2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2014, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
39 #include "name_table.h"
42 #include <linux/export.h>
45 #define SS_LISTENING -1 /* socket is listening */
46 #define SS_READY -2 /* socket is connectionless */
48 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
50 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
51 static void tipc_data_ready(struct sock *sk);
52 static void tipc_write_space(struct sock *sk);
53 static int tipc_release(struct socket *sock);
54 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
56 static const struct proto_ops packet_ops;
57 static const struct proto_ops stream_ops;
58 static const struct proto_ops msg_ops;
60 static struct proto tipc_proto;
61 static struct proto tipc_proto_kern;
64 * Revised TIPC socket locking policy:
66 * Most socket operations take the standard socket lock when they start
67 * and hold it until they finish (or until they need to sleep). Acquiring
68 * this lock grants the owner exclusive access to the fields of the socket
69 * data structures, with the exception of the backlog queue. A few socket
70 * operations can be done without taking the socket lock because they only
71 * read socket information that never changes during the life of the socket.
73 * Socket operations may acquire the lock for the associated TIPC port if they
74 * need to perform an operation on the port. If any routine needs to acquire
75 * both the socket lock and the port lock it must take the socket lock first
76 * to avoid the risk of deadlock.
78 * The dispatcher handling incoming messages cannot grab the socket lock in
79 * the standard fashion, since invoked it runs at the BH level and cannot block.
80 * Instead, it checks to see if the socket lock is currently owned by someone,
81 * and either handles the message itself or adds it to the socket's backlog
82 * queue; in the latter case the queued message is processed once the process
83 * owning the socket lock releases it.
85 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
86 * the problem of a blocked socket operation preventing any other operations
87 * from occurring. However, applications must be careful if they have
88 * multiple threads trying to send (or receive) on the same socket, as these
89 * operations might interfere with each other. For example, doing a connect
90 * and a receive at the same time might allow the receive to consume the
91 * ACK message meant for the connect. While additional work could be done
92 * to try and overcome this, it doesn't seem to be worthwhile at the present.
94 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
95 * that another operation that must be performed in a non-blocking manner is
96 * not delayed for very long because the lock has already been taken.
98 * NOTE: This code assumes that certain fields of a port/socket pair are
99 * constant over its lifetime; such fields can be examined without taking
100 * the socket lock and/or port lock, and do not need to be re-read even
101 * after resuming processing after waiting. These fields include:
103 * - pointer to socket sk structure (aka tipc_sock structure)
104 * - pointer to port structure
111 * advance_rx_queue - discard first buffer in socket receive queue
113 * Caller must hold socket lock
115 static void advance_rx_queue(struct sock *sk)
117 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
121 * reject_rx_queue - reject all buffers in socket receive queue
123 * Caller must hold socket lock
125 static void reject_rx_queue(struct sock *sk)
130 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
131 if (tipc_msg_reverse(buf, &dnode, TIPC_ERR_NO_PORT))
132 tipc_link_xmit2(buf, dnode, 0);
137 * tipc_sk_create - create a TIPC socket
138 * @net: network namespace (must be default network)
139 * @sock: pre-allocated socket structure
140 * @protocol: protocol indicator (must be 0)
141 * @kern: caused by kernel or by userspace?
143 * This routine creates additional data structures used by the TIPC socket,
144 * initializes them, and links them together.
146 * Returns 0 on success, errno otherwise
148 static int tipc_sk_create(struct net *net, struct socket *sock,
149 int protocol, int kern)
151 const struct proto_ops *ops;
154 struct tipc_sock *tsk;
155 struct tipc_port *port;
158 /* Validate arguments */
159 if (unlikely(protocol != 0))
160 return -EPROTONOSUPPORT;
162 switch (sock->type) {
165 state = SS_UNCONNECTED;
169 state = SS_UNCONNECTED;
180 /* Allocate socket's protocol area */
182 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
184 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
192 ref = tipc_port_init(port, TIPC_LOW_IMPORTANCE);
194 pr_warn("Socket registration failed, ref. table exhausted\n");
199 /* Finish initializing socket data structures */
203 sock_init_data(sock, sk);
204 sk->sk_backlog_rcv = tipc_backlog_rcv;
205 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
206 sk->sk_data_ready = tipc_data_ready;
207 sk->sk_write_space = tipc_write_space;
208 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
210 atomic_set(&tsk->dupl_rcvcnt, 0);
211 tipc_port_unlock(port);
213 if (sock->state == SS_READY) {
214 tipc_port_set_unreturnable(port, true);
215 if (sock->type == SOCK_DGRAM)
216 tipc_port_set_unreliable(port, true);
222 * tipc_sock_create_local - create TIPC socket from inside TIPC module
223 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
225 * We cannot use sock_creat_kern here because it bumps module user count.
226 * Since socket owner and creator is the same module we must make sure
227 * that module count remains zero for module local sockets, otherwise
228 * we cannot do rmmod.
230 * Returns 0 on success, errno otherwise
232 int tipc_sock_create_local(int type, struct socket **res)
236 rc = sock_create_lite(AF_TIPC, type, 0, res);
238 pr_err("Failed to create kernel socket\n");
241 tipc_sk_create(&init_net, *res, 0, 1);
247 * tipc_sock_release_local - release socket created by tipc_sock_create_local
248 * @sock: the socket to be released.
250 * Module reference count is not incremented when such sockets are created,
251 * so we must keep it from being decremented when they are released.
253 void tipc_sock_release_local(struct socket *sock)
261 * tipc_sock_accept_local - accept a connection on a socket created
262 * with tipc_sock_create_local. Use this function to avoid that
263 * module reference count is inadvertently incremented.
265 * @sock: the accepting socket
266 * @newsock: reference to the new socket to be created
267 * @flags: socket flags
270 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
273 struct sock *sk = sock->sk;
276 ret = sock_create_lite(sk->sk_family, sk->sk_type,
277 sk->sk_protocol, newsock);
281 ret = tipc_accept(sock, *newsock, flags);
283 sock_release(*newsock);
286 (*newsock)->ops = sock->ops;
291 * tipc_release - destroy a TIPC socket
292 * @sock: socket to destroy
294 * This routine cleans up any messages that are still queued on the socket.
295 * For DGRAM and RDM socket types, all queued messages are rejected.
296 * For SEQPACKET and STREAM socket types, the first message is rejected
297 * and any others are discarded. (If the first message on a STREAM socket
298 * is partially-read, it is discarded and the next one is rejected instead.)
300 * NOTE: Rejected messages are not necessarily returned to the sender! They
301 * are returned or discarded according to the "destination droppable" setting
302 * specified for the message by the sender.
304 * Returns 0 on success, errno otherwise
306 static int tipc_release(struct socket *sock)
308 struct sock *sk = sock->sk;
309 struct tipc_sock *tsk;
310 struct tipc_port *port;
315 * Exit if socket isn't fully initialized (occurs when a failed accept()
316 * releases a pre-allocated child socket that was never used)
326 * Reject all unreceived messages, except on an active connection
327 * (which disconnects locally & sends a 'FIN+' to peer)
329 while (sock->state != SS_DISCONNECTING) {
330 buf = __skb_dequeue(&sk->sk_receive_queue);
333 if (TIPC_SKB_CB(buf)->handle != NULL)
336 if ((sock->state == SS_CONNECTING) ||
337 (sock->state == SS_CONNECTED)) {
338 sock->state = SS_DISCONNECTING;
339 tipc_port_disconnect(port->ref);
341 if (tipc_msg_reverse(buf, &dnode, TIPC_ERR_NO_PORT))
342 tipc_link_xmit2(buf, dnode, 0);
346 /* Destroy TIPC port; also disconnects an active connection and
347 * sends a 'FIN-' to peer.
349 tipc_port_destroy(port);
351 /* Discard any remaining (connection-based) messages in receive queue */
352 __skb_queue_purge(&sk->sk_receive_queue);
354 /* Reject any messages that accumulated in backlog queue */
355 sock->state = SS_DISCONNECTING;
365 * tipc_bind - associate or disassocate TIPC name(s) with a socket
366 * @sock: socket structure
367 * @uaddr: socket address describing name(s) and desired operation
368 * @uaddr_len: size of socket address data structure
370 * Name and name sequence binding is indicated using a positive scope value;
371 * a negative scope value unbinds the specified name. Specifying no name
372 * (i.e. a socket address length of 0) unbinds all names from the socket.
374 * Returns 0 on success, errno otherwise
376 * NOTE: This routine doesn't need to take the socket lock since it doesn't
377 * access any non-constant socket information.
379 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
382 struct sock *sk = sock->sk;
383 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
384 struct tipc_sock *tsk = tipc_sk(sk);
388 if (unlikely(!uaddr_len)) {
389 res = tipc_withdraw(&tsk->port, 0, NULL);
393 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
397 if (addr->family != AF_TIPC) {
402 if (addr->addrtype == TIPC_ADDR_NAME)
403 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
404 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
409 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
410 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
411 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
416 res = (addr->scope > 0) ?
417 tipc_publish(&tsk->port, addr->scope, &addr->addr.nameseq) :
418 tipc_withdraw(&tsk->port, -addr->scope, &addr->addr.nameseq);
425 * tipc_getname - get port ID of socket or peer socket
426 * @sock: socket structure
427 * @uaddr: area for returned socket address
428 * @uaddr_len: area for returned length of socket address
429 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
431 * Returns 0 on success, errno otherwise
433 * NOTE: This routine doesn't need to take the socket lock since it only
434 * accesses socket information that is unchanging (or which changes in
435 * a completely predictable manner).
437 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
438 int *uaddr_len, int peer)
440 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
441 struct tipc_sock *tsk = tipc_sk(sock->sk);
443 memset(addr, 0, sizeof(*addr));
445 if ((sock->state != SS_CONNECTED) &&
446 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
448 addr->addr.id.ref = tipc_port_peerport(&tsk->port);
449 addr->addr.id.node = tipc_port_peernode(&tsk->port);
451 addr->addr.id.ref = tsk->port.ref;
452 addr->addr.id.node = tipc_own_addr;
455 *uaddr_len = sizeof(*addr);
456 addr->addrtype = TIPC_ADDR_ID;
457 addr->family = AF_TIPC;
459 addr->addr.name.domain = 0;
465 * tipc_poll - read and possibly block on pollmask
466 * @file: file structure associated with the socket
467 * @sock: socket for which to calculate the poll bits
470 * Returns pollmask value
473 * It appears that the usual socket locking mechanisms are not useful here
474 * since the pollmask info is potentially out-of-date the moment this routine
475 * exits. TCP and other protocols seem to rely on higher level poll routines
476 * to handle any preventable race conditions, so TIPC will do the same ...
478 * TIPC sets the returned events as follows:
480 * socket state flags set
481 * ------------ ---------
482 * unconnected no read flags
483 * POLLOUT if port is not congested
485 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
488 * connected POLLIN/POLLRDNORM if data in rx queue
489 * POLLOUT if port is not congested
491 * disconnecting POLLIN/POLLRDNORM/POLLHUP
494 * listening POLLIN if SYN in rx queue
497 * ready POLLIN/POLLRDNORM if data in rx queue
498 * [connectionless] POLLOUT (since port cannot be congested)
500 * IMPORTANT: The fact that a read or write operation is indicated does NOT
501 * imply that the operation will succeed, merely that it should be performed
502 * and will not block.
504 static unsigned int tipc_poll(struct file *file, struct socket *sock,
507 struct sock *sk = sock->sk;
508 struct tipc_sock *tsk = tipc_sk(sk);
511 sock_poll_wait(file, sk_sleep(sk), wait);
513 switch ((int)sock->state) {
515 if (!tsk->port.congested)
520 if (!tsk->port.congested)
525 if (!skb_queue_empty(&sk->sk_receive_queue))
526 mask |= (POLLIN | POLLRDNORM);
528 case SS_DISCONNECTING:
529 mask = (POLLIN | POLLRDNORM | POLLHUP);
537 * dest_name_check - verify user is permitted to send to specified port name
538 * @dest: destination address
539 * @m: descriptor for message to be sent
541 * Prevents restricted configuration commands from being issued by
542 * unauthorized users.
544 * Returns 0 if permission is granted, otherwise errno
546 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
548 struct tipc_cfg_msg_hdr hdr;
550 if (unlikely(dest->addrtype == TIPC_ADDR_ID))
552 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
554 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
556 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
559 if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
561 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
563 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
569 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
571 struct sock *sk = sock->sk;
572 struct tipc_sock *tsk = tipc_sk(sk);
577 int err = sock_error(sk);
580 if (sock->state == SS_DISCONNECTING)
584 if (signal_pending(current))
585 return sock_intr_errno(*timeo_p);
587 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
588 done = sk_wait_event(sk, timeo_p, !tsk->port.congested);
589 finish_wait(sk_sleep(sk), &wait);
595 * tipc_sendmcast - send multicast message
596 * @sock: socket structure
597 * @seq: destination address
598 * @iov: message data to send
599 * @dsz: total length of message data
600 * @timeo: timeout to wait for wakeup
602 * Called from function tipc_sendmsg(), which has done all sanity checks
603 * Returns the number of bytes sent on success, or errno
605 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
606 struct iovec *iov, size_t dsz, long timeo)
608 struct sock *sk = sock->sk;
609 struct tipc_sock *tsk = tipc_sk(sk);
613 if (sock->state != SS_READY) {
617 rc = tipc_port_mcast_xmit(&tsk->port, seq, iov, dsz);
618 if (likely(rc >= 0)) {
619 if (sock->state != SS_READY)
620 sock->state = SS_CONNECTING;
623 if (rc != -ELINKCONG)
625 rc = tipc_wait_for_sndmsg(sock, &timeo);
632 * tipc_sendmsg - send message in connectionless manner
633 * @iocb: if NULL, indicates that socket lock is already held
634 * @sock: socket structure
635 * @m: message to send
636 * @dsz: amount of user data to be sent
638 * Message must have an destination specified explicitly.
639 * Used for SOCK_RDM and SOCK_DGRAM messages,
640 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
641 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
643 * Returns the number of bytes sent on success, or errno otherwise
645 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
646 struct msghdr *m, size_t dsz)
648 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
649 struct sock *sk = sock->sk;
650 struct tipc_sock *tsk = tipc_sk(sk);
651 struct tipc_port *port = &tsk->port;
652 struct tipc_msg *mhdr = &port->phdr;
653 struct iovec *iov = m->msg_iov;
656 struct tipc_name_seq *seq = &dest->addr.nameseq;
662 return -EDESTADDRREQ;
664 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
665 (dest->family != AF_TIPC)))
668 if (dsz > TIPC_MAX_USER_MSG_SIZE)
674 if (unlikely(sock->state != SS_READY)) {
675 if (sock->state == SS_LISTENING) {
679 if (sock->state != SS_UNCONNECTED) {
683 if (tsk->port.published) {
687 if (dest->addrtype == TIPC_ADDR_NAME) {
688 tsk->port.conn_type = dest->addr.name.name.type;
689 tsk->port.conn_instance = dest->addr.name.name.instance;
692 rc = dest_name_check(dest, m);
696 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
698 if (dest->addrtype == TIPC_ADDR_MCAST) {
699 rc = tipc_sendmcast(sock, seq, iov, dsz, timeo);
701 } else if (dest->addrtype == TIPC_ADDR_NAME) {
702 u32 type = dest->addr.name.name.type;
703 u32 inst = dest->addr.name.name.instance;
704 u32 domain = dest->addr.name.domain;
707 msg_set_type(mhdr, TIPC_NAMED_MSG);
708 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
709 msg_set_nametype(mhdr, type);
710 msg_set_nameinst(mhdr, inst);
711 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
712 dport = tipc_nametbl_translate(type, inst, &dnode);
713 msg_set_destnode(mhdr, dnode);
714 msg_set_destport(mhdr, dport);
715 if (unlikely(!dport && !dnode)) {
719 } else if (dest->addrtype == TIPC_ADDR_ID) {
720 dnode = dest->addr.id.node;
721 msg_set_type(mhdr, TIPC_DIRECT_MSG);
722 msg_set_lookup_scope(mhdr, 0);
723 msg_set_destnode(mhdr, dnode);
724 msg_set_destport(mhdr, dest->addr.id.ref);
725 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
729 mtu = tipc_node_get_mtu(dnode, tsk->port.ref);
730 rc = tipc_msg_build2(mhdr, iov, 0, dsz, mtu, &buf);
735 rc = tipc_link_xmit2(buf, dnode, tsk->port.ref);
736 if (likely(rc >= 0)) {
737 if (sock->state != SS_READY)
738 sock->state = SS_CONNECTING;
745 if (rc != -ELINKCONG)
748 rc = tipc_wait_for_sndmsg(sock, &timeo);
758 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
760 struct sock *sk = sock->sk;
761 struct tipc_sock *tsk = tipc_sk(sk);
762 struct tipc_port *port = &tsk->port;
767 int err = sock_error(sk);
770 if (sock->state == SS_DISCONNECTING)
772 else if (sock->state != SS_CONNECTED)
776 if (signal_pending(current))
777 return sock_intr_errno(*timeo_p);
779 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
780 done = sk_wait_event(sk, timeo_p,
781 (!port->congested || !port->connected));
782 finish_wait(sk_sleep(sk), &wait);
788 * tipc_send_stream - send stream-oriented data
790 * @sock: socket structure
792 * @dsz: total length of data to be transmitted
794 * Used for SOCK_STREAM data.
796 * Returns the number of bytes sent on success (or partial success),
797 * or errno if no data sent
799 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
800 struct msghdr *m, size_t dsz)
802 struct sock *sk = sock->sk;
803 struct tipc_sock *tsk = tipc_sk(sk);
804 struct tipc_port *port = &tsk->port;
805 struct tipc_msg *mhdr = &port->phdr;
807 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
812 uint mtu, send, sent = 0;
814 /* Handle implied connection establishment */
815 if (unlikely(dest)) {
816 rc = tipc_sendmsg(iocb, sock, m, dsz);
817 if (dsz && (dsz == rc))
821 if (dsz > (uint)INT_MAX)
827 if (unlikely(sock->state != SS_CONNECTED)) {
828 if (sock->state == SS_DISCONNECTING)
835 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
836 dnode = tipc_port_peernode(port);
841 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
842 rc = tipc_msg_build2(mhdr, m->msg_iov, sent, send, mtu, &buf);
843 if (unlikely(rc < 0))
847 if (likely(!tipc_port_congested(port))) {
848 rc = tipc_link_xmit2(buf, dnode, ref);
856 if (rc == -EMSGSIZE) {
857 port->max_pkt = tipc_node_get_mtu(dnode, ref);
860 if (rc != -ELINKCONG)
863 rc = tipc_wait_for_sndpkt(sock, &timeo);
870 return sent ? sent : rc;
874 * tipc_send_packet - send a connection-oriented message
875 * @iocb: if NULL, indicates that socket lock is already held
876 * @sock: socket structure
877 * @m: message to send
878 * @dsz: length of data to be transmitted
880 * Used for SOCK_SEQPACKET messages.
882 * Returns the number of bytes sent on success, or errno otherwise
884 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
885 struct msghdr *m, size_t dsz)
887 if (dsz > TIPC_MAX_USER_MSG_SIZE)
890 return tipc_send_stream(iocb, sock, m, dsz);
894 * auto_connect - complete connection setup to a remote port
895 * @tsk: tipc socket structure
896 * @msg: peer's response message
898 * Returns 0 on success, errno otherwise
900 static int auto_connect(struct tipc_sock *tsk, struct tipc_msg *msg)
902 struct tipc_port *port = &tsk->port;
903 struct socket *sock = tsk->sk.sk_socket;
904 struct tipc_portid peer;
906 peer.ref = msg_origport(msg);
907 peer.node = msg_orignode(msg);
909 __tipc_port_connect(port->ref, port, &peer);
911 if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)
913 msg_set_importance(&port->phdr, (u32)msg_importance(msg));
914 sock->state = SS_CONNECTED;
919 * set_orig_addr - capture sender's address for received message
920 * @m: descriptor for message info
921 * @msg: received message header
923 * Note: Address is not captured if not requested by receiver.
925 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
927 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
930 addr->family = AF_TIPC;
931 addr->addrtype = TIPC_ADDR_ID;
932 memset(&addr->addr, 0, sizeof(addr->addr));
933 addr->addr.id.ref = msg_origport(msg);
934 addr->addr.id.node = msg_orignode(msg);
935 addr->addr.name.domain = 0; /* could leave uninitialized */
936 addr->scope = 0; /* could leave uninitialized */
937 m->msg_namelen = sizeof(struct sockaddr_tipc);
942 * anc_data_recv - optionally capture ancillary data for received message
943 * @m: descriptor for message info
944 * @msg: received message header
945 * @tport: TIPC port associated with message
947 * Note: Ancillary data is not captured if not requested by receiver.
949 * Returns 0 if successful, otherwise errno
951 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
952 struct tipc_port *tport)
960 if (likely(m->msg_controllen == 0))
963 /* Optionally capture errored message object(s) */
964 err = msg ? msg_errcode(msg) : 0;
967 anc_data[1] = msg_data_sz(msg);
968 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
972 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
979 /* Optionally capture message destination object */
980 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
984 anc_data[0] = msg_nametype(msg);
985 anc_data[1] = msg_namelower(msg);
986 anc_data[2] = msg_namelower(msg);
990 anc_data[0] = msg_nametype(msg);
991 anc_data[1] = msg_namelower(msg);
992 anc_data[2] = msg_nameupper(msg);
995 has_name = (tport->conn_type != 0);
996 anc_data[0] = tport->conn_type;
997 anc_data[1] = tport->conn_instance;
998 anc_data[2] = tport->conn_instance;
1004 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1012 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1014 struct sock *sk = sock->sk;
1016 long timeo = *timeop;
1020 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1021 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1022 if (sock->state == SS_DISCONNECTING) {
1027 timeo = schedule_timeout(timeo);
1031 if (!skb_queue_empty(&sk->sk_receive_queue))
1033 err = sock_intr_errno(timeo);
1034 if (signal_pending(current))
1040 finish_wait(sk_sleep(sk), &wait);
1046 * tipc_recvmsg - receive packet-oriented message
1048 * @m: descriptor for message info
1049 * @buf_len: total size of user buffer area
1050 * @flags: receive flags
1052 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1053 * If the complete message doesn't fit in user area, truncate it.
1055 * Returns size of returned message data, errno otherwise
1057 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1058 struct msghdr *m, size_t buf_len, int flags)
1060 struct sock *sk = sock->sk;
1061 struct tipc_sock *tsk = tipc_sk(sk);
1062 struct tipc_port *port = &tsk->port;
1063 struct sk_buff *buf;
1064 struct tipc_msg *msg;
1070 /* Catch invalid receive requests */
1071 if (unlikely(!buf_len))
1076 if (unlikely(sock->state == SS_UNCONNECTED)) {
1081 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1084 /* Look for a message in receive queue; wait if necessary */
1085 res = tipc_wait_for_rcvmsg(sock, &timeo);
1089 /* Look at first message in receive queue */
1090 buf = skb_peek(&sk->sk_receive_queue);
1092 sz = msg_data_sz(msg);
1093 err = msg_errcode(msg);
1095 /* Discard an empty non-errored message & try again */
1096 if ((!sz) && (!err)) {
1097 advance_rx_queue(sk);
1101 /* Capture sender's address (optional) */
1102 set_orig_addr(m, msg);
1104 /* Capture ancillary data (optional) */
1105 res = anc_data_recv(m, msg, port);
1109 /* Capture message data (if valid) & compute return value (always) */
1111 if (unlikely(buf_len < sz)) {
1113 m->msg_flags |= MSG_TRUNC;
1115 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1121 if ((sock->state == SS_READY) ||
1122 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1128 /* Consume received message (optional) */
1129 if (likely(!(flags & MSG_PEEK))) {
1130 if ((sock->state != SS_READY) &&
1131 (++port->conn_unacked >= TIPC_CONNACK_INTV))
1132 tipc_acknowledge(port->ref, port->conn_unacked);
1133 advance_rx_queue(sk);
1141 * tipc_recv_stream - receive stream-oriented data
1143 * @m: descriptor for message info
1144 * @buf_len: total size of user buffer area
1145 * @flags: receive flags
1147 * Used for SOCK_STREAM messages only. If not enough data is available
1148 * will optionally wait for more; never truncates data.
1150 * Returns size of returned message data, errno otherwise
1152 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1153 struct msghdr *m, size_t buf_len, int flags)
1155 struct sock *sk = sock->sk;
1156 struct tipc_sock *tsk = tipc_sk(sk);
1157 struct tipc_port *port = &tsk->port;
1158 struct sk_buff *buf;
1159 struct tipc_msg *msg;
1162 int sz_to_copy, target, needed;
1167 /* Catch invalid receive attempts */
1168 if (unlikely(!buf_len))
1173 if (unlikely(sock->state == SS_UNCONNECTED)) {
1178 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1179 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1182 /* Look for a message in receive queue; wait if necessary */
1183 res = tipc_wait_for_rcvmsg(sock, &timeo);
1187 /* Look at first message in receive queue */
1188 buf = skb_peek(&sk->sk_receive_queue);
1190 sz = msg_data_sz(msg);
1191 err = msg_errcode(msg);
1193 /* Discard an empty non-errored message & try again */
1194 if ((!sz) && (!err)) {
1195 advance_rx_queue(sk);
1199 /* Optionally capture sender's address & ancillary data of first msg */
1200 if (sz_copied == 0) {
1201 set_orig_addr(m, msg);
1202 res = anc_data_recv(m, msg, port);
1207 /* Capture message data (if valid) & compute return value (always) */
1209 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1212 needed = (buf_len - sz_copied);
1213 sz_to_copy = (sz <= needed) ? sz : needed;
1215 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1216 m->msg_iov, sz_to_copy);
1220 sz_copied += sz_to_copy;
1222 if (sz_to_copy < sz) {
1223 if (!(flags & MSG_PEEK))
1224 TIPC_SKB_CB(buf)->handle =
1225 (void *)(unsigned long)(offset + sz_to_copy);
1230 goto exit; /* can't add error msg to valid data */
1232 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1238 /* Consume received message (optional) */
1239 if (likely(!(flags & MSG_PEEK))) {
1240 if (unlikely(++port->conn_unacked >= TIPC_CONNACK_INTV))
1241 tipc_acknowledge(port->ref, port->conn_unacked);
1242 advance_rx_queue(sk);
1245 /* Loop around if more data is required */
1246 if ((sz_copied < buf_len) && /* didn't get all requested data */
1247 (!skb_queue_empty(&sk->sk_receive_queue) ||
1248 (sz_copied < target)) && /* and more is ready or required */
1249 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1250 (!err)) /* and haven't reached a FIN */
1255 return sz_copied ? sz_copied : res;
1259 * tipc_write_space - wake up thread if port congestion is released
1262 static void tipc_write_space(struct sock *sk)
1264 struct socket_wq *wq;
1267 wq = rcu_dereference(sk->sk_wq);
1268 if (wq_has_sleeper(wq))
1269 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1270 POLLWRNORM | POLLWRBAND);
1275 * tipc_data_ready - wake up threads to indicate messages have been received
1277 * @len: the length of messages
1279 static void tipc_data_ready(struct sock *sk)
1281 struct socket_wq *wq;
1284 wq = rcu_dereference(sk->sk_wq);
1285 if (wq_has_sleeper(wq))
1286 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1287 POLLRDNORM | POLLRDBAND);
1292 * filter_connect - Handle all incoming messages for a connection-based socket
1296 * Returns 0 (TIPC_OK) if everyting ok, -TIPC_ERR_NO_PORT otherwise
1298 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1300 struct sock *sk = &tsk->sk;
1301 struct tipc_port *port = &tsk->port;
1302 struct socket *sock = sk->sk_socket;
1303 struct tipc_msg *msg = buf_msg(*buf);
1305 int retval = -TIPC_ERR_NO_PORT;
1311 switch ((int)sock->state) {
1313 /* Accept only connection-based messages sent by peer */
1314 if (msg_connected(msg) && tipc_port_peer_msg(port, msg)) {
1315 if (unlikely(msg_errcode(msg))) {
1316 sock->state = SS_DISCONNECTING;
1317 __tipc_port_disconnect(port);
1323 /* Accept only ACK or NACK message */
1324 if (unlikely(msg_errcode(msg))) {
1325 sock->state = SS_DISCONNECTING;
1326 sk->sk_err = ECONNREFUSED;
1331 if (unlikely(!msg_connected(msg)))
1334 res = auto_connect(tsk, msg);
1336 sock->state = SS_DISCONNECTING;
1342 /* If an incoming message is an 'ACK-', it should be
1343 * discarded here because it doesn't contain useful
1344 * data. In addition, we should try to wake up
1345 * connect() routine if sleeping.
1347 if (msg_data_sz(msg) == 0) {
1350 if (waitqueue_active(sk_sleep(sk)))
1351 wake_up_interruptible(sk_sleep(sk));
1356 case SS_UNCONNECTED:
1357 /* Accept only SYN message */
1358 if (!msg_connected(msg) && !(msg_errcode(msg)))
1361 case SS_DISCONNECTING:
1364 pr_err("Unknown socket state %u\n", sock->state);
1370 * rcvbuf_limit - get proper overload limit of socket receive queue
1374 * For all connection oriented messages, irrespective of importance,
1375 * the default overload value (i.e. 67MB) is set as limit.
1377 * For all connectionless messages, by default new queue limits are
1380 * TIPC_LOW_IMPORTANCE (4 MB)
1381 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1382 * TIPC_HIGH_IMPORTANCE (16 MB)
1383 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1385 * Returns overload limit according to corresponding message importance
1387 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1389 struct tipc_msg *msg = buf_msg(buf);
1391 if (msg_connected(msg))
1392 return sysctl_tipc_rmem[2];
1394 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1395 msg_importance(msg);
1399 * filter_rcv - validate incoming message
1403 * Enqueues message on receive queue if acceptable; optionally handles
1404 * disconnect indication for a connected socket.
1406 * Called with socket lock already taken; port lock may also be taken.
1408 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1411 static int filter_rcv(struct sock *sk, struct sk_buff *buf)
1413 struct socket *sock = sk->sk_socket;
1414 struct tipc_sock *tsk = tipc_sk(sk);
1415 struct tipc_msg *msg = buf_msg(buf);
1416 unsigned int limit = rcvbuf_limit(sk, buf);
1419 /* Reject message if it is wrong sort of message for socket */
1420 if (msg_type(msg) > TIPC_DIRECT_MSG)
1421 return -TIPC_ERR_NO_PORT;
1423 if (sock->state == SS_READY) {
1424 if (msg_connected(msg))
1425 return -TIPC_ERR_NO_PORT;
1427 rc = filter_connect(tsk, &buf);
1428 if (rc != TIPC_OK || buf == NULL)
1432 /* Reject message if there isn't room to queue it */
1433 if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1434 return -TIPC_ERR_OVERLOAD;
1436 /* Enqueue message */
1437 TIPC_SKB_CB(buf)->handle = NULL;
1438 __skb_queue_tail(&sk->sk_receive_queue, buf);
1439 skb_set_owner_r(buf, sk);
1441 sk->sk_data_ready(sk);
1446 * tipc_backlog_rcv - handle incoming message from backlog queue
1450 * Caller must hold socket lock, but not port lock.
1454 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *buf)
1458 struct tipc_sock *tsk = tipc_sk(sk);
1459 uint truesize = buf->truesize;
1461 rc = filter_rcv(sk, buf);
1463 if (unlikely(rc && tipc_msg_reverse(buf, &onode, -rc)))
1464 tipc_link_xmit2(buf, onode, 0);
1465 else if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1466 atomic_add(truesize, &tsk->dupl_rcvcnt);
1472 * tipc_sk_rcv - handle incoming message
1473 * @buf: buffer containing arriving message
1475 * Returns 0 if success, or errno: -EHOSTUNREACH
1477 int tipc_sk_rcv(struct sk_buff *buf)
1479 struct tipc_sock *tsk;
1480 struct tipc_port *port;
1482 u32 dport = msg_destport(buf_msg(buf));
1487 /* Validate destination and message */
1488 port = tipc_port_lock(dport);
1489 if (unlikely(!port)) {
1490 rc = tipc_msg_eval(buf, &dnode);
1494 tsk = tipc_port_to_sock(port);
1500 if (!sock_owned_by_user(sk)) {
1501 rc = filter_rcv(sk, buf);
1503 if (sk->sk_backlog.len == 0)
1504 atomic_set(&tsk->dupl_rcvcnt, 0);
1505 limit = rcvbuf_limit(sk, buf) + atomic_read(&tsk->dupl_rcvcnt);
1506 if (sk_add_backlog(sk, buf, limit))
1507 rc = -TIPC_ERR_OVERLOAD;
1510 tipc_port_unlock(port);
1515 if ((rc < 0) && !tipc_msg_reverse(buf, &dnode, -rc))
1516 return -EHOSTUNREACH;
1518 tipc_link_xmit2(buf, dnode, 0);
1519 return (rc < 0) ? -EHOSTUNREACH : 0;
1522 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1524 struct sock *sk = sock->sk;
1529 int err = sock_error(sk);
1534 if (signal_pending(current))
1535 return sock_intr_errno(*timeo_p);
1537 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1538 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1539 finish_wait(sk_sleep(sk), &wait);
1545 * tipc_connect - establish a connection to another TIPC port
1546 * @sock: socket structure
1547 * @dest: socket address for destination port
1548 * @destlen: size of socket address data structure
1549 * @flags: file-related flags associated with socket
1551 * Returns 0 on success, errno otherwise
1553 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1554 int destlen, int flags)
1556 struct sock *sk = sock->sk;
1557 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1558 struct msghdr m = {NULL,};
1559 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1560 socket_state previous;
1565 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1566 if (sock->state == SS_READY) {
1572 * Reject connection attempt using multicast address
1574 * Note: send_msg() validates the rest of the address fields,
1575 * so there's no need to do it here
1577 if (dst->addrtype == TIPC_ADDR_MCAST) {
1582 previous = sock->state;
1583 switch (sock->state) {
1584 case SS_UNCONNECTED:
1585 /* Send a 'SYN-' to destination */
1587 m.msg_namelen = destlen;
1589 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1590 * indicate send_msg() is never blocked.
1593 m.msg_flags = MSG_DONTWAIT;
1595 res = tipc_sendmsg(NULL, sock, &m, 0);
1596 if ((res < 0) && (res != -EWOULDBLOCK))
1599 /* Just entered SS_CONNECTING state; the only
1600 * difference is that return value in non-blocking
1601 * case is EINPROGRESS, rather than EALREADY.
1605 if (previous == SS_CONNECTING)
1609 timeout = msecs_to_jiffies(timeout);
1610 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1611 res = tipc_wait_for_connect(sock, &timeout);
1626 * tipc_listen - allow socket to listen for incoming connections
1627 * @sock: socket structure
1630 * Returns 0 on success, errno otherwise
1632 static int tipc_listen(struct socket *sock, int len)
1634 struct sock *sk = sock->sk;
1639 if (sock->state != SS_UNCONNECTED)
1642 sock->state = SS_LISTENING;
1650 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1652 struct sock *sk = sock->sk;
1656 /* True wake-one mechanism for incoming connections: only
1657 * one process gets woken up, not the 'whole herd'.
1658 * Since we do not 'race & poll' for established sockets
1659 * anymore, the common case will execute the loop only once.
1662 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1663 TASK_INTERRUPTIBLE);
1664 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1666 timeo = schedule_timeout(timeo);
1670 if (!skb_queue_empty(&sk->sk_receive_queue))
1673 if (sock->state != SS_LISTENING)
1675 err = sock_intr_errno(timeo);
1676 if (signal_pending(current))
1682 finish_wait(sk_sleep(sk), &wait);
1687 * tipc_accept - wait for connection request
1688 * @sock: listening socket
1689 * @newsock: new socket that is to be connected
1690 * @flags: file-related flags associated with socket
1692 * Returns 0 on success, errno otherwise
1694 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1696 struct sock *new_sk, *sk = sock->sk;
1697 struct sk_buff *buf;
1698 struct tipc_port *new_port;
1699 struct tipc_msg *msg;
1700 struct tipc_portid peer;
1707 if (sock->state != SS_LISTENING) {
1711 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1712 res = tipc_wait_for_accept(sock, timeo);
1716 buf = skb_peek(&sk->sk_receive_queue);
1718 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1722 new_sk = new_sock->sk;
1723 new_port = &tipc_sk(new_sk)->port;
1724 new_ref = new_port->ref;
1727 /* we lock on new_sk; but lockdep sees the lock on sk */
1728 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1731 * Reject any stray messages received by new socket
1732 * before the socket lock was taken (very, very unlikely)
1734 reject_rx_queue(new_sk);
1736 /* Connect new socket to it's peer */
1737 peer.ref = msg_origport(msg);
1738 peer.node = msg_orignode(msg);
1739 tipc_port_connect(new_ref, &peer);
1740 new_sock->state = SS_CONNECTED;
1742 tipc_port_set_importance(new_port, msg_importance(msg));
1743 if (msg_named(msg)) {
1744 new_port->conn_type = msg_nametype(msg);
1745 new_port->conn_instance = msg_nameinst(msg);
1749 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1750 * Respond to 'SYN+' by queuing it on new socket.
1752 if (!msg_data_sz(msg)) {
1753 struct msghdr m = {NULL,};
1755 advance_rx_queue(sk);
1756 tipc_send_packet(NULL, new_sock, &m, 0);
1758 __skb_dequeue(&sk->sk_receive_queue);
1759 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1760 skb_set_owner_r(buf, new_sk);
1762 release_sock(new_sk);
1769 * tipc_shutdown - shutdown socket connection
1770 * @sock: socket structure
1771 * @how: direction to close (must be SHUT_RDWR)
1773 * Terminates connection (if necessary), then purges socket's receive queue.
1775 * Returns 0 on success, errno otherwise
1777 static int tipc_shutdown(struct socket *sock, int how)
1779 struct sock *sk = sock->sk;
1780 struct tipc_sock *tsk = tipc_sk(sk);
1781 struct tipc_port *port = &tsk->port;
1782 struct sk_buff *buf;
1786 if (how != SHUT_RDWR)
1791 switch (sock->state) {
1796 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1797 buf = __skb_dequeue(&sk->sk_receive_queue);
1799 if (TIPC_SKB_CB(buf)->handle != NULL) {
1803 tipc_port_disconnect(port->ref);
1804 if (tipc_msg_reverse(buf, &peer, TIPC_CONN_SHUTDOWN))
1805 tipc_link_xmit2(buf, peer, 0);
1807 tipc_port_shutdown(port->ref);
1810 sock->state = SS_DISCONNECTING;
1814 case SS_DISCONNECTING:
1816 /* Discard any unreceived messages */
1817 __skb_queue_purge(&sk->sk_receive_queue);
1819 /* Wake up anyone sleeping in poll */
1820 sk->sk_state_change(sk);
1833 * tipc_setsockopt - set socket option
1834 * @sock: socket structure
1835 * @lvl: option level
1836 * @opt: option identifier
1837 * @ov: pointer to new option value
1838 * @ol: length of option value
1840 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1841 * (to ease compatibility).
1843 * Returns 0 on success, errno otherwise
1845 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
1846 char __user *ov, unsigned int ol)
1848 struct sock *sk = sock->sk;
1849 struct tipc_sock *tsk = tipc_sk(sk);
1850 struct tipc_port *port = &tsk->port;
1854 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1856 if (lvl != SOL_TIPC)
1857 return -ENOPROTOOPT;
1858 if (ol < sizeof(value))
1860 res = get_user(value, (u32 __user *)ov);
1867 case TIPC_IMPORTANCE:
1868 tipc_port_set_importance(port, value);
1870 case TIPC_SRC_DROPPABLE:
1871 if (sock->type != SOCK_STREAM)
1872 tipc_port_set_unreliable(port, value);
1876 case TIPC_DEST_DROPPABLE:
1877 tipc_port_set_unreturnable(port, value);
1879 case TIPC_CONN_TIMEOUT:
1880 tipc_sk(sk)->conn_timeout = value;
1881 /* no need to set "res", since already 0 at this point */
1893 * tipc_getsockopt - get socket option
1894 * @sock: socket structure
1895 * @lvl: option level
1896 * @opt: option identifier
1897 * @ov: receptacle for option value
1898 * @ol: receptacle for length of option value
1900 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1901 * (to ease compatibility).
1903 * Returns 0 on success, errno otherwise
1905 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
1906 char __user *ov, int __user *ol)
1908 struct sock *sk = sock->sk;
1909 struct tipc_sock *tsk = tipc_sk(sk);
1910 struct tipc_port *port = &tsk->port;
1915 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1916 return put_user(0, ol);
1917 if (lvl != SOL_TIPC)
1918 return -ENOPROTOOPT;
1919 res = get_user(len, ol);
1926 case TIPC_IMPORTANCE:
1927 value = tipc_port_importance(port);
1929 case TIPC_SRC_DROPPABLE:
1930 value = tipc_port_unreliable(port);
1932 case TIPC_DEST_DROPPABLE:
1933 value = tipc_port_unreturnable(port);
1935 case TIPC_CONN_TIMEOUT:
1936 value = tipc_sk(sk)->conn_timeout;
1937 /* no need to set "res", since already 0 at this point */
1939 case TIPC_NODE_RECVQ_DEPTH:
1940 value = 0; /* was tipc_queue_size, now obsolete */
1942 case TIPC_SOCK_RECVQ_DEPTH:
1943 value = skb_queue_len(&sk->sk_receive_queue);
1952 return res; /* "get" failed */
1954 if (len < sizeof(value))
1957 if (copy_to_user(ov, &value, sizeof(value)))
1960 return put_user(sizeof(value), ol);
1963 int tipc_ioctl(struct socket *sk, unsigned int cmd, unsigned long arg)
1965 struct tipc_sioc_ln_req lnr;
1966 void __user *argp = (void __user *)arg;
1969 case SIOCGETLINKNAME:
1970 if (copy_from_user(&lnr, argp, sizeof(lnr)))
1972 if (!tipc_node_get_linkname(lnr.bearer_id, lnr.peer,
1973 lnr.linkname, TIPC_MAX_LINK_NAME)) {
1974 if (copy_to_user(argp, &lnr, sizeof(lnr)))
1978 return -EADDRNOTAVAIL;
1981 return -ENOIOCTLCMD;
1985 /* Protocol switches for the various types of TIPC sockets */
1987 static const struct proto_ops msg_ops = {
1988 .owner = THIS_MODULE,
1990 .release = tipc_release,
1992 .connect = tipc_connect,
1993 .socketpair = sock_no_socketpair,
1994 .accept = sock_no_accept,
1995 .getname = tipc_getname,
1997 .ioctl = tipc_ioctl,
1998 .listen = sock_no_listen,
1999 .shutdown = tipc_shutdown,
2000 .setsockopt = tipc_setsockopt,
2001 .getsockopt = tipc_getsockopt,
2002 .sendmsg = tipc_sendmsg,
2003 .recvmsg = tipc_recvmsg,
2004 .mmap = sock_no_mmap,
2005 .sendpage = sock_no_sendpage
2008 static const struct proto_ops packet_ops = {
2009 .owner = THIS_MODULE,
2011 .release = tipc_release,
2013 .connect = tipc_connect,
2014 .socketpair = sock_no_socketpair,
2015 .accept = tipc_accept,
2016 .getname = tipc_getname,
2018 .ioctl = tipc_ioctl,
2019 .listen = tipc_listen,
2020 .shutdown = tipc_shutdown,
2021 .setsockopt = tipc_setsockopt,
2022 .getsockopt = tipc_getsockopt,
2023 .sendmsg = tipc_send_packet,
2024 .recvmsg = tipc_recvmsg,
2025 .mmap = sock_no_mmap,
2026 .sendpage = sock_no_sendpage
2029 static const struct proto_ops stream_ops = {
2030 .owner = THIS_MODULE,
2032 .release = tipc_release,
2034 .connect = tipc_connect,
2035 .socketpair = sock_no_socketpair,
2036 .accept = tipc_accept,
2037 .getname = tipc_getname,
2039 .ioctl = tipc_ioctl,
2040 .listen = tipc_listen,
2041 .shutdown = tipc_shutdown,
2042 .setsockopt = tipc_setsockopt,
2043 .getsockopt = tipc_getsockopt,
2044 .sendmsg = tipc_send_stream,
2045 .recvmsg = tipc_recv_stream,
2046 .mmap = sock_no_mmap,
2047 .sendpage = sock_no_sendpage
2050 static const struct net_proto_family tipc_family_ops = {
2051 .owner = THIS_MODULE,
2053 .create = tipc_sk_create
2056 static struct proto tipc_proto = {
2058 .owner = THIS_MODULE,
2059 .obj_size = sizeof(struct tipc_sock),
2060 .sysctl_rmem = sysctl_tipc_rmem
2063 static struct proto tipc_proto_kern = {
2065 .obj_size = sizeof(struct tipc_sock),
2066 .sysctl_rmem = sysctl_tipc_rmem
2070 * tipc_socket_init - initialize TIPC socket interface
2072 * Returns 0 on success, errno otherwise
2074 int tipc_socket_init(void)
2078 res = proto_register(&tipc_proto, 1);
2080 pr_err("Failed to register TIPC protocol type\n");
2084 res = sock_register(&tipc_family_ops);
2086 pr_err("Failed to register TIPC socket type\n");
2087 proto_unregister(&tipc_proto);
2095 * tipc_socket_stop - stop TIPC socket interface
2097 void tipc_socket_stop(void)
2099 sock_unregister(tipc_family_ops.family);
2100 proto_unregister(&tipc_proto);