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;
69 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
70 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
72 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
73 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
74 static void wakeupdispatch(struct tipc_port *tport);
76 static const struct proto_ops packet_ops;
77 static const struct proto_ops stream_ops;
78 static const struct proto_ops msg_ops;
80 static struct proto tipc_proto;
82 static int sockets_enabled = 0;
84 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
87 * Revised TIPC socket locking policy:
89 * Most socket operations take the standard socket lock when they start
90 * and hold it until they finish (or until they need to sleep). Acquiring
91 * this lock grants the owner exclusive access to the fields of the socket
92 * data structures, with the exception of the backlog queue. A few socket
93 * operations can be done without taking the socket lock because they only
94 * read socket information that never changes during the life of the socket.
96 * Socket operations may acquire the lock for the associated TIPC port if they
97 * need to perform an operation on the port. If any routine needs to acquire
98 * both the socket lock and the port lock it must take the socket lock first
99 * to avoid the risk of deadlock.
101 * The dispatcher handling incoming messages cannot grab the socket lock in
102 * the standard fashion, since invoked it runs at the BH level and cannot block.
103 * Instead, it checks to see if the socket lock is currently owned by someone,
104 * and either handles the message itself or adds it to the socket's backlog
105 * queue; in the latter case the queued message is processed once the process
106 * owning the socket lock releases it.
108 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
109 * the problem of a blocked socket operation preventing any other operations
110 * from occurring. However, applications must be careful if they have
111 * multiple threads trying to send (or receive) on the same socket, as these
112 * operations might interfere with each other. For example, doing a connect
113 * and a receive at the same time might allow the receive to consume the
114 * ACK message meant for the connect. While additional work could be done
115 * to try and overcome this, it doesn't seem to be worthwhile at the present.
117 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
118 * that another operation that must be performed in a non-blocking manner is
119 * not delayed for very long because the lock has already been taken.
121 * NOTE: This code assumes that certain fields of a port/socket pair are
122 * constant over its lifetime; such fields can be examined without taking
123 * the socket lock and/or port lock, and do not need to be re-read even
124 * after resuming processing after waiting. These fields include:
126 * - pointer to socket sk structure (aka tipc_sock structure)
127 * - pointer to port structure
132 * advance_rx_queue - discard first buffer in socket receive queue
134 * Caller must hold socket lock
137 static void advance_rx_queue(struct sock *sk)
139 buf_discard(__skb_dequeue(&sk->sk_receive_queue));
140 atomic_dec(&tipc_queue_size);
144 * discard_rx_queue - discard all buffers in socket receive queue
146 * Caller must hold socket lock
149 static void discard_rx_queue(struct sock *sk)
153 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
154 atomic_dec(&tipc_queue_size);
160 * reject_rx_queue - reject all buffers in socket receive queue
162 * Caller must hold socket lock
165 static void reject_rx_queue(struct sock *sk)
169 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
170 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
171 atomic_dec(&tipc_queue_size);
176 * tipc_create - create a TIPC socket
177 * @net: network namespace (must be default network)
178 * @sock: pre-allocated socket structure
179 * @protocol: protocol indicator (must be 0)
180 * @kern: caused by kernel or by userspace?
182 * This routine creates additional data structures used by the TIPC socket,
183 * initializes them, and links them together.
185 * Returns 0 on success, errno otherwise
188 static int tipc_create(struct net *net, struct socket *sock, int protocol,
191 const struct proto_ops *ops;
194 struct tipc_port *tp_ptr;
196 /* Validate arguments */
198 if (!net_eq(net, &init_net))
199 return -EAFNOSUPPORT;
201 if (unlikely(protocol != 0))
202 return -EPROTONOSUPPORT;
204 switch (sock->type) {
207 state = SS_UNCONNECTED;
211 state = SS_UNCONNECTED;
222 /* Allocate socket's protocol area */
224 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
228 /* Allocate TIPC port for socket to use */
230 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
231 TIPC_LOW_IMPORTANCE);
232 if (unlikely(!tp_ptr)) {
237 /* Finish initializing socket data structures */
242 sock_init_data(sock, sk);
243 sk->sk_rcvtimeo = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
244 sk->sk_backlog_rcv = backlog_rcv;
245 tipc_sk(sk)->p = tp_ptr;
247 spin_unlock_bh(tp_ptr->lock);
249 if (sock->state == SS_READY) {
250 tipc_set_portunreturnable(tp_ptr->ref, 1);
251 if (sock->type == SOCK_DGRAM)
252 tipc_set_portunreliable(tp_ptr->ref, 1);
255 atomic_inc(&tipc_user_count);
260 * release - destroy a TIPC socket
261 * @sock: socket to destroy
263 * This routine cleans up any messages that are still queued on the socket.
264 * For DGRAM and RDM socket types, all queued messages are rejected.
265 * For SEQPACKET and STREAM socket types, the first message is rejected
266 * and any others are discarded. (If the first message on a STREAM socket
267 * is partially-read, it is discarded and the next one is rejected instead.)
269 * NOTE: Rejected messages are not necessarily returned to the sender! They
270 * are returned or discarded according to the "destination droppable" setting
271 * specified for the message by the sender.
273 * Returns 0 on success, errno otherwise
276 static int release(struct socket *sock)
278 struct sock *sk = sock->sk;
279 struct tipc_port *tport;
284 * Exit if socket isn't fully initialized (occurs when a failed accept()
285 * releases a pre-allocated child socket that was never used)
291 tport = tipc_sk_port(sk);
295 * Reject all unreceived messages, except on an active connection
296 * (which disconnects locally & sends a 'FIN+' to peer)
299 while (sock->state != SS_DISCONNECTING) {
300 buf = __skb_dequeue(&sk->sk_receive_queue);
303 atomic_dec(&tipc_queue_size);
304 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
307 if ((sock->state == SS_CONNECTING) ||
308 (sock->state == SS_CONNECTED)) {
309 sock->state = SS_DISCONNECTING;
310 tipc_disconnect(tport->ref);
312 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
317 * Delete TIPC port; this ensures no more messages are queued
318 * (also disconnects an active connection & sends a 'FIN-' to peer)
321 res = tipc_deleteport(tport->ref);
323 /* Discard any remaining (connection-based) messages in receive queue */
325 discard_rx_queue(sk);
327 /* Reject any messages that accumulated in backlog queue */
329 sock->state = SS_DISCONNECTING;
335 atomic_dec(&tipc_user_count);
340 * bind - associate or disassocate TIPC name(s) with a socket
341 * @sock: socket structure
342 * @uaddr: socket address describing name(s) and desired operation
343 * @uaddr_len: size of socket address data structure
345 * Name and name sequence binding is indicated using a positive scope value;
346 * a negative scope value unbinds the specified name. Specifying no name
347 * (i.e. a socket address length of 0) unbinds all names from the socket.
349 * Returns 0 on success, errno otherwise
351 * NOTE: This routine doesn't need to take the socket lock since it doesn't
352 * access any non-constant socket information.
355 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
357 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
358 u32 portref = tipc_sk_port(sock->sk)->ref;
360 if (unlikely(!uaddr_len))
361 return tipc_withdraw(portref, 0, NULL);
363 if (uaddr_len < sizeof(struct sockaddr_tipc))
365 if (addr->family != AF_TIPC)
366 return -EAFNOSUPPORT;
368 if (addr->addrtype == TIPC_ADDR_NAME)
369 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
370 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
371 return -EAFNOSUPPORT;
373 return (addr->scope > 0) ?
374 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
375 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
379 * get_name - get port ID of socket or peer socket
380 * @sock: socket structure
381 * @uaddr: area for returned socket address
382 * @uaddr_len: area for returned length of socket address
383 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
385 * Returns 0 on success, errno otherwise
387 * NOTE: This routine doesn't need to take the socket lock since it only
388 * accesses socket information that is unchanging (or which changes in
389 * a completely predictable manner).
392 static int get_name(struct socket *sock, struct sockaddr *uaddr,
393 int *uaddr_len, int peer)
395 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
396 struct tipc_sock *tsock = tipc_sk(sock->sk);
399 if ((sock->state != SS_CONNECTED) &&
400 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
402 addr->addr.id.ref = tsock->peer_name.ref;
403 addr->addr.id.node = tsock->peer_name.node;
405 tipc_ownidentity(tsock->p->ref, &addr->addr.id);
408 *uaddr_len = sizeof(*addr);
409 addr->addrtype = TIPC_ADDR_ID;
410 addr->family = AF_TIPC;
412 addr->addr.name.domain = 0;
418 * poll - read and possibly block on pollmask
419 * @file: file structure associated with the socket
420 * @sock: socket for which to calculate the poll bits
423 * Returns pollmask value
426 * It appears that the usual socket locking mechanisms are not useful here
427 * since the pollmask info is potentially out-of-date the moment this routine
428 * exits. TCP and other protocols seem to rely on higher level poll routines
429 * to handle any preventable race conditions, so TIPC will do the same ...
431 * TIPC sets the returned events as follows:
432 * a) POLLRDNORM and POLLIN are set if the socket's receive queue is non-empty
433 * or if a connection-oriented socket is does not have an active connection
434 * (i.e. a read operation will not block).
435 * b) POLLOUT is set except when a socket's connection has been terminated
436 * (i.e. a write operation will not block).
437 * c) POLLHUP is set when a socket's connection has been terminated.
439 * IMPORTANT: The fact that a read or write operation will not block does NOT
440 * imply that the operation will succeed!
443 static unsigned int poll(struct file *file, struct socket *sock,
446 struct sock *sk = sock->sk;
449 poll_wait(file, sk_sleep(sk), wait);
451 if (!skb_queue_empty(&sk->sk_receive_queue) ||
452 (sock->state == SS_UNCONNECTED) ||
453 (sock->state == SS_DISCONNECTING))
454 mask = (POLLRDNORM | POLLIN);
458 if (sock->state == SS_DISCONNECTING)
467 * dest_name_check - verify user is permitted to send to specified port name
468 * @dest: destination address
469 * @m: descriptor for message to be sent
471 * Prevents restricted configuration commands from being issued by
472 * unauthorized users.
474 * Returns 0 if permission is granted, otherwise errno
477 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
479 struct tipc_cfg_msg_hdr hdr;
481 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
483 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
485 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
488 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
490 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
497 * send_msg - send message in connectionless manner
498 * @iocb: if NULL, indicates that socket lock is already held
499 * @sock: socket structure
500 * @m: message to send
501 * @total_len: length of message
503 * Message must have an destination specified explicitly.
504 * Used for SOCK_RDM and SOCK_DGRAM messages,
505 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
506 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
508 * Returns the number of bytes sent on success, or errno otherwise
511 static int send_msg(struct kiocb *iocb, struct socket *sock,
512 struct msghdr *m, size_t total_len)
514 struct sock *sk = sock->sk;
515 struct tipc_port *tport = tipc_sk_port(sk);
516 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
521 return -EDESTADDRREQ;
522 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
523 (dest->family != AF_TIPC)))
529 needs_conn = (sock->state != SS_READY);
530 if (unlikely(needs_conn)) {
531 if (sock->state == SS_LISTENING) {
535 if (sock->state != SS_UNCONNECTED) {
539 if ((tport->published) ||
540 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
544 if (dest->addrtype == TIPC_ADDR_NAME) {
545 tport->conn_type = dest->addr.name.name.type;
546 tport->conn_instance = dest->addr.name.name.instance;
549 /* Abort any pending connection attempts (very unlikely) */
555 if (dest->addrtype == TIPC_ADDR_NAME) {
556 if ((res = dest_name_check(dest, m)))
558 res = tipc_send2name(tport->ref,
559 &dest->addr.name.name,
560 dest->addr.name.domain,
564 else if (dest->addrtype == TIPC_ADDR_ID) {
565 res = tipc_send2port(tport->ref,
570 else if (dest->addrtype == TIPC_ADDR_MCAST) {
575 if ((res = dest_name_check(dest, m)))
577 res = tipc_multicast(tport->ref,
583 if (likely(res != -ELINKCONG)) {
584 if (needs_conn && (res >= 0)) {
585 sock->state = SS_CONNECTING;
589 if (m->msg_flags & MSG_DONTWAIT) {
594 res = wait_event_interruptible(*sk_sleep(sk),
608 * send_packet - send a connection-oriented message
609 * @iocb: if NULL, indicates that socket lock is already held
610 * @sock: socket structure
611 * @m: message to send
612 * @total_len: length of message
614 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
616 * Returns the number of bytes sent on success, or errno otherwise
619 static int send_packet(struct kiocb *iocb, struct socket *sock,
620 struct msghdr *m, size_t total_len)
622 struct sock *sk = sock->sk;
623 struct tipc_port *tport = tipc_sk_port(sk);
624 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
627 /* Handle implied connection establishment */
630 return send_msg(iocb, sock, m, total_len);
636 if (unlikely(sock->state != SS_CONNECTED)) {
637 if (sock->state == SS_DISCONNECTING)
644 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
645 if (likely(res != -ELINKCONG)) {
648 if (m->msg_flags & MSG_DONTWAIT) {
653 res = wait_event_interruptible(*sk_sleep(sk),
654 (!tport->congested || !tport->connected));
666 * send_stream - send stream-oriented data
668 * @sock: socket structure
670 * @total_len: total length of data to be sent
672 * Used for SOCK_STREAM data.
674 * Returns the number of bytes sent on success (or partial success),
675 * or errno if no data sent
678 static int send_stream(struct kiocb *iocb, struct socket *sock,
679 struct msghdr *m, size_t total_len)
681 struct sock *sk = sock->sk;
682 struct tipc_port *tport = tipc_sk_port(sk);
683 struct msghdr my_msg;
685 struct iovec *curr_iov;
687 char __user *curr_start;
696 /* Handle special cases where there is no connection */
698 if (unlikely(sock->state != SS_CONNECTED)) {
699 if (sock->state == SS_UNCONNECTED) {
700 res = send_packet(NULL, sock, m, total_len);
702 } else if (sock->state == SS_DISCONNECTING) {
711 if (unlikely(m->msg_name)) {
717 * Send each iovec entry using one or more messages
719 * Note: This algorithm is good for the most likely case
720 * (i.e. one large iovec entry), but could be improved to pass sets
721 * of small iovec entries into send_packet().
724 curr_iov = m->msg_iov;
725 curr_iovlen = m->msg_iovlen;
726 my_msg.msg_iov = &my_iov;
727 my_msg.msg_iovlen = 1;
728 my_msg.msg_flags = m->msg_flags;
729 my_msg.msg_name = NULL;
732 hdr_size = msg_hdr_sz(&tport->phdr);
734 while (curr_iovlen--) {
735 curr_start = curr_iov->iov_base;
736 curr_left = curr_iov->iov_len;
739 bytes_to_send = tport->max_pkt - hdr_size;
740 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
741 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
742 if (curr_left < bytes_to_send)
743 bytes_to_send = curr_left;
744 my_iov.iov_base = curr_start;
745 my_iov.iov_len = bytes_to_send;
746 if ((res = send_packet(NULL, sock, &my_msg, 0)) < 0) {
751 curr_left -= bytes_to_send;
752 curr_start += bytes_to_send;
753 bytes_sent += bytes_to_send;
765 * auto_connect - complete connection setup to a remote port
766 * @sock: socket structure
767 * @msg: peer's response message
769 * Returns 0 on success, errno otherwise
772 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
774 struct tipc_sock *tsock = tipc_sk(sock->sk);
776 if (msg_errcode(msg)) {
777 sock->state = SS_DISCONNECTING;
778 return -ECONNREFUSED;
781 tsock->peer_name.ref = msg_origport(msg);
782 tsock->peer_name.node = msg_orignode(msg);
783 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
784 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
785 sock->state = SS_CONNECTED;
790 * set_orig_addr - capture sender's address for received message
791 * @m: descriptor for message info
792 * @msg: received message header
794 * Note: Address is not captured if not requested by receiver.
797 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
799 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
802 addr->family = AF_TIPC;
803 addr->addrtype = TIPC_ADDR_ID;
804 addr->addr.id.ref = msg_origport(msg);
805 addr->addr.id.node = msg_orignode(msg);
806 addr->addr.name.domain = 0; /* could leave uninitialized */
807 addr->scope = 0; /* could leave uninitialized */
808 m->msg_namelen = sizeof(struct sockaddr_tipc);
813 * anc_data_recv - optionally capture ancillary data for received message
814 * @m: descriptor for message info
815 * @msg: received message header
816 * @tport: TIPC port associated with message
818 * Note: Ancillary data is not captured if not requested by receiver.
820 * Returns 0 if successful, otherwise errno
823 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
824 struct tipc_port *tport)
832 if (likely(m->msg_controllen == 0))
835 /* Optionally capture errored message object(s) */
837 err = msg ? msg_errcode(msg) : 0;
840 anc_data[1] = msg_data_sz(msg);
841 if ((res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data)))
844 (res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
849 /* Optionally capture message destination object */
851 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
855 anc_data[0] = msg_nametype(msg);
856 anc_data[1] = msg_namelower(msg);
857 anc_data[2] = msg_namelower(msg);
861 anc_data[0] = msg_nametype(msg);
862 anc_data[1] = msg_namelower(msg);
863 anc_data[2] = msg_nameupper(msg);
866 has_name = (tport->conn_type != 0);
867 anc_data[0] = tport->conn_type;
868 anc_data[1] = tport->conn_instance;
869 anc_data[2] = tport->conn_instance;
875 (res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data)))
882 * recv_msg - receive packet-oriented message
884 * @m: descriptor for message info
885 * @buf_len: total size of user buffer area
886 * @flags: receive flags
888 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
889 * If the complete message doesn't fit in user area, truncate it.
891 * Returns size of returned message data, errno otherwise
894 static int recv_msg(struct kiocb *iocb, struct socket *sock,
895 struct msghdr *m, size_t buf_len, int flags)
897 struct sock *sk = sock->sk;
898 struct tipc_port *tport = tipc_sk_port(sk);
900 struct tipc_msg *msg;
905 /* Catch invalid receive requests */
907 if (m->msg_iovlen != 1)
908 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */
910 if (unlikely(!buf_len))
915 if (unlikely(sock->state == SS_UNCONNECTED)) {
922 /* Look for a message in receive queue; wait if necessary */
924 while (skb_queue_empty(&sk->sk_receive_queue)) {
925 if (sock->state == SS_DISCONNECTING) {
929 if (flags & MSG_DONTWAIT) {
934 res = wait_event_interruptible(*sk_sleep(sk),
935 (!skb_queue_empty(&sk->sk_receive_queue) ||
936 (sock->state == SS_DISCONNECTING)));
942 /* Look at first message in receive queue */
944 buf = skb_peek(&sk->sk_receive_queue);
946 sz = msg_data_sz(msg);
947 err = msg_errcode(msg);
949 /* Complete connection setup for an implied connect */
951 if (unlikely(sock->state == SS_CONNECTING)) {
952 res = auto_connect(sock, msg);
957 /* Discard an empty non-errored message & try again */
959 if ((!sz) && (!err)) {
960 advance_rx_queue(sk);
964 /* Capture sender's address (optional) */
966 set_orig_addr(m, msg);
968 /* Capture ancillary data (optional) */
970 res = anc_data_recv(m, msg, tport);
974 /* Capture message data (if valid) & compute return value (always) */
977 if (unlikely(buf_len < sz)) {
979 m->msg_flags |= MSG_TRUNC;
981 if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
988 if ((sock->state == SS_READY) ||
989 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
995 /* Consume received message (optional) */
997 if (likely(!(flags & MSG_PEEK))) {
998 if ((sock->state != SS_READY) &&
999 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1000 tipc_acknowledge(tport->ref, tport->conn_unacked);
1001 advance_rx_queue(sk);
1009 * recv_stream - receive stream-oriented data
1011 * @m: descriptor for message info
1012 * @buf_len: total size of user buffer area
1013 * @flags: receive flags
1015 * Used for SOCK_STREAM messages only. If not enough data is available
1016 * will optionally wait for more; never truncates data.
1018 * Returns size of returned message data, errno otherwise
1021 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1022 struct msghdr *m, size_t buf_len, int flags)
1024 struct sock *sk = sock->sk;
1025 struct tipc_port *tport = tipc_sk_port(sk);
1026 struct sk_buff *buf;
1027 struct tipc_msg *msg;
1029 int sz_to_copy, target, needed;
1031 char __user *crs = m->msg_iov->iov_base;
1032 unsigned char *buf_crs;
1036 /* Catch invalid receive attempts */
1038 if (m->msg_iovlen != 1)
1039 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */
1041 if (unlikely(!buf_len))
1046 if (unlikely((sock->state == SS_UNCONNECTED) ||
1047 (sock->state == SS_CONNECTING))) {
1052 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1056 /* Look for a message in receive queue; wait if necessary */
1058 while (skb_queue_empty(&sk->sk_receive_queue)) {
1059 if (sock->state == SS_DISCONNECTING) {
1063 if (flags & MSG_DONTWAIT) {
1068 res = wait_event_interruptible(*sk_sleep(sk),
1069 (!skb_queue_empty(&sk->sk_receive_queue) ||
1070 (sock->state == SS_DISCONNECTING)));
1076 /* Look at first message in receive queue */
1078 buf = skb_peek(&sk->sk_receive_queue);
1080 sz = msg_data_sz(msg);
1081 err = msg_errcode(msg);
1083 /* Discard an empty non-errored message & try again */
1085 if ((!sz) && (!err)) {
1086 advance_rx_queue(sk);
1090 /* Optionally capture sender's address & ancillary data of first msg */
1092 if (sz_copied == 0) {
1093 set_orig_addr(m, msg);
1094 res = anc_data_recv(m, msg, tport);
1099 /* Capture message data (if valid) & compute return value (always) */
1102 buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1103 sz = (unsigned char *)msg + msg_size(msg) - buf_crs;
1105 needed = (buf_len - sz_copied);
1106 sz_to_copy = (sz <= needed) ? sz : needed;
1107 if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1111 sz_copied += sz_to_copy;
1113 if (sz_to_copy < sz) {
1114 if (!(flags & MSG_PEEK))
1115 TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1122 goto exit; /* can't add error msg to valid data */
1124 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1130 /* Consume received message (optional) */
1132 if (likely(!(flags & MSG_PEEK))) {
1133 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1134 tipc_acknowledge(tport->ref, tport->conn_unacked);
1135 advance_rx_queue(sk);
1138 /* Loop around if more data is required */
1140 if ((sz_copied < buf_len) && /* didn't get all requested data */
1141 (!skb_queue_empty(&sk->sk_receive_queue) ||
1142 (sz_copied < target)) && /* and more is ready or required */
1143 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1144 (!err)) /* and haven't reached a FIN */
1149 return sz_copied ? sz_copied : res;
1153 * rx_queue_full - determine if receive queue can accept another message
1154 * @msg: message to be added to queue
1155 * @queue_size: current size of queue
1156 * @base: nominal maximum size of queue
1158 * Returns 1 if queue is unable to accept message, 0 otherwise
1161 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1164 u32 imp = msg_importance(msg);
1166 if (imp == TIPC_LOW_IMPORTANCE)
1168 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1169 threshold = base * 2;
1170 else if (imp == TIPC_HIGH_IMPORTANCE)
1171 threshold = base * 100;
1175 if (msg_connected(msg))
1178 return (queue_size >= threshold);
1182 * filter_rcv - validate incoming message
1186 * Enqueues message on receive queue if acceptable; optionally handles
1187 * disconnect indication for a connected socket.
1189 * Called with socket lock already taken; port lock may also be taken.
1191 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1194 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1196 struct socket *sock = sk->sk_socket;
1197 struct tipc_msg *msg = buf_msg(buf);
1200 /* Reject message if it is wrong sort of message for socket */
1203 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1204 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1205 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1208 if (sock->state == SS_READY) {
1209 if (msg_connected(msg)) {
1210 msg_dbg(msg, "dispatch filter 1\n");
1211 return TIPC_ERR_NO_PORT;
1214 if (msg_mcast(msg)) {
1215 msg_dbg(msg, "dispatch filter 2\n");
1216 return TIPC_ERR_NO_PORT;
1218 if (sock->state == SS_CONNECTED) {
1219 if (!msg_connected(msg)) {
1220 msg_dbg(msg, "dispatch filter 3\n");
1221 return TIPC_ERR_NO_PORT;
1224 else if (sock->state == SS_CONNECTING) {
1225 if (!msg_connected(msg) && (msg_errcode(msg) == 0)) {
1226 msg_dbg(msg, "dispatch filter 4\n");
1227 return TIPC_ERR_NO_PORT;
1230 else if (sock->state == SS_LISTENING) {
1231 if (msg_connected(msg) || msg_errcode(msg)) {
1232 msg_dbg(msg, "dispatch filter 5\n");
1233 return TIPC_ERR_NO_PORT;
1236 else if (sock->state == SS_DISCONNECTING) {
1237 msg_dbg(msg, "dispatch filter 6\n");
1238 return TIPC_ERR_NO_PORT;
1240 else /* (sock->state == SS_UNCONNECTED) */ {
1241 if (msg_connected(msg) || msg_errcode(msg)) {
1242 msg_dbg(msg, "dispatch filter 7\n");
1243 return TIPC_ERR_NO_PORT;
1248 /* Reject message if there isn't room to queue it */
1250 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1251 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1252 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1253 return TIPC_ERR_OVERLOAD;
1255 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1256 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1257 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1258 return TIPC_ERR_OVERLOAD;
1261 /* Enqueue message (finally!) */
1263 msg_dbg(msg, "<DISP<: ");
1264 TIPC_SKB_CB(buf)->handle = msg_data(msg);
1265 atomic_inc(&tipc_queue_size);
1266 __skb_queue_tail(&sk->sk_receive_queue, buf);
1268 /* Initiate connection termination for an incoming 'FIN' */
1270 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1271 sock->state = SS_DISCONNECTING;
1272 tipc_disconnect_port(tipc_sk_port(sk));
1275 if (waitqueue_active(sk_sleep(sk)))
1276 wake_up_interruptible(sk_sleep(sk));
1281 * backlog_rcv - handle incoming message from backlog queue
1285 * Caller must hold socket lock, but not port lock.
1290 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1294 res = filter_rcv(sk, buf);
1296 tipc_reject_msg(buf, res);
1301 * dispatch - handle incoming message
1302 * @tport: TIPC port that received message
1305 * Called with port lock already taken.
1307 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1310 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1312 struct sock *sk = (struct sock *)tport->usr_handle;
1316 * Process message if socket is unlocked; otherwise add to backlog queue
1318 * This code is based on sk_receive_skb(), but must be distinct from it
1319 * since a TIPC-specific filter/reject mechanism is utilized
1323 if (!sock_owned_by_user(sk)) {
1324 res = filter_rcv(sk, buf);
1326 if (sk_add_backlog(sk, buf))
1327 res = TIPC_ERR_OVERLOAD;
1337 * wakeupdispatch - wake up port after congestion
1338 * @tport: port to wakeup
1340 * Called with port lock already taken.
1343 static void wakeupdispatch(struct tipc_port *tport)
1345 struct sock *sk = (struct sock *)tport->usr_handle;
1347 if (waitqueue_active(sk_sleep(sk)))
1348 wake_up_interruptible(sk_sleep(sk));
1352 * connect - establish a connection to another TIPC port
1353 * @sock: socket structure
1354 * @dest: socket address for destination port
1355 * @destlen: size of socket address data structure
1356 * @flags: file-related flags associated with socket
1358 * Returns 0 on success, errno otherwise
1361 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1364 struct sock *sk = sock->sk;
1365 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1366 struct msghdr m = {NULL,};
1367 struct sk_buff *buf;
1368 struct tipc_msg *msg;
1373 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1375 if (sock->state == SS_READY) {
1380 /* For now, TIPC does not support the non-blocking form of connect() */
1382 if (flags & O_NONBLOCK) {
1387 /* Issue Posix-compliant error code if socket is in the wrong state */
1389 if (sock->state == SS_LISTENING) {
1393 if (sock->state == SS_CONNECTING) {
1397 if (sock->state != SS_UNCONNECTED) {
1403 * Reject connection attempt using multicast address
1405 * Note: send_msg() validates the rest of the address fields,
1406 * so there's no need to do it here
1409 if (dst->addrtype == TIPC_ADDR_MCAST) {
1414 /* Reject any messages already in receive queue (very unlikely) */
1416 reject_rx_queue(sk);
1418 /* Send a 'SYN-' to destination */
1421 m.msg_namelen = destlen;
1422 res = send_msg(NULL, sock, &m, 0);
1427 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1430 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1431 (!skb_queue_empty(&sk->sk_receive_queue) ||
1432 (sock->state != SS_CONNECTING)),
1437 buf = skb_peek(&sk->sk_receive_queue);
1440 res = auto_connect(sock, msg);
1442 if (!msg_data_sz(msg))
1443 advance_rx_queue(sk);
1446 if (sock->state == SS_CONNECTED) {
1449 res = -ECONNREFUSED;
1456 ; /* leave "res" unchanged */
1457 sock->state = SS_DISCONNECTING;
1466 * listen - allow socket to listen for incoming connections
1467 * @sock: socket structure
1470 * Returns 0 on success, errno otherwise
1473 static int listen(struct socket *sock, int len)
1475 struct sock *sk = sock->sk;
1480 if (sock->state == SS_READY)
1482 else if (sock->state != SS_UNCONNECTED)
1485 sock->state = SS_LISTENING;
1494 * accept - wait for connection request
1495 * @sock: listening socket
1496 * @newsock: new socket that is to be connected
1497 * @flags: file-related flags associated with socket
1499 * Returns 0 on success, errno otherwise
1502 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1504 struct sock *sk = sock->sk;
1505 struct sk_buff *buf;
1510 if (sock->state == SS_READY) {
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 msg_dbg(msg,"<ACC<: ");
1570 if (!msg_data_sz(msg)) {
1571 struct msghdr m = {NULL,};
1573 advance_rx_queue(sk);
1574 send_packet(NULL, new_sock, &m, 0);
1576 __skb_dequeue(&sk->sk_receive_queue);
1577 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1579 release_sock(new_sk);
1587 * shutdown - shutdown socket connection
1588 * @sock: socket structure
1589 * @how: direction to close (must be SHUT_RDWR)
1591 * Terminates connection (if necessary), then purges socket's receive queue.
1593 * Returns 0 on success, errno otherwise
1596 static int shutdown(struct socket *sock, int how)
1598 struct sock *sk = sock->sk;
1599 struct tipc_port *tport = tipc_sk_port(sk);
1600 struct sk_buff *buf;
1603 if (how != SHUT_RDWR)
1608 switch (sock->state) {
1612 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1614 buf = __skb_dequeue(&sk->sk_receive_queue);
1616 atomic_dec(&tipc_queue_size);
1617 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1621 tipc_disconnect(tport->ref);
1622 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1624 tipc_shutdown(tport->ref);
1627 sock->state = SS_DISCONNECTING;
1631 case SS_DISCONNECTING:
1633 /* Discard any unreceived messages; wake up sleeping tasks */
1635 discard_rx_queue(sk);
1636 if (waitqueue_active(sk_sleep(sk)))
1637 wake_up_interruptible(sk_sleep(sk));
1650 * setsockopt - set socket option
1651 * @sock: socket structure
1652 * @lvl: option level
1653 * @opt: option identifier
1654 * @ov: pointer to new option value
1655 * @ol: length of option value
1657 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1658 * (to ease compatibility).
1660 * Returns 0 on success, errno otherwise
1663 static int setsockopt(struct socket *sock,
1664 int lvl, int opt, char __user *ov, unsigned int ol)
1666 struct sock *sk = sock->sk;
1667 struct tipc_port *tport = tipc_sk_port(sk);
1671 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1673 if (lvl != SOL_TIPC)
1674 return -ENOPROTOOPT;
1675 if (ol < sizeof(value))
1677 if ((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 sk->sk_rcvtimeo = msecs_to_jiffies(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 if ((res = get_user(len, ol)))
1741 case TIPC_IMPORTANCE:
1742 res = tipc_portimportance(tport->ref, &value);
1744 case TIPC_SRC_DROPPABLE:
1745 res = tipc_portunreliable(tport->ref, &value);
1747 case TIPC_DEST_DROPPABLE:
1748 res = tipc_portunreturnable(tport->ref, &value);
1750 case TIPC_CONN_TIMEOUT:
1751 value = jiffies_to_msecs(sk->sk_rcvtimeo);
1752 /* no need to set "res", since already 0 at this point */
1754 case TIPC_NODE_RECVQ_DEPTH:
1755 value = (u32)atomic_read(&tipc_queue_size);
1757 case TIPC_SOCK_RECVQ_DEPTH:
1758 value = skb_queue_len(&sk->sk_receive_queue);
1769 else if (len < sizeof(value)) {
1772 else if (copy_to_user(ov, &value, sizeof(value))) {
1776 res = 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);