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"
47 #define SS_LISTENING -1 /* socket is listening */
48 #define SS_READY -2 /* socket is connectionless */
50 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
52 #define TIPC_FWD_MSG 1
53 #define TIPC_CONN_OK 0
54 #define TIPC_CONN_PROBING 1
55 #define TIPC_MAX_PORT 0xffffffff
56 #define TIPC_MIN_PORT 1
59 * struct tipc_sock - TIPC socket structure
60 * @sk: socket - interacts with 'port' and with user via the socket API
61 * @connected: non-zero if port is currently connected to a peer port
62 * @conn_type: TIPC type used when connection was established
63 * @conn_instance: TIPC instance used when connection was established
64 * @published: non-zero if port has one or more associated names
65 * @max_pkt: maximum packet size "hint" used when building messages sent by port
66 * @portid: unique port identity in TIPC socket hash table
67 * @phdr: preformatted message header used when sending messages
68 * @port_list: adjacent ports in TIPC's global list of ports
69 * @publications: list of publications for port
70 * @pub_count: total # of publications port has made during its lifetime
73 * @conn_timeout: the time we can wait for an unresponded setup request
74 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
75 * @link_cong: non-zero if owner must sleep because of link congestion
76 * @sent_unacked: # messages sent by socket, and not yet acked by peer
77 * @rcv_unacked: # messages read by user, but not yet acked back to peer
78 * @remote: 'connected' peer for dgram/rdm
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
91 struct list_head sock_list
;
92 struct list_head publications
;
95 unsigned long probing_intv
;
104 struct sockaddr_tipc remote
;
105 struct rhash_head node
;
109 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
110 static void tipc_data_ready(struct sock
*sk
);
111 static void tipc_write_space(struct sock
*sk
);
112 static void tipc_sock_destruct(struct sock
*sk
);
113 static int tipc_release(struct socket
*sock
);
114 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
);
115 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
);
116 static void tipc_sk_timeout(unsigned long data
);
117 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
118 struct tipc_name_seq
const *seq
);
119 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
120 struct tipc_name_seq
const *seq
);
121 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
);
122 static int tipc_sk_insert(struct tipc_sock
*tsk
);
123 static void tipc_sk_remove(struct tipc_sock
*tsk
);
124 static int __tipc_send_stream(struct socket
*sock
, struct msghdr
*m
,
126 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dsz
);
128 static const struct proto_ops packet_ops
;
129 static const struct proto_ops stream_ops
;
130 static const struct proto_ops msg_ops
;
131 static struct proto tipc_proto
;
133 static const struct rhashtable_params tsk_rht_params
;
136 * Revised TIPC socket locking policy:
138 * Most socket operations take the standard socket lock when they start
139 * and hold it until they finish (or until they need to sleep). Acquiring
140 * this lock grants the owner exclusive access to the fields of the socket
141 * data structures, with the exception of the backlog queue. A few socket
142 * operations can be done without taking the socket lock because they only
143 * read socket information that never changes during the life of the socket.
145 * Socket operations may acquire the lock for the associated TIPC port if they
146 * need to perform an operation on the port. If any routine needs to acquire
147 * both the socket lock and the port lock it must take the socket lock first
148 * to avoid the risk of deadlock.
150 * The dispatcher handling incoming messages cannot grab the socket lock in
151 * the standard fashion, since invoked it runs at the BH level and cannot block.
152 * Instead, it checks to see if the socket lock is currently owned by someone,
153 * and either handles the message itself or adds it to the socket's backlog
154 * queue; in the latter case the queued message is processed once the process
155 * owning the socket lock releases it.
157 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
158 * the problem of a blocked socket operation preventing any other operations
159 * from occurring. However, applications must be careful if they have
160 * multiple threads trying to send (or receive) on the same socket, as these
161 * operations might interfere with each other. For example, doing a connect
162 * and a receive at the same time might allow the receive to consume the
163 * ACK message meant for the connect. While additional work could be done
164 * to try and overcome this, it doesn't seem to be worthwhile at the present.
166 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
167 * that another operation that must be performed in a non-blocking manner is
168 * not delayed for very long because the lock has already been taken.
170 * NOTE: This code assumes that certain fields of a port/socket pair are
171 * constant over its lifetime; such fields can be examined without taking
172 * the socket lock and/or port lock, and do not need to be re-read even
173 * after resuming processing after waiting. These fields include:
175 * - pointer to socket sk structure (aka tipc_sock structure)
176 * - pointer to port structure
180 static u32
tsk_own_node(struct tipc_sock
*tsk
)
182 return msg_prevnode(&tsk
->phdr
);
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 bool tsk_conn_cong(struct tipc_sock
*tsk
)
235 return tsk
->snt_unacked
>= tsk
->snd_win
;
238 /* tsk_blocks(): translate a buffer size in bytes to number of
239 * advertisable blocks, taking into account the ratio truesize(len)/len
240 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
242 static u16
tsk_adv_blocks(int len
)
244 return len
/ FLOWCTL_BLK_SZ
/ 4;
247 /* tsk_inc(): increment counter for sent or received data
248 * - If block based flow control is not supported by peer we
249 * fall back to message based ditto, incrementing the counter
251 static u16
tsk_inc(struct tipc_sock
*tsk
, int msglen
)
253 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
254 return ((msglen
/ FLOWCTL_BLK_SZ
) + 1);
259 * tsk_advance_rx_queue - discard first buffer in socket receive queue
261 * Caller must hold socket lock
263 static void tsk_advance_rx_queue(struct sock
*sk
)
265 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
268 /* tipc_sk_respond() : send response message back to sender
270 static void tipc_sk_respond(struct sock
*sk
, struct sk_buff
*skb
, int err
)
274 u32 onode
= tipc_own_addr(sock_net(sk
));
276 if (!tipc_msg_reverse(onode
, &skb
, err
))
279 dnode
= msg_destnode(buf_msg(skb
));
280 selector
= msg_origport(buf_msg(skb
));
281 tipc_node_xmit_skb(sock_net(sk
), skb
, dnode
, selector
);
285 * tsk_rej_rx_queue - reject all buffers in socket receive queue
287 * Caller must hold socket lock
289 static void tsk_rej_rx_queue(struct sock
*sk
)
293 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)))
294 tipc_sk_respond(sk
, skb
, TIPC_ERR_NO_PORT
);
297 /* tsk_peer_msg - verify if message was sent by connected port's peer
299 * Handles cases where the node's network address has changed from
300 * the default of <0.0.0> to its configured setting.
302 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
304 struct tipc_net
*tn
= net_generic(sock_net(&tsk
->sk
), tipc_net_id
);
305 u32 peer_port
= tsk_peer_port(tsk
);
309 if (unlikely(!tsk
->connected
))
312 if (unlikely(msg_origport(msg
) != peer_port
))
315 orig_node
= msg_orignode(msg
);
316 peer_node
= tsk_peer_node(tsk
);
318 if (likely(orig_node
== peer_node
))
321 if (!orig_node
&& (peer_node
== tn
->own_addr
))
324 if (!peer_node
&& (orig_node
== tn
->own_addr
))
331 * tipc_sk_create - create a TIPC socket
332 * @net: network namespace (must be default network)
333 * @sock: pre-allocated socket structure
334 * @protocol: protocol indicator (must be 0)
335 * @kern: caused by kernel or by userspace?
337 * This routine creates additional data structures used by the TIPC socket,
338 * initializes them, and links them together.
340 * Returns 0 on success, errno otherwise
342 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
343 int protocol
, int kern
)
346 const struct proto_ops
*ops
;
349 struct tipc_sock
*tsk
;
350 struct tipc_msg
*msg
;
352 /* Validate arguments */
353 if (unlikely(protocol
!= 0))
354 return -EPROTONOSUPPORT
;
356 switch (sock
->type
) {
359 state
= SS_UNCONNECTED
;
363 state
= SS_UNCONNECTED
;
374 /* Allocate socket's protocol area */
375 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
, kern
);
380 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
381 INIT_LIST_HEAD(&tsk
->publications
);
383 tn
= net_generic(sock_net(sk
), tipc_net_id
);
384 tipc_msg_init(tn
->own_addr
, msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
387 /* Finish initializing socket data structures */
390 sock_init_data(sock
, sk
);
391 if (tipc_sk_insert(tsk
)) {
392 pr_warn("Socket create failed; port number exhausted\n");
395 msg_set_origport(msg
, tsk
->portid
);
396 setup_timer(&sk
->sk_timer
, tipc_sk_timeout
, (unsigned long)tsk
);
397 sk
->sk_backlog_rcv
= tipc_backlog_rcv
;
398 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
399 sk
->sk_data_ready
= tipc_data_ready
;
400 sk
->sk_write_space
= tipc_write_space
;
401 sk
->sk_destruct
= tipc_sock_destruct
;
402 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
403 atomic_set(&tsk
->dupl_rcvcnt
, 0);
405 /* Start out with safe limits until we receive an advertised window */
406 tsk
->snd_win
= tsk_adv_blocks(RCVBUF_MIN
);
407 tsk
->rcv_win
= tsk
->snd_win
;
409 if (sock
->state
== SS_READY
) {
410 tsk_set_unreturnable(tsk
, true);
411 if (sock
->type
== SOCK_DGRAM
)
412 tsk_set_unreliable(tsk
, true);
417 static void tipc_sk_callback(struct rcu_head
*head
)
419 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
425 * tipc_release - destroy a TIPC socket
426 * @sock: socket to destroy
428 * This routine cleans up any messages that are still queued on the socket.
429 * For DGRAM and RDM socket types, all queued messages are rejected.
430 * For SEQPACKET and STREAM socket types, the first message is rejected
431 * and any others are discarded. (If the first message on a STREAM socket
432 * is partially-read, it is discarded and the next one is rejected instead.)
434 * NOTE: Rejected messages are not necessarily returned to the sender! They
435 * are returned or discarded according to the "destination droppable" setting
436 * specified for the message by the sender.
438 * Returns 0 on success, errno otherwise
440 static int tipc_release(struct socket
*sock
)
442 struct sock
*sk
= sock
->sk
;
444 struct tipc_sock
*tsk
;
449 * Exit if socket isn't fully initialized (occurs when a failed accept()
450 * releases a pre-allocated child socket that was never used)
460 * Reject all unreceived messages, except on an active connection
461 * (which disconnects locally & sends a 'FIN+' to peer)
463 dnode
= tsk_peer_node(tsk
);
464 while (sock
->state
!= SS_DISCONNECTING
) {
465 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
468 if (TIPC_SKB_CB(skb
)->handle
!= NULL
)
471 if ((sock
->state
== SS_CONNECTING
) ||
472 (sock
->state
== SS_CONNECTED
)) {
473 sock
->state
= SS_DISCONNECTING
;
475 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
477 tipc_sk_respond(sk
, skb
, TIPC_ERR_NO_PORT
);
481 tipc_sk_withdraw(tsk
, 0, NULL
);
482 sk_stop_timer(sk
, &sk
->sk_timer
);
484 if (tsk
->connected
) {
485 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
486 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0, dnode
,
487 tsk_own_node(tsk
), tsk_peer_port(tsk
),
488 tsk
->portid
, TIPC_ERR_NO_PORT
);
490 tipc_node_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
491 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
494 /* Reject any messages that accumulated in backlog queue */
495 sock
->state
= SS_DISCONNECTING
;
498 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
505 * tipc_bind - associate or disassocate TIPC name(s) with a socket
506 * @sock: socket structure
507 * @uaddr: socket address describing name(s) and desired operation
508 * @uaddr_len: size of socket address data structure
510 * Name and name sequence binding is indicated using a positive scope value;
511 * a negative scope value unbinds the specified name. Specifying no name
512 * (i.e. a socket address length of 0) unbinds all names from the socket.
514 * Returns 0 on success, errno otherwise
516 * NOTE: This routine doesn't need to take the socket lock since it doesn't
517 * access any non-constant socket information.
519 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
522 struct sock
*sk
= sock
->sk
;
523 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
524 struct tipc_sock
*tsk
= tipc_sk(sk
);
528 if (unlikely(!uaddr_len
)) {
529 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
533 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
537 if (addr
->family
!= AF_TIPC
) {
542 if (addr
->addrtype
== TIPC_ADDR_NAME
)
543 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
544 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
549 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
550 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
551 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
556 res
= (addr
->scope
> 0) ?
557 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
558 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
565 * tipc_getname - get port ID of socket or peer socket
566 * @sock: socket structure
567 * @uaddr: area for returned socket address
568 * @uaddr_len: area for returned length of socket address
569 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
571 * Returns 0 on success, errno otherwise
573 * NOTE: This routine doesn't need to take the socket lock since it only
574 * accesses socket information that is unchanging (or which changes in
575 * a completely predictable manner).
577 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
578 int *uaddr_len
, int peer
)
580 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
581 struct tipc_sock
*tsk
= tipc_sk(sock
->sk
);
582 struct tipc_net
*tn
= net_generic(sock_net(sock
->sk
), tipc_net_id
);
584 memset(addr
, 0, sizeof(*addr
));
586 if ((sock
->state
!= SS_CONNECTED
) &&
587 ((peer
!= 2) || (sock
->state
!= SS_DISCONNECTING
)))
589 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
590 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
592 addr
->addr
.id
.ref
= tsk
->portid
;
593 addr
->addr
.id
.node
= tn
->own_addr
;
596 *uaddr_len
= sizeof(*addr
);
597 addr
->addrtype
= TIPC_ADDR_ID
;
598 addr
->family
= AF_TIPC
;
600 addr
->addr
.name
.domain
= 0;
606 * tipc_poll - read and possibly block on pollmask
607 * @file: file structure associated with the socket
608 * @sock: socket for which to calculate the poll bits
611 * Returns pollmask value
614 * It appears that the usual socket locking mechanisms are not useful here
615 * since the pollmask info is potentially out-of-date the moment this routine
616 * exits. TCP and other protocols seem to rely on higher level poll routines
617 * to handle any preventable race conditions, so TIPC will do the same ...
619 * TIPC sets the returned events as follows:
621 * socket state flags set
622 * ------------ ---------
623 * unconnected no read flags
624 * POLLOUT if port is not congested
626 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
629 * connected POLLIN/POLLRDNORM if data in rx queue
630 * POLLOUT if port is not congested
632 * disconnecting POLLIN/POLLRDNORM/POLLHUP
635 * listening POLLIN if SYN in rx queue
638 * ready POLLIN/POLLRDNORM if data in rx queue
639 * [connectionless] POLLOUT (since port cannot be congested)
641 * IMPORTANT: The fact that a read or write operation is indicated does NOT
642 * imply that the operation will succeed, merely that it should be performed
643 * and will not block.
645 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
648 struct sock
*sk
= sock
->sk
;
649 struct tipc_sock
*tsk
= tipc_sk(sk
);
652 sock_poll_wait(file
, sk_sleep(sk
), wait
);
654 switch ((int)sock
->state
) {
661 if (!tsk
->link_cong
&& !tsk_conn_cong(tsk
))
666 if (!skb_queue_empty(&sk
->sk_receive_queue
))
667 mask
|= (POLLIN
| POLLRDNORM
);
669 case SS_DISCONNECTING
:
670 mask
= (POLLIN
| POLLRDNORM
| POLLHUP
);
678 * tipc_sendmcast - send multicast message
679 * @sock: socket structure
680 * @seq: destination address
681 * @msg: message to send
682 * @dsz: total length of message data
683 * @timeo: timeout to wait for wakeup
685 * Called from function tipc_sendmsg(), which has done all sanity checks
686 * Returns the number of bytes sent on success, or errno
688 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
689 struct msghdr
*msg
, size_t dsz
, long timeo
)
691 struct sock
*sk
= sock
->sk
;
692 struct tipc_sock
*tsk
= tipc_sk(sk
);
693 struct net
*net
= sock_net(sk
);
694 struct tipc_msg
*mhdr
= &tsk
->phdr
;
695 struct sk_buff_head pktchain
;
696 struct iov_iter save
= msg
->msg_iter
;
700 msg_set_type(mhdr
, TIPC_MCAST_MSG
);
701 msg_set_lookup_scope(mhdr
, TIPC_CLUSTER_SCOPE
);
702 msg_set_destport(mhdr
, 0);
703 msg_set_destnode(mhdr
, 0);
704 msg_set_nametype(mhdr
, seq
->type
);
705 msg_set_namelower(mhdr
, seq
->lower
);
706 msg_set_nameupper(mhdr
, seq
->upper
);
707 msg_set_hdr_sz(mhdr
, MCAST_H_SIZE
);
709 skb_queue_head_init(&pktchain
);
712 mtu
= tipc_bcast_get_mtu(net
);
713 rc
= tipc_msg_build(mhdr
, msg
, 0, dsz
, mtu
, &pktchain
);
714 if (unlikely(rc
< 0))
718 rc
= tipc_bcast_xmit(net
, &pktchain
);
722 if (rc
== -ELINKCONG
) {
724 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
728 __skb_queue_purge(&pktchain
);
729 if (rc
== -EMSGSIZE
) {
730 msg
->msg_iter
= save
;
739 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
740 * @arrvq: queue with arriving messages, to be cloned after destination lookup
741 * @inputq: queue with cloned messages, delivered to socket after dest lookup
743 * Multi-threaded: parallel calls with reference to same queues may occur
745 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
746 struct sk_buff_head
*inputq
)
748 struct tipc_msg
*msg
;
749 struct tipc_plist dports
;
751 u32 scope
= TIPC_CLUSTER_SCOPE
;
752 struct sk_buff_head tmpq
;
754 struct sk_buff
*skb
, *_skb
;
756 __skb_queue_head_init(&tmpq
);
757 tipc_plist_init(&dports
);
759 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
760 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
762 hsz
= skb_headroom(skb
) + msg_hdr_sz(msg
);
764 if (in_own_node(net
, msg_orignode(msg
)))
765 scope
= TIPC_NODE_SCOPE
;
767 /* Create destination port list and message clones: */
768 tipc_nametbl_mc_translate(net
,
769 msg_nametype(msg
), msg_namelower(msg
),
770 msg_nameupper(msg
), scope
, &dports
);
771 portid
= tipc_plist_pop(&dports
);
772 for (; portid
; portid
= tipc_plist_pop(&dports
)) {
773 _skb
= __pskb_copy(skb
, hsz
, GFP_ATOMIC
);
775 msg_set_destport(buf_msg(_skb
), portid
);
776 __skb_queue_tail(&tmpq
, _skb
);
779 pr_warn("Failed to clone mcast rcv buffer\n");
781 /* Append to inputq if not already done by other thread */
782 spin_lock_bh(&inputq
->lock
);
783 if (skb_peek(arrvq
) == skb
) {
784 skb_queue_splice_tail_init(&tmpq
, inputq
);
785 kfree_skb(__skb_dequeue(arrvq
));
787 spin_unlock_bh(&inputq
->lock
);
788 __skb_queue_purge(&tmpq
);
791 tipc_sk_rcv(net
, inputq
);
795 * tipc_sk_proto_rcv - receive a connection mng protocol message
796 * @tsk: receiving socket
797 * @skb: pointer to message buffer.
799 static void tipc_sk_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
801 struct sock
*sk
= &tsk
->sk
;
802 struct tipc_msg
*hdr
= buf_msg(skb
);
803 int mtyp
= msg_type(hdr
);
806 /* Ignore if connection cannot be validated: */
807 if (!tsk_peer_msg(tsk
, hdr
))
810 tsk
->probing_state
= TIPC_CONN_OK
;
812 if (mtyp
== CONN_PROBE
) {
813 msg_set_type(hdr
, CONN_PROBE_REPLY
);
814 tipc_sk_respond(sk
, skb
, TIPC_OK
);
816 } else if (mtyp
== CONN_ACK
) {
817 conn_cong
= tsk_conn_cong(tsk
);
818 tsk
->snt_unacked
-= msg_conn_ack(hdr
);
819 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
820 tsk
->snd_win
= msg_adv_win(hdr
);
822 sk
->sk_write_space(sk
);
823 } else if (mtyp
!= CONN_PROBE_REPLY
) {
824 pr_warn("Received unknown CONN_PROTO msg\n");
830 static int tipc_wait_for_sndmsg(struct socket
*sock
, long *timeo_p
)
832 struct sock
*sk
= sock
->sk
;
833 struct tipc_sock
*tsk
= tipc_sk(sk
);
838 int err
= sock_error(sk
);
841 if (sock
->state
== SS_DISCONNECTING
)
845 if (signal_pending(current
))
846 return sock_intr_errno(*timeo_p
);
848 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
849 done
= sk_wait_event(sk
, timeo_p
, !tsk
->link_cong
);
850 finish_wait(sk_sleep(sk
), &wait
);
856 * tipc_sendmsg - send message in connectionless manner
857 * @sock: socket structure
858 * @m: message to send
859 * @dsz: amount of user data to be sent
861 * Message must have an destination specified explicitly.
862 * Used for SOCK_RDM and SOCK_DGRAM messages,
863 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
864 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
866 * Returns the number of bytes sent on success, or errno otherwise
868 static int tipc_sendmsg(struct socket
*sock
,
869 struct msghdr
*m
, size_t dsz
)
871 struct sock
*sk
= sock
->sk
;
875 ret
= __tipc_sendmsg(sock
, m
, dsz
);
881 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
883 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
884 struct sock
*sk
= sock
->sk
;
885 struct tipc_sock
*tsk
= tipc_sk(sk
);
886 struct net
*net
= sock_net(sk
);
887 struct tipc_msg
*mhdr
= &tsk
->phdr
;
889 struct sk_buff_head pktchain
;
891 struct tipc_name_seq
*seq
;
892 struct iov_iter save
;
897 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
899 if (unlikely(!dest
)) {
900 if (tsk
->connected
&& sock
->state
== SS_READY
)
903 return -EDESTADDRREQ
;
904 } else if (unlikely(m
->msg_namelen
< sizeof(*dest
)) ||
905 dest
->family
!= AF_TIPC
) {
908 if (unlikely(sock
->state
!= SS_READY
)) {
909 if (sock
->state
== SS_LISTENING
)
911 if (sock
->state
!= SS_UNCONNECTED
)
915 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
916 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
917 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
920 seq
= &dest
->addr
.nameseq
;
921 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
923 if (dest
->addrtype
== TIPC_ADDR_MCAST
) {
924 return tipc_sendmcast(sock
, seq
, m
, dsz
, timeo
);
925 } else if (dest
->addrtype
== TIPC_ADDR_NAME
) {
926 u32 type
= dest
->addr
.name
.name
.type
;
927 u32 inst
= dest
->addr
.name
.name
.instance
;
928 u32 domain
= dest
->addr
.name
.domain
;
931 msg_set_type(mhdr
, TIPC_NAMED_MSG
);
932 msg_set_hdr_sz(mhdr
, NAMED_H_SIZE
);
933 msg_set_nametype(mhdr
, type
);
934 msg_set_nameinst(mhdr
, inst
);
935 msg_set_lookup_scope(mhdr
, tipc_addr_scope(domain
));
936 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
937 msg_set_destnode(mhdr
, dnode
);
938 msg_set_destport(mhdr
, dport
);
939 if (unlikely(!dport
&& !dnode
))
940 return -EHOSTUNREACH
;
941 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
942 dnode
= dest
->addr
.id
.node
;
943 msg_set_type(mhdr
, TIPC_DIRECT_MSG
);
944 msg_set_lookup_scope(mhdr
, 0);
945 msg_set_destnode(mhdr
, dnode
);
946 msg_set_destport(mhdr
, dest
->addr
.id
.ref
);
947 msg_set_hdr_sz(mhdr
, BASIC_H_SIZE
);
950 skb_queue_head_init(&pktchain
);
953 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
954 rc
= tipc_msg_build(mhdr
, m
, 0, dsz
, mtu
, &pktchain
);
959 skb
= skb_peek(&pktchain
);
960 TIPC_SKB_CB(skb
)->wakeup_pending
= tsk
->link_cong
;
961 rc
= tipc_node_xmit(net
, &pktchain
, dnode
, tsk
->portid
);
963 if (sock
->state
!= SS_READY
)
964 sock
->state
= SS_CONNECTING
;
967 if (rc
== -ELINKCONG
) {
969 rc
= tipc_wait_for_sndmsg(sock
, &timeo
);
973 __skb_queue_purge(&pktchain
);
974 if (rc
== -EMSGSIZE
) {
984 static int tipc_wait_for_sndpkt(struct socket
*sock
, long *timeo_p
)
986 struct sock
*sk
= sock
->sk
;
987 struct tipc_sock
*tsk
= tipc_sk(sk
);
992 int err
= sock_error(sk
);
995 if (sock
->state
== SS_DISCONNECTING
)
997 else if (sock
->state
!= SS_CONNECTED
)
1001 if (signal_pending(current
))
1002 return sock_intr_errno(*timeo_p
);
1004 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1005 done
= sk_wait_event(sk
, timeo_p
,
1007 !tsk_conn_cong(tsk
)) ||
1009 finish_wait(sk_sleep(sk
), &wait
);
1015 * tipc_send_stream - send stream-oriented data
1016 * @sock: socket structure
1018 * @dsz: total length of data to be transmitted
1020 * Used for SOCK_STREAM data.
1022 * Returns the number of bytes sent on success (or partial success),
1023 * or errno if no data sent
1025 static int tipc_send_stream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1027 struct sock
*sk
= sock
->sk
;
1031 ret
= __tipc_send_stream(sock
, m
, dsz
);
1037 static int __tipc_send_stream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1039 struct sock
*sk
= sock
->sk
;
1040 struct net
*net
= sock_net(sk
);
1041 struct tipc_sock
*tsk
= tipc_sk(sk
);
1042 struct tipc_msg
*mhdr
= &tsk
->phdr
;
1043 struct sk_buff_head pktchain
;
1044 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1045 u32 portid
= tsk
->portid
;
1049 uint mtu
, send
, sent
= 0;
1050 struct iov_iter save
;
1051 int hlen
= MIN_H_SIZE
;
1053 /* Handle implied connection establishment */
1054 if (unlikely(dest
)) {
1055 rc
= __tipc_sendmsg(sock
, m
, dsz
);
1056 hlen
= msg_hdr_sz(mhdr
);
1057 if (dsz
&& (dsz
== rc
))
1058 tsk
->snt_unacked
= tsk_inc(tsk
, dsz
+ hlen
);
1061 if (dsz
> (uint
)INT_MAX
)
1064 if (unlikely(sock
->state
!= SS_CONNECTED
)) {
1065 if (sock
->state
== SS_DISCONNECTING
)
1071 timeo
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1072 dnode
= tsk_peer_node(tsk
);
1073 skb_queue_head_init(&pktchain
);
1078 send
= min_t(uint
, dsz
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1079 rc
= tipc_msg_build(mhdr
, m
, sent
, send
, mtu
, &pktchain
);
1080 if (unlikely(rc
< 0))
1084 if (likely(!tsk_conn_cong(tsk
))) {
1085 rc
= tipc_node_xmit(net
, &pktchain
, dnode
, portid
);
1087 tsk
->snt_unacked
+= tsk_inc(tsk
, send
+ hlen
);
1093 if (rc
== -EMSGSIZE
) {
1094 __skb_queue_purge(&pktchain
);
1095 tsk
->max_pkt
= tipc_node_get_mtu(net
, dnode
,
1100 if (rc
!= -ELINKCONG
)
1105 rc
= tipc_wait_for_sndpkt(sock
, &timeo
);
1108 __skb_queue_purge(&pktchain
);
1109 return sent
? sent
: rc
;
1113 * tipc_send_packet - send a connection-oriented message
1114 * @sock: socket structure
1115 * @m: message to send
1116 * @dsz: length of data to be transmitted
1118 * Used for SOCK_SEQPACKET messages.
1120 * Returns the number of bytes sent on success, or errno otherwise
1122 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1124 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1127 return tipc_send_stream(sock
, m
, dsz
);
1130 /* tipc_sk_finish_conn - complete the setup of a connection
1132 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1135 struct sock
*sk
= &tsk
->sk
;
1136 struct net
*net
= sock_net(sk
);
1137 struct tipc_msg
*msg
= &tsk
->phdr
;
1139 msg_set_destnode(msg
, peer_node
);
1140 msg_set_destport(msg
, peer_port
);
1141 msg_set_type(msg
, TIPC_CONN_MSG
);
1142 msg_set_lookup_scope(msg
, 0);
1143 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1145 tsk
->probing_intv
= CONN_PROBING_INTERVAL
;
1146 tsk
->probing_state
= TIPC_CONN_OK
;
1148 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ tsk
->probing_intv
);
1149 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1150 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1151 tsk
->peer_caps
= tipc_node_get_capabilities(net
, peer_node
);
1152 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1155 /* Fall back to message based flow control */
1156 tsk
->rcv_win
= FLOWCTL_MSG_WIN
;
1157 tsk
->snd_win
= FLOWCTL_MSG_WIN
;
1161 * set_orig_addr - capture sender's address for received message
1162 * @m: descriptor for message info
1163 * @msg: received message header
1165 * Note: Address is not captured if not requested by receiver.
1167 static void set_orig_addr(struct msghdr
*m
, struct tipc_msg
*msg
)
1169 DECLARE_SOCKADDR(struct sockaddr_tipc
*, addr
, m
->msg_name
);
1172 addr
->family
= AF_TIPC
;
1173 addr
->addrtype
= TIPC_ADDR_ID
;
1174 memset(&addr
->addr
, 0, sizeof(addr
->addr
));
1175 addr
->addr
.id
.ref
= msg_origport(msg
);
1176 addr
->addr
.id
.node
= msg_orignode(msg
);
1177 addr
->addr
.name
.domain
= 0; /* could leave uninitialized */
1178 addr
->scope
= 0; /* could leave uninitialized */
1179 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1184 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1185 * @m: descriptor for message info
1186 * @msg: received message header
1187 * @tsk: TIPC port associated with message
1189 * Note: Ancillary data is not captured if not requested by receiver.
1191 * Returns 0 if successful, otherwise errno
1193 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1194 struct tipc_sock
*tsk
)
1202 if (likely(m
->msg_controllen
== 0))
1205 /* Optionally capture errored message object(s) */
1206 err
= msg
? msg_errcode(msg
) : 0;
1207 if (unlikely(err
)) {
1209 anc_data
[1] = msg_data_sz(msg
);
1210 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1214 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1221 /* Optionally capture message destination object */
1222 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1223 switch (dest_type
) {
1224 case TIPC_NAMED_MSG
:
1226 anc_data
[0] = msg_nametype(msg
);
1227 anc_data
[1] = msg_namelower(msg
);
1228 anc_data
[2] = msg_namelower(msg
);
1230 case TIPC_MCAST_MSG
:
1232 anc_data
[0] = msg_nametype(msg
);
1233 anc_data
[1] = msg_namelower(msg
);
1234 anc_data
[2] = msg_nameupper(msg
);
1237 has_name
= (tsk
->conn_type
!= 0);
1238 anc_data
[0] = tsk
->conn_type
;
1239 anc_data
[1] = tsk
->conn_instance
;
1240 anc_data
[2] = tsk
->conn_instance
;
1246 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1254 static void tipc_sk_send_ack(struct tipc_sock
*tsk
)
1256 struct net
*net
= sock_net(&tsk
->sk
);
1257 struct sk_buff
*skb
= NULL
;
1258 struct tipc_msg
*msg
;
1259 u32 peer_port
= tsk_peer_port(tsk
);
1260 u32 dnode
= tsk_peer_node(tsk
);
1262 if (!tsk
->connected
)
1264 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1265 dnode
, tsk_own_node(tsk
), peer_port
,
1266 tsk
->portid
, TIPC_OK
);
1270 msg_set_conn_ack(msg
, tsk
->rcv_unacked
);
1271 tsk
->rcv_unacked
= 0;
1273 /* Adjust to and advertize the correct window limit */
1274 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
) {
1275 tsk
->rcv_win
= tsk_adv_blocks(tsk
->sk
.sk_rcvbuf
);
1276 msg_set_adv_win(msg
, tsk
->rcv_win
);
1278 tipc_node_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1281 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1283 struct sock
*sk
= sock
->sk
;
1285 long timeo
= *timeop
;
1289 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1290 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1291 if (sock
->state
== SS_DISCONNECTING
) {
1296 timeo
= schedule_timeout(timeo
);
1300 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1305 err
= sock_intr_errno(timeo
);
1306 if (signal_pending(current
))
1309 finish_wait(sk_sleep(sk
), &wait
);
1315 * tipc_recvmsg - receive packet-oriented message
1316 * @m: descriptor for message info
1317 * @buf_len: total size of user buffer area
1318 * @flags: receive flags
1320 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1321 * If the complete message doesn't fit in user area, truncate it.
1323 * Returns size of returned message data, errno otherwise
1325 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
, size_t buf_len
,
1328 struct sock
*sk
= sock
->sk
;
1329 struct tipc_sock
*tsk
= tipc_sk(sk
);
1330 struct sk_buff
*buf
;
1331 struct tipc_msg
*msg
;
1337 /* Catch invalid receive requests */
1338 if (unlikely(!buf_len
))
1343 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1348 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1351 /* Look for a message in receive queue; wait if necessary */
1352 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1356 /* Look at first message in receive queue */
1357 buf
= skb_peek(&sk
->sk_receive_queue
);
1359 sz
= msg_data_sz(msg
);
1360 hlen
= msg_hdr_sz(msg
);
1361 err
= msg_errcode(msg
);
1363 /* Discard an empty non-errored message & try again */
1364 if ((!sz
) && (!err
)) {
1365 tsk_advance_rx_queue(sk
);
1369 /* Capture sender's address (optional) */
1370 set_orig_addr(m
, msg
);
1372 /* Capture ancillary data (optional) */
1373 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1377 /* Capture message data (if valid) & compute return value (always) */
1379 if (unlikely(buf_len
< sz
)) {
1381 m
->msg_flags
|= MSG_TRUNC
;
1383 res
= skb_copy_datagram_msg(buf
, hlen
, m
, sz
);
1388 if ((sock
->state
== SS_READY
) ||
1389 ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
))
1395 if (unlikely(flags
& MSG_PEEK
))
1398 if (likely(sock
->state
!= SS_READY
)) {
1399 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ sz
);
1400 if (unlikely(tsk
->rcv_unacked
>= (tsk
->rcv_win
/ 4)))
1401 tipc_sk_send_ack(tsk
);
1403 tsk_advance_rx_queue(sk
);
1410 * tipc_recv_stream - receive stream-oriented data
1411 * @m: descriptor for message info
1412 * @buf_len: total size of user buffer area
1413 * @flags: receive flags
1415 * Used for SOCK_STREAM messages only. If not enough data is available
1416 * will optionally wait for more; never truncates data.
1418 * Returns size of returned message data, errno otherwise
1420 static int tipc_recv_stream(struct socket
*sock
, struct msghdr
*m
,
1421 size_t buf_len
, int flags
)
1423 struct sock
*sk
= sock
->sk
;
1424 struct tipc_sock
*tsk
= tipc_sk(sk
);
1425 struct sk_buff
*buf
;
1426 struct tipc_msg
*msg
;
1429 int sz_to_copy
, target
, needed
;
1434 /* Catch invalid receive attempts */
1435 if (unlikely(!buf_len
))
1440 if (unlikely(sock
->state
== SS_UNCONNECTED
)) {
1445 target
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buf_len
);
1446 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1449 /* Look for a message in receive queue; wait if necessary */
1450 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1454 /* Look at first message in receive queue */
1455 buf
= skb_peek(&sk
->sk_receive_queue
);
1457 sz
= msg_data_sz(msg
);
1458 hlen
= msg_hdr_sz(msg
);
1459 err
= msg_errcode(msg
);
1461 /* Discard an empty non-errored message & try again */
1462 if ((!sz
) && (!err
)) {
1463 tsk_advance_rx_queue(sk
);
1467 /* Optionally capture sender's address & ancillary data of first msg */
1468 if (sz_copied
== 0) {
1469 set_orig_addr(m
, msg
);
1470 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1475 /* Capture message data (if valid) & compute return value (always) */
1477 u32 offset
= (u32
)(unsigned long)(TIPC_SKB_CB(buf
)->handle
);
1480 needed
= (buf_len
- sz_copied
);
1481 sz_to_copy
= (sz
<= needed
) ? sz
: needed
;
1483 res
= skb_copy_datagram_msg(buf
, hlen
+ offset
, m
, sz_to_copy
);
1487 sz_copied
+= sz_to_copy
;
1489 if (sz_to_copy
< sz
) {
1490 if (!(flags
& MSG_PEEK
))
1491 TIPC_SKB_CB(buf
)->handle
=
1492 (void *)(unsigned long)(offset
+ sz_to_copy
);
1497 goto exit
; /* can't add error msg to valid data */
1499 if ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
)
1505 if (unlikely(flags
& MSG_PEEK
))
1508 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ sz
);
1509 if (unlikely(tsk
->rcv_unacked
>= (tsk
->rcv_win
/ 4)))
1510 tipc_sk_send_ack(tsk
);
1511 tsk_advance_rx_queue(sk
);
1513 /* Loop around if more data is required */
1514 if ((sz_copied
< buf_len
) && /* didn't get all requested data */
1515 (!skb_queue_empty(&sk
->sk_receive_queue
) ||
1516 (sz_copied
< target
)) && /* and more is ready or required */
1517 (!err
)) /* and haven't reached a FIN */
1522 return sz_copied
? sz_copied
: res
;
1526 * tipc_write_space - wake up thread if port congestion is released
1529 static void tipc_write_space(struct sock
*sk
)
1531 struct socket_wq
*wq
;
1534 wq
= rcu_dereference(sk
->sk_wq
);
1535 if (skwq_has_sleeper(wq
))
1536 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1537 POLLWRNORM
| POLLWRBAND
);
1542 * tipc_data_ready - wake up threads to indicate messages have been received
1544 * @len: the length of messages
1546 static void tipc_data_ready(struct sock
*sk
)
1548 struct socket_wq
*wq
;
1551 wq
= rcu_dereference(sk
->sk_wq
);
1552 if (skwq_has_sleeper(wq
))
1553 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1554 POLLRDNORM
| POLLRDBAND
);
1558 static void tipc_sock_destruct(struct sock
*sk
)
1560 __skb_queue_purge(&sk
->sk_receive_queue
);
1564 * filter_connect - Handle all incoming messages for a connection-based socket
1566 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1568 * Returns true if everything ok, false otherwise
1570 static bool filter_connect(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
1572 struct sock
*sk
= &tsk
->sk
;
1573 struct net
*net
= sock_net(sk
);
1574 struct socket
*sock
= sk
->sk_socket
;
1575 struct tipc_msg
*hdr
= buf_msg(skb
);
1577 if (unlikely(msg_mcast(hdr
)))
1580 switch ((int)sock
->state
) {
1583 /* Accept only connection-based messages sent by peer */
1584 if (unlikely(!tsk_peer_msg(tsk
, hdr
)))
1587 if (unlikely(msg_errcode(hdr
))) {
1588 sock
->state
= SS_DISCONNECTING
;
1590 /* Let timer expire on it's own */
1591 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
1598 /* Accept only ACK or NACK message */
1599 if (unlikely(!msg_connected(hdr
)))
1602 if (unlikely(msg_errcode(hdr
))) {
1603 sock
->state
= SS_DISCONNECTING
;
1604 sk
->sk_err
= ECONNREFUSED
;
1608 if (unlikely(!msg_isdata(hdr
))) {
1609 sock
->state
= SS_DISCONNECTING
;
1610 sk
->sk_err
= EINVAL
;
1614 tipc_sk_finish_conn(tsk
, msg_origport(hdr
), msg_orignode(hdr
));
1615 msg_set_importance(&tsk
->phdr
, msg_importance(hdr
));
1616 sock
->state
= SS_CONNECTED
;
1618 /* If 'ACK+' message, add to socket receive queue */
1619 if (msg_data_sz(hdr
))
1622 /* If empty 'ACK-' message, wake up sleeping connect() */
1623 if (waitqueue_active(sk_sleep(sk
)))
1624 wake_up_interruptible(sk_sleep(sk
));
1626 /* 'ACK-' message is neither accepted nor rejected: */
1627 msg_set_dest_droppable(hdr
, 1);
1631 case SS_UNCONNECTED
:
1633 /* Accept only SYN message */
1634 if (!msg_connected(hdr
) && !(msg_errcode(hdr
)))
1637 case SS_DISCONNECTING
:
1640 pr_err("Unknown socket state %u\n", sock
->state
);
1646 * rcvbuf_limit - get proper overload limit of socket receive queue
1650 * For connection oriented messages, irrespective of importance,
1651 * default queue limit is 2 MB.
1653 * For connectionless messages, queue limits are based on message
1654 * importance as follows:
1656 * TIPC_LOW_IMPORTANCE (2 MB)
1657 * TIPC_MEDIUM_IMPORTANCE (4 MB)
1658 * TIPC_HIGH_IMPORTANCE (8 MB)
1659 * TIPC_CRITICAL_IMPORTANCE (16 MB)
1661 * Returns overload limit according to corresponding message importance
1663 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*skb
)
1665 struct tipc_sock
*tsk
= tipc_sk(sk
);
1666 struct tipc_msg
*hdr
= buf_msg(skb
);
1668 if (unlikely(!msg_connected(hdr
)))
1669 return sk
->sk_rcvbuf
<< msg_importance(hdr
);
1671 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
1672 return sk
->sk_rcvbuf
;
1674 return FLOWCTL_MSG_LIM
;
1678 * filter_rcv - validate incoming message
1680 * @skb: pointer to message.
1682 * Enqueues message on receive queue if acceptable; optionally handles
1683 * disconnect indication for a connected socket.
1685 * Called with socket lock already taken
1687 * Returns true if message was added to socket receive queue, otherwise false
1689 static bool filter_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1691 struct socket
*sock
= sk
->sk_socket
;
1692 struct tipc_sock
*tsk
= tipc_sk(sk
);
1693 struct tipc_msg
*hdr
= buf_msg(skb
);
1694 unsigned int limit
= rcvbuf_limit(sk
, skb
);
1696 int usr
= msg_user(hdr
);
1698 if (unlikely(msg_user(hdr
) == CONN_MANAGER
)) {
1699 tipc_sk_proto_rcv(tsk
, skb
);
1703 if (unlikely(usr
== SOCK_WAKEUP
)) {
1706 sk
->sk_write_space(sk
);
1710 /* Drop if illegal message type */
1711 if (unlikely(msg_type(hdr
) > TIPC_DIRECT_MSG
)) {
1716 /* Reject if wrong message type for current socket state */
1717 if (unlikely(sock
->state
== SS_READY
)) {
1718 if (msg_connected(hdr
)) {
1719 err
= TIPC_ERR_NO_PORT
;
1722 } else if (unlikely(!filter_connect(tsk
, skb
))) {
1723 err
= TIPC_ERR_NO_PORT
;
1727 /* Reject message if there isn't room to queue it */
1728 if (unlikely(sk_rmem_alloc_get(sk
) + skb
->truesize
>= limit
)) {
1729 err
= TIPC_ERR_OVERLOAD
;
1733 /* Enqueue message */
1734 TIPC_SKB_CB(skb
)->handle
= NULL
;
1735 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1736 skb_set_owner_r(skb
, sk
);
1738 sk
->sk_data_ready(sk
);
1742 tipc_sk_respond(sk
, skb
, err
);
1747 * tipc_backlog_rcv - handle incoming message from backlog queue
1751 * Caller must hold socket lock
1755 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1757 unsigned int truesize
= skb
->truesize
;
1759 if (likely(filter_rcv(sk
, skb
)))
1760 atomic_add(truesize
, &tipc_sk(sk
)->dupl_rcvcnt
);
1765 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1766 * inputq and try adding them to socket or backlog queue
1767 * @inputq: list of incoming buffers with potentially different destinations
1768 * @sk: socket where the buffers should be enqueued
1769 * @dport: port number for the socket
1771 * Caller must hold socket lock
1773 static void tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
1778 struct sk_buff
*skb
;
1779 unsigned long time_limit
= jiffies
+ 2;
1781 while (skb_queue_len(inputq
)) {
1782 if (unlikely(time_after_eq(jiffies
, time_limit
)))
1785 skb
= tipc_skb_dequeue(inputq
, dport
);
1789 /* Add message directly to receive queue if possible */
1790 if (!sock_owned_by_user(sk
)) {
1791 filter_rcv(sk
, skb
);
1795 /* Try backlog, compensating for double-counted bytes */
1796 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
1797 if (!sk
->sk_backlog
.len
)
1798 atomic_set(dcnt
, 0);
1799 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
1800 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
1803 /* Overload => reject message back to sender */
1804 tipc_sk_respond(sk
, skb
, TIPC_ERR_OVERLOAD
);
1810 * tipc_sk_rcv - handle a chain of incoming buffers
1811 * @inputq: buffer list containing the buffers
1812 * Consumes all buffers in list until inputq is empty
1813 * Note: may be called in multiple threads referring to the same queue
1815 void tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
1817 u32 dnode
, dport
= 0;
1819 struct tipc_sock
*tsk
;
1821 struct sk_buff
*skb
;
1823 while (skb_queue_len(inputq
)) {
1824 dport
= tipc_skb_peek_port(inputq
, dport
);
1825 tsk
= tipc_sk_lookup(net
, dport
);
1829 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
1830 tipc_sk_enqueue(inputq
, sk
, dport
);
1831 spin_unlock_bh(&sk
->sk_lock
.slock
);
1837 /* No destination socket => dequeue skb if still there */
1838 skb
= tipc_skb_dequeue(inputq
, dport
);
1842 /* Try secondary lookup if unresolved named message */
1843 err
= TIPC_ERR_NO_PORT
;
1844 if (tipc_msg_lookup_dest(net
, skb
, &err
))
1847 /* Prepare for message rejection */
1848 if (!tipc_msg_reverse(tipc_own_addr(net
), &skb
, err
))
1851 dnode
= msg_destnode(buf_msg(skb
));
1852 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
1856 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
1858 struct sock
*sk
= sock
->sk
;
1863 int err
= sock_error(sk
);
1868 if (signal_pending(current
))
1869 return sock_intr_errno(*timeo_p
);
1871 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1872 done
= sk_wait_event(sk
, timeo_p
, sock
->state
!= SS_CONNECTING
);
1873 finish_wait(sk_sleep(sk
), &wait
);
1879 * tipc_connect - establish a connection to another TIPC port
1880 * @sock: socket structure
1881 * @dest: socket address for destination port
1882 * @destlen: size of socket address data structure
1883 * @flags: file-related flags associated with socket
1885 * Returns 0 on success, errno otherwise
1887 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
1888 int destlen
, int flags
)
1890 struct sock
*sk
= sock
->sk
;
1891 struct tipc_sock
*tsk
= tipc_sk(sk
);
1892 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
1893 struct msghdr m
= {NULL
,};
1894 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
1895 socket_state previous
;
1900 /* DGRAM/RDM connect(), just save the destaddr */
1901 if (sock
->state
== SS_READY
) {
1902 if (dst
->family
== AF_UNSPEC
) {
1903 memset(&tsk
->remote
, 0, sizeof(struct sockaddr_tipc
));
1905 } else if (destlen
!= sizeof(struct sockaddr_tipc
)) {
1908 memcpy(&tsk
->remote
, dest
, destlen
);
1915 * Reject connection attempt using multicast address
1917 * Note: send_msg() validates the rest of the address fields,
1918 * so there's no need to do it here
1920 if (dst
->addrtype
== TIPC_ADDR_MCAST
) {
1925 previous
= sock
->state
;
1926 switch (sock
->state
) {
1927 case SS_UNCONNECTED
:
1928 /* Send a 'SYN-' to destination */
1930 m
.msg_namelen
= destlen
;
1932 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1933 * indicate send_msg() is never blocked.
1936 m
.msg_flags
= MSG_DONTWAIT
;
1938 res
= __tipc_sendmsg(sock
, &m
, 0);
1939 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
1942 /* Just entered SS_CONNECTING state; the only
1943 * difference is that return value in non-blocking
1944 * case is EINPROGRESS, rather than EALREADY.
1948 if (previous
== SS_CONNECTING
)
1952 timeout
= msecs_to_jiffies(timeout
);
1953 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1954 res
= tipc_wait_for_connect(sock
, &timeout
);
1969 * tipc_listen - allow socket to listen for incoming connections
1970 * @sock: socket structure
1973 * Returns 0 on success, errno otherwise
1975 static int tipc_listen(struct socket
*sock
, int len
)
1977 struct sock
*sk
= sock
->sk
;
1982 if (sock
->state
!= SS_UNCONNECTED
)
1985 sock
->state
= SS_LISTENING
;
1993 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
1995 struct sock
*sk
= sock
->sk
;
1999 /* True wake-one mechanism for incoming connections: only
2000 * one process gets woken up, not the 'whole herd'.
2001 * Since we do not 'race & poll' for established sockets
2002 * anymore, the common case will execute the loop only once.
2005 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
2006 TASK_INTERRUPTIBLE
);
2007 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
2009 timeo
= schedule_timeout(timeo
);
2013 if (!skb_queue_empty(&sk
->sk_receive_queue
))
2016 if (sock
->state
!= SS_LISTENING
)
2021 err
= sock_intr_errno(timeo
);
2022 if (signal_pending(current
))
2025 finish_wait(sk_sleep(sk
), &wait
);
2030 * tipc_accept - wait for connection request
2031 * @sock: listening socket
2032 * @newsock: new socket that is to be connected
2033 * @flags: file-related flags associated with socket
2035 * Returns 0 on success, errno otherwise
2037 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
)
2039 struct sock
*new_sk
, *sk
= sock
->sk
;
2040 struct sk_buff
*buf
;
2041 struct tipc_sock
*new_tsock
;
2042 struct tipc_msg
*msg
;
2048 if (sock
->state
!= SS_LISTENING
) {
2052 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2053 res
= tipc_wait_for_accept(sock
, timeo
);
2057 buf
= skb_peek(&sk
->sk_receive_queue
);
2059 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, 1);
2062 security_sk_clone(sock
->sk
, new_sock
->sk
);
2064 new_sk
= new_sock
->sk
;
2065 new_tsock
= tipc_sk(new_sk
);
2068 /* we lock on new_sk; but lockdep sees the lock on sk */
2069 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2072 * Reject any stray messages received by new socket
2073 * before the socket lock was taken (very, very unlikely)
2075 tsk_rej_rx_queue(new_sk
);
2077 /* Connect new socket to it's peer */
2078 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2079 new_sock
->state
= SS_CONNECTED
;
2081 tsk_set_importance(new_tsock
, msg_importance(msg
));
2082 if (msg_named(msg
)) {
2083 new_tsock
->conn_type
= msg_nametype(msg
);
2084 new_tsock
->conn_instance
= msg_nameinst(msg
);
2088 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2089 * Respond to 'SYN+' by queuing it on new socket.
2091 if (!msg_data_sz(msg
)) {
2092 struct msghdr m
= {NULL
,};
2094 tsk_advance_rx_queue(sk
);
2095 __tipc_send_stream(new_sock
, &m
, 0);
2097 __skb_dequeue(&sk
->sk_receive_queue
);
2098 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2099 skb_set_owner_r(buf
, new_sk
);
2101 release_sock(new_sk
);
2108 * tipc_shutdown - shutdown socket connection
2109 * @sock: socket structure
2110 * @how: direction to close (must be SHUT_RDWR)
2112 * Terminates connection (if necessary), then purges socket's receive queue.
2114 * Returns 0 on success, errno otherwise
2116 static int tipc_shutdown(struct socket
*sock
, int how
)
2118 struct sock
*sk
= sock
->sk
;
2119 struct net
*net
= sock_net(sk
);
2120 struct tipc_sock
*tsk
= tipc_sk(sk
);
2121 struct sk_buff
*skb
;
2122 u32 dnode
= tsk_peer_node(tsk
);
2123 u32 dport
= tsk_peer_port(tsk
);
2124 u32 onode
= tipc_own_addr(net
);
2125 u32 oport
= tsk
->portid
;
2128 if (how
!= SHUT_RDWR
)
2133 switch (sock
->state
) {
2138 dnode
= tsk_peer_node(tsk
);
2140 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2141 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
2143 if (TIPC_SKB_CB(skb
)->handle
!= NULL
) {
2147 tipc_sk_respond(sk
, skb
, TIPC_CONN_SHUTDOWN
);
2149 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
2150 TIPC_CONN_MSG
, SHORT_H_SIZE
,
2151 0, dnode
, onode
, dport
, oport
,
2152 TIPC_CONN_SHUTDOWN
);
2153 tipc_node_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
2156 sock
->state
= SS_DISCONNECTING
;
2157 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
2160 case SS_DISCONNECTING
:
2162 /* Discard any unreceived messages */
2163 __skb_queue_purge(&sk
->sk_receive_queue
);
2165 /* Wake up anyone sleeping in poll */
2166 sk
->sk_state_change(sk
);
2178 static void tipc_sk_timeout(unsigned long data
)
2180 struct tipc_sock
*tsk
= (struct tipc_sock
*)data
;
2181 struct sock
*sk
= &tsk
->sk
;
2182 struct sk_buff
*skb
= NULL
;
2183 u32 peer_port
, peer_node
;
2184 u32 own_node
= tsk_own_node(tsk
);
2187 if (!tsk
->connected
) {
2191 peer_port
= tsk_peer_port(tsk
);
2192 peer_node
= tsk_peer_node(tsk
);
2194 if (tsk
->probing_state
== TIPC_CONN_PROBING
) {
2195 if (!sock_owned_by_user(sk
)) {
2196 sk
->sk_socket
->state
= SS_DISCONNECTING
;
2198 tipc_node_remove_conn(sock_net(sk
), tsk_peer_node(tsk
),
2199 tsk_peer_port(tsk
));
2200 sk
->sk_state_change(sk
);
2202 /* Try again later */
2203 sk_reset_timer(sk
, &sk
->sk_timer
, (HZ
/ 20));
2207 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
,
2208 INT_H_SIZE
, 0, peer_node
, own_node
,
2209 peer_port
, tsk
->portid
, TIPC_OK
);
2210 tsk
->probing_state
= TIPC_CONN_PROBING
;
2211 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ tsk
->probing_intv
);
2215 tipc_node_xmit_skb(sock_net(sk
), skb
, peer_node
, tsk
->portid
);
2220 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2221 struct tipc_name_seq
const *seq
)
2223 struct net
*net
= sock_net(&tsk
->sk
);
2224 struct publication
*publ
;
2229 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2230 if (key
== tsk
->portid
)
2233 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2234 scope
, tsk
->portid
, key
);
2235 if (unlikely(!publ
))
2238 list_add(&publ
->pport_list
, &tsk
->publications
);
2244 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2245 struct tipc_name_seq
const *seq
)
2247 struct net
*net
= sock_net(&tsk
->sk
);
2248 struct publication
*publ
;
2249 struct publication
*safe
;
2252 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2254 if (publ
->scope
!= scope
)
2256 if (publ
->type
!= seq
->type
)
2258 if (publ
->lower
!= seq
->lower
)
2260 if (publ
->upper
!= seq
->upper
)
2262 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2263 publ
->ref
, publ
->key
);
2267 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2268 publ
->ref
, publ
->key
);
2271 if (list_empty(&tsk
->publications
))
2276 /* tipc_sk_reinit: set non-zero address in all existing sockets
2277 * when we go from standalone to network mode.
2279 void tipc_sk_reinit(struct net
*net
)
2281 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2282 const struct bucket_table
*tbl
;
2283 struct rhash_head
*pos
;
2284 struct tipc_sock
*tsk
;
2285 struct tipc_msg
*msg
;
2289 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2290 for (i
= 0; i
< tbl
->size
; i
++) {
2291 rht_for_each_entry_rcu(tsk
, pos
, tbl
, i
, node
) {
2292 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2294 msg_set_prevnode(msg
, tn
->own_addr
);
2295 msg_set_orignode(msg
, tn
->own_addr
);
2296 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2302 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2304 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2305 struct tipc_sock
*tsk
;
2308 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2310 sock_hold(&tsk
->sk
);
2316 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2318 struct sock
*sk
= &tsk
->sk
;
2319 struct net
*net
= sock_net(sk
);
2320 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2321 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2322 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2324 while (remaining
--) {
2326 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2327 portid
= TIPC_MIN_PORT
;
2328 tsk
->portid
= portid
;
2329 sock_hold(&tsk
->sk
);
2330 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2339 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2341 struct sock
*sk
= &tsk
->sk
;
2342 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2344 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2345 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
2350 static const struct rhashtable_params tsk_rht_params
= {
2352 .head_offset
= offsetof(struct tipc_sock
, node
),
2353 .key_offset
= offsetof(struct tipc_sock
, portid
),
2354 .key_len
= sizeof(u32
), /* portid */
2355 .max_size
= 1048576,
2357 .automatic_shrinking
= true,
2360 int tipc_sk_rht_init(struct net
*net
)
2362 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2364 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2367 void tipc_sk_rht_destroy(struct net
*net
)
2369 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2371 /* Wait for socket readers to complete */
2374 rhashtable_destroy(&tn
->sk_rht
);
2378 * tipc_setsockopt - set socket option
2379 * @sock: socket structure
2380 * @lvl: option level
2381 * @opt: option identifier
2382 * @ov: pointer to new option value
2383 * @ol: length of option value
2385 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2386 * (to ease compatibility).
2388 * Returns 0 on success, errno otherwise
2390 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2391 char __user
*ov
, unsigned int ol
)
2393 struct sock
*sk
= sock
->sk
;
2394 struct tipc_sock
*tsk
= tipc_sk(sk
);
2398 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2400 if (lvl
!= SOL_TIPC
)
2401 return -ENOPROTOOPT
;
2402 if (ol
< sizeof(value
))
2404 res
= get_user(value
, (u32 __user
*)ov
);
2411 case TIPC_IMPORTANCE
:
2412 res
= tsk_set_importance(tsk
, value
);
2414 case TIPC_SRC_DROPPABLE
:
2415 if (sock
->type
!= SOCK_STREAM
)
2416 tsk_set_unreliable(tsk
, value
);
2420 case TIPC_DEST_DROPPABLE
:
2421 tsk_set_unreturnable(tsk
, value
);
2423 case TIPC_CONN_TIMEOUT
:
2424 tipc_sk(sk
)->conn_timeout
= value
;
2425 /* no need to set "res", since already 0 at this point */
2437 * tipc_getsockopt - get socket option
2438 * @sock: socket structure
2439 * @lvl: option level
2440 * @opt: option identifier
2441 * @ov: receptacle for option value
2442 * @ol: receptacle for length of option value
2444 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2445 * (to ease compatibility).
2447 * Returns 0 on success, errno otherwise
2449 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2450 char __user
*ov
, int __user
*ol
)
2452 struct sock
*sk
= sock
->sk
;
2453 struct tipc_sock
*tsk
= tipc_sk(sk
);
2458 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2459 return put_user(0, ol
);
2460 if (lvl
!= SOL_TIPC
)
2461 return -ENOPROTOOPT
;
2462 res
= get_user(len
, ol
);
2469 case TIPC_IMPORTANCE
:
2470 value
= tsk_importance(tsk
);
2472 case TIPC_SRC_DROPPABLE
:
2473 value
= tsk_unreliable(tsk
);
2475 case TIPC_DEST_DROPPABLE
:
2476 value
= tsk_unreturnable(tsk
);
2478 case TIPC_CONN_TIMEOUT
:
2479 value
= tsk
->conn_timeout
;
2480 /* no need to set "res", since already 0 at this point */
2482 case TIPC_NODE_RECVQ_DEPTH
:
2483 value
= 0; /* was tipc_queue_size, now obsolete */
2485 case TIPC_SOCK_RECVQ_DEPTH
:
2486 value
= skb_queue_len(&sk
->sk_receive_queue
);
2495 return res
; /* "get" failed */
2497 if (len
< sizeof(value
))
2500 if (copy_to_user(ov
, &value
, sizeof(value
)))
2503 return put_user(sizeof(value
), ol
);
2506 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2508 struct sock
*sk
= sock
->sk
;
2509 struct tipc_sioc_ln_req lnr
;
2510 void __user
*argp
= (void __user
*)arg
;
2513 case SIOCGETLINKNAME
:
2514 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2516 if (!tipc_node_get_linkname(sock_net(sk
),
2517 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2518 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2519 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2523 return -EADDRNOTAVAIL
;
2525 return -ENOIOCTLCMD
;
2529 /* Protocol switches for the various types of TIPC sockets */
2531 static const struct proto_ops msg_ops
= {
2532 .owner
= THIS_MODULE
,
2534 .release
= tipc_release
,
2536 .connect
= tipc_connect
,
2537 .socketpair
= sock_no_socketpair
,
2538 .accept
= sock_no_accept
,
2539 .getname
= tipc_getname
,
2541 .ioctl
= tipc_ioctl
,
2542 .listen
= sock_no_listen
,
2543 .shutdown
= tipc_shutdown
,
2544 .setsockopt
= tipc_setsockopt
,
2545 .getsockopt
= tipc_getsockopt
,
2546 .sendmsg
= tipc_sendmsg
,
2547 .recvmsg
= tipc_recvmsg
,
2548 .mmap
= sock_no_mmap
,
2549 .sendpage
= sock_no_sendpage
2552 static const struct proto_ops packet_ops
= {
2553 .owner
= THIS_MODULE
,
2555 .release
= tipc_release
,
2557 .connect
= tipc_connect
,
2558 .socketpair
= sock_no_socketpair
,
2559 .accept
= tipc_accept
,
2560 .getname
= tipc_getname
,
2562 .ioctl
= tipc_ioctl
,
2563 .listen
= tipc_listen
,
2564 .shutdown
= tipc_shutdown
,
2565 .setsockopt
= tipc_setsockopt
,
2566 .getsockopt
= tipc_getsockopt
,
2567 .sendmsg
= tipc_send_packet
,
2568 .recvmsg
= tipc_recvmsg
,
2569 .mmap
= sock_no_mmap
,
2570 .sendpage
= sock_no_sendpage
2573 static const struct proto_ops stream_ops
= {
2574 .owner
= THIS_MODULE
,
2576 .release
= tipc_release
,
2578 .connect
= tipc_connect
,
2579 .socketpair
= sock_no_socketpair
,
2580 .accept
= tipc_accept
,
2581 .getname
= tipc_getname
,
2583 .ioctl
= tipc_ioctl
,
2584 .listen
= tipc_listen
,
2585 .shutdown
= tipc_shutdown
,
2586 .setsockopt
= tipc_setsockopt
,
2587 .getsockopt
= tipc_getsockopt
,
2588 .sendmsg
= tipc_send_stream
,
2589 .recvmsg
= tipc_recv_stream
,
2590 .mmap
= sock_no_mmap
,
2591 .sendpage
= sock_no_sendpage
2594 static const struct net_proto_family tipc_family_ops
= {
2595 .owner
= THIS_MODULE
,
2597 .create
= tipc_sk_create
2600 static struct proto tipc_proto
= {
2602 .owner
= THIS_MODULE
,
2603 .obj_size
= sizeof(struct tipc_sock
),
2604 .sysctl_rmem
= sysctl_tipc_rmem
2608 * tipc_socket_init - initialize TIPC socket interface
2610 * Returns 0 on success, errno otherwise
2612 int tipc_socket_init(void)
2616 res
= proto_register(&tipc_proto
, 1);
2618 pr_err("Failed to register TIPC protocol type\n");
2622 res
= sock_register(&tipc_family_ops
);
2624 pr_err("Failed to register TIPC socket type\n");
2625 proto_unregister(&tipc_proto
);
2633 * tipc_socket_stop - stop TIPC socket interface
2635 void tipc_socket_stop(void)
2637 sock_unregister(tipc_family_ops
.family
);
2638 proto_unregister(&tipc_proto
);
2641 /* Caller should hold socket lock for the passed tipc socket. */
2642 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
2646 struct nlattr
*nest
;
2648 peer_node
= tsk_peer_node(tsk
);
2649 peer_port
= tsk_peer_port(tsk
);
2651 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
2653 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
2655 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
2658 if (tsk
->conn_type
!= 0) {
2659 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
2661 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
2663 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
2666 nla_nest_end(skb
, nest
);
2671 nla_nest_cancel(skb
, nest
);
2676 /* Caller should hold socket lock for the passed tipc socket. */
2677 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
2678 struct tipc_sock
*tsk
)
2682 struct nlattr
*attrs
;
2683 struct net
*net
= sock_net(skb
->sk
);
2684 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2686 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2687 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
2691 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
2693 goto genlmsg_cancel
;
2694 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
2695 goto attr_msg_cancel
;
2696 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tn
->own_addr
))
2697 goto attr_msg_cancel
;
2699 if (tsk
->connected
) {
2700 err
= __tipc_nl_add_sk_con(skb
, tsk
);
2702 goto attr_msg_cancel
;
2703 } else if (!list_empty(&tsk
->publications
)) {
2704 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
2705 goto attr_msg_cancel
;
2707 nla_nest_end(skb
, attrs
);
2708 genlmsg_end(skb
, hdr
);
2713 nla_nest_cancel(skb
, attrs
);
2715 genlmsg_cancel(skb
, hdr
);
2720 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2723 struct tipc_sock
*tsk
;
2724 const struct bucket_table
*tbl
;
2725 struct rhash_head
*pos
;
2726 struct net
*net
= sock_net(skb
->sk
);
2727 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2728 u32 tbl_id
= cb
->args
[0];
2729 u32 prev_portid
= cb
->args
[1];
2732 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2733 for (; tbl_id
< tbl
->size
; tbl_id
++) {
2734 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
2735 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2736 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
2737 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2741 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
2743 prev_portid
= tsk
->portid
;
2744 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2748 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2753 cb
->args
[0] = tbl_id
;
2754 cb
->args
[1] = prev_portid
;
2759 /* Caller should hold socket lock for the passed tipc socket. */
2760 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
2761 struct netlink_callback
*cb
,
2762 struct publication
*publ
)
2765 struct nlattr
*attrs
;
2767 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2768 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
2772 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
2774 goto genlmsg_cancel
;
2776 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
2777 goto attr_msg_cancel
;
2778 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
2779 goto attr_msg_cancel
;
2780 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
2781 goto attr_msg_cancel
;
2782 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
2783 goto attr_msg_cancel
;
2785 nla_nest_end(skb
, attrs
);
2786 genlmsg_end(skb
, hdr
);
2791 nla_nest_cancel(skb
, attrs
);
2793 genlmsg_cancel(skb
, hdr
);
2798 /* Caller should hold socket lock for the passed tipc socket. */
2799 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
2800 struct netlink_callback
*cb
,
2801 struct tipc_sock
*tsk
, u32
*last_publ
)
2804 struct publication
*p
;
2807 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
2808 if (p
->key
== *last_publ
)
2811 if (p
->key
!= *last_publ
) {
2812 /* We never set seq or call nl_dump_check_consistent()
2813 * this means that setting prev_seq here will cause the
2814 * consistence check to fail in the netlink callback
2815 * handler. Resulting in the last NLMSG_DONE message
2816 * having the NLM_F_DUMP_INTR flag set.
2823 p
= list_first_entry(&tsk
->publications
, struct publication
,
2827 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
2828 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
2830 *last_publ
= p
->key
;
2839 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2842 u32 tsk_portid
= cb
->args
[0];
2843 u32 last_publ
= cb
->args
[1];
2844 u32 done
= cb
->args
[2];
2845 struct net
*net
= sock_net(skb
->sk
);
2846 struct tipc_sock
*tsk
;
2849 struct nlattr
**attrs
;
2850 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
2852 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
2856 if (!attrs
[TIPC_NLA_SOCK
])
2859 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
2860 attrs
[TIPC_NLA_SOCK
],
2861 tipc_nl_sock_policy
);
2865 if (!sock
[TIPC_NLA_SOCK_REF
])
2868 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
2874 tsk
= tipc_sk_lookup(net
, tsk_portid
);
2878 lock_sock(&tsk
->sk
);
2879 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
2882 release_sock(&tsk
->sk
);
2885 cb
->args
[0] = tsk_portid
;
2886 cb
->args
[1] = last_publ
;