2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, Ericsson AB
5 * Copyright (c) 2004-2008, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/module.h>
38 #include <linux/types.h>
39 #include <linux/net.h>
40 #include <linux/socket.h>
41 #include <linux/errno.h>
43 #include <linux/poll.h>
44 #include <linux/fcntl.h>
45 #include <linux/gfp.h>
46 #include <asm/string.h>
47 #include <asm/atomic.h>
50 #include <linux/tipc.h>
51 #include <linux/tipc_config.h>
52 #include <net/tipc/tipc_msg.h>
53 #include <net/tipc/tipc_port.h>
57 #define SS_LISTENING -1 /* socket is listening */
58 #define SS_READY -2 /* socket is connectionless */
60 #define OVERLOAD_LIMIT_BASE 5000
61 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
66 struct tipc_portid peer_name;
70 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
71 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
73 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
74 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
75 static void wakeupdispatch(struct tipc_port *tport);
77 static const struct proto_ops packet_ops;
78 static const struct proto_ops stream_ops;
79 static const struct proto_ops msg_ops;
81 static struct proto tipc_proto;
83 static int sockets_enabled = 0;
85 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
88 * Revised TIPC socket locking policy:
90 * Most socket operations take the standard socket lock when they start
91 * and hold it until they finish (or until they need to sleep). Acquiring
92 * this lock grants the owner exclusive access to the fields of the socket
93 * data structures, with the exception of the backlog queue. A few socket
94 * operations can be done without taking the socket lock because they only
95 * read socket information that never changes during the life of the socket.
97 * Socket operations may acquire the lock for the associated TIPC port if they
98 * need to perform an operation on the port. If any routine needs to acquire
99 * both the socket lock and the port lock it must take the socket lock first
100 * to avoid the risk of deadlock.
102 * The dispatcher handling incoming messages cannot grab the socket lock in
103 * the standard fashion, since invoked it runs at the BH level and cannot block.
104 * Instead, it checks to see if the socket lock is currently owned by someone,
105 * and either handles the message itself or adds it to the socket's backlog
106 * queue; in the latter case the queued message is processed once the process
107 * owning the socket lock releases it.
109 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
110 * the problem of a blocked socket operation preventing any other operations
111 * from occurring. However, applications must be careful if they have
112 * multiple threads trying to send (or receive) on the same socket, as these
113 * operations might interfere with each other. For example, doing a connect
114 * and a receive at the same time might allow the receive to consume the
115 * ACK message meant for the connect. While additional work could be done
116 * to try and overcome this, it doesn't seem to be worthwhile at the present.
118 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
119 * that another operation that must be performed in a non-blocking manner is
120 * not delayed for very long because the lock has already been taken.
122 * NOTE: This code assumes that certain fields of a port/socket pair are
123 * constant over its lifetime; such fields can be examined without taking
124 * the socket lock and/or port lock, and do not need to be re-read even
125 * after resuming processing after waiting. These fields include:
127 * - pointer to socket sk structure (aka tipc_sock structure)
128 * - pointer to port structure
133 * advance_rx_queue - discard first buffer in socket receive queue
135 * Caller must hold socket lock
138 static void advance_rx_queue(struct sock *sk)
140 buf_discard(__skb_dequeue(&sk->sk_receive_queue));
141 atomic_dec(&tipc_queue_size);
145 * discard_rx_queue - discard all buffers in socket receive queue
147 * Caller must hold socket lock
150 static void discard_rx_queue(struct sock *sk)
154 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
155 atomic_dec(&tipc_queue_size);
161 * reject_rx_queue - reject all buffers in socket receive queue
163 * Caller must hold socket lock
166 static void reject_rx_queue(struct sock *sk)
170 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
171 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
172 atomic_dec(&tipc_queue_size);
177 * tipc_create - create a TIPC socket
178 * @net: network namespace (must be default network)
179 * @sock: pre-allocated socket structure
180 * @protocol: protocol indicator (must be 0)
181 * @kern: caused by kernel or by userspace?
183 * This routine creates additional data structures used by the TIPC socket,
184 * initializes them, and links them together.
186 * Returns 0 on success, errno otherwise
189 static int tipc_create(struct net *net, struct socket *sock, int protocol,
192 const struct proto_ops *ops;
195 struct tipc_port *tp_ptr;
197 /* Validate arguments */
199 if (!net_eq(net, &init_net))
200 return -EAFNOSUPPORT;
202 if (unlikely(protocol != 0))
203 return -EPROTONOSUPPORT;
205 switch (sock->type) {
208 state = SS_UNCONNECTED;
212 state = SS_UNCONNECTED;
223 /* Allocate socket's protocol area */
225 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
229 /* Allocate TIPC port for socket to use */
231 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
232 TIPC_LOW_IMPORTANCE);
233 if (unlikely(!tp_ptr)) {
238 /* Finish initializing socket data structures */
243 sock_init_data(sock, sk);
244 sk->sk_backlog_rcv = backlog_rcv;
245 tipc_sk(sk)->p = tp_ptr;
246 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
248 spin_unlock_bh(tp_ptr->lock);
250 if (sock->state == SS_READY) {
251 tipc_set_portunreturnable(tp_ptr->ref, 1);
252 if (sock->type == SOCK_DGRAM)
253 tipc_set_portunreliable(tp_ptr->ref, 1);
256 atomic_inc(&tipc_user_count);
261 * release - destroy a TIPC socket
262 * @sock: socket to destroy
264 * This routine cleans up any messages that are still queued on the socket.
265 * For DGRAM and RDM socket types, all queued messages are rejected.
266 * For SEQPACKET and STREAM socket types, the first message is rejected
267 * and any others are discarded. (If the first message on a STREAM socket
268 * is partially-read, it is discarded and the next one is rejected instead.)
270 * NOTE: Rejected messages are not necessarily returned to the sender! They
271 * are returned or discarded according to the "destination droppable" setting
272 * specified for the message by the sender.
274 * Returns 0 on success, errno otherwise
277 static int release(struct socket *sock)
279 struct sock *sk = sock->sk;
280 struct tipc_port *tport;
285 * Exit if socket isn't fully initialized (occurs when a failed accept()
286 * releases a pre-allocated child socket that was never used)
292 tport = tipc_sk_port(sk);
296 * Reject all unreceived messages, except on an active connection
297 * (which disconnects locally & sends a 'FIN+' to peer)
300 while (sock->state != SS_DISCONNECTING) {
301 buf = __skb_dequeue(&sk->sk_receive_queue);
304 atomic_dec(&tipc_queue_size);
305 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
308 if ((sock->state == SS_CONNECTING) ||
309 (sock->state == SS_CONNECTED)) {
310 sock->state = SS_DISCONNECTING;
311 tipc_disconnect(tport->ref);
313 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
318 * Delete TIPC port; this ensures no more messages are queued
319 * (also disconnects an active connection & sends a 'FIN-' to peer)
322 res = tipc_deleteport(tport->ref);
324 /* Discard any remaining (connection-based) messages in receive queue */
326 discard_rx_queue(sk);
328 /* Reject any messages that accumulated in backlog queue */
330 sock->state = SS_DISCONNECTING;
336 atomic_dec(&tipc_user_count);
341 * bind - associate or disassocate TIPC name(s) with a socket
342 * @sock: socket structure
343 * @uaddr: socket address describing name(s) and desired operation
344 * @uaddr_len: size of socket address data structure
346 * Name and name sequence binding is indicated using a positive scope value;
347 * a negative scope value unbinds the specified name. Specifying no name
348 * (i.e. a socket address length of 0) unbinds all names from the socket.
350 * Returns 0 on success, errno otherwise
352 * NOTE: This routine doesn't need to take the socket lock since it doesn't
353 * access any non-constant socket information.
356 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
358 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
359 u32 portref = tipc_sk_port(sock->sk)->ref;
361 if (unlikely(!uaddr_len))
362 return tipc_withdraw(portref, 0, NULL);
364 if (uaddr_len < sizeof(struct sockaddr_tipc))
366 if (addr->family != AF_TIPC)
367 return -EAFNOSUPPORT;
369 if (addr->addrtype == TIPC_ADDR_NAME)
370 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
371 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
372 return -EAFNOSUPPORT;
374 return (addr->scope > 0) ?
375 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
376 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
380 * get_name - get port ID of socket or peer socket
381 * @sock: socket structure
382 * @uaddr: area for returned socket address
383 * @uaddr_len: area for returned length of socket address
384 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
386 * Returns 0 on success, errno otherwise
388 * NOTE: This routine doesn't need to take the socket lock since it only
389 * accesses socket information that is unchanging (or which changes in
390 * a completely predictable manner).
393 static int get_name(struct socket *sock, struct sockaddr *uaddr,
394 int *uaddr_len, int peer)
396 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
397 struct tipc_sock *tsock = tipc_sk(sock->sk);
400 if ((sock->state != SS_CONNECTED) &&
401 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
403 addr->addr.id.ref = tsock->peer_name.ref;
404 addr->addr.id.node = tsock->peer_name.node;
406 tipc_ownidentity(tsock->p->ref, &addr->addr.id);
409 *uaddr_len = sizeof(*addr);
410 addr->addrtype = TIPC_ADDR_ID;
411 addr->family = AF_TIPC;
413 addr->addr.name.domain = 0;
419 * poll - read and possibly block on pollmask
420 * @file: file structure associated with the socket
421 * @sock: socket for which to calculate the poll bits
424 * Returns pollmask value
427 * It appears that the usual socket locking mechanisms are not useful here
428 * since the pollmask info is potentially out-of-date the moment this routine
429 * exits. TCP and other protocols seem to rely on higher level poll routines
430 * to handle any preventable race conditions, so TIPC will do the same ...
432 * TIPC sets the returned events as follows:
434 * socket state flags set
435 * ------------ ---------
436 * unconnected no read flags
439 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
442 * connected POLLIN/POLLRDNORM if data in rx queue
443 * POLLOUT if port is not congested
445 * disconnecting POLLIN/POLLRDNORM/POLLHUP
448 * listening POLLIN if SYN in rx queue
451 * ready POLLIN/POLLRDNORM if data in rx queue
452 * [connectionless] POLLOUT (since port cannot be congested)
454 * IMPORTANT: The fact that a read or write operation is indicated does NOT
455 * imply that the operation will succeed, merely that it should be performed
456 * and will not block.
459 static unsigned int poll(struct file *file, struct socket *sock,
462 struct sock *sk = sock->sk;
465 poll_wait(file, sk_sleep(sk), wait);
467 switch ((int)sock->state) {
470 if (!tipc_sk_port(sk)->congested)
475 if (!skb_queue_empty(&sk->sk_receive_queue))
476 mask |= (POLLIN | POLLRDNORM);
478 case SS_DISCONNECTING:
479 mask = (POLLIN | POLLRDNORM | POLLHUP);
487 * dest_name_check - verify user is permitted to send to specified port name
488 * @dest: destination address
489 * @m: descriptor for message to be sent
491 * Prevents restricted configuration commands from being issued by
492 * unauthorized users.
494 * Returns 0 if permission is granted, otherwise errno
497 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
499 struct tipc_cfg_msg_hdr hdr;
501 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
503 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
505 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
508 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
510 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
517 * send_msg - send message in connectionless manner
518 * @iocb: if NULL, indicates that socket lock is already held
519 * @sock: socket structure
520 * @m: message to send
521 * @total_len: length of message
523 * Message must have an destination specified explicitly.
524 * Used for SOCK_RDM and SOCK_DGRAM messages,
525 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
526 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
528 * Returns the number of bytes sent on success, or errno otherwise
531 static int send_msg(struct kiocb *iocb, struct socket *sock,
532 struct msghdr *m, size_t total_len)
534 struct sock *sk = sock->sk;
535 struct tipc_port *tport = tipc_sk_port(sk);
536 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
541 return -EDESTADDRREQ;
542 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
543 (dest->family != AF_TIPC)))
549 needs_conn = (sock->state != SS_READY);
550 if (unlikely(needs_conn)) {
551 if (sock->state == SS_LISTENING) {
555 if (sock->state != SS_UNCONNECTED) {
559 if ((tport->published) ||
560 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
564 if (dest->addrtype == TIPC_ADDR_NAME) {
565 tport->conn_type = dest->addr.name.name.type;
566 tport->conn_instance = dest->addr.name.name.instance;
569 /* Abort any pending connection attempts (very unlikely) */
575 if (dest->addrtype == TIPC_ADDR_NAME) {
576 if ((res = dest_name_check(dest, m)))
578 res = tipc_send2name(tport->ref,
579 &dest->addr.name.name,
580 dest->addr.name.domain,
584 else if (dest->addrtype == TIPC_ADDR_ID) {
585 res = tipc_send2port(tport->ref,
590 else if (dest->addrtype == TIPC_ADDR_MCAST) {
595 if ((res = dest_name_check(dest, m)))
597 res = tipc_multicast(tport->ref,
603 if (likely(res != -ELINKCONG)) {
604 if (needs_conn && (res >= 0)) {
605 sock->state = SS_CONNECTING;
609 if (m->msg_flags & MSG_DONTWAIT) {
614 res = wait_event_interruptible(*sk_sleep(sk),
628 * send_packet - send a connection-oriented message
629 * @iocb: if NULL, indicates that socket lock is already held
630 * @sock: socket structure
631 * @m: message to send
632 * @total_len: length of message
634 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
636 * Returns the number of bytes sent on success, or errno otherwise
639 static int send_packet(struct kiocb *iocb, struct socket *sock,
640 struct msghdr *m, size_t total_len)
642 struct sock *sk = sock->sk;
643 struct tipc_port *tport = tipc_sk_port(sk);
644 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
647 /* Handle implied connection establishment */
650 return send_msg(iocb, sock, m, total_len);
656 if (unlikely(sock->state != SS_CONNECTED)) {
657 if (sock->state == SS_DISCONNECTING)
664 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
665 if (likely(res != -ELINKCONG)) {
668 if (m->msg_flags & MSG_DONTWAIT) {
673 res = wait_event_interruptible(*sk_sleep(sk),
674 (!tport->congested || !tport->connected));
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)) {
737 * Send each iovec entry using one or more messages
739 * Note: This algorithm is good for the most likely case
740 * (i.e. one large iovec entry), but could be improved to pass sets
741 * of small iovec entries into send_packet().
744 curr_iov = m->msg_iov;
745 curr_iovlen = m->msg_iovlen;
746 my_msg.msg_iov = &my_iov;
747 my_msg.msg_iovlen = 1;
748 my_msg.msg_flags = m->msg_flags;
749 my_msg.msg_name = NULL;
752 hdr_size = msg_hdr_sz(&tport->phdr);
754 while (curr_iovlen--) {
755 curr_start = curr_iov->iov_base;
756 curr_left = curr_iov->iov_len;
759 bytes_to_send = tport->max_pkt - hdr_size;
760 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
761 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
762 if (curr_left < bytes_to_send)
763 bytes_to_send = curr_left;
764 my_iov.iov_base = curr_start;
765 my_iov.iov_len = bytes_to_send;
766 if ((res = send_packet(NULL, sock, &my_msg, 0)) < 0) {
771 curr_left -= bytes_to_send;
772 curr_start += bytes_to_send;
773 bytes_sent += bytes_to_send;
785 * auto_connect - complete connection setup to a remote port
786 * @sock: socket structure
787 * @msg: peer's response message
789 * Returns 0 on success, errno otherwise
792 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
794 struct tipc_sock *tsock = tipc_sk(sock->sk);
796 if (msg_errcode(msg)) {
797 sock->state = SS_DISCONNECTING;
798 return -ECONNREFUSED;
801 tsock->peer_name.ref = msg_origport(msg);
802 tsock->peer_name.node = msg_orignode(msg);
803 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
804 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
805 sock->state = SS_CONNECTED;
810 * set_orig_addr - capture sender's address for received message
811 * @m: descriptor for message info
812 * @msg: received message header
814 * Note: Address is not captured if not requested by receiver.
817 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
819 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
822 addr->family = AF_TIPC;
823 addr->addrtype = TIPC_ADDR_ID;
824 addr->addr.id.ref = msg_origport(msg);
825 addr->addr.id.node = msg_orignode(msg);
826 addr->addr.name.domain = 0; /* could leave uninitialized */
827 addr->scope = 0; /* could leave uninitialized */
828 m->msg_namelen = sizeof(struct sockaddr_tipc);
833 * anc_data_recv - optionally capture ancillary data for received message
834 * @m: descriptor for message info
835 * @msg: received message header
836 * @tport: TIPC port associated with message
838 * Note: Ancillary data is not captured if not requested by receiver.
840 * Returns 0 if successful, otherwise errno
843 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
844 struct tipc_port *tport)
852 if (likely(m->msg_controllen == 0))
855 /* Optionally capture errored message object(s) */
857 err = msg ? msg_errcode(msg) : 0;
860 anc_data[1] = msg_data_sz(msg);
861 if ((res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data)))
864 (res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
869 /* Optionally capture message destination object */
871 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
875 anc_data[0] = msg_nametype(msg);
876 anc_data[1] = msg_namelower(msg);
877 anc_data[2] = msg_namelower(msg);
881 anc_data[0] = msg_nametype(msg);
882 anc_data[1] = msg_namelower(msg);
883 anc_data[2] = msg_nameupper(msg);
886 has_name = (tport->conn_type != 0);
887 anc_data[0] = tport->conn_type;
888 anc_data[1] = tport->conn_instance;
889 anc_data[2] = tport->conn_instance;
895 (res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data)))
902 * recv_msg - receive packet-oriented message
904 * @m: descriptor for message info
905 * @buf_len: total size of user buffer area
906 * @flags: receive flags
908 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
909 * If the complete message doesn't fit in user area, truncate it.
911 * Returns size of returned message data, errno otherwise
914 static int recv_msg(struct kiocb *iocb, struct socket *sock,
915 struct msghdr *m, size_t buf_len, int flags)
917 struct sock *sk = sock->sk;
918 struct tipc_port *tport = tipc_sk_port(sk);
920 struct tipc_msg *msg;
925 /* Catch invalid receive requests */
927 if (m->msg_iovlen != 1)
928 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */
930 if (unlikely(!buf_len))
935 if (unlikely(sock->state == SS_UNCONNECTED)) {
942 /* Look for a message in receive queue; wait if necessary */
944 while (skb_queue_empty(&sk->sk_receive_queue)) {
945 if (sock->state == SS_DISCONNECTING) {
949 if (flags & MSG_DONTWAIT) {
954 res = wait_event_interruptible(*sk_sleep(sk),
955 (!skb_queue_empty(&sk->sk_receive_queue) ||
956 (sock->state == SS_DISCONNECTING)));
962 /* Look at first message in receive queue */
964 buf = skb_peek(&sk->sk_receive_queue);
966 sz = msg_data_sz(msg);
967 err = msg_errcode(msg);
969 /* Complete connection setup for an implied connect */
971 if (unlikely(sock->state == SS_CONNECTING)) {
972 res = auto_connect(sock, msg);
977 /* Discard an empty non-errored message & try again */
979 if ((!sz) && (!err)) {
980 advance_rx_queue(sk);
984 /* Capture sender's address (optional) */
986 set_orig_addr(m, msg);
988 /* Capture ancillary data (optional) */
990 res = anc_data_recv(m, msg, tport);
994 /* Capture message data (if valid) & compute return value (always) */
997 if (unlikely(buf_len < sz)) {
999 m->msg_flags |= MSG_TRUNC;
1001 if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
1008 if ((sock->state == SS_READY) ||
1009 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1015 /* Consume received message (optional) */
1017 if (likely(!(flags & MSG_PEEK))) {
1018 if ((sock->state != SS_READY) &&
1019 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1020 tipc_acknowledge(tport->ref, tport->conn_unacked);
1021 advance_rx_queue(sk);
1029 * recv_stream - receive stream-oriented data
1031 * @m: descriptor for message info
1032 * @buf_len: total size of user buffer area
1033 * @flags: receive flags
1035 * Used for SOCK_STREAM messages only. If not enough data is available
1036 * will optionally wait for more; never truncates data.
1038 * Returns size of returned message data, errno otherwise
1041 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1042 struct msghdr *m, size_t buf_len, int flags)
1044 struct sock *sk = sock->sk;
1045 struct tipc_port *tport = tipc_sk_port(sk);
1046 struct sk_buff *buf;
1047 struct tipc_msg *msg;
1049 int sz_to_copy, target, needed;
1051 char __user *crs = m->msg_iov->iov_base;
1052 unsigned char *buf_crs;
1056 /* Catch invalid receive attempts */
1058 if (m->msg_iovlen != 1)
1059 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */
1061 if (unlikely(!buf_len))
1066 if (unlikely((sock->state == SS_UNCONNECTED) ||
1067 (sock->state == SS_CONNECTING))) {
1072 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1076 /* Look for a message in receive queue; wait if necessary */
1078 while (skb_queue_empty(&sk->sk_receive_queue)) {
1079 if (sock->state == SS_DISCONNECTING) {
1083 if (flags & MSG_DONTWAIT) {
1088 res = wait_event_interruptible(*sk_sleep(sk),
1089 (!skb_queue_empty(&sk->sk_receive_queue) ||
1090 (sock->state == SS_DISCONNECTING)));
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 buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1123 sz = (unsigned char *)msg + msg_size(msg) - buf_crs;
1125 needed = (buf_len - sz_copied);
1126 sz_to_copy = (sz <= needed) ? sz : needed;
1127 if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1131 sz_copied += sz_to_copy;
1133 if (sz_to_copy < sz) {
1134 if (!(flags & MSG_PEEK))
1135 TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1142 goto exit; /* can't add error msg to valid data */
1144 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1150 /* Consume received message (optional) */
1152 if (likely(!(flags & MSG_PEEK))) {
1153 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1154 tipc_acknowledge(tport->ref, tport->conn_unacked);
1155 advance_rx_queue(sk);
1158 /* Loop around if more data is required */
1160 if ((sz_copied < buf_len) && /* didn't get all requested data */
1161 (!skb_queue_empty(&sk->sk_receive_queue) ||
1162 (sz_copied < target)) && /* and more is ready or required */
1163 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1164 (!err)) /* and haven't reached a FIN */
1169 return sz_copied ? sz_copied : res;
1173 * rx_queue_full - determine if receive queue can accept another message
1174 * @msg: message to be added to queue
1175 * @queue_size: current size of queue
1176 * @base: nominal maximum size of queue
1178 * Returns 1 if queue is unable to accept message, 0 otherwise
1181 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1184 u32 imp = msg_importance(msg);
1186 if (imp == TIPC_LOW_IMPORTANCE)
1188 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1189 threshold = base * 2;
1190 else if (imp == TIPC_HIGH_IMPORTANCE)
1191 threshold = base * 100;
1195 if (msg_connected(msg))
1198 return queue_size >= threshold;
1202 * filter_rcv - validate incoming message
1206 * Enqueues message on receive queue if acceptable; optionally handles
1207 * disconnect indication for a connected socket.
1209 * Called with socket lock already taken; port lock may also be taken.
1211 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1214 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1216 struct socket *sock = sk->sk_socket;
1217 struct tipc_msg *msg = buf_msg(buf);
1220 /* Reject message if it is wrong sort of message for socket */
1223 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1224 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1225 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1228 if (sock->state == SS_READY) {
1229 if (msg_connected(msg)) {
1230 msg_dbg(msg, "dispatch filter 1\n");
1231 return TIPC_ERR_NO_PORT;
1234 if (msg_mcast(msg)) {
1235 msg_dbg(msg, "dispatch filter 2\n");
1236 return TIPC_ERR_NO_PORT;
1238 if (sock->state == SS_CONNECTED) {
1239 if (!msg_connected(msg)) {
1240 msg_dbg(msg, "dispatch filter 3\n");
1241 return TIPC_ERR_NO_PORT;
1244 else if (sock->state == SS_CONNECTING) {
1245 if (!msg_connected(msg) && (msg_errcode(msg) == 0)) {
1246 msg_dbg(msg, "dispatch filter 4\n");
1247 return TIPC_ERR_NO_PORT;
1250 else if (sock->state == SS_LISTENING) {
1251 if (msg_connected(msg) || msg_errcode(msg)) {
1252 msg_dbg(msg, "dispatch filter 5\n");
1253 return TIPC_ERR_NO_PORT;
1256 else if (sock->state == SS_DISCONNECTING) {
1257 msg_dbg(msg, "dispatch filter 6\n");
1258 return TIPC_ERR_NO_PORT;
1260 else /* (sock->state == SS_UNCONNECTED) */ {
1261 if (msg_connected(msg) || msg_errcode(msg)) {
1262 msg_dbg(msg, "dispatch filter 7\n");
1263 return TIPC_ERR_NO_PORT;
1268 /* Reject message if there isn't room to queue it */
1270 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1271 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1272 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1273 return TIPC_ERR_OVERLOAD;
1275 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1276 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1277 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1278 return TIPC_ERR_OVERLOAD;
1281 /* Enqueue message (finally!) */
1283 msg_dbg(msg, "<DISP<: ");
1284 TIPC_SKB_CB(buf)->handle = msg_data(msg);
1285 atomic_inc(&tipc_queue_size);
1286 __skb_queue_tail(&sk->sk_receive_queue, buf);
1288 /* Initiate connection termination for an incoming 'FIN' */
1290 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1291 sock->state = SS_DISCONNECTING;
1292 tipc_disconnect_port(tipc_sk_port(sk));
1295 if (waitqueue_active(sk_sleep(sk)))
1296 wake_up_interruptible(sk_sleep(sk));
1301 * backlog_rcv - handle incoming message from backlog queue
1305 * Caller must hold socket lock, but not port lock.
1310 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1314 res = filter_rcv(sk, buf);
1316 tipc_reject_msg(buf, res);
1321 * dispatch - handle incoming message
1322 * @tport: TIPC port that received message
1325 * Called with port lock already taken.
1327 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1330 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1332 struct sock *sk = (struct sock *)tport->usr_handle;
1336 * Process message if socket is unlocked; otherwise add to backlog queue
1338 * This code is based on sk_receive_skb(), but must be distinct from it
1339 * since a TIPC-specific filter/reject mechanism is utilized
1343 if (!sock_owned_by_user(sk)) {
1344 res = filter_rcv(sk, buf);
1346 if (sk_add_backlog(sk, buf))
1347 res = TIPC_ERR_OVERLOAD;
1357 * wakeupdispatch - wake up port after congestion
1358 * @tport: port to wakeup
1360 * Called with port lock already taken.
1363 static void wakeupdispatch(struct tipc_port *tport)
1365 struct sock *sk = (struct sock *)tport->usr_handle;
1367 if (waitqueue_active(sk_sleep(sk)))
1368 wake_up_interruptible(sk_sleep(sk));
1372 * connect - establish a connection to another TIPC port
1373 * @sock: socket structure
1374 * @dest: socket address for destination port
1375 * @destlen: size of socket address data structure
1376 * @flags: file-related flags associated with socket
1378 * Returns 0 on success, errno otherwise
1381 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1384 struct sock *sk = sock->sk;
1385 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1386 struct msghdr m = {NULL,};
1387 struct sk_buff *buf;
1388 struct tipc_msg *msg;
1394 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1396 if (sock->state == SS_READY) {
1401 /* For now, TIPC does not support the non-blocking form of connect() */
1403 if (flags & O_NONBLOCK) {
1408 /* Issue Posix-compliant error code if socket is in the wrong state */
1410 if (sock->state == SS_LISTENING) {
1414 if (sock->state == SS_CONNECTING) {
1418 if (sock->state != SS_UNCONNECTED) {
1424 * Reject connection attempt using multicast address
1426 * Note: send_msg() validates the rest of the address fields,
1427 * so there's no need to do it here
1430 if (dst->addrtype == TIPC_ADDR_MCAST) {
1435 /* Reject any messages already in receive queue (very unlikely) */
1437 reject_rx_queue(sk);
1439 /* Send a 'SYN-' to destination */
1442 m.msg_namelen = destlen;
1443 res = send_msg(NULL, sock, &m, 0);
1448 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1450 timeout = tipc_sk(sk)->conn_timeout;
1452 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1453 (!skb_queue_empty(&sk->sk_receive_queue) ||
1454 (sock->state != SS_CONNECTING)),
1455 timeout ? timeout : MAX_SCHEDULE_TIMEOUT);
1459 buf = skb_peek(&sk->sk_receive_queue);
1462 res = auto_connect(sock, msg);
1464 if (!msg_data_sz(msg))
1465 advance_rx_queue(sk);
1468 if (sock->state == SS_CONNECTED) {
1471 res = -ECONNREFUSED;
1478 ; /* leave "res" unchanged */
1479 sock->state = SS_DISCONNECTING;
1488 * listen - allow socket to listen for incoming connections
1489 * @sock: socket structure
1492 * Returns 0 on success, errno otherwise
1495 static int listen(struct socket *sock, int len)
1497 struct sock *sk = sock->sk;
1502 if (sock->state == SS_READY)
1504 else if (sock->state != SS_UNCONNECTED)
1507 sock->state = SS_LISTENING;
1516 * accept - wait for connection request
1517 * @sock: listening socket
1518 * @newsock: new socket that is to be connected
1519 * @flags: file-related flags associated with socket
1521 * Returns 0 on success, errno otherwise
1524 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1526 struct sock *sk = sock->sk;
1527 struct sk_buff *buf;
1532 if (sock->state == SS_READY) {
1536 if (sock->state != SS_LISTENING) {
1541 while (skb_queue_empty(&sk->sk_receive_queue)) {
1542 if (flags & O_NONBLOCK) {
1547 res = wait_event_interruptible(*sk_sleep(sk),
1548 (!skb_queue_empty(&sk->sk_receive_queue)));
1554 buf = skb_peek(&sk->sk_receive_queue);
1556 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1558 struct sock *new_sk = new_sock->sk;
1559 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1560 struct tipc_port *new_tport = new_tsock->p;
1561 u32 new_ref = new_tport->ref;
1562 struct tipc_msg *msg = buf_msg(buf);
1567 * Reject any stray messages received by new socket
1568 * before the socket lock was taken (very, very unlikely)
1571 reject_rx_queue(new_sk);
1573 /* Connect new socket to it's peer */
1575 new_tsock->peer_name.ref = msg_origport(msg);
1576 new_tsock->peer_name.node = msg_orignode(msg);
1577 tipc_connect2port(new_ref, &new_tsock->peer_name);
1578 new_sock->state = SS_CONNECTED;
1580 tipc_set_portimportance(new_ref, msg_importance(msg));
1581 if (msg_named(msg)) {
1582 new_tport->conn_type = msg_nametype(msg);
1583 new_tport->conn_instance = msg_nameinst(msg);
1587 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1588 * Respond to 'SYN+' by queuing it on new socket.
1591 msg_dbg(msg,"<ACC<: ");
1592 if (!msg_data_sz(msg)) {
1593 struct msghdr m = {NULL,};
1595 advance_rx_queue(sk);
1596 send_packet(NULL, new_sock, &m, 0);
1598 __skb_dequeue(&sk->sk_receive_queue);
1599 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1601 release_sock(new_sk);
1609 * shutdown - shutdown socket connection
1610 * @sock: socket structure
1611 * @how: direction to close (must be SHUT_RDWR)
1613 * Terminates connection (if necessary), then purges socket's receive queue.
1615 * Returns 0 on success, errno otherwise
1618 static int shutdown(struct socket *sock, int how)
1620 struct sock *sk = sock->sk;
1621 struct tipc_port *tport = tipc_sk_port(sk);
1622 struct sk_buff *buf;
1625 if (how != SHUT_RDWR)
1630 switch (sock->state) {
1634 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1636 buf = __skb_dequeue(&sk->sk_receive_queue);
1638 atomic_dec(&tipc_queue_size);
1639 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1643 tipc_disconnect(tport->ref);
1644 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1646 tipc_shutdown(tport->ref);
1649 sock->state = SS_DISCONNECTING;
1653 case SS_DISCONNECTING:
1655 /* Discard any unreceived messages; wake up sleeping tasks */
1657 discard_rx_queue(sk);
1658 if (waitqueue_active(sk_sleep(sk)))
1659 wake_up_interruptible(sk_sleep(sk));
1672 * setsockopt - set socket option
1673 * @sock: socket structure
1674 * @lvl: option level
1675 * @opt: option identifier
1676 * @ov: pointer to new option value
1677 * @ol: length of option value
1679 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1680 * (to ease compatibility).
1682 * Returns 0 on success, errno otherwise
1685 static int setsockopt(struct socket *sock,
1686 int lvl, int opt, char __user *ov, unsigned int ol)
1688 struct sock *sk = sock->sk;
1689 struct tipc_port *tport = tipc_sk_port(sk);
1693 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1695 if (lvl != SOL_TIPC)
1696 return -ENOPROTOOPT;
1697 if (ol < sizeof(value))
1699 if ((res = get_user(value, (u32 __user *)ov)))
1705 case TIPC_IMPORTANCE:
1706 res = tipc_set_portimportance(tport->ref, value);
1708 case TIPC_SRC_DROPPABLE:
1709 if (sock->type != SOCK_STREAM)
1710 res = tipc_set_portunreliable(tport->ref, value);
1714 case TIPC_DEST_DROPPABLE:
1715 res = tipc_set_portunreturnable(tport->ref, value);
1717 case TIPC_CONN_TIMEOUT:
1718 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value);
1719 /* no need to set "res", since already 0 at this point */
1731 * getsockopt - get socket option
1732 * @sock: socket structure
1733 * @lvl: option level
1734 * @opt: option identifier
1735 * @ov: receptacle for option value
1736 * @ol: receptacle for length of option value
1738 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1739 * (to ease compatibility).
1741 * Returns 0 on success, errno otherwise
1744 static int getsockopt(struct socket *sock,
1745 int lvl, int opt, char __user *ov, int __user *ol)
1747 struct sock *sk = sock->sk;
1748 struct tipc_port *tport = tipc_sk_port(sk);
1753 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1754 return put_user(0, ol);
1755 if (lvl != SOL_TIPC)
1756 return -ENOPROTOOPT;
1757 if ((res = get_user(len, ol)))
1763 case TIPC_IMPORTANCE:
1764 res = tipc_portimportance(tport->ref, &value);
1766 case TIPC_SRC_DROPPABLE:
1767 res = tipc_portunreliable(tport->ref, &value);
1769 case TIPC_DEST_DROPPABLE:
1770 res = tipc_portunreturnable(tport->ref, &value);
1772 case TIPC_CONN_TIMEOUT:
1773 value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout);
1774 /* no need to set "res", since already 0 at this point */
1776 case TIPC_NODE_RECVQ_DEPTH:
1777 value = (u32)atomic_read(&tipc_queue_size);
1779 case TIPC_SOCK_RECVQ_DEPTH:
1780 value = skb_queue_len(&sk->sk_receive_queue);
1791 else if (len < sizeof(value)) {
1794 else if (copy_to_user(ov, &value, sizeof(value))) {
1798 res = put_user(sizeof(value), ol);
1805 * Protocol switches for the various types of TIPC sockets
1808 static const struct proto_ops msg_ops = {
1809 .owner = THIS_MODULE,
1814 .socketpair = sock_no_socketpair,
1816 .getname = get_name,
1818 .ioctl = sock_no_ioctl,
1820 .shutdown = shutdown,
1821 .setsockopt = setsockopt,
1822 .getsockopt = getsockopt,
1823 .sendmsg = send_msg,
1824 .recvmsg = recv_msg,
1825 .mmap = sock_no_mmap,
1826 .sendpage = sock_no_sendpage
1829 static const struct proto_ops packet_ops = {
1830 .owner = THIS_MODULE,
1835 .socketpair = sock_no_socketpair,
1837 .getname = get_name,
1839 .ioctl = sock_no_ioctl,
1841 .shutdown = shutdown,
1842 .setsockopt = setsockopt,
1843 .getsockopt = getsockopt,
1844 .sendmsg = send_packet,
1845 .recvmsg = recv_msg,
1846 .mmap = sock_no_mmap,
1847 .sendpage = sock_no_sendpage
1850 static const struct proto_ops stream_ops = {
1851 .owner = THIS_MODULE,
1856 .socketpair = sock_no_socketpair,
1858 .getname = get_name,
1860 .ioctl = sock_no_ioctl,
1862 .shutdown = shutdown,
1863 .setsockopt = setsockopt,
1864 .getsockopt = getsockopt,
1865 .sendmsg = send_stream,
1866 .recvmsg = recv_stream,
1867 .mmap = sock_no_mmap,
1868 .sendpage = sock_no_sendpage
1871 static const struct net_proto_family tipc_family_ops = {
1872 .owner = THIS_MODULE,
1874 .create = tipc_create
1877 static struct proto tipc_proto = {
1879 .owner = THIS_MODULE,
1880 .obj_size = sizeof(struct tipc_sock)
1884 * tipc_socket_init - initialize TIPC socket interface
1886 * Returns 0 on success, errno otherwise
1888 int tipc_socket_init(void)
1892 res = proto_register(&tipc_proto, 1);
1894 err("Failed to register TIPC protocol type\n");
1898 res = sock_register(&tipc_family_ops);
1900 err("Failed to register TIPC socket type\n");
1901 proto_unregister(&tipc_proto);
1905 sockets_enabled = 1;
1911 * tipc_socket_stop - stop TIPC socket interface
1914 void tipc_socket_stop(void)
1916 if (!sockets_enabled)
1919 sockets_enabled = 0;
1920 sock_unregister(tipc_family_ops.family);
1921 proto_unregister(&tipc_proto);