN
));
2263 rep_tlv
= (struct tlv_desc
*)buf
->data
;
2264 pb
= TLV_DATA(rep_tlv
);
2265 pb_len
= ULTRA_STRING_MAX_LEN
;
2268 tbl
= rht_dereference_rcu((&tipc_sk_rht
)->tbl
, &tipc_sk_rht
);
2269 for (i
= 0; i
< tbl
->size
; i
++) {
2270 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2271 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2272 str_len
+= tipc_sk_show(tsk
, pb
+ str_len
,
2273 pb_len
- str_len
, 0);
2274 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2279 str_len
+= 1; /* for "\0" */
2280 skb_put(buf
, TLV_SPACE(str_len
));
2281 TLV_SET(rep_tlv
, TIPC_TLV_ULTRA_STRING
, NULL
, str_len
);
2286 /* tipc_sk_reinit: set non-zero address in all existing sockets
2287 * when we go from standalone to network mode.
2289 void tipc_sk_reinit(void)
2291 const struct bucket_table
*tbl
;
2292 struct rhash_head
*pos
;
2293 struct tipc_sock
*tsk
;
2294 struct tipc_msg
*msg
;
2298 tbl
= rht_dereference_rcu((&tipc_sk_rht
)->tbl
, &tipc_sk_rht
);
2299 for (i
= 0; i
< tbl
->size
; i
++) {
2300 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2301 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2303 msg_set_prevnode(msg
, tipc_own_addr
);
2304 msg_set_orignode(msg
, tipc_own_addr
);
2305 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2311 static struct tipc_sock
*tipc_sk_lookup(u32 portid
)
2313 struct tipc_sock
*tsk
;
2316 tsk
= rhashtable_lookup(&tipc_sk_rht
, &portid
);
2318 sock_hold(&tsk
->sk
);
2324 static int tipc_sk_insert(struct tipc_sock
*tsk
)
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(&tipc_sk_rht
, &tsk
->node
))
2343 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2345 struct sock
*sk
= &tsk
->sk
;
2347 if (rhashtable_remove(&tipc_sk_rht
, &tsk
->node
)) {
2348 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
2353 int tipc_sk_rht_init(void)
2355 struct rhashtable_params rht_params
= {
2357 .head_offset
= offsetof(struct tipc_sock
, node
),
2358 .key_offset
= offsetof(struct tipc_sock
, portid
),
2359 .key_len
= sizeof(u32
), /* portid */
2361 .max_shift
= 20, /* 1M */
2362 .min_shift
= 8, /* 256 */
2363 .grow_decision
= rht_grow_above_75
,
2364 .shrink_decision
= rht_shrink_below_30
,
2367 return rhashtable_init(&tipc_sk_rht
, &rht_params
);
2370 void tipc_sk_rht_destroy(void)
2372 /* Wait for socket readers to complete */
2375 rhashtable_destroy(&tipc_sk_rht
);
2379 * tipc_setsockopt - set socket option
2380 * @sock: socket structure
2381 * @lvl: option level
2382 * @opt: option identifier
2383 * @ov: pointer to new option value
2384 * @ol: length of option value
2386 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2387 * (to ease compatibility).
2389 * Returns 0 on success, errno otherwise
2391 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2392 char __user
*ov
, unsigned int ol
)
2394 struct sock
*sk
= sock
->sk
;
2395 struct tipc_sock
*tsk
= tipc_sk(sk
);
2399 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2401 if (lvl
!= SOL_TIPC
)
2402 return -ENOPROTOOPT
;
2403 if (ol
< sizeof(value
))
2405 res
= get_user(value
, (u32 __user
*)ov
);
2412 case TIPC_IMPORTANCE
:
2413 res
= tsk_set_importance(tsk
, value
);
2415 case TIPC_SRC_DROPPABLE
:
2416 if (sock
->type
!= SOCK_STREAM
)
2417 tsk_set_unreliable(tsk
, value
);
2421 case TIPC_DEST_DROPPABLE
:
2422 tsk_set_unreturnable(tsk
, value
);
2424 case TIPC_CONN_TIMEOUT
:
2425 tipc_sk(sk
)->conn_timeout
= value
;
2426 /* no need to set "res", since already 0 at this point */
2438 * tipc_getsockopt - get socket option
2439 * @sock: socket structure
2440 * @lvl: option level
2441 * @opt: option identifier
2442 * @ov: receptacle for option value
2443 * @ol: receptacle for length of option value
2445 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2446 * (to ease compatibility).
2448 * Returns 0 on success, errno otherwise
2450 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2451 char __user
*ov
, int __user
*ol
)
2453 struct sock
*sk
= sock
->sk
;
2454 struct tipc_sock
*tsk
= tipc_sk(sk
);
2459 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2460 return put_user(0, ol
);
2461 if (lvl
!= SOL_TIPC
)
2462 return -ENOPROTOOPT
;
2463 res
= get_user(len
, ol
);
2470 case TIPC_IMPORTANCE
:
2471 value
= tsk_importance(tsk
);
2473 case TIPC_SRC_DROPPABLE
:
2474 value
= tsk_unreliable(tsk
);
2476 case TIPC_DEST_DROPPABLE
:
2477 value
= tsk_unreturnable(tsk
);
2479 case TIPC_CONN_TIMEOUT
:
2480 value
= tsk
->conn_timeout
;
2481 /* no need to set "res", since already 0 at this point */
2483 case TIPC_NODE_RECVQ_DEPTH
:
2484 value
= 0; /* was tipc_queue_size, now obsolete */
2486 case TIPC_SOCK_RECVQ_DEPTH
:
2487 value
= skb_queue_len(&sk
->sk_receive_queue
);
2496 return res
; /* "get" failed */
2498 if (len
< sizeof(value
))
2501 if (copy_to_user(ov
, &value
, sizeof(value
)))
2504 return put_user(sizeof(value
), ol
);
2507 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2509 struct sock
*sk
= sock
->sk
;
2510 struct tipc_sioc_ln_req lnr
;
2511 void __user
*argp
= (void __user
*)arg
;
2514 case SIOCGETLINKNAME
:
2515 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2517 if (!tipc_node_get_linkname(sock_net(sk
),
2518 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2519 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2520 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2524 return -EADDRNOTAVAIL
;
2526 return -ENOIOCTLCMD
;
2530 /* Protocol switches for the various types of TIPC sockets */
2532 static const struct proto_ops msg_ops
= {
2533 .owner
= THIS_MODULE
,
2535 .release
= tipc_release
,
2537 .connect
= tipc_connect
,
2538 .socketpair
= sock_no_socketpair
,
2539 .accept
= sock_no_accept
,
2540 .getname
= tipc_getname
,
2542 .ioctl
= tipc_ioctl
,
2543 .listen
= sock_no_listen
,
2544 .shutdown
= tipc_shutdown
,
2545 .setsockopt
= tipc_setsockopt
,
2546 .getsockopt
= tipc_getsockopt
,
2547 .sendmsg
= tipc_sendmsg
,
2548 .recvmsg
= tipc_recvmsg
,
2549 .mmap
= sock_no_mmap
,
2550 .sendpage
= sock_no_sendpage
2553 static const struct proto_ops packet_ops
= {
2554 .owner
= THIS_MODULE
,
2556 .release
= tipc_release
,
2558 .connect
= tipc_connect
,
2559 .socketpair
= sock_no_socketpair
,
2560 .accept
= tipc_accept
,
2561 .getname
= tipc_getname
,
2563 .ioctl
= tipc_ioctl
,
2564 .listen
= tipc_listen
,
2565 .shutdown
= tipc_shutdown
,
2566 .setsockopt
= tipc_setsockopt
,
2567 .getsockopt
= tipc_getsockopt
,
2568 .sendmsg
= tipc_send_packet
,
2569 .recvmsg
= tipc_recvmsg
,
2570 .mmap
= sock_no_mmap
,
2571 .sendpage
= sock_no_sendpage
2574 static const struct proto_ops stream_ops
= {
2575 .owner
= THIS_MODULE
,
2577 .release
= tipc_release
,
2579 .connect
= tipc_connect
,
2580 .socketpair
= sock_no_socketpair
,
2581 .accept
= tipc_accept
,
2582 .getname
= tipc_getname
,
2584 .ioctl
= tipc_ioctl
,
2585 .listen
= tipc_listen
,
2586 .shutdown
= tipc_shutdown
,
2587 .setsockopt
= tipc_setsockopt
,
2588 .getsockopt
= tipc_getsockopt
,
2589 .sendmsg
= tipc_send_stream
,
2590 .recvmsg
= tipc_recv_stream
,
2591 .mmap
= sock_no_mmap
,
2592 .sendpage
= sock_no_sendpage
2595 static const struct net_proto_family tipc_family_ops
= {
2596 .owner
= THIS_MODULE
,
2598 .create
= tipc_sk_create
2601 static struct proto tipc_proto
= {
2603 .owner
= THIS_MODULE
,
2604 .obj_size
= sizeof(struct tipc_sock
),
2605 .sysctl_rmem
= sysctl_tipc_rmem
2608 static struct proto tipc_proto_kern
= {
2610 .obj_size
= sizeof(struct tipc_sock
),
2611 .sysctl_rmem
= sysctl_tipc_rmem
2615 * tipc_socket_init - initialize TIPC socket interface
2617 * Returns 0 on success, errno otherwise
2619 int tipc_socket_init(void)
2623 res
= proto_register(&tipc_proto
, 1);
2625 pr_err("Failed to register TIPC protocol type\n");
2629 res
= sock_register(&tipc_family_ops
);
2631 pr_err("Failed to register TIPC socket type\n");
2632 proto_unregister(&tipc_proto
);
2640 * tipc_socket_stop - stop TIPC socket interface
2642 void tipc_socket_stop(void)
2644 sock_unregister(tipc_family_ops
.family
);
2645 proto_unregister(&tipc_proto
);
2648 /* Caller should hold socket lock for the passed tipc socket. */
2649 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
2653 struct nlattr
*nest
;
2655 peer_node
= tsk_peer_node(tsk
);
2656 peer_port
= tsk_peer_port(tsk
);
2658 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
2660 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
2662 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
2665 if (tsk
->conn_type
!= 0) {
2666 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
2668 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
2670 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
2673 nla_nest_end(skb
, nest
);
2678 nla_nest_cancel(skb
, nest
);
2683 /* Caller should hold socket lock for the passed tipc socket. */
2684 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
2685 struct tipc_sock
*tsk
)
2689 struct nlattr
*attrs
;
2691 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2692 &tipc_genl_v2_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
2696 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
2698 goto genlmsg_cancel
;
2699 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
2700 goto attr_msg_cancel
;
2701 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tipc_own_addr
))
2702 goto attr_msg_cancel
;
2704 if (tsk
->connected
) {
2705 err
= __tipc_nl_add_sk_con(skb
, tsk
);
2707 goto attr_msg_cancel
;
2708 } else if (!list_empty(&tsk
->publications
)) {
2709 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
2710 goto attr_msg_cancel
;
2712 nla_nest_end(skb
, attrs
);
2713 genlmsg_end(skb
, hdr
);
2718 nla_nest_cancel(skb
, attrs
);
2720 genlmsg_cancel(skb
, hdr
);
2725 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2728 struct tipc_sock
*tsk
;
2729 const struct bucket_table
*tbl
;
2730 struct rhash_head
*pos
;
2731 u32 prev_portid
= cb
->args
[0];
2732 u32 portid
= prev_portid
;
2736 tbl
= rht_dereference_rcu((&tipc_sk_rht
)->tbl
, &tipc_sk_rht
);
2737 for (i
= 0; i
< tbl
->size
; i
++) {
2738 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2739 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2740 portid
= tsk
->portid
;
2741 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
2742 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2746 prev_portid
= portid
;
2751 cb
->args
[0] = prev_portid
;
2756 /* Caller should hold socket lock for the passed tipc socket. */
2757 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
2758 struct netlink_callback
*cb
,
2759 struct publication
*publ
)
2762 struct nlattr
*attrs
;
2764 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2765 &tipc_genl_v2_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
2769 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
2771 goto genlmsg_cancel
;
2773 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
2774 goto attr_msg_cancel
;
2775 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
2776 goto attr_msg_cancel
;
2777 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
2778 goto attr_msg_cancel
;
2779 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
2780 goto attr_msg_cancel
;
2782 nla_nest_end(skb
, attrs
);
2783 genlmsg_end(skb
, hdr
);
2788 nla_nest_cancel(skb
, attrs
);
2790 genlmsg_cancel(skb
, hdr
);
2795 /* Caller should hold socket lock for the passed tipc socket. */
2796 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
2797 struct netlink_callback
*cb
,
2798 struct tipc_sock
*tsk
, u32
*last_publ
)
2801 struct publication
*p
;
2804 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
2805 if (p
->key
== *last_publ
)
2808 if (p
->key
!= *last_publ
) {
2809 /* We never set seq or call nl_dump_check_consistent()
2810 * this means that setting prev_seq here will cause the
2811 * consistence check to fail in the netlink callback
2812 * handler. Resulting in the last NLMSG_DONE message
2813 * having the NLM_F_DUMP_INTR flag set.
2820 p
= list_first_entry(&tsk
->publications
, struct publication
,
2824 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
2825 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
2827 *last_publ
= p
->key
;
2836 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2839 u32 tsk_portid
= cb
->args
[0];
2840 u32 last_publ
= cb
->args
[1];
2841 u32 done
= cb
->args
[2];
2842 struct tipc_sock
*tsk
;
2845 struct nlattr
**attrs
;
2846 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
2848 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
2852 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
2853 attrs
[TIPC_NLA_SOCK
],
2854 tipc_nl_sock_policy
);
2858 if (!sock
[TIPC_NLA_SOCK_REF
])
2861 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
2867 tsk
= tipc_sk_lookup(tsk_portid
);
2871 lock_sock(&tsk
->sk
);
2872 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
2875 release_sock(&tsk
->sk
);
2878 cb
->args
[0] = tsk_portid
;
2879 cb
->args
[1] = last_publ
;