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;
52 unsigned int conn_timeout;
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 = 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;
532 return -EDESTADDRREQ;
533 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
534 (dest->family != AF_TIPC)))
536 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
537 (m->msg_iovlen > (unsigned)INT_MAX))
543 needs_conn = (sock->state != SS_READY);
544 if (unlikely(needs_conn)) {
545 if (sock->state == SS_LISTENING) {
549 if (sock->state != SS_UNCONNECTED) {
553 if ((tport->published) ||
554 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
558 if (dest->addrtype == TIPC_ADDR_NAME) {
559 tport->conn_type = dest->addr.name.name.type;
560 tport->conn_instance = dest->addr.name.name.instance;
563 /* Abort any pending connection attempts (very unlikely) */
568 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
571 if (dest->addrtype == TIPC_ADDR_NAME) {
572 res = dest_name_check(dest, m);
575 res = tipc_send2name(tport->ref,
576 &dest->addr.name.name,
577 dest->addr.name.domain,
581 } else if (dest->addrtype == TIPC_ADDR_ID) {
582 res = tipc_send2port(tport->ref,
587 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
592 res = dest_name_check(dest, m);
595 res = tipc_multicast(tport->ref,
601 if (likely(res != -ELINKCONG)) {
602 if (needs_conn && (res >= 0))
603 sock->state = SS_CONNECTING;
606 if (timeout_val <= 0L) {
607 res = timeout_val ? timeout_val : -EWOULDBLOCK;
611 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
612 !tport->congested, timeout_val);
623 * send_packet - send a connection-oriented message
624 * @iocb: if NULL, indicates that socket lock is already held
625 * @sock: socket structure
626 * @m: message to send
627 * @total_len: length of message
629 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
631 * Returns the number of bytes sent on success, or errno otherwise
634 static int send_packet(struct kiocb *iocb, struct socket *sock,
635 struct msghdr *m, size_t total_len)
637 struct sock *sk = sock->sk;
638 struct tipc_port *tport = tipc_sk_port(sk);
639 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
643 /* Handle implied connection establishment */
646 return send_msg(iocb, sock, m, total_len);
648 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
649 (m->msg_iovlen > (unsigned)INT_MAX))
655 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
658 if (unlikely(sock->state != SS_CONNECTED)) {
659 if (sock->state == SS_DISCONNECTING)
666 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
668 if (likely(res != -ELINKCONG))
670 if (timeout_val <= 0L) {
671 res = timeout_val ? timeout_val : -EWOULDBLOCK;
675 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
676 (!tport->congested || !tport->connected), timeout_val);
686 * send_stream - send stream-oriented data
688 * @sock: socket structure
690 * @total_len: total length of data to be sent
692 * Used for SOCK_STREAM data.
694 * Returns the number of bytes sent on success (or partial success),
695 * or errno if no data sent
698 static int send_stream(struct kiocb *iocb, struct socket *sock,
699 struct msghdr *m, size_t total_len)
701 struct sock *sk = sock->sk;
702 struct tipc_port *tport = tipc_sk_port(sk);
703 struct msghdr my_msg;
705 struct iovec *curr_iov;
707 char __user *curr_start;
716 /* Handle special cases where there is no connection */
718 if (unlikely(sock->state != SS_CONNECTED)) {
719 if (sock->state == SS_UNCONNECTED) {
720 res = send_packet(NULL, sock, m, total_len);
722 } else if (sock->state == SS_DISCONNECTING) {
731 if (unlikely(m->msg_name)) {
736 if ((total_len > (unsigned)INT_MAX) ||
737 (m->msg_iovlen > (unsigned)INT_MAX)) {
743 * Send each iovec entry using one or more messages
745 * Note: This algorithm is good for the most likely case
746 * (i.e. one large iovec entry), but could be improved to pass sets
747 * of small iovec entries into send_packet().
750 curr_iov = m->msg_iov;
751 curr_iovlen = m->msg_iovlen;
752 my_msg.msg_iov = &my_iov;
753 my_msg.msg_iovlen = 1;
754 my_msg.msg_flags = m->msg_flags;
755 my_msg.msg_name = NULL;
758 hdr_size = msg_hdr_sz(&tport->phdr);
760 while (curr_iovlen--) {
761 curr_start = curr_iov->iov_base;
762 curr_left = curr_iov->iov_len;
765 bytes_to_send = tport->max_pkt - hdr_size;
766 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
767 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
768 if (curr_left < bytes_to_send)
769 bytes_to_send = curr_left;
770 my_iov.iov_base = curr_start;
771 my_iov.iov_len = bytes_to_send;
772 res = send_packet(NULL, sock, &my_msg, bytes_to_send);
778 curr_left -= bytes_to_send;
779 curr_start += bytes_to_send;
780 bytes_sent += bytes_to_send;
792 * auto_connect - complete connection setup to a remote port
793 * @sock: socket structure
794 * @msg: peer's response message
796 * Returns 0 on success, errno otherwise
799 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
801 struct tipc_sock *tsock = tipc_sk(sock->sk);
803 if (msg_errcode(msg)) {
804 sock->state = SS_DISCONNECTING;
805 return -ECONNREFUSED;
808 tsock->peer_name.ref = msg_origport(msg);
809 tsock->peer_name.node = msg_orignode(msg);
810 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
811 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
812 sock->state = SS_CONNECTED;
817 * set_orig_addr - capture sender's address for received message
818 * @m: descriptor for message info
819 * @msg: received message header
821 * Note: Address is not captured if not requested by receiver.
824 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
826 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
829 addr->family = AF_TIPC;
830 addr->addrtype = TIPC_ADDR_ID;
831 addr->addr.id.ref = msg_origport(msg);
832 addr->addr.id.node = msg_orignode(msg);
833 addr->addr.name.domain = 0; /* could leave uninitialized */
834 addr->scope = 0; /* could leave uninitialized */
835 m->msg_namelen = sizeof(struct sockaddr_tipc);
840 * anc_data_recv - optionally capture ancillary data for received message
841 * @m: descriptor for message info
842 * @msg: received message header
843 * @tport: TIPC port associated with message
845 * Note: Ancillary data is not captured if not requested by receiver.
847 * Returns 0 if successful, otherwise errno
850 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
851 struct tipc_port *tport)
859 if (likely(m->msg_controllen == 0))
862 /* Optionally capture errored message object(s) */
864 err = msg ? msg_errcode(msg) : 0;
867 anc_data[1] = msg_data_sz(msg);
868 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
872 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
879 /* Optionally capture message destination object */
881 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
885 anc_data[0] = msg_nametype(msg);
886 anc_data[1] = msg_namelower(msg);
887 anc_data[2] = msg_namelower(msg);
891 anc_data[0] = msg_nametype(msg);
892 anc_data[1] = msg_namelower(msg);
893 anc_data[2] = msg_nameupper(msg);
896 has_name = (tport->conn_type != 0);
897 anc_data[0] = tport->conn_type;
898 anc_data[1] = tport->conn_instance;
899 anc_data[2] = tport->conn_instance;
905 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
914 * recv_msg - receive packet-oriented message
916 * @m: descriptor for message info
917 * @buf_len: total size of user buffer area
918 * @flags: receive flags
920 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
921 * If the complete message doesn't fit in user area, truncate it.
923 * Returns size of returned message data, errno otherwise
926 static int recv_msg(struct kiocb *iocb, struct socket *sock,
927 struct msghdr *m, size_t buf_len, int flags)
929 struct sock *sk = sock->sk;
930 struct tipc_port *tport = tipc_sk_port(sk);
932 struct tipc_msg *msg;
938 /* Catch invalid receive requests */
940 if (unlikely(!buf_len))
945 if (unlikely(sock->state == SS_UNCONNECTED)) {
950 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
953 /* Look for a message in receive queue; wait if necessary */
955 while (skb_queue_empty(&sk->sk_receive_queue)) {
956 if (sock->state == SS_DISCONNECTING) {
961 res = timeout ? timeout : -EWOULDBLOCK;
965 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
971 /* Look at first message in receive queue */
973 buf = skb_peek(&sk->sk_receive_queue);
975 sz = msg_data_sz(msg);
976 err = msg_errcode(msg);
978 /* Complete connection setup for an implied connect */
980 if (unlikely(sock->state == SS_CONNECTING)) {
981 res = auto_connect(sock, msg);
986 /* Discard an empty non-errored message & try again */
988 if ((!sz) && (!err)) {
989 advance_rx_queue(sk);
993 /* Capture sender's address (optional) */
995 set_orig_addr(m, msg);
997 /* Capture ancillary data (optional) */
999 res = anc_data_recv(m, msg, tport);
1003 /* Capture message data (if valid) & compute return value (always) */
1006 if (unlikely(buf_len < sz)) {
1008 m->msg_flags |= MSG_TRUNC;
1010 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1016 if ((sock->state == SS_READY) ||
1017 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1023 /* Consume received message (optional) */
1025 if (likely(!(flags & MSG_PEEK))) {
1026 if ((sock->state != SS_READY) &&
1027 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1028 tipc_acknowledge(tport->ref, tport->conn_unacked);
1029 advance_rx_queue(sk);
1037 * recv_stream - receive stream-oriented data
1039 * @m: descriptor for message info
1040 * @buf_len: total size of user buffer area
1041 * @flags: receive flags
1043 * Used for SOCK_STREAM messages only. If not enough data is available
1044 * will optionally wait for more; never truncates data.
1046 * Returns size of returned message data, errno otherwise
1049 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1050 struct msghdr *m, size_t buf_len, int flags)
1052 struct sock *sk = sock->sk;
1053 struct tipc_port *tport = tipc_sk_port(sk);
1054 struct sk_buff *buf;
1055 struct tipc_msg *msg;
1058 int sz_to_copy, target, needed;
1063 /* Catch invalid receive attempts */
1065 if (unlikely(!buf_len))
1070 if (unlikely((sock->state == SS_UNCONNECTED) ||
1071 (sock->state == SS_CONNECTING))) {
1076 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1077 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1080 /* Look for a message in receive queue; wait if necessary */
1082 while (skb_queue_empty(&sk->sk_receive_queue)) {
1083 if (sock->state == SS_DISCONNECTING) {
1087 if (timeout <= 0L) {
1088 res = timeout ? timeout : -EWOULDBLOCK;
1092 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1093 tipc_rx_ready(sock),
1098 /* Look at first message in receive queue */
1100 buf = skb_peek(&sk->sk_receive_queue);
1102 sz = msg_data_sz(msg);
1103 err = msg_errcode(msg);
1105 /* Discard an empty non-errored message & try again */
1107 if ((!sz) && (!err)) {
1108 advance_rx_queue(sk);
1112 /* Optionally capture sender's address & ancillary data of first msg */
1114 if (sz_copied == 0) {
1115 set_orig_addr(m, msg);
1116 res = anc_data_recv(m, msg, tport);
1121 /* Capture message data (if valid) & compute return value (always) */
1124 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1127 needed = (buf_len - sz_copied);
1128 sz_to_copy = (sz <= needed) ? sz : needed;
1130 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1131 m->msg_iov, sz_to_copy);
1135 sz_copied += sz_to_copy;
1137 if (sz_to_copy < sz) {
1138 if (!(flags & MSG_PEEK))
1139 TIPC_SKB_CB(buf)->handle =
1140 (void *)(unsigned long)(offset + sz_to_copy);
1145 goto exit; /* can't add error msg to valid data */
1147 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1153 /* Consume received message (optional) */
1155 if (likely(!(flags & MSG_PEEK))) {
1156 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1157 tipc_acknowledge(tport->ref, tport->conn_unacked);
1158 advance_rx_queue(sk);
1161 /* Loop around if more data is required */
1163 if ((sz_copied < buf_len) && /* didn't get all requested data */
1164 (!skb_queue_empty(&sk->sk_receive_queue) ||
1165 (sz_copied < target)) && /* and more is ready or required */
1166 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1167 (!err)) /* and haven't reached a FIN */
1172 return sz_copied ? sz_copied : res;
1176 * rx_queue_full - determine if receive queue can accept another message
1177 * @msg: message to be added to queue
1178 * @queue_size: current size of queue
1179 * @base: nominal maximum size of queue
1181 * Returns 1 if queue is unable to accept message, 0 otherwise
1184 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1187 u32 imp = msg_importance(msg);
1189 if (imp == TIPC_LOW_IMPORTANCE)
1191 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1192 threshold = base * 2;
1193 else if (imp == TIPC_HIGH_IMPORTANCE)
1194 threshold = base * 100;
1198 if (msg_connected(msg))
1201 return queue_size >= threshold;
1205 * filter_rcv - validate incoming message
1209 * Enqueues message on receive queue if acceptable; optionally handles
1210 * disconnect indication for a connected socket.
1212 * Called with socket lock already taken; port lock may also be taken.
1214 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1217 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1219 struct socket *sock = sk->sk_socket;
1220 struct tipc_msg *msg = buf_msg(buf);
1223 /* Reject message if it is wrong sort of message for socket */
1226 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1227 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1228 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1231 if (sock->state == SS_READY) {
1232 if (msg_connected(msg))
1233 return TIPC_ERR_NO_PORT;
1236 return TIPC_ERR_NO_PORT;
1237 if (sock->state == SS_CONNECTED) {
1238 if (!msg_connected(msg))
1239 return TIPC_ERR_NO_PORT;
1240 } else if (sock->state == SS_CONNECTING) {
1241 if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1242 return TIPC_ERR_NO_PORT;
1243 } else if (sock->state == SS_LISTENING) {
1244 if (msg_connected(msg) || msg_errcode(msg))
1245 return TIPC_ERR_NO_PORT;
1246 } else if (sock->state == SS_DISCONNECTING) {
1247 return TIPC_ERR_NO_PORT;
1248 } else /* (sock->state == SS_UNCONNECTED) */ {
1249 if (msg_connected(msg) || msg_errcode(msg))
1250 return TIPC_ERR_NO_PORT;
1254 /* Reject message if there isn't room to queue it */
1256 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1257 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1258 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1259 return TIPC_ERR_OVERLOAD;
1261 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1262 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1263 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1264 return TIPC_ERR_OVERLOAD;
1267 /* Enqueue message (finally!) */
1269 TIPC_SKB_CB(buf)->handle = 0;
1270 atomic_inc(&tipc_queue_size);
1271 __skb_queue_tail(&sk->sk_receive_queue, buf);
1273 /* Initiate connection termination for an incoming 'FIN' */
1275 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1276 sock->state = SS_DISCONNECTING;
1277 tipc_disconnect_port(tipc_sk_port(sk));
1280 if (waitqueue_active(sk_sleep(sk)))
1281 wake_up_interruptible(sk_sleep(sk));
1286 * backlog_rcv - handle incoming message from backlog queue
1290 * Caller must hold socket lock, but not port lock.
1295 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1299 res = filter_rcv(sk, buf);
1301 tipc_reject_msg(buf, res);
1306 * dispatch - handle incoming message
1307 * @tport: TIPC port that received message
1310 * Called with port lock already taken.
1312 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1315 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1317 struct sock *sk = (struct sock *)tport->usr_handle;
1321 * Process message if socket is unlocked; otherwise add to backlog queue
1323 * This code is based on sk_receive_skb(), but must be distinct from it
1324 * since a TIPC-specific filter/reject mechanism is utilized
1328 if (!sock_owned_by_user(sk)) {
1329 res = filter_rcv(sk, buf);
1331 if (sk_add_backlog(sk, buf))
1332 res = TIPC_ERR_OVERLOAD;
1342 * wakeupdispatch - wake up port after congestion
1343 * @tport: port to wakeup
1345 * Called with port lock already taken.
1348 static void wakeupdispatch(struct tipc_port *tport)
1350 struct sock *sk = (struct sock *)tport->usr_handle;
1352 if (waitqueue_active(sk_sleep(sk)))
1353 wake_up_interruptible(sk_sleep(sk));
1357 * connect - establish a connection to another TIPC port
1358 * @sock: socket structure
1359 * @dest: socket address for destination port
1360 * @destlen: size of socket address data structure
1361 * @flags: file-related flags associated with socket
1363 * Returns 0 on success, errno otherwise
1366 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1369 struct sock *sk = sock->sk;
1370 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1371 struct msghdr m = {NULL,};
1372 struct sk_buff *buf;
1373 struct tipc_msg *msg;
1374 unsigned int timeout;
1379 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1381 if (sock->state == SS_READY) {
1386 /* For now, TIPC does not support the non-blocking form of connect() */
1388 if (flags & O_NONBLOCK) {
1393 /* Issue Posix-compliant error code if socket is in the wrong state */
1395 if (sock->state == SS_LISTENING) {
1399 if (sock->state == SS_CONNECTING) {
1403 if (sock->state != SS_UNCONNECTED) {
1409 * Reject connection attempt using multicast address
1411 * Note: send_msg() validates the rest of the address fields,
1412 * so there's no need to do it here
1415 if (dst->addrtype == TIPC_ADDR_MCAST) {
1420 /* Reject any messages already in receive queue (very unlikely) */
1422 reject_rx_queue(sk);
1424 /* Send a 'SYN-' to destination */
1427 m.msg_namelen = destlen;
1428 res = send_msg(NULL, sock, &m, 0);
1432 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1434 timeout = tipc_sk(sk)->conn_timeout;
1436 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1437 (!skb_queue_empty(&sk->sk_receive_queue) ||
1438 (sock->state != SS_CONNECTING)),
1439 timeout ? (long)msecs_to_jiffies(timeout)
1440 : MAX_SCHEDULE_TIMEOUT);
1444 buf = skb_peek(&sk->sk_receive_queue);
1447 res = auto_connect(sock, msg);
1449 if (!msg_data_sz(msg))
1450 advance_rx_queue(sk);
1453 if (sock->state == SS_CONNECTED)
1456 res = -ECONNREFUSED;
1462 ; /* leave "res" unchanged */
1463 sock->state = SS_DISCONNECTING;
1472 * listen - allow socket to listen for incoming connections
1473 * @sock: socket structure
1476 * Returns 0 on success, errno otherwise
1479 static int listen(struct socket *sock, int len)
1481 struct sock *sk = sock->sk;
1486 if (sock->state != SS_UNCONNECTED)
1489 sock->state = SS_LISTENING;
1498 * accept - wait for connection request
1499 * @sock: listening socket
1500 * @newsock: new socket that is to be connected
1501 * @flags: file-related flags associated with socket
1503 * Returns 0 on success, errno otherwise
1506 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1508 struct sock *sk = sock->sk;
1509 struct sk_buff *buf;
1514 if (sock->state != SS_LISTENING) {
1519 while (skb_queue_empty(&sk->sk_receive_queue)) {
1520 if (flags & O_NONBLOCK) {
1525 res = wait_event_interruptible(*sk_sleep(sk),
1526 (!skb_queue_empty(&sk->sk_receive_queue)));
1532 buf = skb_peek(&sk->sk_receive_queue);
1534 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1536 struct sock *new_sk = new_sock->sk;
1537 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1538 struct tipc_port *new_tport = new_tsock->p;
1539 u32 new_ref = new_tport->ref;
1540 struct tipc_msg *msg = buf_msg(buf);
1545 * Reject any stray messages received by new socket
1546 * before the socket lock was taken (very, very unlikely)
1549 reject_rx_queue(new_sk);
1551 /* Connect new socket to it's peer */
1553 new_tsock->peer_name.ref = msg_origport(msg);
1554 new_tsock->peer_name.node = msg_orignode(msg);
1555 tipc_connect2port(new_ref, &new_tsock->peer_name);
1556 new_sock->state = SS_CONNECTED;
1558 tipc_set_portimportance(new_ref, msg_importance(msg));
1559 if (msg_named(msg)) {
1560 new_tport->conn_type = msg_nametype(msg);
1561 new_tport->conn_instance = msg_nameinst(msg);
1565 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1566 * Respond to 'SYN+' by queuing it on new socket.
1569 if (!msg_data_sz(msg)) {
1570 struct msghdr m = {NULL,};
1572 advance_rx_queue(sk);
1573 send_packet(NULL, new_sock, &m, 0);
1575 __skb_dequeue(&sk->sk_receive_queue);
1576 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1578 release_sock(new_sk);
1586 * shutdown - shutdown socket connection
1587 * @sock: socket structure
1588 * @how: direction to close (must be SHUT_RDWR)
1590 * Terminates connection (if necessary), then purges socket's receive queue.
1592 * Returns 0 on success, errno otherwise
1595 static int shutdown(struct socket *sock, int how)
1597 struct sock *sk = sock->sk;
1598 struct tipc_port *tport = tipc_sk_port(sk);
1599 struct sk_buff *buf;
1602 if (how != SHUT_RDWR)
1607 switch (sock->state) {
1611 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1613 buf = __skb_dequeue(&sk->sk_receive_queue);
1615 atomic_dec(&tipc_queue_size);
1616 if (TIPC_SKB_CB(buf)->handle != 0) {
1620 tipc_disconnect(tport->ref);
1621 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1623 tipc_shutdown(tport->ref);
1626 sock->state = SS_DISCONNECTING;
1630 case SS_DISCONNECTING:
1632 /* Discard any unreceived messages; wake up sleeping tasks */
1634 discard_rx_queue(sk);
1635 if (waitqueue_active(sk_sleep(sk)))
1636 wake_up_interruptible(sk_sleep(sk));
1649 * setsockopt - set socket option
1650 * @sock: socket structure
1651 * @lvl: option level
1652 * @opt: option identifier
1653 * @ov: pointer to new option value
1654 * @ol: length of option value
1656 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1657 * (to ease compatibility).
1659 * Returns 0 on success, errno otherwise
1662 static int setsockopt(struct socket *sock,
1663 int lvl, int opt, char __user *ov, unsigned int ol)
1665 struct sock *sk = sock->sk;
1666 struct tipc_port *tport = tipc_sk_port(sk);
1670 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1672 if (lvl != SOL_TIPC)
1673 return -ENOPROTOOPT;
1674 if (ol < sizeof(value))
1676 res = get_user(value, (u32 __user *)ov);
1683 case TIPC_IMPORTANCE:
1684 res = tipc_set_portimportance(tport->ref, value);
1686 case TIPC_SRC_DROPPABLE:
1687 if (sock->type != SOCK_STREAM)
1688 res = tipc_set_portunreliable(tport->ref, value);
1692 case TIPC_DEST_DROPPABLE:
1693 res = tipc_set_portunreturnable(tport->ref, value);
1695 case TIPC_CONN_TIMEOUT:
1696 tipc_sk(sk)->conn_timeout = value;
1697 /* no need to set "res", since already 0 at this point */
1709 * getsockopt - get socket option
1710 * @sock: socket structure
1711 * @lvl: option level
1712 * @opt: option identifier
1713 * @ov: receptacle for option value
1714 * @ol: receptacle for length of option value
1716 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1717 * (to ease compatibility).
1719 * Returns 0 on success, errno otherwise
1722 static int getsockopt(struct socket *sock,
1723 int lvl, int opt, char __user *ov, int __user *ol)
1725 struct sock *sk = sock->sk;
1726 struct tipc_port *tport = tipc_sk_port(sk);
1731 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1732 return put_user(0, ol);
1733 if (lvl != SOL_TIPC)
1734 return -ENOPROTOOPT;
1735 res = get_user(len, ol);
1742 case TIPC_IMPORTANCE:
1743 res = tipc_portimportance(tport->ref, &value);
1745 case TIPC_SRC_DROPPABLE:
1746 res = tipc_portunreliable(tport->ref, &value);
1748 case TIPC_DEST_DROPPABLE:
1749 res = tipc_portunreturnable(tport->ref, &value);
1751 case TIPC_CONN_TIMEOUT:
1752 value = tipc_sk(sk)->conn_timeout;
1753 /* no need to set "res", since already 0 at this point */
1755 case TIPC_NODE_RECVQ_DEPTH:
1756 value = (u32)atomic_read(&tipc_queue_size);
1758 case TIPC_SOCK_RECVQ_DEPTH:
1759 value = skb_queue_len(&sk->sk_receive_queue);
1768 return res; /* "get" failed */
1770 if (len < sizeof(value))
1773 if (copy_to_user(ov, &value, sizeof(value)))
1776 return put_user(sizeof(value), ol);
1780 * Protocol switches for the various types of TIPC sockets
1783 static const struct proto_ops msg_ops = {
1784 .owner = THIS_MODULE,
1789 .socketpair = sock_no_socketpair,
1790 .accept = sock_no_accept,
1791 .getname = get_name,
1793 .ioctl = sock_no_ioctl,
1794 .listen = sock_no_listen,
1795 .shutdown = shutdown,
1796 .setsockopt = setsockopt,
1797 .getsockopt = getsockopt,
1798 .sendmsg = send_msg,
1799 .recvmsg = recv_msg,
1800 .mmap = sock_no_mmap,
1801 .sendpage = sock_no_sendpage
1804 static const struct proto_ops packet_ops = {
1805 .owner = THIS_MODULE,
1810 .socketpair = sock_no_socketpair,
1812 .getname = get_name,
1814 .ioctl = sock_no_ioctl,
1816 .shutdown = shutdown,
1817 .setsockopt = setsockopt,
1818 .getsockopt = getsockopt,
1819 .sendmsg = send_packet,
1820 .recvmsg = recv_msg,
1821 .mmap = sock_no_mmap,
1822 .sendpage = sock_no_sendpage
1825 static const struct proto_ops stream_ops = {
1826 .owner = THIS_MODULE,
1831 .socketpair = sock_no_socketpair,
1833 .getname = get_name,
1835 .ioctl = sock_no_ioctl,
1837 .shutdown = shutdown,
1838 .setsockopt = setsockopt,
1839 .getsockopt = getsockopt,
1840 .sendmsg = send_stream,
1841 .recvmsg = recv_stream,
1842 .mmap = sock_no_mmap,
1843 .sendpage = sock_no_sendpage
1846 static const struct net_proto_family tipc_family_ops = {
1847 .owner = THIS_MODULE,
1849 .create = tipc_create
1852 static struct proto tipc_proto = {
1854 .owner = THIS_MODULE,
1855 .obj_size = sizeof(struct tipc_sock)
1859 * tipc_socket_init - initialize TIPC socket interface
1861 * Returns 0 on success, errno otherwise
1863 int tipc_socket_init(void)
1867 res = proto_register(&tipc_proto, 1);
1869 err("Failed to register TIPC protocol type\n");
1873 res = sock_register(&tipc_family_ops);
1875 err("Failed to register TIPC socket type\n");
1876 proto_unregister(&tipc_proto);
1880 sockets_enabled = 1;
1886 * tipc_socket_stop - stop TIPC socket interface
1889 void tipc_socket_stop(void)
1891 if (!sockets_enabled)
1894 sockets_enabled = 0;
1895 sock_unregister(tipc_family_ops.family);
1896 proto_unregister(&tipc_proto);