2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2011, 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.
42 #define SS_LISTENING -1 /* socket is listening */
43 #define SS_READY -2 /* socket is connectionless */
45 #define OVERLOAD_LIMIT_BASE 5000
46 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51 struct tipc_portid peer_name;
55 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
56 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
58 #define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
59 (sock->state == SS_DISCONNECTING))
61 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
62 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
63 static void wakeupdispatch(struct tipc_port *tport);
65 static const struct proto_ops packet_ops;
66 static const struct proto_ops stream_ops;
67 static const struct proto_ops msg_ops;
69 static struct proto tipc_proto;
71 static int sockets_enabled;
73 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
76 * Revised TIPC socket locking policy:
78 * Most socket operations take the standard socket lock when they start
79 * and hold it until they finish (or until they need to sleep). Acquiring
80 * this lock grants the owner exclusive access to the fields of the socket
81 * data structures, with the exception of the backlog queue. A few socket
82 * operations can be done without taking the socket lock because they only
83 * read socket information that never changes during the life of the socket.
85 * Socket operations may acquire the lock for the associated TIPC port if they
86 * need to perform an operation on the port. If any routine needs to acquire
87 * both the socket lock and the port lock it must take the socket lock first
88 * to avoid the risk of deadlock.
90 * The dispatcher handling incoming messages cannot grab the socket lock in
91 * the standard fashion, since invoked it runs at the BH level and cannot block.
92 * Instead, it checks to see if the socket lock is currently owned by someone,
93 * and either handles the message itself or adds it to the socket's backlog
94 * queue; in the latter case the queued message is processed once the process
95 * owning the socket lock releases it.
97 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
98 * the problem of a blocked socket operation preventing any other operations
99 * from occurring. However, applications must be careful if they have
100 * multiple threads trying to send (or receive) on the same socket, as these
101 * operations might interfere with each other. For example, doing a connect
102 * and a receive at the same time might allow the receive to consume the
103 * ACK message meant for the connect. While additional work could be done
104 * to try and overcome this, it doesn't seem to be worthwhile at the present.
106 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
107 * that another operation that must be performed in a non-blocking manner is
108 * not delayed for very long because the lock has already been taken.
110 * NOTE: This code assumes that certain fields of a port/socket pair are
111 * constant over its lifetime; such fields can be examined without taking
112 * the socket lock and/or port lock, and do not need to be re-read even
113 * after resuming processing after waiting. These fields include:
115 * - pointer to socket sk structure (aka tipc_sock structure)
116 * - pointer to port structure
121 * advance_rx_queue - discard first buffer in socket receive queue
123 * Caller must hold socket lock
126 static void advance_rx_queue(struct sock *sk)
128 buf_discard(__skb_dequeue(&sk->sk_receive_queue));
129 atomic_dec(&tipc_queue_size);
133 * discard_rx_queue - discard all buffers in socket receive queue
135 * Caller must hold socket lock
138 static void discard_rx_queue(struct sock *sk)
142 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
143 atomic_dec(&tipc_queue_size);
149 * reject_rx_queue - reject all buffers in socket receive queue
151 * Caller must hold socket lock
154 static void reject_rx_queue(struct sock *sk)
158 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
159 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
160 atomic_dec(&tipc_queue_size);
165 * tipc_create - create a TIPC socket
166 * @net: network namespace (must be default network)
167 * @sock: pre-allocated socket structure
168 * @protocol: protocol indicator (must be 0)
169 * @kern: caused by kernel or by userspace?
171 * This routine creates additional data structures used by the TIPC socket,
172 * initializes them, and links them together.
174 * Returns 0 on success, errno otherwise
177 static int tipc_create(struct net *net, struct socket *sock, int protocol,
180 const struct proto_ops *ops;
183 struct tipc_port *tp_ptr;
185 /* Validate arguments */
187 if (!net_eq(net, &init_net))
188 return -EAFNOSUPPORT;
190 if (unlikely(protocol != 0))
191 return -EPROTONOSUPPORT;
193 switch (sock->type) {
196 state = SS_UNCONNECTED;
200 state = SS_UNCONNECTED;
211 /* Allocate socket's protocol area */
213 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
217 /* Allocate TIPC port for socket to use */
219 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
220 TIPC_LOW_IMPORTANCE);
221 if (unlikely(!tp_ptr)) {
226 /* Finish initializing socket data structures */
231 sock_init_data(sock, sk);
232 sk->sk_backlog_rcv = backlog_rcv;
233 tipc_sk(sk)->p = tp_ptr;
234 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
236 spin_unlock_bh(tp_ptr->lock);
238 if (sock->state == SS_READY) {
239 tipc_set_portunreturnable(tp_ptr->ref, 1);
240 if (sock->type == SOCK_DGRAM)
241 tipc_set_portunreliable(tp_ptr->ref, 1);
248 * release - destroy a TIPC socket
249 * @sock: socket to destroy
251 * This routine cleans up any messages that are still queued on the socket.
252 * For DGRAM and RDM socket types, all queued messages are rejected.
253 * For SEQPACKET and STREAM socket types, the first message is rejected
254 * and any others are discarded. (If the first message on a STREAM socket
255 * is partially-read, it is discarded and the next one is rejected instead.)
257 * NOTE: Rejected messages are not necessarily returned to the sender! They
258 * are returned or discarded according to the "destination droppable" setting
259 * specified for the message by the sender.
261 * Returns 0 on success, errno otherwise
264 static int release(struct socket *sock)
266 struct sock *sk = sock->sk;
267 struct tipc_port *tport;
272 * Exit if socket isn't fully initialized (occurs when a failed accept()
273 * releases a pre-allocated child socket that was never used)
279 tport = tipc_sk_port(sk);
283 * Reject all unreceived messages, except on an active connection
284 * (which disconnects locally & sends a 'FIN+' to peer)
287 while (sock->state != SS_DISCONNECTING) {
288 buf = __skb_dequeue(&sk->sk_receive_queue);
291 atomic_dec(&tipc_queue_size);
292 if (TIPC_SKB_CB(buf)->handle != 0)
295 if ((sock->state == SS_CONNECTING) ||
296 (sock->state == SS_CONNECTED)) {
297 sock->state = SS_DISCONNECTING;
298 tipc_disconnect(tport->ref);
300 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
305 * Delete TIPC port; this ensures no more messages are queued
306 * (also disconnects an active connection & sends a 'FIN-' to peer)
309 res = tipc_deleteport(tport->ref);
311 /* Discard any remaining (connection-based) messages in receive queue */
313 discard_rx_queue(sk);
315 /* Reject any messages that accumulated in backlog queue */
317 sock->state = SS_DISCONNECTING;
327 * bind - associate or disassocate TIPC name(s) with a socket
328 * @sock: socket structure
329 * @uaddr: socket address describing name(s) and desired operation
330 * @uaddr_len: size of socket address data structure
332 * Name and name sequence binding is indicated using a positive scope value;
333 * a negative scope value unbinds the specified name. Specifying no name
334 * (i.e. a socket address length of 0) unbinds all names from the socket.
336 * Returns 0 on success, errno otherwise
338 * NOTE: This routine doesn't need to take the socket lock since it doesn't
339 * access any non-constant socket information.
342 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
344 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
345 u32 portref = tipc_sk_port(sock->sk)->ref;
347 if (unlikely(!uaddr_len))
348 return tipc_withdraw(portref, 0, NULL);
350 if (uaddr_len < sizeof(struct sockaddr_tipc))
352 if (addr->family != AF_TIPC)
353 return -EAFNOSUPPORT;
355 if (addr->addrtype == TIPC_ADDR_NAME)
356 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
357 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
358 return -EAFNOSUPPORT;
360 return (addr->scope > 0) ?
361 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
362 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
366 * get_name - get port ID of socket or peer socket
367 * @sock: socket structure
368 * @uaddr: area for returned socket address
369 * @uaddr_len: area for returned length of socket address
370 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
372 * Returns 0 on success, errno otherwise
374 * NOTE: This routine doesn't need to take the socket lock since it only
375 * accesses socket information that is unchanging (or which changes in
376 * a completely predictable manner).
379 static int get_name(struct socket *sock, struct sockaddr *uaddr,
380 int *uaddr_len, int peer)
382 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
383 struct tipc_sock *tsock = tipc_sk(sock->sk);
385 memset(addr, 0, sizeof(*addr));
387 if ((sock->state != SS_CONNECTED) &&
388 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
390 addr->addr.id.ref = tsock->peer_name.ref;
391 addr->addr.id.node = tsock->peer_name.node;
393 addr->addr.id.ref = tsock->p->ref;
394 addr->addr.id.node = tipc_own_addr;
397 *uaddr_len = sizeof(*addr);
398 addr->addrtype = TIPC_ADDR_ID;
399 addr->family = AF_TIPC;
401 addr->addr.name.domain = 0;
407 * poll - read and possibly block on pollmask
408 * @file: file structure associated with the socket
409 * @sock: socket for which to calculate the poll bits
412 * Returns pollmask value
415 * It appears that the usual socket locking mechanisms are not useful here
416 * since the pollmask info is potentially out-of-date the moment this routine
417 * exits. TCP and other protocols seem to rely on higher level poll routines
418 * to handle any preventable race conditions, so TIPC will do the same ...
420 * TIPC sets the returned events as follows:
422 * socket state flags set
423 * ------------ ---------
424 * unconnected no read flags
427 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
430 * connected POLLIN/POLLRDNORM if data in rx queue
431 * POLLOUT if port is not congested
433 * disconnecting POLLIN/POLLRDNORM/POLLHUP
436 * listening POLLIN if SYN in rx queue
439 * ready POLLIN/POLLRDNORM if data in rx queue
440 * [connectionless] POLLOUT (since port cannot be congested)
442 * IMPORTANT: The fact that a read or write operation is indicated does NOT
443 * imply that the operation will succeed, merely that it should be performed
444 * and will not block.
447 static unsigned int poll(struct file *file, struct socket *sock,
450 struct sock *sk = sock->sk;
453 poll_wait(file, sk_sleep(sk), wait);
455 switch ((int)sock->state) {
458 if (!tipc_sk_port(sk)->congested)
463 if (!skb_queue_empty(&sk->sk_receive_queue))
464 mask |= (POLLIN | POLLRDNORM);
466 case SS_DISCONNECTING:
467 mask = (POLLIN | POLLRDNORM | POLLHUP);
475 * dest_name_check - verify user is permitted to send to specified port name
476 * @dest: destination address
477 * @m: descriptor for message to be sent
479 * Prevents restricted configuration commands from being issued by
480 * unauthorized users.
482 * Returns 0 if permission is granted, otherwise errno
485 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
487 struct tipc_cfg_msg_hdr hdr;
489 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
491 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
493 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
496 if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
498 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
500 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
507 * send_msg - send message in connectionless manner
508 * @iocb: if NULL, indicates that socket lock is already held
509 * @sock: socket structure
510 * @m: message to send
511 * @total_len: length of message
513 * Message must have an destination specified explicitly.
514 * Used for SOCK_RDM and SOCK_DGRAM messages,
515 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
516 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
518 * Returns the number of bytes sent on success, or errno otherwise
521 static int send_msg(struct kiocb *iocb, struct socket *sock,
522 struct msghdr *m, size_t total_len)
524 struct sock *sk = sock->sk;
525 struct tipc_port *tport = tipc_sk_port(sk);
526 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
531 return -EDESTADDRREQ;
532 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
533 (dest->family != AF_TIPC)))
535 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
536 (m->msg_iovlen > (unsigned)INT_MAX))
542 needs_conn = (sock->state != SS_READY);
543 if (unlikely(needs_conn)) {
544 if (sock->state == SS_LISTENING) {
548 if (sock->state != SS_UNCONNECTED) {
552 if ((tport->published) ||
553 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
557 if (dest->addrtype == TIPC_ADDR_NAME) {
558 tport->conn_type = dest->addr.name.name.type;
559 tport->conn_instance = dest->addr.name.name.instance;
562 /* Abort any pending connection attempts (very unlikely) */
568 if (dest->addrtype == TIPC_ADDR_NAME) {
569 res = dest_name_check(dest, m);
572 res = tipc_send2name(tport->ref,
573 &dest->addr.name.name,
574 dest->addr.name.domain,
578 } else if (dest->addrtype == TIPC_ADDR_ID) {
579 res = tipc_send2port(tport->ref,
584 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
589 res = dest_name_check(dest, m);
592 res = tipc_multicast(tport->ref,
598 if (likely(res != -ELINKCONG)) {
599 if (needs_conn && (res >= 0))
600 sock->state = SS_CONNECTING;
603 if (m->msg_flags & MSG_DONTWAIT) {
608 res = wait_event_interruptible(*sk_sleep(sk),
622 * send_packet - send a connection-oriented message
623 * @iocb: if NULL, indicates that socket lock is already held
624 * @sock: socket structure
625 * @m: message to send
626 * @total_len: length of message
628 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
630 * Returns the number of bytes sent on success, or errno otherwise
633 static int send_packet(struct kiocb *iocb, struct socket *sock,
634 struct msghdr *m, size_t total_len)
636 struct sock *sk = sock->sk;
637 struct tipc_port *tport = tipc_sk_port(sk);
638 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
641 /* Handle implied connection establishment */
644 return send_msg(iocb, sock, m, total_len);
646 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
647 (m->msg_iovlen > (unsigned)INT_MAX))
654 if (unlikely(sock->state != SS_CONNECTED)) {
655 if (sock->state == SS_DISCONNECTING)
662 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
664 if (likely(res != -ELINKCONG))
666 if (m->msg_flags & MSG_DONTWAIT) {
671 res = wait_event_interruptible(*sk_sleep(sk),
672 (!tport->congested || !tport->connected));
684 * send_stream - send stream-oriented data
686 * @sock: socket structure
688 * @total_len: total length of data to be sent
690 * Used for SOCK_STREAM data.
692 * Returns the number of bytes sent on success (or partial success),
693 * or errno if no data sent
696 static int send_stream(struct kiocb *iocb, struct socket *sock,
697 struct msghdr *m, size_t total_len)
699 struct sock *sk = sock->sk;
700 struct tipc_port *tport = tipc_sk_port(sk);
701 struct msghdr my_msg;
703 struct iovec *curr_iov;
705 char __user *curr_start;
714 /* Handle special cases where there is no connection */
716 if (unlikely(sock->state != SS_CONNECTED)) {
717 if (sock->state == SS_UNCONNECTED) {
718 res = send_packet(NULL, sock, m, total_len);
720 } else if (sock->state == SS_DISCONNECTING) {
729 if (unlikely(m->msg_name)) {
734 if ((total_len > (unsigned)INT_MAX) ||
735 (m->msg_iovlen > (unsigned)INT_MAX)) {
741 * Send each iovec entry using one or more messages
743 * Note: This algorithm is good for the most likely case
744 * (i.e. one large iovec entry), but could be improved to pass sets
745 * of small iovec entries into send_packet().
748 curr_iov = m->msg_iov;
749 curr_iovlen = m->msg_iovlen;
750 my_msg.msg_iov = &my_iov;
751 my_msg.msg_iovlen = 1;
752 my_msg.msg_flags = m->msg_flags;
753 my_msg.msg_name = NULL;
756 hdr_size = msg_hdr_sz(&tport->phdr);
758 while (curr_iovlen--) {
759 curr_start = curr_iov->iov_base;
760 curr_left = curr_iov->iov_len;
763 bytes_to_send = tport->max_pkt - hdr_size;
764 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
765 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
766 if (curr_left < bytes_to_send)
767 bytes_to_send = curr_left;
768 my_iov.iov_base = curr_start;
769 my_iov.iov_len = bytes_to_send;
770 res = send_packet(NULL, sock, &my_msg, bytes_to_send);
776 curr_left -= bytes_to_send;
777 curr_start += bytes_to_send;
778 bytes_sent += bytes_to_send;
790 * auto_connect - complete connection setup to a remote port
791 * @sock: socket structure
792 * @msg: peer's response message
794 * Returns 0 on success, errno otherwise
797 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
799 struct tipc_sock *tsock = tipc_sk(sock->sk);
801 if (msg_errcode(msg)) {
802 sock->state = SS_DISCONNECTING;
803 return -ECONNREFUSED;
806 tsock->peer_name.ref = msg_origport(msg);
807 tsock->peer_name.node = msg_orignode(msg);
808 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
809 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
810 sock->state = SS_CONNECTED;
815 * set_orig_addr - capture sender's address for received message
816 * @m: descriptor for message info
817 * @msg: received message header
819 * Note: Address is not captured if not requested by receiver.
822 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
824 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
827 addr->family = AF_TIPC;
828 addr->addrtype = TIPC_ADDR_ID;
829 addr->addr.id.ref = msg_origport(msg);
830 addr->addr.id.node = msg_orignode(msg);
831 addr->addr.name.domain = 0; /* could leave uninitialized */
832 addr->scope = 0; /* could leave uninitialized */
833 m->msg_namelen = sizeof(struct sockaddr_tipc);
838 * anc_data_recv - optionally capture ancillary data for received message
839 * @m: descriptor for message info
840 * @msg: received message header
841 * @tport: TIPC port associated with message
843 * Note: Ancillary data is not captured if not requested by receiver.
845 * Returns 0 if successful, otherwise errno
848 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
849 struct tipc_port *tport)
857 if (likely(m->msg_controllen == 0))
860 /* Optionally capture errored message object(s) */
862 err = msg ? msg_errcode(msg) : 0;
865 anc_data[1] = msg_data_sz(msg);
866 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
870 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
877 /* Optionally capture message destination object */
879 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
883 anc_data[0] = msg_nametype(msg);
884 anc_data[1] = msg_namelower(msg);
885 anc_data[2] = msg_namelower(msg);
889 anc_data[0] = msg_nametype(msg);
890 anc_data[1] = msg_namelower(msg);
891 anc_data[2] = msg_nameupper(msg);
894 has_name = (tport->conn_type != 0);
895 anc_data[0] = tport->conn_type;
896 anc_data[1] = tport->conn_instance;
897 anc_data[2] = tport->conn_instance;
903 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
912 * recv_msg - receive packet-oriented message
914 * @m: descriptor for message info
915 * @buf_len: total size of user buffer area
916 * @flags: receive flags
918 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
919 * If the complete message doesn't fit in user area, truncate it.
921 * Returns size of returned message data, errno otherwise
924 static int recv_msg(struct kiocb *iocb, struct socket *sock,
925 struct msghdr *m, size_t buf_len, int flags)
927 struct sock *sk = sock->sk;
928 struct tipc_port *tport = tipc_sk_port(sk);
930 struct tipc_msg *msg;
936 /* Catch invalid receive requests */
938 if (unlikely(!buf_len))
943 if (unlikely(sock->state == SS_UNCONNECTED)) {
948 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
951 /* Look for a message in receive queue; wait if necessary */
953 while (skb_queue_empty(&sk->sk_receive_queue)) {
954 if (sock->state == SS_DISCONNECTING) {
959 res = timeout ? timeout : -EWOULDBLOCK;
963 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
969 /* Look at first message in receive queue */
971 buf = skb_peek(&sk->sk_receive_queue);
973 sz = msg_data_sz(msg);
974 err = msg_errcode(msg);
976 /* Complete connection setup for an implied connect */
978 if (unlikely(sock->state == SS_CONNECTING)) {
979 res = auto_connect(sock, msg);
984 /* Discard an empty non-errored message & try again */
986 if ((!sz) && (!err)) {
987 advance_rx_queue(sk);
991 /* Capture sender's address (optional) */
993 set_orig_addr(m, msg);
995 /* Capture ancillary data (optional) */
997 res = anc_data_recv(m, msg, tport);
1001 /* Capture message data (if valid) & compute return value (always) */
1004 if (unlikely(buf_len < sz)) {
1006 m->msg_flags |= MSG_TRUNC;
1008 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1014 if ((sock->state == SS_READY) ||
1015 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1021 /* Consume received message (optional) */
1023 if (likely(!(flags & MSG_PEEK))) {
1024 if ((sock->state != SS_READY) &&
1025 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1026 tipc_acknowledge(tport->ref, tport->conn_unacked);
1027 advance_rx_queue(sk);
1035 * recv_stream - receive stream-oriented data
1037 * @m: descriptor for message info
1038 * @buf_len: total size of user buffer area
1039 * @flags: receive flags
1041 * Used for SOCK_STREAM messages only. If not enough data is available
1042 * will optionally wait for more; never truncates data.
1044 * Returns size of returned message data, errno otherwise
1047 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1048 struct msghdr *m, size_t buf_len, int flags)
1050 struct sock *sk = sock->sk;
1051 struct tipc_port *tport = tipc_sk_port(sk);
1052 struct sk_buff *buf;
1053 struct tipc_msg *msg;
1056 int sz_to_copy, target, needed;
1061 /* Catch invalid receive attempts */
1063 if (unlikely(!buf_len))
1068 if (unlikely((sock->state == SS_UNCONNECTED) ||
1069 (sock->state == SS_CONNECTING))) {
1074 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1075 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1078 /* Look for a message in receive queue; wait if necessary */
1080 while (skb_queue_empty(&sk->sk_receive_queue)) {
1081 if (sock->state == SS_DISCONNECTING) {
1085 if (timeout <= 0L) {
1086 res = timeout ? timeout : -EWOULDBLOCK;
1090 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1091 tipc_rx_ready(sock),
1096 /* Look at first message in receive queue */
1098 buf = skb_peek(&sk->sk_receive_queue);
1100 sz = msg_data_sz(msg);
1101 err = msg_errcode(msg);
1103 /* Discard an empty non-errored message & try again */
1105 if ((!sz) && (!err)) {
1106 advance_rx_queue(sk);
1110 /* Optionally capture sender's address & ancillary data of first msg */
1112 if (sz_copied == 0) {
1113 set_orig_addr(m, msg);
1114 res = anc_data_recv(m, msg, tport);
1119 /* Capture message data (if valid) & compute return value (always) */
1122 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1125 needed = (buf_len - sz_copied);
1126 sz_to_copy = (sz <= needed) ? sz : needed;
1128 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1129 m->msg_iov, sz_to_copy);
1133 sz_copied += sz_to_copy;
1135 if (sz_to_copy < sz) {
1136 if (!(flags & MSG_PEEK))
1137 TIPC_SKB_CB(buf)->handle =
1138 (void *)(unsigned long)(offset + sz_to_copy);
1143 goto exit; /* can't add error msg to valid data */
1145 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1151 /* Consume received message (optional) */
1153 if (likely(!(flags & MSG_PEEK))) {
1154 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1155 tipc_acknowledge(tport->ref, tport->conn_unacked);
1156 advance_rx_queue(sk);
1159 /* Loop around if more data is required */
1161 if ((sz_copied < buf_len) && /* didn't get all requested data */
1162 (!skb_queue_empty(&sk->sk_receive_queue) ||
1163 (sz_copied < target)) && /* and more is ready or required */
1164 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1165 (!err)) /* and haven't reached a FIN */
1170 return sz_copied ? sz_copied : res;
1174 * rx_queue_full - determine if receive queue can accept another message
1175 * @msg: message to be added to queue
1176 * @queue_size: current size of queue
1177 * @base: nominal maximum size of queue
1179 * Returns 1 if queue is unable to accept message, 0 otherwise
1182 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1185 u32 imp = msg_importance(msg);
1187 if (imp == TIPC_LOW_IMPORTANCE)
1189 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1190 threshold = base * 2;
1191 else if (imp == TIPC_HIGH_IMPORTANCE)
1192 threshold = base * 100;
1196 if (msg_connected(msg))
1199 return queue_size >= threshold;
1203 * filter_rcv - validate incoming message
1207 * Enqueues message on receive queue if acceptable; optionally handles
1208 * disconnect indication for a connected socket.
1210 * Called with socket lock already taken; port lock may also be taken.
1212 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1215 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1217 struct socket *sock = sk->sk_socket;
1218 struct tipc_msg *msg = buf_msg(buf);
1221 /* Reject message if it is wrong sort of message for socket */
1224 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1225 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1226 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1229 if (sock->state == SS_READY) {
1230 if (msg_connected(msg))
1231 return TIPC_ERR_NO_PORT;
1234 return TIPC_ERR_NO_PORT;
1235 if (sock->state == SS_CONNECTED) {
1236 if (!msg_connected(msg))
1237 return TIPC_ERR_NO_PORT;
1238 } else if (sock->state == SS_CONNECTING) {
1239 if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1240 return TIPC_ERR_NO_PORT;
1241 } else if (sock->state == SS_LISTENING) {
1242 if (msg_connected(msg) || msg_errcode(msg))
1243 return TIPC_ERR_NO_PORT;
1244 } else if (sock->state == SS_DISCONNECTING) {
1245 return TIPC_ERR_NO_PORT;
1246 } else /* (sock->state == SS_UNCONNECTED) */ {
1247 if (msg_connected(msg) || msg_errcode(msg))
1248 return TIPC_ERR_NO_PORT;
1252 /* Reject message if there isn't room to queue it */
1254 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1255 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1256 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1257 return TIPC_ERR_OVERLOAD;
1259 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1260 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1261 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1262 return TIPC_ERR_OVERLOAD;
1265 /* Enqueue message (finally!) */
1267 TIPC_SKB_CB(buf)->handle = 0;
1268 atomic_inc(&tipc_queue_size);
1269 __skb_queue_tail(&sk->sk_receive_queue, buf);
1271 /* Initiate connection termination for an incoming 'FIN' */
1273 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1274 sock->state = SS_DISCONNECTING;
1275 tipc_disconnect_port(tipc_sk_port(sk));
1278 if (waitqueue_active(sk_sleep(sk)))
1279 wake_up_interruptible(sk_sleep(sk));
1284 * backlog_rcv - handle incoming message from backlog queue
1288 * Caller must hold socket lock, but not port lock.
1293 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1297 res = filter_rcv(sk, buf);
1299 tipc_reject_msg(buf, res);
1304 * dispatch - handle incoming message
1305 * @tport: TIPC port that received message
1308 * Called with port lock already taken.
1310 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1313 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1315 struct sock *sk = (struct sock *)tport->usr_handle;
1319 * Process message if socket is unlocked; otherwise add to backlog queue
1321 * This code is based on sk_receive_skb(), but must be distinct from it
1322 * since a TIPC-specific filter/reject mechanism is utilized
1326 if (!sock_owned_by_user(sk)) {
1327 res = filter_rcv(sk, buf);
1329 if (sk_add_backlog(sk, buf))
1330 res = TIPC_ERR_OVERLOAD;
1340 * wakeupdispatch - wake up port after congestion
1341 * @tport: port to wakeup
1343 * Called with port lock already taken.
1346 static void wakeupdispatch(struct tipc_port *tport)
1348 struct sock *sk = (struct sock *)tport->usr_handle;
1350 if (waitqueue_active(sk_sleep(sk)))
1351 wake_up_interruptible(sk_sleep(sk));
1355 * connect - establish a connection to another TIPC port
1356 * @sock: socket structure
1357 * @dest: socket address for destination port
1358 * @destlen: size of socket address data structure
1359 * @flags: file-related flags associated with socket
1361 * Returns 0 on success, errno otherwise
1364 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1367 struct sock *sk = sock->sk;
1368 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1369 struct msghdr m = {NULL,};
1370 struct sk_buff *buf;
1371 struct tipc_msg *msg;
1377 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1379 if (sock->state == SS_READY) {
1384 /* For now, TIPC does not support the non-blocking form of connect() */
1386 if (flags & O_NONBLOCK) {
1391 /* Issue Posix-compliant error code if socket is in the wrong state */
1393 if (sock->state == SS_LISTENING) {
1397 if (sock->state == SS_CONNECTING) {
1401 if (sock->state != SS_UNCONNECTED) {
1407 * Reject connection attempt using multicast address
1409 * Note: send_msg() validates the rest of the address fields,
1410 * so there's no need to do it here
1413 if (dst->addrtype == TIPC_ADDR_MCAST) {
1418 /* Reject any messages already in receive queue (very unlikely) */
1420 reject_rx_queue(sk);
1422 /* Send a 'SYN-' to destination */
1425 m.msg_namelen = destlen;
1426 res = send_msg(NULL, sock, &m, 0);
1430 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1432 timeout = tipc_sk(sk)->conn_timeout;
1434 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1435 (!skb_queue_empty(&sk->sk_receive_queue) ||
1436 (sock->state != SS_CONNECTING)),
1437 timeout ? timeout : MAX_SCHEDULE_TIMEOUT);
1441 buf = skb_peek(&sk->sk_receive_queue);
1444 res = auto_connect(sock, msg);
1446 if (!msg_data_sz(msg))
1447 advance_rx_queue(sk);
1450 if (sock->state == SS_CONNECTED)
1453 res = -ECONNREFUSED;
1459 ; /* leave "res" unchanged */
1460 sock->state = SS_DISCONNECTING;
1469 * listen - allow socket to listen for incoming connections
1470 * @sock: socket structure
1473 * Returns 0 on success, errno otherwise
1476 static int listen(struct socket *sock, int len)
1478 struct sock *sk = sock->sk;
1483 if (sock->state == SS_READY)
1485 else if (sock->state != SS_UNCONNECTED)
1488 sock->state = SS_LISTENING;
1497 * accept - wait for connection request
1498 * @sock: listening socket
1499 * @newsock: new socket that is to be connected
1500 * @flags: file-related flags associated with socket
1502 * Returns 0 on success, errno otherwise
1505 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1507 struct sock *sk = sock->sk;
1508 struct sk_buff *buf;
1513 if (sock->state == SS_READY) {
1517 if (sock->state != SS_LISTENING) {
1522 while (skb_queue_empty(&sk->sk_receive_queue)) {
1523 if (flags & O_NONBLOCK) {
1528 res = wait_event_interruptible(*sk_sleep(sk),
1529 (!skb_queue_empty(&sk->sk_receive_queue)));
1535 buf = skb_peek(&sk->sk_receive_queue);
1537 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1539 struct sock *new_sk = new_sock->sk;
1540 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1541 struct tipc_port *new_tport = new_tsock->p;
1542 u32 new_ref = new_tport->ref;
1543 struct tipc_msg *msg = buf_msg(buf);
1548 * Reject any stray messages received by new socket
1549 * before the socket lock was taken (very, very unlikely)
1552 reject_rx_queue(new_sk);
1554 /* Connect new socket to it's peer */
1556 new_tsock->peer_name.ref = msg_origport(msg);
1557 new_tsock->peer_name.node = msg_orignode(msg);
1558 tipc_connect2port(new_ref, &new_tsock->peer_name);
1559 new_sock->state = SS_CONNECTED;
1561 tipc_set_portimportance(new_ref, msg_importance(msg));
1562 if (msg_named(msg)) {
1563 new_tport->conn_type = msg_nametype(msg);
1564 new_tport->conn_instance = msg_nameinst(msg);
1568 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1569 * Respond to 'SYN+' by queuing it on new socket.
1572 if (!msg_data_sz(msg)) {
1573 struct msghdr m = {NULL,};
1575 advance_rx_queue(sk);
1576 send_packet(NULL, new_sock, &m, 0);
1578 __skb_dequeue(&sk->sk_receive_queue);
1579 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1581 release_sock(new_sk);
1589 * shutdown - shutdown socket connection
1590 * @sock: socket structure
1591 * @how: direction to close (must be SHUT_RDWR)
1593 * Terminates connection (if necessary), then purges socket's receive queue.
1595 * Returns 0 on success, errno otherwise
1598 static int shutdown(struct socket *sock, int how)
1600 struct sock *sk = sock->sk;
1601 struct tipc_port *tport = tipc_sk_port(sk);
1602 struct sk_buff *buf;
1605 if (how != SHUT_RDWR)
1610 switch (sock->state) {
1614 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1616 buf = __skb_dequeue(&sk->sk_receive_queue);
1618 atomic_dec(&tipc_queue_size);
1619 if (TIPC_SKB_CB(buf)->handle != 0) {
1623 tipc_disconnect(tport->ref);
1624 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1626 tipc_shutdown(tport->ref);
1629 sock->state = SS_DISCONNECTING;
1633 case SS_DISCONNECTING:
1635 /* Discard any unreceived messages; wake up sleeping tasks */
1637 discard_rx_queue(sk);
1638 if (waitqueue_active(sk_sleep(sk)))
1639 wake_up_interruptible(sk_sleep(sk));
1652 * setsockopt - set socket option
1653 * @sock: socket structure
1654 * @lvl: option level
1655 * @opt: option identifier
1656 * @ov: pointer to new option value
1657 * @ol: length of option value
1659 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1660 * (to ease compatibility).
1662 * Returns 0 on success, errno otherwise
1665 static int setsockopt(struct socket *sock,
1666 int lvl, int opt, char __user *ov, unsigned int ol)
1668 struct sock *sk = sock->sk;
1669 struct tipc_port *tport = tipc_sk_port(sk);
1673 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1675 if (lvl != SOL_TIPC)
1676 return -ENOPROTOOPT;
1677 if (ol < sizeof(value))
1679 res = get_user(value, (u32 __user *)ov);
1686 case TIPC_IMPORTANCE:
1687 res = tipc_set_portimportance(tport->ref, value);
1689 case TIPC_SRC_DROPPABLE:
1690 if (sock->type != SOCK_STREAM)
1691 res = tipc_set_portunreliable(tport->ref, value);
1695 case TIPC_DEST_DROPPABLE:
1696 res = tipc_set_portunreturnable(tport->ref, value);
1698 case TIPC_CONN_TIMEOUT:
1699 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value);
1700 /* no need to set "res", since already 0 at this point */
1712 * getsockopt - get socket option
1713 * @sock: socket structure
1714 * @lvl: option level
1715 * @opt: option identifier
1716 * @ov: receptacle for option value
1717 * @ol: receptacle for length of option value
1719 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1720 * (to ease compatibility).
1722 * Returns 0 on success, errno otherwise
1725 static int getsockopt(struct socket *sock,
1726 int lvl, int opt, char __user *ov, int __user *ol)
1728 struct sock *sk = sock->sk;
1729 struct tipc_port *tport = tipc_sk_port(sk);
1734 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1735 return put_user(0, ol);
1736 if (lvl != SOL_TIPC)
1737 return -ENOPROTOOPT;
1738 res = get_user(len, ol);
1745 case TIPC_IMPORTANCE:
1746 res = tipc_portimportance(tport->ref, &value);
1748 case TIPC_SRC_DROPPABLE:
1749 res = tipc_portunreliable(tport->ref, &value);
1751 case TIPC_DEST_DROPPABLE:
1752 res = tipc_portunreturnable(tport->ref, &value);
1754 case TIPC_CONN_TIMEOUT:
1755 value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout);
1756 /* no need to set "res", since already 0 at this point */
1758 case TIPC_NODE_RECVQ_DEPTH:
1759 value = (u32)atomic_read(&tipc_queue_size);
1761 case TIPC_SOCK_RECVQ_DEPTH:
1762 value = skb_queue_len(&sk->sk_receive_queue);
1771 return res; /* "get" failed */
1773 if (len < sizeof(value))
1776 if (copy_to_user(ov, &value, sizeof(value)))
1779 return put_user(sizeof(value), ol);
1783 * Protocol switches for the various types of TIPC sockets
1786 static const struct proto_ops msg_ops = {
1787 .owner = THIS_MODULE,
1792 .socketpair = sock_no_socketpair,
1794 .getname = get_name,
1796 .ioctl = sock_no_ioctl,
1798 .shutdown = shutdown,
1799 .setsockopt = setsockopt,
1800 .getsockopt = getsockopt,
1801 .sendmsg = send_msg,
1802 .recvmsg = recv_msg,
1803 .mmap = sock_no_mmap,
1804 .sendpage = sock_no_sendpage
1807 static const struct proto_ops packet_ops = {
1808 .owner = THIS_MODULE,
1813 .socketpair = sock_no_socketpair,
1815 .getname = get_name,
1817 .ioctl = sock_no_ioctl,
1819 .shutdown = shutdown,
1820 .setsockopt = setsockopt,
1821 .getsockopt = getsockopt,
1822 .sendmsg = send_packet,
1823 .recvmsg = recv_msg,
1824 .mmap = sock_no_mmap,
1825 .sendpage = sock_no_sendpage
1828 static const struct proto_ops stream_ops = {
1829 .owner = THIS_MODULE,
1834 .socketpair = sock_no_socketpair,
1836 .getname = get_name,
1838 .ioctl = sock_no_ioctl,
1840 .shutdown = shutdown,
1841 .setsockopt = setsockopt,
1842 .getsockopt = getsockopt,
1843 .sendmsg = send_stream,
1844 .recvmsg = recv_stream,
1845 .mmap = sock_no_mmap,
1846 .sendpage = sock_no_sendpage
1849 static const struct net_proto_family tipc_family_ops = {
1850 .owner = THIS_MODULE,
1852 .create = tipc_create
1855 static struct proto tipc_proto = {
1857 .owner = THIS_MODULE,
1858 .obj_size = sizeof(struct tipc_sock)
1862 * tipc_socket_init - initialize TIPC socket interface
1864 * Returns 0 on success, errno otherwise
1866 int tipc_socket_init(void)
1870 res = proto_register(&tipc_proto, 1);
1872 err("Failed to register TIPC protocol type\n");
1876 res = sock_register(&tipc_family_ops);
1878 err("Failed to register TIPC socket type\n");
1879 proto_unregister(&tipc_proto);
1883 sockets_enabled = 1;
1889 * tipc_socket_stop - stop TIPC socket interface
1892 void tipc_socket_stop(void)
1894 if (!sockets_enabled)
1897 sockets_enabled = 0;
1898 sock_unregister(tipc_family_ops.family);
1899 proto_unregister(&tipc_proto);