2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2015, 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>
39 #include "name_table.h"
42 #include "name_distr.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 msecs_to_jiffies(3600000) /* [ms] => 1 h */
50 #define TIPC_FWD_MSG 1
51 #define TIPC_CONN_OK 0
52 #define TIPC_CONN_PROBING 1
53 #define TIPC_MAX_PORT 0xffffffff
54 #define TIPC_MIN_PORT 1
57 * struct tipc_sock - TIPC socket structure
58 * @sk: socket - interacts with 'port' and with user via the socket API
59 * @connected: non-zero if port is currently connected to a peer port
60 * @conn_type: TIPC type used when connection was established
61 * @conn_instance: TIPC instance used when connection was established
62 * @published: non-zero if port has one or more associated names
63 * @max_pkt: maximum packet size "hint" used when building messages sent by port
64 * @portid: unique port identity in TIPC socket hash table
65 * @phdr: preformatted message header used when sending messages
66 * @port_list: adjacent ports in TIPC's global list of ports
67 * @publications: list of publications for port
68 * @pub_count: total # of publications port has made during its lifetime
71 * @conn_timeout: the time we can wait for an unresponded setup request
72 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
73 * @link_cong: non-zero if owner must sleep because of link congestion
74 * @sent_unacked: # messages sent by socket, and not yet acked by peer
75 * @rcv_unacked: # messages read by user, but not yet acked back to peer
76 * @remote: 'connected' peer for dgram/rdm
77 * @node: hash table node
78 * @rcu: rcu struct for tipc_sock
89 struct list_head sock_list
;
90 struct list_head publications
;
93 unsigned long probing_intv
;
99 struct sockaddr_tipc remote
;
100 struct rhash_head node
;
104 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
105 static void tipc_data_ready(struct sock
*sk
);
106 static void tipc_write_space(struct sock
*sk
);
107 static int tipc_release(struct socket
*sock
);
108 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
);
109 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
);
110 static void tipc_sk_timeout(unsigned long data
);
111 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
112 struct tipc_name_seq
const *seq
);
113 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
114 struct tipc_name_seq
const *seq
);
115 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
);
116 static int tipc_sk_insert(struct tipc_sock
*tsk
);
117 static void tipc_sk_remove(struct tipc_sock
*tsk
);
118 static int __tipc_send_stream(struct socket
*sock
, struct msghdr
*m
,
120 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dsz
);
122 static const struct proto_ops packet_ops
;
123 static const struct proto_ops stream_ops
;
124 static const struct proto_ops msg_ops
;
125 static struct proto tipc_proto
;
127 static const struct nla_policy tipc_nl_sock_policy
[TIPC_NLA_SOCK_MAX
+ 1] = {
128 [TIPC_NLA_SOCK_UNSPEC
] = { .type
= NLA_UNSPEC
},
129 [TIPC_NLA_SOCK_ADDR
] = { .type
= NLA_U32
},
130 [TIPC_NLA_SOCK_REF
] = { .type
= NLA_U32
},
131 [TIPC_NLA_SOCK_CON
] = { .type
= NLA_NESTED
},
132 [TIPC_NLA_SOCK_HAS_PUBL
] = { .type
= NLA_FLAG
}
135 static const struct rhashtable_params tsk_rht_params
;
138 * Revised TIPC socket locking policy:
140 * Most socket operations take the standard socket lock when they start
141 * and hold it until they finish (or until they need to sleep). Acquiring
142 * this lock grants the owner exclusive access to the fields of the socket
143 * data structures, with the exception of the backlog queue. A few socket
144 * operations can be done without taking the socket lock because they only
145 * read socket information that never changes during the life of the socket.
147 * Socket operations may acquire the lock for the associated TIPC port if they
148 * need to perform an operation on the port. If any routine needs to acquire
149 * both the socket lock and the port lock it must take the socket lock first
150 * to avoid the risk of deadlock.
152 * The dispatcher handling incoming messages cannot grab the socket lock in
153 * the standard fashion, since invoked it runs at the BH level and cannot block.
154 * Instead, it checks to see if the socket lock is currently owned by someone,
155 * and either handles the message itself or adds it to the socket's backlog
156 * queue; in the latter case the queued message is processed once the process
157 * owning the socket lock releases it.
159 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
160 * the problem of a blocked socket operation preventing any other operations
161 * from occurring. However, applications must be careful if they have
162 * multiple threads trying to send (or receive) on the same socket, as these
163 * operations might interfere with each other. For example, doing a connect
164 * and a receive at the same time might allow the receive to consume the
165 * ACK message meant for the connect. While additional work could be done
166 * to try and overcome this, it doesn't seem to be worthwhile at the present.
168 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
169 * that another operation that must be performed in a non-blocking manner is
170 * not delayed for very long because the lock has already been taken.
172 * NOTE: This code assumes that certain fields of a port/socket pair are
173 * constant over its lifetime; such fields can be examined without taking
174 * the socket lock and/or port lock, and do not need to be re-read even
175 * after resuming processing after waiting. These fields include:
177 * - pointer to socket sk structure (aka tipc_sock structure)
178 * - pointer to port structure
182 static u32
tsk_own_node(struct tipc_sock
*tsk
)
184 return msg_prevnode(&tsk
->phdr
);
187 static u32
tsk_peer_node(struct tipc_sock
*tsk
)
189 return msg_destnode(&tsk
->phdr
);
192 static u32
tsk_peer_port(struct tipc_sock
*tsk
)
194 return msg_destport(&tsk
->phdr
);
197 static bool tsk_unreliable(struct tipc_sock
*tsk
)
199 return msg_src_droppable(&tsk
->phdr
) != 0;
202 static void tsk_set_unreliable(struct tipc_sock
*tsk
, bool unreliable
)
204 msg_set_src_droppable(&tsk
->phdr
, unreliable
? 1 : 0);
207 static bool tsk_unreturnable(struct tipc_sock
*tsk
)
209 return msg_dest_droppable(&tsk
->phdr
) != 0;
212 static void tsk_set_unreturnable(struct tipc_sock
*tsk
, bool unreturnable
)
214 msg_set_dest_droppable(&tsk
->phdr
, unreturnable
? 1 : 0);
217 static int tsk_importance(struct tipc_sock
*tsk
)
219 return msg_importance(&tsk
->phdr
);
222 static int tsk_set_importance(struct tipc_sock
*tsk
, int imp
)
224 if (imp
> TIPC_CRITICAL_IMPORTANCE
)
226 msg_set_importance(&tsk
->phdr
, (u32
)imp
);
230 static struct tipc_sock
*tipc_sk(const struct sock
*sk
)
232 return container_of(sk
, struct tipc_sock
, sk
);
235 static int tsk_conn_cong(struct tipc_sock
*tsk
)
237 return tsk
->sent_unacked
>= TIPC_FLOWCTRL_WIN
;
241 * tsk_advance_rx_queue - discard first buffer in socket receive queue
243 * Caller must hold socket lock
245 static void tsk_advance_rx_queue(struct sock
*sk
)
247 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
251 * tsk_rej_rx_queue - reject all buffers in socket receive queue
253 * Caller must hold socket lock
255 static void tsk_rej_rx_queue(struct sock
*sk
)
259 u32 own_node
= tsk_own_node(tipc_sk(sk
));
261 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
))) {
262 if (tipc_msg_reverse(own_node
, skb
, &dnode
, TIPC_ERR_NO_PORT
))
263 tipc_link_xmit_skb(sock_net(sk
), skb
, dnode
, 0);
267 /* tsk_peer_msg - verify if message was sent by connected port's peer
269 * Handles cases where the node's network address has changed from
270 * the default of <0.0.0> to its configured setting.
272 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
274 struct tipc_net
*tn
= net_generic(sock_net(&tsk
->sk
), tipc_net_id
);
275 u32 peer_port
= tsk_peer_port(tsk
);
279 if (unlikely(!tsk
->connected
))
282 if (unlikely(msg_origport(msg
) != peer_port
))
285 orig_node
= msg_orignode(msg
);
286 peer_node
= tsk_peer_node(tsk
);
288 if (likely(orig_node
== peer_node
))
291 if (!orig_node
&& (peer_node
== tn
->own_addr
))
294 if (!peer_node
&& (orig_node
== tn
->own_addr
))
301 * tipc_sk_create - create a TIPC socket
302 * @net: network namespace (must be default network)
303 * @sock: pre-allocated socket structure
304 * @protocol: protocol indicator (must be 0)
305 * @kern: caused by kernel or by userspace?
307 * This routine creates additional data structures used by the TIPC socket,
308 * initializes them, and links them together.
310 * Returns 0 on success, errno otherwise
312 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
313 int protocol
, int kern
)
316 const struct proto_ops
*ops
;
319 struct tipc_sock
*tsk
;
320 struct tipc_msg
*msg
;
322 /* Validate arguments */
323 if (unlikely(protocol
!= 0))
324 return -EPROTONOSUPPORT
;
326 switch (sock
->type
) {
329 state
= SS_UNCONNECTED
;
333 state
= SS_UNCONNECTED
;
344 /* Allocate socket's protocol area */
345 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
);
350 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
351 INIT_LIST_HEAD(&tsk
->publications
);
353 tn
= net_generic(sock_net(sk
), tipc_net_id
);
354 tipc_msg_init(tn
->own_addr
, msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
357 /* Finish initializing socket data structures */
360 sock_init_data(sock
, sk
);
361 if (tipc_sk_insert(tsk
)) {
362 pr_warn("Socket create failed; port numbrer exhausted\n");
365 msg_set_origport(msg
, tsk
->portid
);
366 setup_timer(&sk
->sk_timer
, tipc_sk_timeout
, (unsigned long)tsk
);
367 sk
->sk_backlog_rcv
= tipc_backlog_rcv
;
368 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
369 sk
->sk_data_ready
= tipc_data_ready
;
370 sk
->sk_write_space
= tipc_write_space
;
371 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
372 tsk
->sent_unacked
= 0;
373 atomic_set(&tsk
->dupl_rcvcnt
, 0);
375 if (sock
->state
== SS_READY
) {
376 tsk_set_unreturnable(tsk
, true);
377 if (sock
->type
== SOCK_DGRAM
)
378 tsk_set_unreliable(tsk
, true);
383 static void tipc_sk_callback(struct rcu_head
*head
)
385 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
391 * tipc_release - destroy a TIPC socket
392 * @sock: socket to destroy
394 * This routine cleans up any messages that are still queued on the socket.
395 * For DGRAM and RDM socket types, all queued messages are rejected.
396 * For SEQPACKET and STREAM socket types, the first message is rejected
397 * and any others are discarded. (If the first message on a STREAM socket
398 * is partially-read, it is discarded and the next one is rejected instead.)
400 * NOTE: Rejected messages are not necessarily returned to the sender! They
401 * are returned or discarded according to the "destination droppable" setting
402 * specified for the message by the sender.
404 * Returns 0 on success, errno otherwise
406 static int tipc_release(struct socket
*sock
)
408 struct sock
*sk
= sock
->sk
;
410 struct tipc_sock
*tsk
;
412 u32 dnode
, probing_state
;
415 * Exit if socket isn't fully initialized (occurs when a failed accept()
416 * releases a pre-allocated child socket that was never used)
426 * Reject all unreceived messages, except on an active connection
427 * (which disconnects locally & sends a 'FIN+' to peer)
429 dnode
= tsk_peer_node(tsk
);
430 while (sock
->state
!= SS_DISCONNECTING
) {
431 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
434 if (TIPC_SKB_CB(skb
)->handle
!= NULL
)
437 if ((sock
->state
== SS_CONNECTING
) ||
438 (sock
->state
== SS_CONNECTED
)) {
439 sock
->state
= SS_DISCONNECTING
;
441 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
443 if (tipc_msg_reverse(tsk_own_node(tsk
), skb
, &dnode
,
445 tipc_link_xmit_skb(net
, skb
, dnode
, 0);
449 tipc_sk_withdraw(tsk
, 0, NULL
);
450 probing_state
= tsk
->probing_state
;
451 if (del_timer_sync(&sk
->sk_timer
) &&
452 probing_state
!= TIPC_CONN_PROBING
)
455 if (tsk
->connected
) {
456 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
457 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0, dnode
,
458 tsk_own_node(tsk
), tsk_peer_port(tsk
),
459 tsk
->portid
, TIPC_ERR_NO_PORT
);
461 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
462 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
465 /* Discard any remaining (connection-based) messages in receive queue */
466 __skb_queue_purge(&sk
->sk_receive_queue
);
468 /* Reject any messages that accumulated in backlog queue */
469 sock
->state
= SS_DISCONNECTING
;
472 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
479 * tipc_bind - associate or disassocate TIPC name(s) with a socket
480 * @sock: socket structure
481 * @uaddr: socket address describing name(s) and desired operation
482 * @uaddr_len: size of socket address data structure
484 * Name and name sequence binding is indicated using a positive scope value;
485 * a negative scope value unbinds the specified name. Specifying no name
486 * (i.e. a socket address length of 0) unbinds all names from the socket.
488 * Returns 0 on success, errno otherwise
490 * NOTE: This routine doesn't need to take the socket lock since it doesn't
491 * access any non-constant socket information.
493 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
496 struct sock
*sk
= sock
->sk
;
497 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
498 struct tipc_sock
*tsk
= tipc_sk(sk
);
502 if (unlikely(!uaddr_len
)) {
503 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
507 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
511 if (addr
->family
!= AF_TIPC
) {
516 if (addr
->addrtype
== TIPC_ADDR_NAME
)
517 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
518 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
523 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
524 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
525 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
530 res
= (addr
->scope
> 0) ?
531 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
532 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
539 * tipc_getname - get port ID of socket or peer socket
540 * @sock: socket structure
541 * @uaddr: area for returned socket address
542 * @uaddr_len: area for returned length of socket address
543 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
545 * Returns 0 on success, errno otherwise
547 * NOTE: This routine doesn't need to take the socket lock since it only
548 * accesses socket information that is unchanging (or which changes in
549 * a completely predictable manner).
551 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
552 int *uaddr_len
, int peer
)
554 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
555 struct tipc_sock
*tsk
= tipc_sk(sock
->sk
);
556 struct tipc_net
*tn
= net_generic(sock_net(sock
->sk
), tipc_net_id
);
558 memset(addr
, 0, sizeof(*addr
));
560 if ((sock
->state
!= SS_CONNECTED
) &&
561 ((peer
!= 2) || (sock
->state
!= SS_DISCONNECTING
)))
563 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
564 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
566 addr
->addr
.id
.ref
= tsk
->portid
;
567 addr
->addr
.id
.node
= tn
->own_addr
;
570 *uaddr_len
= sizeof(*addr
);
571 addr
->addrtype
= TIPC_ADDR_ID
;
572 addr
->family
= AF_TIPC
;
574 addr
->addr
.name
.domain
= 0;
580 * tipc_poll - read and possibly block on pollmask
581 * @file: file structure associated with the socket
582 * @sock: socket for which to calculate the poll bits
585 * Returns pollmask value
588 * It appears that the usual socket locking mechanisms are not useful here
589 * since the pollmask info is potentially out-of-date the moment this routine
590 * exits. TCP and other protocols seem to rely on higher level poll routines
591 * to handle any preventable race conditions, so TIPC will do the same ...
593 * TIPC sets the returned events as follows:
595 * socket state flags set
596 * ------------ ---------
597 * unconnected no read flags
598 * POLLOUT if port is not congested
600 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
603 * connected POLLIN/POLLRDNORM if data in rx queue
604 * POLLOUT if port is not congested
606 * disconnecting POLLIN/POLLRDNORM/POLLHUP
609 * listening POLLIN if SYN in rx queue
612 * ready POLLIN/POLLRDNORM if data in rx queue
613 * [connectionless] POLLOUT (since port cannot be congested)
615 * IMPORTANT: The fact that a read or write operation is indicated does NOT
616 * imply that the operation will succeed, merely that it should be performed
617 * and will not block.
619 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
622 struct sock
*sk
= sock
->sk
;
623 struct tipc_sock
*tsk
= tipc_sk(sk
);
626 sock_poll_wait(file
, sk_sleep(sk
), wait
);
628 switch ((int)sock
->state
) {
635 if (!tsk
->link_cong
&& !tsk_conn_cong(tsk
))
640 if (!skb_queue_empty(&sk
->sk_receive_queue
))
641 mask
|= (POLLIN
| POLLRDNORM
);
643 case SS_DISCONNECTING
:
644 mask
= (POLLIN
| POLLRDNORM
| POLLHUP
);
652 * tipc_sendmcast - send multicast message
653 * @sock: socket structure
654 * @seq: destination address
655 * @msg: message to send
656 * @dsz: total length of message data
657 * @timeo: timeout to wait for wakeup
659 * Called from function tipc_sendmsg(), which has done all sanity checks
660 * Returns the number of bytes sent on success, or errno
662 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
663 struct msghdr
*msg
, size_t dsz
, long timeo
)
665 struct sock
*sk
= sock
->sk
;
666 struct tipc_sock
*tsk
= tipc_sk(sk
);
667 struct net
*net
= sock_net(sk
);
668 struct tipc_msg
*mhdr
= &tsk
->phdr
;
669 struct sk_buff_head
*pktchain
= &sk
->sk_write_queue
;
670 struct iov_iter save
= msg
->msg_iter
;
674 msg_set_type(mhdr
, TIPC_MCAST_MSG
);
675 msg_set_lookup_scope(mhdr
, TIPC_CLUSTER_SCOPE
);
676 msg_set_destport(mhdr
, 0);
677 msg_set_destnode(mhdr
, 0);
678 msg_set_nametype(mhdr
, seq
->type
);
679 msg_set_namelower(mhdr
, seq
->lower
);
680 msg_set_nameupper(mhdr
, seq
->upper
);
681 msg_set_hdr_sz(mhdr
, MCAST_H_SIZE
);
684 mtu
= tipc_bclink_get_mtu();
685 rc
= tipc_msg_build(mhdr
, msg
, 0, dsz
, mtu
, pktchain
);
686 if (unlikely(rc
< 0))
690 rc
= tipc_bclink_xmit(net
, pktchain
);
691 if (likely(rc
>= 0)) {
695 if (rc
== -EMSGSIZE
) {
696 msg
->msg_iter
= save
;
699 if (rc
!= -ELINKCONG
)
701 tipc_sk(sk
)->link_cong
= 1;
702 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
704 __skb_queue_purge(pktchain
);
710 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
711 * @arrvq: queue with arriving messages, to be cloned after destination lookup
712 * @inputq: queue with cloned messages, delivered to socket after dest lookup
714 * Multi-threaded: parallel calls with reference to same queues may occur
716 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
717 struct sk_buff_head
*inputq
)
719 struct tipc_msg
*msg
;
720 struct tipc_plist dports
;
722 u32 scope
= TIPC_CLUSTER_SCOPE
;
723 struct sk_buff_head tmpq
;
725 struct sk_buff
*skb
, *_skb
;
727 __skb_queue_head_init(&tmpq
);
728 tipc_plist_init(&dports
);
730 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
731 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
733 hsz
= skb_headroom(skb
) + msg_hdr_sz(msg
);
735 if (in_own_node(net
, msg_orignode(msg
)))
736 scope
= TIPC_NODE_SCOPE
;
738 /* Create destination port list and message clones: */
739 tipc_nametbl_mc_translate(net
,
740 msg_nametype(msg
), msg_namelower(msg
),
741 msg_nameupper(msg
), scope
, &dports
);
742 portid
= tipc_plist_pop(&dports
);
743 for (; portid
; portid
= tipc_plist_pop(&dports
)) {
744 _skb
= __pskb_copy(skb
, hsz
, GFP_ATOMIC
);
746 msg_set_destport(buf_msg(_skb
), portid
);
747 __skb_queue_tail(&tmpq
, _skb
);
750 pr_warn("Failed to clone mcast rcv buffer\n");
752 /* Append to inputq if not already done by other thread */
753 spin_lock_bh(&inputq
->lock
);
754 if (skb_peek(arrvq
) == skb
) {
755 skb_queue_splice_tail_init(&tmpq
, inputq
);
756 kfree_skb(__skb_dequeue(arrvq
));
758 spin_unlock_bh(&inputq
->lock
);
759 __skb_queue_purge(&tmpq
);
762 tipc_sk_rcv(net
, inputq
);
766 * tipc_sk_proto_rcv - receive a connection mng protocol message
767 * @tsk: receiving socket
768 * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
770 static void tipc_sk_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
**skb
)
772 struct tipc_msg
*msg
= buf_msg(*skb
);
775 u32 own_node
= tsk_own_node(tsk
);
776 /* Ignore if connection cannot be validated: */
777 if (!tsk_peer_msg(tsk
, msg
))
780 tsk
->probing_state
= TIPC_CONN_OK
;
782 if (msg_type(msg
) == CONN_ACK
) {
783 conn_cong
= tsk_conn_cong(tsk
);
784 tsk
->sent_unacked
-= msg_msgcnt(msg
);
786 tsk
->sk
.sk_write_space(&tsk
->sk
);
787 } else if (msg_type(msg
) == CONN_PROBE
) {
788 if (tipc_msg_reverse(own_node
, *skb
, &dnode
, TIPC_OK
)) {
789 msg_set_type(msg
, CONN_PROBE_REPLY
);
793 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
799 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
)
801 struct sock
*sk
= sock
->sk
;
802 struct tipc_sock
*tsk
= tipc_sk(sk
);
807 int err
= sock_error(sk
);
810 if (sock
->state
== SS_DISCONNECTING
)
814 if (signal_pending(current
))
815 return sock_intr_errno(*timeo_p
);
817 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
818 done
= sk_wait_event(sk
, timeo_p
, !tsk
->link_cong
);
819 finish_wait(sk_sleep(sk
), &wait
);
825 * tipc_sendmsg - send message in connectionless manner
826 * @sock: socket structure
827 * @m: message to send
828 * @dsz: amount of user data to be sent
830 * Message must have an destination specified explicitly.
831 * Used for SOCK_RDM and SOCK_DGRAM messages,
832 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
833 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
835 * Returns the number of bytes sent on success, or errno otherwise
837 static int tipc_sendmsg(struct socket
*sock
,
838 struct msghdr
*m
, size_t dsz
)
840 struct sock
*sk
= sock
->sk
;
844 ret
= __tipc_sendmsg(sock
, m
, dsz
);
850 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
852 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
853 struct sock
*sk
= sock
->sk
;
854 struct tipc_sock
*tsk
= tipc_sk(sk
);
855 struct net
*net
= sock_net(sk
);
856 struct tipc_msg
*mhdr
= &tsk
->phdr
;
858 struct sk_buff_head
*pktchain
= &sk
->sk_write_queue
;
860 struct tipc_name_seq
*seq
;
861 struct iov_iter save
;
866 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
868 if (unlikely(!dest
)) {
869 if (tsk
->connected
&& sock
->state
== SS_READY
)
872 return -EDESTADDRREQ
;
873 } else if (unlikely(m
->msg_namelen
< sizeof(*dest
)) ||
874 dest
->family
!= AF_TIPC
) {
877 if (unlikely(sock
->state
!= SS_READY
)) {
878 if (sock
->state
== SS_LISTENING
)
880 if (sock
->state
!= SS_UNCONNECTED
)
884 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
885 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
886 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
889 seq
= &dest
->addr
.nameseq
;
890 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
892 if (dest
->addrtype
== TIPC_ADDR_MCAST
) {
893 return tipc_sendmcast(sock
, seq
, m
, dsz
, timeo
);
894 } else if (dest
->addrtype
== TIPC_ADDR_NAME
) {
895 u32 type
= dest
->addr
.name
.name
.type
;
896 u32 inst
= dest
->addr
.name
.name
.instance
;
897 u32 domain
= dest
->addr
.name
.domain
;
900 msg_set_type(mhdr
, TIPC_NAMED_MSG
);
901 msg_set_hdr_sz(mhdr
, NAMED_H_SIZE
);
902 msg_set_nametype(mhdr
, type
);
903 msg_set_nameinst(mhdr
, inst
);
904 msg_set_lookup_scope(mhdr
, tipc_addr_scope(domain
));
905 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
906 msg_set_destnode(mhdr
, dnode
);
907 msg_set_destport(mhdr
, dport
);
908 if (unlikely(!dport
&& !dnode
))
909 return -EHOSTUNREACH
;
910 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
911 dnode
= dest
->addr
.id
.node
;
912 msg_set_type(mhdr
, TIPC_DIRECT_MSG
);
913 msg_set_lookup_scope(mhdr
, 0);
914 msg_set_destnode(mhdr
, dnode
);
915 msg_set_destport(mhdr
, dest
->addr
.id
.ref
);
916 msg_set_hdr_sz(mhdr
, BASIC_H_SIZE
);
921 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
922 rc
= tipc_msg_build(mhdr
, m
, 0, dsz
, mtu
, pktchain
);
927 skb
= skb_peek(pktchain
);
928 TIPC_SKB_CB(skb
)->wakeup_pending
= tsk
->link_cong
;
929 rc
= tipc_link_xmit(net
, pktchain
, dnode
, tsk
->portid
);
930 if (likely(rc
>= 0)) {
931 if (sock
->state
!= SS_READY
)
932 sock
->state
= SS_CONNECTING
;
936 if (rc
== -EMSGSIZE
) {
940 if (rc
!= -ELINKCONG
)
943 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
945 __skb_queue_purge(pktchain
);
951 static int tipc_wait_for_sndpkt(struct socket
*sock
, long *timeo_p
)
953 struct sock
*sk
= sock
->sk
;
954 struct tipc_sock
*tsk
= tipc_sk(sk
);
959 int err
= sock_error(sk
);
962 if (sock
->state
== SS_DISCONNECTING
)
964 else if (sock
->state
!= SS_CONNECTED
)
968 if (signal_pending(current
))
969 return sock_intr_errno(*timeo_p
);
971 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
972 done
= sk_wait_event(sk
, timeo_p
,
974 !tsk_conn_cong(tsk
)) ||
976 finish_wait(sk_sleep(sk
), &wait
);
982 * tipc_send_stream - send stream-oriented data
983 * @sock: socket structure
985 * @dsz: total length of data to be transmitted
987 * Used for SOCK_STREAM data.
989 * Returns the number of bytes sent on success (or partial success),
990 * or errno if no data sent
992 static int tipc_send_stream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
994 struct sock
*sk
= sock
->sk
;
998 ret
= __tipc_send_stream(sock
, m
, dsz
);
1004 static int __tipc_send_stream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1006 struct sock
*sk
= sock
->sk
;
1007 struct net
*net
= sock_net(sk
);
1008 struct tipc_sock
*tsk
= tipc_sk(sk
);
1009 struct tipc_msg
*mhdr
= &tsk
->phdr
;
1010 struct sk_buff_head
*pktchain
= &sk
->sk_write_queue
;
1011 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1012 u32 portid
= tsk
->portid
;
1016 uint mtu
, send
, sent
= 0;
1017 struct iov_iter save
;
1019 /* Handle implied connection establishment */
1020 if (unlikely(dest
)) {
1021 rc
= __tipc_sendmsg(sock
, m
, dsz
);
1022 if (dsz
&& (dsz
== rc
))
1023 tsk
->sent_unacked
= 1;
1026 if (dsz
> (uint
)INT_MAX
)
1029 if (unlikely(sock
->state
!= SS_CONNECTED
)) {
1030 if (sock
->state
== SS_DISCONNECTING
)
1036 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1037 dnode
= tsk_peer_node(tsk
);
1042 send
= min_t(uint
, dsz
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1043 rc
= tipc_msg_build(mhdr
, m
, sent
, send
, mtu
, pktchain
);
1044 if (unlikely(rc
< 0))
1047 if (likely(!tsk_conn_cong(tsk
))) {
1048 rc
= tipc_link_xmit(net
, pktchain
, dnode
, portid
);
1050 tsk
->sent_unacked
++;
1056 if (rc
== -EMSGSIZE
) {
1057 tsk
->max_pkt
= tipc_node_get_mtu(net
, dnode
,
1062 if (rc
!= -ELINKCONG
)
1066 rc
= tipc_wait_for_sndpkt(sock
, &timeo
);
1068 __skb_queue_purge(pktchain
);
1071 return sent
? sent
: rc
;
1075 * tipc_send_packet - send a connection-oriented message
1076 * @sock: socket structure
1077 * @m: message to send
1078 * @dsz: length of data to be transmitted
1080 * Used for SOCK_SEQPACKET messages.
1082 * Returns the number of bytes sent on success, or errno otherwise
1084 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1086 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1089 return tipc_send_stream(sock
, m
, dsz
);
1092 /* tipc_sk_finish_conn - complete the setup of a connection
1094 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1097 struct sock
*sk
= &tsk
->sk
;
1098 struct net
*net
= sock_net(sk
);
1099 struct tipc_msg
*msg
= &tsk
->phdr
;
1101 msg_set_destnode(msg
, peer_node
);
1102 msg_set_destport(msg
, peer_port
);
1103 msg_set_type(msg
, TIPC_CONN_MSG
);
1104 msg_set_lookup_scope(msg
, 0);
1105 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1107 tsk
->probing_intv
= CONN_PROBING_INTERVAL
;
1108 tsk
->probing_state
= TIPC_CONN_OK
;
1110 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ tsk
->probing_intv
);
1111 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1112 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1116 * set_orig_addr - capture sender's address for received message
1117 * @m: descriptor for message info
1118 * @msg: received message header
1120 * Note: Address is not captured if not requested by receiver.
1122 static void set_orig_addr(struct msghdr
*m
, struct tipc_msg
*msg
)
1124 DECLARE_SOCKADDR(struct sockaddr_tipc
*, addr
, m
->msg_name
);
1127 addr
->family
= AF_TIPC
;
1128 addr
->addrtype
= TIPC_ADDR_ID
;
1129 memset(&addr
->addr
, 0, sizeof(addr
->addr
));
1130 addr
->addr
.id
.ref
= msg_origport(msg
);
1131 addr
->addr
.id
.node
= msg_orignode(msg
);
1132 addr
->addr
.name
.domain
= 0; /* could leave uninitialized */
1133 addr
->scope
= 0; /* could leave uninitialized */
1134 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1139 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1140 * @m: descriptor for message info
1141 * @msg: received message header
1142 * @tsk: TIPC port associated with message
1144 * Note: Ancillary data is not captured if not requested by receiver.
1146 * Returns 0 if successful, otherwise errno
1148 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1149 struct tipc_sock
*tsk
)
1157 if (likely(m
->msg_controllen
== 0))
1160 /* Optionally capture errored message object(s) */
1161 err
= msg
? msg_errcode(msg
) : 0;
1162 if (unlikely(err
)) {
1164 anc_data
[1] = msg_data_sz(msg
);
1165 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1169 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1176 /* Optionally capture message destination object */
1177 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1178 switch (dest_type
) {
1179 case TIPC_NAMED_MSG
:
1181 anc_data
[0] = msg_nametype(msg
);
1182 anc_data
[1] = msg_namelower(msg
);
1183 anc_data
[2] = msg_namelower(msg
);
1185 case TIPC_MCAST_MSG
:
1187 anc_data
[0] = msg_nametype(msg
);
1188 anc_data
[1] = msg_namelower(msg
);
1189 anc_data
[2] = msg_nameupper(msg
);
1192 has_name
= (tsk
->conn_type
!= 0);
1193 anc_data
[0] = tsk
->conn_type
;
1194 anc_data
[1] = tsk
->conn_instance
;
1195 anc_data
[2] = tsk
->conn_instance
;
1201 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1209 static void tipc_sk_send_ack(struct tipc_sock
*tsk
, uint ack
)
1211 struct net
*net
= sock_net(&tsk
->sk
);
1212 struct sk_buff
*skb
= NULL
;
1213 struct tipc_msg
*msg
;
1214 u32 peer_port
= tsk_peer_port(tsk
);
1215 u32 dnode
= tsk_peer_node(tsk
);
1217 if (!tsk
->connected
)
1219 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1220 dnode
, tsk_own_node(tsk
), peer_port
,
1221 tsk
->portid
, TIPC_OK
);
1225 msg_set_msgcnt(msg
, ack
);
1226 tipc_link_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1229 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1231 struct sock
*sk
= sock
->sk
;
1233 long timeo
= *timeop
;
1237 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1238 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1239 if (sock
->state
== SS_DISCONNECTING
) {
1244 timeo
= schedule_timeout(timeo
);
1248 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1253 err
= sock_intr_errno(timeo
);
1254 if (signal_pending(current
))
1257 finish_wait(sk_sleep(sk
), &wait
);
1263 * tipc_recvmsg - receive packet-oriented message
1264 * @m: descriptor for message info
1265 * @buf_len: total size of user buffer area
1266 * @flags: receive flags
1268 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1269 * If the complete message doesn't fit in user area, truncate it.
1271 * Returns size of returned message data, errno otherwise
1273 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
, size_t buf_len
,
1276 struct sock
*sk
= sock
->sk
;
1277 struct tipc_sock
*tsk
= tipc_sk(sk
);
1278 struct sk_buff
*buf
;
1279 struct tipc_msg
*msg
;
1285 /* Catch invalid receive requests */
1286 if (unlikely(!buf_len
))
1291 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1296 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1299 /* Look for a message in receive queue; wait if necessary */
1300 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1304 /* Look at first message in receive queue */
1305 buf
= skb_peek(&sk
->sk_receive_queue
);
1307 sz
= msg_data_sz(msg
);
1308 err
= msg_errcode(msg
);
1310 /* Discard an empty non-errored message & try again */
1311 if ((!sz
) && (!err
)) {
1312 tsk_advance_rx_queue(sk
);
1316 /* Capture sender's address (optional) */
1317 set_orig_addr(m
, msg
);
1319 /* Capture ancillary data (optional) */
1320 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1324 /* Capture message data (if valid) & compute return value (always) */
1326 if (unlikely(buf_len
< sz
)) {
1328 m
->msg_flags
|= MSG_TRUNC
;
1330 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
), m
, sz
);
1335 if ((sock
->state
== SS_READY
) ||
1336 ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
))
1342 /* Consume received message (optional) */
1343 if (likely(!(flags
& MSG_PEEK
))) {
1344 if ((sock
->state
!= SS_READY
) &&
1345 (++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1346 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1347 tsk
->rcv_unacked
= 0;
1349 tsk_advance_rx_queue(sk
);
1357 * tipc_recv_stream - receive stream-oriented data
1358 * @m: descriptor for message info
1359 * @buf_len: total size of user buffer area
1360 * @flags: receive flags
1362 * Used for SOCK_STREAM messages only. If not enough data is available
1363 * will optionally wait for more; never truncates data.
1365 * Returns size of returned message data, errno otherwise
1367 static int tipc_recv_stream(struct socket
*sock
, struct msghdr
*m
,
1368 size_t buf_len
, int flags
)
1370 struct sock
*sk
= sock
->sk
;
1371 struct tipc_sock
*tsk
= tipc_sk(sk
);
1372 struct sk_buff
*buf
;
1373 struct tipc_msg
*msg
;
1376 int sz_to_copy
, target
, needed
;
1381 /* Catch invalid receive attempts */
1382 if (unlikely(!buf_len
))
1387 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1392 target
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buf_len
);
1393 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1396 /* Look for a message in receive queue; wait if necessary */
1397 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1401 /* Look at first message in receive queue */
1402 buf
= skb_peek(&sk
->sk_receive_queue
);
1404 sz
= msg_data_sz(msg
);
1405 err
= msg_errcode(msg
);
1407 /* Discard an empty non-errored message & try again */
1408 if ((!sz
) && (!err
)) {
1409 tsk_advance_rx_queue(sk
);
1413 /* Optionally capture sender's address & ancillary data of first msg */
1414 if (sz_copied
== 0) {
1415 set_orig_addr(m
, msg
);
1416 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1421 /* Capture message data (if valid) & compute return value (always) */
1423 u32 offset
= (u32
)(unsigned long)(TIPC_SKB_CB(buf
)->handle
);
1426 needed
= (buf_len
- sz_copied
);
1427 sz_to_copy
= (sz
<= needed
) ? sz
: needed
;
1429 res
= skb_copy_datagram_msg(buf
, msg_hdr_sz(msg
) + offset
,
1434 sz_copied
+= sz_to_copy
;
1436 if (sz_to_copy
< sz
) {
1437 if (!(flags
& MSG_PEEK
))
1438 TIPC_SKB_CB(buf
)->handle
=
1439 (void *)(unsigned long)(offset
+ sz_to_copy
);
1444 goto exit
; /* can't add error msg to valid data */
1446 if ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
)
1452 /* Consume received message (optional) */
1453 if (likely(!(flags
& MSG_PEEK
))) {
1454 if (unlikely(++tsk
->rcv_unacked
>= TIPC_CONNACK_INTV
)) {
1455 tipc_sk_send_ack(tsk
, tsk
->rcv_unacked
);
1456 tsk
->rcv_unacked
= 0;
1458 tsk_advance_rx_queue(sk
);
1461 /* Loop around if more data is required */
1462 if ((sz_copied
< buf_len
) && /* didn't get all requested data */
1463 (!skb_queue_empty(&sk
->sk_receive_queue
) ||
1464 (sz_copied
< target
)) && /* and more is ready or required */
1465 (!(flags
& MSG_PEEK
)) && /* and aren't just peeking at data */
1466 (!err
)) /* and haven't reached a FIN */
1471 return sz_copied
? sz_copied
: res
;
1475 * tipc_write_space - wake up thread if port congestion is released
1478 static void tipc_write_space(struct sock
*sk
)
1480 struct socket_wq
*wq
;
1483 wq
= rcu_dereference(sk
->sk_wq
);
1484 if (wq_has_sleeper(wq
))
1485 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1486 POLLWRNORM
| POLLWRBAND
);
1491 * tipc_data_ready - wake up threads to indicate messages have been received
1493 * @len: the length of messages
1495 static void tipc_data_ready(struct sock
*sk
)
1497 struct socket_wq
*wq
;
1500 wq
= rcu_dereference(sk
->sk_wq
);
1501 if (wq_has_sleeper(wq
))
1502 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1503 POLLRDNORM
| POLLRDBAND
);
1508 * filter_connect - Handle all incoming messages for a connection-based socket
1510 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1512 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1514 static int filter_connect(struct tipc_sock
*tsk
, struct sk_buff
**skb
)
1516 struct sock
*sk
= &tsk
->sk
;
1517 struct net
*net
= sock_net(sk
);
1518 struct socket
*sock
= sk
->sk_socket
;
1519 struct tipc_msg
*msg
= buf_msg(*skb
);
1520 int retval
= -TIPC_ERR_NO_PORT
;
1525 switch ((int)sock
->state
) {
1527 /* Accept only connection-based messages sent by peer */
1528 if (tsk_peer_msg(tsk
, msg
)) {
1529 if (unlikely(msg_errcode(msg
))) {
1530 sock
->state
= SS_DISCONNECTING
;
1532 /* let timer expire on it's own */
1533 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
1540 /* Accept only ACK or NACK message */
1542 if (unlikely(!msg_connected(msg
)))
1545 if (unlikely(msg_errcode(msg
))) {
1546 sock
->state
= SS_DISCONNECTING
;
1547 sk
->sk_err
= ECONNREFUSED
;
1552 if (unlikely(msg_importance(msg
) > TIPC_CRITICAL_IMPORTANCE
)) {
1553 sock
->state
= SS_DISCONNECTING
;
1554 sk
->sk_err
= EINVAL
;
1559 tipc_sk_finish_conn(tsk
, msg_origport(msg
), msg_orignode(msg
));
1560 msg_set_importance(&tsk
->phdr
, msg_importance(msg
));
1561 sock
->state
= SS_CONNECTED
;
1563 /* If an incoming message is an 'ACK-', it should be
1564 * discarded here because it doesn't contain useful
1565 * data. In addition, we should try to wake up
1566 * connect() routine if sleeping.
1568 if (msg_data_sz(msg
) == 0) {
1571 if (waitqueue_active(sk_sleep(sk
)))
1572 wake_up_interruptible(sk_sleep(sk
));
1577 case SS_UNCONNECTED
:
1578 /* Accept only SYN message */
1579 if (!msg_connected(msg
) && !(msg_errcode(msg
)))
1582 case SS_DISCONNECTING
:
1585 pr_err("Unknown socket state %u\n", sock
->state
);
1591 * rcvbuf_limit - get proper overload limit of socket receive queue
1595 * For all connection oriented messages, irrespective of importance,
1596 * the default overload value (i.e. 67MB) is set as limit.
1598 * For all connectionless messages, by default new queue limits are
1601 * TIPC_LOW_IMPORTANCE (4 MB)
1602 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1603 * TIPC_HIGH_IMPORTANCE (16 MB)
1604 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1606 * Returns overload limit according to corresponding message importance
1608 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*buf
)
1610 struct tipc_msg
*msg
= buf_msg(buf
);
1612 if (msg_connected(msg
))
1613 return sysctl_tipc_rmem
[2];
1615 return sk
->sk_rcvbuf
>> TIPC_CRITICAL_IMPORTANCE
<<
1616 msg_importance(msg
);
1620 * filter_rcv - validate incoming message
1622 * @skb: pointer to message. Set to NULL if buffer is consumed.
1624 * Enqueues message on receive queue if acceptable; optionally handles
1625 * disconnect indication for a connected socket.
1627 * Called with socket lock already taken
1629 * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1631 static int filter_rcv(struct sock
*sk
, struct sk_buff
**skb
)
1633 struct socket
*sock
= sk
->sk_socket
;
1634 struct tipc_sock
*tsk
= tipc_sk(sk
);
1635 struct tipc_msg
*msg
= buf_msg(*skb
);
1636 unsigned int limit
= rcvbuf_limit(sk
, *skb
);
1639 if (unlikely(msg_user(msg
) == CONN_MANAGER
)) {
1640 tipc_sk_proto_rcv(tsk
, skb
);
1644 if (unlikely(msg_user(msg
) == SOCK_WAKEUP
)) {
1647 sk
->sk_write_space(sk
);
1652 /* Reject message if it is wrong sort of message for socket */
1653 if (msg_type(msg
) > TIPC_DIRECT_MSG
)
1654 return -TIPC_ERR_NO_PORT
;
1656 if (sock
->state
== SS_READY
) {
1657 if (msg_connected(msg
))
1658 return -TIPC_ERR_NO_PORT
;
1660 rc
= filter_connect(tsk
, skb
);
1661 if (rc
!= TIPC_OK
|| !*skb
)
1665 /* Reject message if there isn't room to queue it */
1666 if (sk_rmem_alloc_get(sk
) + (*skb
)->truesize
>= limit
)
1667 return -TIPC_ERR_OVERLOAD
;
1669 /* Enqueue message */
1670 TIPC_SKB_CB(*skb
)->handle
= NULL
;
1671 __skb_queue_tail(&sk
->sk_receive_queue
, *skb
);
1672 skb_set_owner_r(*skb
, sk
);
1674 sk
->sk_data_ready(sk
);
1680 * tipc_backlog_rcv - handle incoming message from backlog queue
1684 * Caller must hold socket lock
1688 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1693 struct tipc_sock
*tsk
= tipc_sk(sk
);
1694 struct net
*net
= sock_net(sk
);
1695 uint truesize
= skb
->truesize
;
1697 err
= filter_rcv(sk
, &skb
);
1699 dcnt
= &tsk
->dupl_rcvcnt
;
1700 if (atomic_read(dcnt
) < TIPC_CONN_OVERLOAD_LIMIT
)
1701 atomic_add(truesize
, dcnt
);
1704 if (!err
|| tipc_msg_reverse(tsk_own_node(tsk
), skb
, &dnode
, -err
))
1705 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
1710 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1711 * inputq and try adding them to socket or backlog queue
1712 * @inputq: list of incoming buffers with potentially different destinations
1713 * @sk: socket where the buffers should be enqueued
1714 * @dport: port number for the socket
1715 * @_skb: returned buffer to be forwarded or rejected, if applicable
1717 * Caller must hold socket lock
1719 * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1720 * or -TIPC_ERR_NO_PORT
1722 static int tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
1723 u32 dport
, struct sk_buff
**_skb
)
1728 struct sk_buff
*skb
;
1729 unsigned long time_limit
= jiffies
+ 2;
1731 while (skb_queue_len(inputq
)) {
1732 if (unlikely(time_after_eq(jiffies
, time_limit
)))
1734 skb
= tipc_skb_dequeue(inputq
, dport
);
1737 if (!sock_owned_by_user(sk
)) {
1738 err
= filter_rcv(sk
, &skb
);
1744 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
1745 if (sk
->sk_backlog
.len
)
1746 atomic_set(dcnt
, 0);
1747 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
1748 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
1751 return -TIPC_ERR_OVERLOAD
;
1757 * tipc_sk_rcv - handle a chain of incoming buffers
1758 * @inputq: buffer list containing the buffers
1759 * Consumes all buffers in list until inputq is empty
1760 * Note: may be called in multiple threads referring to the same queue
1761 * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1762 * Only node local calls check the return value, sending single-buffer queues
1764 int tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
1766 u32 dnode
, dport
= 0;
1767 int err
= -TIPC_ERR_NO_PORT
;
1768 struct sk_buff
*skb
;
1769 struct tipc_sock
*tsk
;
1770 struct tipc_net
*tn
;
1773 while (skb_queue_len(inputq
)) {
1775 dport
= tipc_skb_peek_port(inputq
, dport
);
1776 tsk
= tipc_sk_lookup(net
, dport
);
1779 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
1780 err
= tipc_sk_enqueue(inputq
, sk
, dport
, &skb
);
1781 spin_unlock_bh(&sk
->sk_lock
.slock
);
1786 skb
= tipc_skb_dequeue(inputq
, dport
);
1790 if (tipc_msg_lookup_dest(net
, skb
, &dnode
, &err
))
1793 dnode
= msg_destnode(buf_msg(skb
));
1796 tn
= net_generic(net
, tipc_net_id
);
1797 if (!tipc_msg_reverse(tn
->own_addr
, skb
, &dnode
, -err
))
1800 tipc_link_xmit_skb(net
, skb
, dnode
, dport
);
1802 return err
? -EHOSTUNREACH
: 0;
1805 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
1807 struct sock
*sk
= sock
->sk
;
1812 int err
= sock_error(sk
);
1817 if (signal_pending(current
))
1818 return sock_intr_errno(*timeo_p
);
1820 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1821 done
= sk_wait_event(sk
, timeo_p
, sock
->state
!= SS_CONNECTING
);
1822 finish_wait(sk_sleep(sk
), &wait
);
1828 * tipc_connect - establish a connection to another TIPC port
1829 * @sock: socket structure
1830 * @dest: socket address for destination port
1831 * @destlen: size of socket address data structure
1832 * @flags: file-related flags associated with socket
1834 * Returns 0 on success, errno otherwise
1836 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
1837 int destlen
, int flags
)
1839 struct sock
*sk
= sock
->sk
;
1840 struct tipc_sock
*tsk
= tipc_sk(sk
);
1841 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
1842 struct msghdr m
= {NULL
,};
1843 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
1844 socket_state previous
;
1849 /* DGRAM/RDM connect(), just save the destaddr */
1850 if (sock
->state
== SS_READY
) {
1851 if (dst
->family
== AF_UNSPEC
) {
1852 memset(&tsk
->remote
, 0, sizeof(struct sockaddr_tipc
));
1854 } else if (destlen
!= sizeof(struct sockaddr_tipc
)) {
1857 memcpy(&tsk
->remote
, dest
, destlen
);
1864 * Reject connection attempt using multicast address
1866 * Note: send_msg() validates the rest of the address fields,
1867 * so there's no need to do it here
1869 if (dst
->addrtype
== TIPC_ADDR_MCAST
) {
1874 previous
= sock
->state
;
1875 switch (sock
->state
) {
1876 case SS_UNCONNECTED
:
1877 /* Send a 'SYN-' to destination */
1879 m
.msg_namelen
= destlen
;
1881 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1882 * indicate send_msg() is never blocked.
1885 m
.msg_flags
= MSG_DONTWAIT
;
1887 res
= __tipc_sendmsg(sock
, &m
, 0);
1888 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
1891 /* Just entered SS_CONNECTING state; the only
1892 * difference is that return value in non-blocking
1893 * case is EINPROGRESS, rather than EALREADY.
1897 if (previous
== SS_CONNECTING
)
1901 timeout
= msecs_to_jiffies(timeout
);
1902 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1903 res
= tipc_wait_for_connect(sock
, &timeout
);
1918 * tipc_listen - allow socket to listen for incoming connections
1919 * @sock: socket structure
1922 * Returns 0 on success, errno otherwise
1924 static int tipc_listen(struct socket
*sock
, int len
)
1926 struct sock
*sk
= sock
->sk
;
1931 if (sock
->state
!= SS_UNCONNECTED
)
1934 sock
->state
= SS_LISTENING
;
1942 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
1944 struct sock
*sk
= sock
->sk
;
1948 /* True wake-one mechanism for incoming connections: only
1949 * one process gets woken up, not the 'whole herd'.
1950 * Since we do not 'race & poll' for established sockets
1951 * anymore, the common case will execute the loop only once.
1954 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
1955 TASK_INTERRUPTIBLE
);
1956 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1958 timeo
= schedule_timeout(timeo
);
1962 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1965 if (sock
->state
!= SS_LISTENING
)
1970 err
= sock_intr_errno(timeo
);
1971 if (signal_pending(current
))
1974 finish_wait(sk_sleep(sk
), &wait
);
1979 * tipc_accept - wait for connection request
1980 * @sock: listening socket
1981 * @newsock: new socket that is to be connected
1982 * @flags: file-related flags associated with socket
1984 * Returns 0 on success, errno otherwise
1986 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
)
1988 struct sock
*new_sk
, *sk
= sock
->sk
;
1989 struct sk_buff
*buf
;
1990 struct tipc_sock
*new_tsock
;
1991 struct tipc_msg
*msg
;
1997 if (sock
->state
!= SS_LISTENING
) {
2001 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2002 res
= tipc_wait_for_accept(sock
, timeo
);
2006 buf
= skb_peek(&sk
->sk_receive_queue
);
2008 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, 1);
2012 new_sk
= new_sock
->sk
;
2013 new_tsock
= tipc_sk(new_sk
);
2016 /* we lock on new_sk; but lockdep sees the lock on sk */
2017 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2020 * Reject any stray messages received by new socket
2021 * before the socket lock was taken (very, very unlikely)
2023 tsk_rej_rx_queue(new_sk
);
2025 /* Connect new socket to it's peer */
2026 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2027 new_sock
->state
= SS_CONNECTED
;
2029 tsk_set_importance(new_tsock
, msg_importance(msg
));
2030 if (msg_named(msg
)) {
2031 new_tsock
->conn_type
= msg_nametype(msg
);
2032 new_tsock
->conn_instance
= msg_nameinst(msg
);
2036 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2037 * Respond to 'SYN+' by queuing it on new socket.
2039 if (!msg_data_sz(msg
)) {
2040 struct msghdr m
= {NULL
,};
2042 tsk_advance_rx_queue(sk
);
2043 __tipc_send_stream(new_sock
, &m
, 0);
2045 __skb_dequeue(&sk
->sk_receive_queue
);
2046 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2047 skb_set_owner_r(buf
, new_sk
);
2049 release_sock(new_sk
);
2056 * tipc_shutdown - shutdown socket connection
2057 * @sock: socket structure
2058 * @how: direction to close (must be SHUT_RDWR)
2060 * Terminates connection (if necessary), then purges socket's receive queue.
2062 * Returns 0 on success, errno otherwise
2064 static int tipc_shutdown(struct socket
*sock
, int how
)
2066 struct sock
*sk
= sock
->sk
;
2067 struct net
*net
= sock_net(sk
);
2068 struct tipc_sock
*tsk
= tipc_sk(sk
);
2069 struct sk_buff
*skb
;
2073 if (how
!= SHUT_RDWR
)
2078 switch (sock
->state
) {
2083 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2084 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
2086 if (TIPC_SKB_CB(skb
)->handle
!= NULL
) {
2090 if (tipc_msg_reverse(tsk_own_node(tsk
), skb
, &dnode
,
2091 TIPC_CONN_SHUTDOWN
))
2092 tipc_link_xmit_skb(net
, skb
, dnode
,
2095 dnode
= tsk_peer_node(tsk
);
2097 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
2098 TIPC_CONN_MSG
, SHORT_H_SIZE
,
2099 0, dnode
, tsk_own_node(tsk
),
2101 tsk
->portid
, TIPC_CONN_SHUTDOWN
);
2102 tipc_link_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
2105 sock
->state
= SS_DISCONNECTING
;
2106 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
2109 case SS_DISCONNECTING
:
2111 /* Discard any unreceived messages */
2112 __skb_queue_purge(&sk
->sk_receive_queue
);
2114 /* Wake up anyone sleeping in poll */
2115 sk
->sk_state_change(sk
);
2127 static void tipc_sk_timeout(unsigned long data
)
2129 struct tipc_sock
*tsk
= (struct tipc_sock
*)data
;
2130 struct sock
*sk
= &tsk
->sk
;
2131 struct sk_buff
*skb
= NULL
;
2132 u32 peer_port
, peer_node
;
2133 u32 own_node
= tsk_own_node(tsk
);
2136 if (!tsk
->connected
) {
2140 peer_port
= tsk_peer_port(tsk
);
2141 peer_node
= tsk_peer_node(tsk
);
2143 if (tsk
->probing_state
== TIPC_CONN_PROBING
) {
2144 /* Previous probe not answered -> self abort */
2145 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
2146 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0,
2147 own_node
, peer_node
, tsk
->portid
,
2148 peer_port
, TIPC_ERR_NO_PORT
);
2150 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
,
2151 INT_H_SIZE
, 0, peer_node
, own_node
,
2152 peer_port
, tsk
->portid
, TIPC_OK
);
2153 tsk
->probing_state
= TIPC_CONN_PROBING
;
2154 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ tsk
->probing_intv
);
2158 tipc_link_xmit_skb(sock_net(sk
), skb
, peer_node
, tsk
->portid
);
2163 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2164 struct tipc_name_seq
const *seq
)
2166 struct net
*net
= sock_net(&tsk
->sk
);
2167 struct publication
*publ
;
2172 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2173 if (key
== tsk
->portid
)
2176 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2177 scope
, tsk
->portid
, key
);
2178 if (unlikely(!publ
))
2181 list_add(&publ
->pport_list
, &tsk
->publications
);
2187 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2188 struct tipc_name_seq
const *seq
)
2190 struct net
*net
= sock_net(&tsk
->sk
);
2191 struct publication
*publ
;
2192 struct publication
*safe
;
2195 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2197 if (publ
->scope
!= scope
)
2199 if (publ
->type
!= seq
->type
)
2201 if (publ
->lower
!= seq
->lower
)
2203 if (publ
->upper
!= seq
->upper
)
2205 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2206 publ
->ref
, publ
->key
);
2210 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2211 publ
->ref
, publ
->key
);
2214 if (list_empty(&tsk
->publications
))
2219 /* tipc_sk_reinit: set non-zero address in all existing sockets
2220 * when we go from standalone to network mode.
2222 void tipc_sk_reinit(struct net
*net
)
2224 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2225 const struct bucket_table
*tbl
;
2226 struct rhash_head
*pos
;
2227 struct tipc_sock
*tsk
;
2228 struct tipc_msg
*msg
;
2232 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2233 for (i
= 0; i
< tbl
->size
; i
++) {
2234 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2235 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2237 msg_set_prevnode(msg
, tn
->own_addr
);
2238 msg_set_orignode(msg
, tn
->own_addr
);
2239 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2245 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2247 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2248 struct tipc_sock
*tsk
;
2251 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2253 sock_hold(&tsk
->sk
);
2259 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2261 struct sock
*sk
= &tsk
->sk
;
2262 struct net
*net
= sock_net(sk
);
2263 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2264 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2265 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2267 while (remaining
--) {
2269 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2270 portid
= TIPC_MIN_PORT
;
2271 tsk
->portid
= portid
;
2272 sock_hold(&tsk
->sk
);
2273 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2282 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2284 struct sock
*sk
= &tsk
->sk
;
2285 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2287 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2288 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
2293 static const struct rhashtable_params tsk_rht_params
= {
2295 .head_offset
= offsetof(struct tipc_sock
, node
),
2296 .key_offset
= offsetof(struct tipc_sock
, portid
),
2297 .key_len
= sizeof(u32
), /* portid */
2298 .max_size
= 1048576,
2300 .automatic_shrinking
= true,
2303 int tipc_sk_rht_init(struct net
*net
)
2305 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2307 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2310 void tipc_sk_rht_destroy(struct net
*net
)
2312 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2314 /* Wait for socket readers to complete */
2317 rhashtable_destroy(&tn
->sk_rht
);
2321 * tipc_setsockopt - set socket option
2322 * @sock: socket structure
2323 * @lvl: option level
2324 * @opt: option identifier
2325 * @ov: pointer to new option value
2326 * @ol: length of option value
2328 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2329 * (to ease compatibility).
2331 * Returns 0 on success, errno otherwise
2333 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2334 char __user
*ov
, unsigned int ol
)
2336 struct sock
*sk
= sock
->sk
;
2337 struct tipc_sock
*tsk
= tipc_sk(sk
);
2341 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2343 if (lvl
!= SOL_TIPC
)
2344 return -ENOPROTOOPT
;
2345 if (ol
< sizeof(value
))
2347 res
= get_user(value
, (u32 __user
*)ov
);
2354 case TIPC_IMPORTANCE
:
2355 res
= tsk_set_importance(tsk
, value
);
2357 case TIPC_SRC_DROPPABLE
:
2358 if (sock
->type
!= SOCK_STREAM
)
2359 tsk_set_unreliable(tsk
, value
);
2363 case TIPC_DEST_DROPPABLE
:
2364 tsk_set_unreturnable(tsk
, value
);
2366 case TIPC_CONN_TIMEOUT
:
2367 tipc_sk(sk
)->conn_timeout
= value
;
2368 /* no need to set "res", since already 0 at this point */
2380 * tipc_getsockopt - get socket option
2381 * @sock: socket structure
2382 * @lvl: option level
2383 * @opt: option identifier
2384 * @ov: receptacle for option value
2385 * @ol: receptacle for length of option value
2387 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2388 * (to ease compatibility).
2390 * Returns 0 on success, errno otherwise
2392 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2393 char __user
*ov
, int __user
*ol
)
2395 struct sock
*sk
= sock
->sk
;
2396 struct tipc_sock
*tsk
= tipc_sk(sk
);
2401 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2402 return put_user(0, ol
);
2403 if (lvl
!= SOL_TIPC
)
2404 return -ENOPROTOOPT
;
2405 res
= get_user(len
, ol
);
2412 case TIPC_IMPORTANCE
:
2413 value
= tsk_importance(tsk
);
2415 case TIPC_SRC_DROPPABLE
:
2416 value
= tsk_unreliable(tsk
);
2418 case TIPC_DEST_DROPPABLE
:
2419 value
= tsk_unreturnable(tsk
);
2421 case TIPC_CONN_TIMEOUT
:
2422 value
= tsk
->conn_timeout
;
2423 /* no need to set "res", since already 0 at this point */
2425 case TIPC_NODE_RECVQ_DEPTH
:
2426 value
= 0; /* was tipc_queue_size, now obsolete */
2428 case TIPC_SOCK_RECVQ_DEPTH
:
2429 value
= skb_queue_len(&sk
->sk_receive_queue
);
2438 return res
; /* "get" failed */
2440 if (len
< sizeof(value
))
2443 if (copy_to_user(ov
, &value
, sizeof(value
)))
2446 return put_user(sizeof(value
), ol
);
2449 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2451 struct sock
*sk
= sock
->sk
;
2452 struct tipc_sioc_ln_req lnr
;
2453 void __user
*argp
= (void __user
*)arg
;
2456 case SIOCGETLINKNAME
:
2457 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2459 if (!tipc_node_get_linkname(sock_net(sk
),
2460 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2461 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2462 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2466 return -EADDRNOTAVAIL
;
2468 return -ENOIOCTLCMD
;
2472 /* Protocol switches for the various types of TIPC sockets */
2474 static const struct proto_ops msg_ops
= {
2475 .owner
= THIS_MODULE
,
2477 .release
= tipc_release
,
2479 .connect
= tipc_connect
,
2480 .socketpair
= sock_no_socketpair
,
2481 .accept
= sock_no_accept
,
2482 .getname
= tipc_getname
,
2484 .ioctl
= tipc_ioctl
,
2485 .listen
= sock_no_listen
,
2486 .shutdown
= tipc_shutdown
,
2487 .setsockopt
= tipc_setsockopt
,
2488 .getsockopt
= tipc_getsockopt
,
2489 .sendmsg
= tipc_sendmsg
,
2490 .recvmsg
= tipc_recvmsg
,
2491 .mmap
= sock_no_mmap
,
2492 .sendpage
= sock_no_sendpage
2495 static const struct proto_ops packet_ops
= {
2496 .owner
= THIS_MODULE
,
2498 .release
= tipc_release
,
2500 .connect
= tipc_connect
,
2501 .socketpair
= sock_no_socketpair
,
2502 .accept
= tipc_accept
,
2503 .getname
= tipc_getname
,
2505 .ioctl
= tipc_ioctl
,
2506 .listen
= tipc_listen
,
2507 .shutdown
= tipc_shutdown
,
2508 .setsockopt
= tipc_setsockopt
,
2509 .getsockopt
= tipc_getsockopt
,
2510 .sendmsg
= tipc_send_packet
,
2511 .recvmsg
= tipc_recvmsg
,
2512 .mmap
= sock_no_mmap
,
2513 .sendpage
= sock_no_sendpage
2516 static const struct proto_ops stream_ops
= {
2517 .owner
= THIS_MODULE
,
2519 .release
= tipc_release
,
2521 .connect
= tipc_connect
,
2522 .socketpair
= sock_no_socketpair
,
2523 .accept
= tipc_accept
,
2524 .getname
= tipc_getname
,
2526 .ioctl
= tipc_ioctl
,
2527 .listen
= tipc_listen
,
2528 .shutdown
= tipc_shutdown
,
2529 .setsockopt
= tipc_setsockopt
,
2530 .getsockopt
= tipc_getsockopt
,
2531 .sendmsg
= tipc_send_stream
,
2532 .recvmsg
= tipc_recv_stream
,
2533 .mmap
= sock_no_mmap
,
2534 .sendpage
= sock_no_sendpage
2537 static const struct net_proto_family tipc_family_ops
= {
2538 .owner
= THIS_MODULE
,
2540 .create
= tipc_sk_create
2543 static struct proto tipc_proto
= {
2545 .owner
= THIS_MODULE
,
2546 .obj_size
= sizeof(struct tipc_sock
),
2547 .sysctl_rmem
= sysctl_tipc_rmem
2551 * tipc_socket_init - initialize TIPC socket interface
2553 * Returns 0 on success, errno otherwise
2555 int tipc_socket_init(void)
2559 res
= proto_register(&tipc_proto
, 1);
2561 pr_err("Failed to register TIPC protocol type\n");
2565 res
= sock_register(&tipc_family_ops
);
2567 pr_err("Failed to register TIPC socket type\n");
2568 proto_unregister(&tipc_proto
);
2576 * tipc_socket_stop - stop TIPC socket interface
2578 void tipc_socket_stop(void)
2580 sock_unregister(tipc_family_ops
.family
);
2581 proto_unregister(&tipc_proto
);
2584 /* Caller should hold socket lock for the passed tipc socket. */
2585 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
2589 struct nlattr
*nest
;
2591 peer_node
= tsk_peer_node(tsk
);
2592 peer_port
= tsk_peer_port(tsk
);
2594 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
2596 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
2598 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
2601 if (tsk
->conn_type
!= 0) {
2602 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
2604 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
2606 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
2609 nla_nest_end(skb
, nest
);
2614 nla_nest_cancel(skb
, nest
);
2619 /* Caller should hold socket lock for the passed tipc socket. */
2620 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
2621 struct tipc_sock
*tsk
)
2625 struct nlattr
*attrs
;
2626 struct net
*net
= sock_net(skb
->sk
);
2627 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2629 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2630 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
2634 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
2636 goto genlmsg_cancel
;
2637 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
2638 goto attr_msg_cancel
;
2639 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tn
->own_addr
))
2640 goto attr_msg_cancel
;
2642 if (tsk
->connected
) {
2643 err
= __tipc_nl_add_sk_con(skb
, tsk
);
2645 goto attr_msg_cancel
;
2646 } else if (!list_empty(&tsk
->publications
)) {
2647 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
2648 goto attr_msg_cancel
;
2650 nla_nest_end(skb
, attrs
);
2651 genlmsg_end(skb
, hdr
);
2656 nla_nest_cancel(skb
, attrs
);
2658 genlmsg_cancel(skb
, hdr
);
2663 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2666 struct tipc_sock
*tsk
;
2667 const struct bucket_table
*tbl
;
2668 struct rhash_head
*pos
;
2669 struct net
*net
= sock_net(skb
->sk
);
2670 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2671 u32 tbl_id
= cb
->args
[0];
2672 u32 prev_portid
= cb
->args
[1];
2675 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2676 for (; tbl_id
< tbl
->size
; tbl_id
++) {
2677 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
2678 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2679 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
2680 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2684 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
2686 prev_portid
= tsk
->portid
;
2687 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2691 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2696 cb
->args
[0] = tbl_id
;
2697 cb
->args
[1] = prev_portid
;
2702 /* Caller should hold socket lock for the passed tipc socket. */
2703 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
2704 struct netlink_callback
*cb
,
2705 struct publication
*publ
)
2708 struct nlattr
*attrs
;
2710 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2711 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
2715 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
2717 goto genlmsg_cancel
;
2719 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
2720 goto attr_msg_cancel
;
2721 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
2722 goto attr_msg_cancel
;
2723 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
2724 goto attr_msg_cancel
;
2725 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
2726 goto attr_msg_cancel
;
2728 nla_nest_end(skb
, attrs
);
2729 genlmsg_end(skb
, hdr
);
2734 nla_nest_cancel(skb
, attrs
);
2736 genlmsg_cancel(skb
, hdr
);
2741 /* Caller should hold socket lock for the passed tipc socket. */
2742 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
2743 struct netlink_callback
*cb
,
2744 struct tipc_sock
*tsk
, u32
*last_publ
)
2747 struct publication
*p
;
2750 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
2751 if (p
->key
== *last_publ
)
2754 if (p
->key
!= *last_publ
) {
2755 /* We never set seq or call nl_dump_check_consistent()
2756 * this means that setting prev_seq here will cause the
2757 * consistence check to fail in the netlink callback
2758 * handler. Resulting in the last NLMSG_DONE message
2759 * having the NLM_F_DUMP_INTR flag set.
2766 p
= list_first_entry(&tsk
->publications
, struct publication
,
2770 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
2771 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
2773 *last_publ
= p
->key
;
2782 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2785 u32 tsk_portid
= cb
->args
[0];
2786 u32 last_publ
= cb
->args
[1];
2787 u32 done
= cb
->args
[2];
2788 struct net
*net
= sock_net(skb
->sk
);
2789 struct tipc_sock
*tsk
;
2792 struct nlattr
**attrs
;
2793 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
2795 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
2799 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
2800 attrs
[TIPC_NLA_SOCK
],
2801 tipc_nl_sock_policy
);
2805 if (!sock
[TIPC_NLA_SOCK_REF
])
2808 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
2814 tsk
= tipc_sk_lookup(net
, tsk_portid
);
2818 lock_sock(&tsk
->sk
);
2819 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
2822 release_sock(&tsk
->sk
);
2825 cb
->args
[0] = tsk_portid
;
2826 cb
->args
[1] = last_publ
;