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.
39 #include <linux/tipc.h>
40 #include <linux/tipc_config.h>
45 #define SS_LISTENING -1 /* socket is listening */
46 #define SS_READY -2 /* socket is connectionless */
48 #define OVERLOAD_LIMIT_BASE 5000
49 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
54 struct tipc_portid peer_name;
58 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
59 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
61 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
62 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
63 static void wakeupdispatch(struct tipc_port *tport);
65 static const struct proto_ops packet_ops;
66 static const struct proto_ops stream_ops;
67 static const struct proto_ops msg_ops;
69 static struct proto tipc_proto;
71 static int sockets_enabled;
73 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
76 * Revised TIPC socket locking policy:
78 * Most socket operations take the standard socket lock when they start
79 * and hold it until they finish (or until they need to sleep). Acquiring
80 * this lock grants the owner exclusive access to the fields of the socket
81 * data structures, with the exception of the backlog queue. A few socket
82 * operations can be done without taking the socket lock because they only
83 * read socket information that never changes during the life of the socket.
85 * Socket operations may acquire the lock for the associated TIPC port if they
86 * need to perform an operation on the port. If any routine needs to acquire
87 * both the socket lock and the port lock it must take the socket lock first
88 * to avoid the risk of deadlock.
90 * The dispatcher handling incoming messages cannot grab the socket lock in
91 * the standard fashion, since invoked it runs at the BH level and cannot block.
92 * Instead, it checks to see if the socket lock is currently owned by someone,
93 * and either handles the message itself or adds it to the socket's backlog
94 * queue; in the latter case the queued message is processed once the process
95 * owning the socket lock releases it.
97 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
98 * the problem of a blocked socket operation preventing any other operations
99 * from occurring. However, applications must be careful if they have
100 * multiple threads trying to send (or receive) on the same socket, as these
101 * operations might interfere with each other. For example, doing a connect
102 * and a receive at the same time might allow the receive to consume the
103 * ACK message meant for the connect. While additional work could be done
104 * to try and overcome this, it doesn't seem to be worthwhile at the present.
106 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
107 * that another operation that must be performed in a non-blocking manner is
108 * not delayed for very long because the lock has already been taken.
110 * NOTE: This code assumes that certain fields of a port/socket pair are
111 * constant over its lifetime; such fields can be examined without taking
112 * the socket lock and/or port lock, and do not need to be re-read even
113 * after resuming processing after waiting. These fields include:
115 * - pointer to socket sk structure (aka tipc_sock structure)
116 * - pointer to port structure
121 * advance_rx_queue - discard first buffer in socket receive queue
123 * Caller must hold socket lock
126 static void advance_rx_queue(struct sock *sk)
128 buf_discard(__skb_dequeue(&sk->sk_receive_queue));
129 atomic_dec(&tipc_queue_size);
133 * discard_rx_queue - discard all buffers in socket receive queue
135 * Caller must hold socket lock
138 static void discard_rx_queue(struct sock *sk)
142 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
143 atomic_dec(&tipc_queue_size);
149 * reject_rx_queue - reject all buffers in socket receive queue
151 * Caller must hold socket lock
154 static void reject_rx_queue(struct sock *sk)
158 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
159 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
160 atomic_dec(&tipc_queue_size);
165 * tipc_create - create a TIPC socket
166 * @net: network namespace (must be default network)
167 * @sock: pre-allocated socket structure
168 * @protocol: protocol indicator (must be 0)
169 * @kern: caused by kernel or by userspace?
171 * This routine creates additional data structures used by the TIPC socket,
172 * initializes them, and links them together.
174 * Returns 0 on success, errno otherwise
177 static int tipc_create(struct net *net, struct socket *sock, int protocol,
180 const struct proto_ops *ops;
183 struct tipc_port *tp_ptr;
185 /* Validate arguments */
187 if (!net_eq(net, &init_net))
188 return -EAFNOSUPPORT;
190 if (unlikely(protocol != 0))
191 return -EPROTONOSUPPORT;
193 switch (sock->type) {
196 state = SS_UNCONNECTED;
200 state = SS_UNCONNECTED;
211 /* Allocate socket's protocol area */
213 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
217 /* Allocate TIPC port for socket to use */
219 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
220 TIPC_LOW_IMPORTANCE);
221 if (unlikely(!tp_ptr)) {
226 /* Finish initializing socket data structures */
231 sock_init_data(sock, sk);
232 sk->sk_backlog_rcv = backlog_rcv;
233 tipc_sk(sk)->p = tp_ptr;
234 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
236 spin_unlock_bh(tp_ptr->lock);
238 if (sock->state == SS_READY) {
239 tipc_set_portunreturnable(tp_ptr->ref, 1);
240 if (sock->type == SOCK_DGRAM)
241 tipc_set_portunreliable(tp_ptr->ref, 1);
244 atomic_inc(&tipc_user_count);
249 * release - destroy a TIPC socket
250 * @sock: socket to destroy
252 * This routine cleans up any messages that are still queued on the socket.
253 * For DGRAM and RDM socket types, all queued messages are rejected.
254 * For SEQPACKET and STREAM socket types, the first message is rejected
255 * and any others are discarded. (If the first message on a STREAM socket
256 * is partially-read, it is discarded and the next one is rejected instead.)
258 * NOTE: Rejected messages are not necessarily returned to the sender! They
259 * are returned or discarded according to the "destination droppable" setting
260 * specified for the message by the sender.
262 * Returns 0 on success, errno otherwise
265 static int release(struct socket *sock)
267 struct sock *sk = sock->sk;
268 struct tipc_port *tport;
273 * Exit if socket isn't fully initialized (occurs when a failed accept()
274 * releases a pre-allocated child socket that was never used)
280 tport = tipc_sk_port(sk);
284 * Reject all unreceived messages, except on an active connection
285 * (which disconnects locally & sends a 'FIN+' to peer)
288 while (sock->state != SS_DISCONNECTING) {
289 buf = __skb_dequeue(&sk->sk_receive_queue);
292 atomic_dec(&tipc_queue_size);
293 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf)))
296 if ((sock->state == SS_CONNECTING) ||
297 (sock->state == SS_CONNECTED)) {
298 sock->state = SS_DISCONNECTING;
299 tipc_disconnect(tport->ref);
301 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
306 * Delete TIPC port; this ensures no more messages are queued
307 * (also disconnects an active connection & sends a 'FIN-' to peer)
310 res = tipc_deleteport(tport->ref);
312 /* Discard any remaining (connection-based) messages in receive queue */
314 discard_rx_queue(sk);
316 /* Reject any messages that accumulated in backlog queue */
318 sock->state = SS_DISCONNECTING;
324 atomic_dec(&tipc_user_count);
329 * bind - associate or disassocate TIPC name(s) with a socket
330 * @sock: socket structure
331 * @uaddr: socket address describing name(s) and desired operation
332 * @uaddr_len: size of socket address data structure
334 * Name and name sequence binding is indicated using a positive scope value;
335 * a negative scope value unbinds the specified name. Specifying no name
336 * (i.e. a socket address length of 0) unbinds all names from the socket.
338 * Returns 0 on success, errno otherwise
340 * NOTE: This routine doesn't need to take the socket lock since it doesn't
341 * access any non-constant socket information.
344 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
346 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
347 u32 portref = tipc_sk_port(sock->sk)->ref;
349 if (unlikely(!uaddr_len))
350 return tipc_withdraw(portref, 0, NULL);
352 if (uaddr_len < sizeof(struct sockaddr_tipc))
354 if (addr->family != AF_TIPC)
355 return -EAFNOSUPPORT;
357 if (addr->addrtype == TIPC_ADDR_NAME)
358 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
359 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
360 return -EAFNOSUPPORT;
362 return (addr->scope > 0) ?
363 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
364 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
368 * get_name - get port ID of socket or peer socket
369 * @sock: socket structure
370 * @uaddr: area for returned socket address
371 * @uaddr_len: area for returned length of socket address
372 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
374 * Returns 0 on success, errno otherwise
376 * NOTE: This routine doesn't need to take the socket lock since it only
377 * accesses socket information that is unchanging (or which changes in
378 * a completely predictable manner).
381 static int get_name(struct socket *sock, struct sockaddr *uaddr,
382 int *uaddr_len, int peer)
384 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
385 struct tipc_sock *tsock = tipc_sk(sock->sk);
387 memset(addr, 0, sizeof(*addr));
389 if ((sock->state != SS_CONNECTED) &&
390 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
392 addr->addr.id.ref = tsock->peer_name.ref;
393 addr->addr.id.node = tsock->peer_name.node;
395 addr->addr.id.ref = tsock->p->ref;
396 addr->addr.id.node = tipc_own_addr;
399 *uaddr_len = sizeof(*addr);
400 addr->addrtype = TIPC_ADDR_ID;
401 addr->family = AF_TIPC;
403 addr->addr.name.domain = 0;
409 * poll - read and possibly block on pollmask
410 * @file: file structure associated with the socket
411 * @sock: socket for which to calculate the poll bits
414 * Returns pollmask value
417 * It appears that the usual socket locking mechanisms are not useful here
418 * since the pollmask info is potentially out-of-date the moment this routine
419 * exits. TCP and other protocols seem to rely on higher level poll routines
420 * to handle any preventable race conditions, so TIPC will do the same ...
422 * TIPC sets the returned events as follows:
424 * socket state flags set
425 * ------------ ---------
426 * unconnected no read flags
429 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
432 * connected POLLIN/POLLRDNORM if data in rx queue
433 * POLLOUT if port is not congested
435 * disconnecting POLLIN/POLLRDNORM/POLLHUP
438 * listening POLLIN if SYN in rx queue
441 * ready POLLIN/POLLRDNORM if data in rx queue
442 * [connectionless] POLLOUT (since port cannot be congested)
444 * IMPORTANT: The fact that a read or write operation is indicated does NOT
445 * imply that the operation will succeed, merely that it should be performed
446 * and will not block.
449 static unsigned int poll(struct file *file, struct socket *sock,
452 struct sock *sk = sock->sk;
455 poll_wait(file, sk_sleep(sk), wait);
457 switch ((int)sock->state) {
460 if (!tipc_sk_port(sk)->congested)
465 if (!skb_queue_empty(&sk->sk_receive_queue))
466 mask |= (POLLIN | POLLRDNORM);
468 case SS_DISCONNECTING:
469 mask = (POLLIN | POLLRDNORM | POLLHUP);
477 * dest_name_check - verify user is permitted to send to specified port name
478 * @dest: destination address
479 * @m: descriptor for message to be sent
481 * Prevents restricted configuration commands from being issued by
482 * unauthorized users.
484 * Returns 0 if permission is granted, otherwise errno
487 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
489 struct tipc_cfg_msg_hdr hdr;
491 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
493 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
495 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
498 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
500 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
507 * send_msg - send message in connectionless manner
508 * @iocb: if NULL, indicates that socket lock is already held
509 * @sock: socket structure
510 * @m: message to send
511 * @total_len: length of message
513 * Message must have an destination specified explicitly.
514 * Used for SOCK_RDM and SOCK_DGRAM messages,
515 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
516 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
518 * Returns the number of bytes sent on success, or errno otherwise
521 static int send_msg(struct kiocb *iocb, struct socket *sock,
522 struct msghdr *m, size_t total_len)
524 struct sock *sk = sock->sk;
525 struct tipc_port *tport = tipc_sk_port(sk);
526 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
531 return -EDESTADDRREQ;
532 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
533 (dest->family != AF_TIPC)))
539 needs_conn = (sock->state != SS_READY);
540 if (unlikely(needs_conn)) {
541 if (sock->state == SS_LISTENING) {
545 if (sock->state != SS_UNCONNECTED) {
549 if ((tport->published) ||
550 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
554 if (dest->addrtype == TIPC_ADDR_NAME) {
555 tport->conn_type = dest->addr.name.name.type;
556 tport->conn_instance = dest->addr.name.name.instance;
559 /* Abort any pending connection attempts (very unlikely) */
565 if (dest->addrtype == TIPC_ADDR_NAME) {
566 res = dest_name_check(dest, m);
569 res = tipc_send2name(tport->ref,
570 &dest->addr.name.name,
571 dest->addr.name.domain,
574 } else if (dest->addrtype == TIPC_ADDR_ID) {
575 res = tipc_send2port(tport->ref,
579 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
584 res = dest_name_check(dest, m);
587 res = tipc_multicast(tport->ref,
592 if (likely(res != -ELINKCONG)) {
593 if (needs_conn && (res >= 0))
594 sock->state = SS_CONNECTING;
597 if (m->msg_flags & MSG_DONTWAIT) {
602 res = wait_event_interruptible(*sk_sleep(sk),
616 * send_packet - send a connection-oriented message
617 * @iocb: if NULL, indicates that socket lock is already held
618 * @sock: socket structure
619 * @m: message to send
620 * @total_len: length of message
622 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
624 * Returns the number of bytes sent on success, or errno otherwise
627 static int send_packet(struct kiocb *iocb, struct socket *sock,
628 struct msghdr *m, size_t total_len)
630 struct sock *sk = sock->sk;
631 struct tipc_port *tport = tipc_sk_port(sk);
632 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
635 /* Handle implied connection establishment */
638 return send_msg(iocb, sock, m, total_len);
644 if (unlikely(sock->state != SS_CONNECTED)) {
645 if (sock->state == SS_DISCONNECTING)
652 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov);
653 if (likely(res != -ELINKCONG))
655 if (m->msg_flags & MSG_DONTWAIT) {
660 res = wait_event_interruptible(*sk_sleep(sk),
661 (!tport->congested || !tport->connected));
673 * send_stream - send stream-oriented data
675 * @sock: socket structure
677 * @total_len: total length of data to be sent
679 * Used for SOCK_STREAM data.
681 * Returns the number of bytes sent on success (or partial success),
682 * or errno if no data sent
685 static int send_stream(struct kiocb *iocb, struct socket *sock,
686 struct msghdr *m, size_t total_len)
688 struct sock *sk = sock->sk;
689 struct tipc_port *tport = tipc_sk_port(sk);
690 struct msghdr my_msg;
692 struct iovec *curr_iov;
694 char __user *curr_start;
703 /* Handle special cases where there is no connection */
705 if (unlikely(sock->state != SS_CONNECTED)) {
706 if (sock->state == SS_UNCONNECTED) {
707 res = send_packet(NULL, sock, m, total_len);
709 } else if (sock->state == SS_DISCONNECTING) {
718 if (unlikely(m->msg_name)) {
724 * Send each iovec entry using one or more messages
726 * Note: This algorithm is good for the most likely case
727 * (i.e. one large iovec entry), but could be improved to pass sets
728 * of small iovec entries into send_packet().
731 curr_iov = m->msg_iov;
732 curr_iovlen = m->msg_iovlen;
733 my_msg.msg_iov = &my_iov;
734 my_msg.msg_iovlen = 1;
735 my_msg.msg_flags = m->msg_flags;
736 my_msg.msg_name = NULL;
739 hdr_size = msg_hdr_sz(&tport->phdr);
741 while (curr_iovlen--) {
742 curr_start = curr_iov->iov_base;
743 curr_left = curr_iov->iov_len;
746 bytes_to_send = tport->max_pkt - hdr_size;
747 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
748 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
749 if (curr_left < bytes_to_send)
750 bytes_to_send = curr_left;
751 my_iov.iov_base = curr_start;
752 my_iov.iov_len = bytes_to_send;
753 res = send_packet(NULL, sock, &my_msg, 0);
759 curr_left -= bytes_to_send;
760 curr_start += bytes_to_send;
761 bytes_sent += bytes_to_send;
773 * auto_connect - complete connection setup to a remote port
774 * @sock: socket structure
775 * @msg: peer's response message
777 * Returns 0 on success, errno otherwise
780 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
782 struct tipc_sock *tsock = tipc_sk(sock->sk);
784 if (msg_errcode(msg)) {
785 sock->state = SS_DISCONNECTING;
786 return -ECONNREFUSED;
789 tsock->peer_name.ref = msg_origport(msg);
790 tsock->peer_name.node = msg_orignode(msg);
791 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
792 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
793 sock->state = SS_CONNECTED;
798 * set_orig_addr - capture sender's address for received message
799 * @m: descriptor for message info
800 * @msg: received message header
802 * Note: Address is not captured if not requested by receiver.
805 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
807 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
810 addr->family = AF_TIPC;
811 addr->addrtype = TIPC_ADDR_ID;
812 addr->addr.id.ref = msg_origport(msg);
813 addr->addr.id.node = msg_orignode(msg);
814 addr->addr.name.domain = 0; /* could leave uninitialized */
815 addr->scope = 0; /* could leave uninitialized */
816 m->msg_namelen = sizeof(struct sockaddr_tipc);
821 * anc_data_recv - optionally capture ancillary data for received message
822 * @m: descriptor for message info
823 * @msg: received message header
824 * @tport: TIPC port associated with message
826 * Note: Ancillary data is not captured if not requested by receiver.
828 * Returns 0 if successful, otherwise errno
831 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
832 struct tipc_port *tport)
840 if (likely(m->msg_controllen == 0))
843 /* Optionally capture errored message object(s) */
845 err = msg ? msg_errcode(msg) : 0;
848 anc_data[1] = msg_data_sz(msg);
849 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
853 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
860 /* Optionally capture message destination object */
862 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
866 anc_data[0] = msg_nametype(msg);
867 anc_data[1] = msg_namelower(msg);
868 anc_data[2] = msg_namelower(msg);
872 anc_data[0] = msg_nametype(msg);
873 anc_data[1] = msg_namelower(msg);
874 anc_data[2] = msg_nameupper(msg);
877 has_name = (tport->conn_type != 0);
878 anc_data[0] = tport->conn_type;
879 anc_data[1] = tport->conn_instance;
880 anc_data[2] = tport->conn_instance;
886 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
895 * recv_msg - receive packet-oriented message
897 * @m: descriptor for message info
898 * @buf_len: total size of user buffer area
899 * @flags: receive flags
901 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
902 * If the complete message doesn't fit in user area, truncate it.
904 * Returns size of returned message data, errno otherwise
907 static int recv_msg(struct kiocb *iocb, struct socket *sock,
908 struct msghdr *m, size_t buf_len, int flags)
910 struct sock *sk = sock->sk;
911 struct tipc_port *tport = tipc_sk_port(sk);
913 struct tipc_msg *msg;
918 /* Catch invalid receive requests */
920 if (m->msg_iovlen != 1)
921 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */
923 if (unlikely(!buf_len))
928 if (unlikely(sock->state == SS_UNCONNECTED)) {
935 /* Look for a message in receive queue; wait if necessary */
937 while (skb_queue_empty(&sk->sk_receive_queue)) {
938 if (sock->state == SS_DISCONNECTING) {
942 if (flags & MSG_DONTWAIT) {
947 res = wait_event_interruptible(*sk_sleep(sk),
948 (!skb_queue_empty(&sk->sk_receive_queue) ||
949 (sock->state == SS_DISCONNECTING)));
955 /* Look at first message in receive queue */
957 buf = skb_peek(&sk->sk_receive_queue);
959 sz = msg_data_sz(msg);
960 err = msg_errcode(msg);
962 /* Complete connection setup for an implied connect */
964 if (unlikely(sock->state == SS_CONNECTING)) {
965 res = auto_connect(sock, msg);
970 /* Discard an empty non-errored message & try again */
972 if ((!sz) && (!err)) {
973 advance_rx_queue(sk);
977 /* Capture sender's address (optional) */
979 set_orig_addr(m, msg);
981 /* Capture ancillary data (optional) */
983 res = anc_data_recv(m, msg, tport);
987 /* Capture message data (if valid) & compute return value (always) */
990 if (unlikely(buf_len < sz)) {
992 m->msg_flags |= MSG_TRUNC;
994 if (unlikely(copy_to_user(m->msg_iov->iov_base, msg_data(msg),
1001 if ((sock->state == SS_READY) ||
1002 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1008 /* Consume received message (optional) */
1010 if (likely(!(flags & MSG_PEEK))) {
1011 if ((sock->state != SS_READY) &&
1012 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1013 tipc_acknowledge(tport->ref, tport->conn_unacked);
1014 advance_rx_queue(sk);
1022 * recv_stream - receive stream-oriented data
1024 * @m: descriptor for message info
1025 * @buf_len: total size of user buffer area
1026 * @flags: receive flags
1028 * Used for SOCK_STREAM messages only. If not enough data is available
1029 * will optionally wait for more; never truncates data.
1031 * Returns size of returned message data, errno otherwise
1034 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1035 struct msghdr *m, size_t buf_len, int flags)
1037 struct sock *sk = sock->sk;
1038 struct tipc_port *tport = tipc_sk_port(sk);
1039 struct sk_buff *buf;
1040 struct tipc_msg *msg;
1042 int sz_to_copy, target, needed;
1044 char __user *crs = m->msg_iov->iov_base;
1045 unsigned char *buf_crs;
1049 /* Catch invalid receive attempts */
1051 if (m->msg_iovlen != 1)
1052 return -EOPNOTSUPP; /* Don't do multiple iovec entries yet */
1054 if (unlikely(!buf_len))
1059 if (unlikely((sock->state == SS_UNCONNECTED) ||
1060 (sock->state == SS_CONNECTING))) {
1065 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1069 /* Look for a message in receive queue; wait if necessary */
1071 while (skb_queue_empty(&sk->sk_receive_queue)) {
1072 if (sock->state == SS_DISCONNECTING) {
1076 if (flags & MSG_DONTWAIT) {
1081 res = wait_event_interruptible(*sk_sleep(sk),
1082 (!skb_queue_empty(&sk->sk_receive_queue) ||
1083 (sock->state == SS_DISCONNECTING)));
1089 /* Look at first message in receive queue */
1091 buf = skb_peek(&sk->sk_receive_queue);
1093 sz = msg_data_sz(msg);
1094 err = msg_errcode(msg);
1096 /* Discard an empty non-errored message & try again */
1098 if ((!sz) && (!err)) {
1099 advance_rx_queue(sk);
1103 /* Optionally capture sender's address & ancillary data of first msg */
1105 if (sz_copied == 0) {
1106 set_orig_addr(m, msg);
1107 res = anc_data_recv(m, msg, tport);
1112 /* Capture message data (if valid) & compute return value (always) */
1115 buf_crs = (unsigned char *)(TIPC_SKB_CB(buf)->handle);
1116 sz = (unsigned char *)msg + msg_size(msg) - buf_crs;
1118 needed = (buf_len - sz_copied);
1119 sz_to_copy = (sz <= needed) ? sz : needed;
1120 if (unlikely(copy_to_user(crs, buf_crs, sz_to_copy))) {
1124 sz_copied += sz_to_copy;
1126 if (sz_to_copy < sz) {
1127 if (!(flags & MSG_PEEK))
1128 TIPC_SKB_CB(buf)->handle = buf_crs + sz_to_copy;
1135 goto exit; /* can't add error msg to valid data */
1137 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1143 /* Consume received message (optional) */
1145 if (likely(!(flags & MSG_PEEK))) {
1146 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1147 tipc_acknowledge(tport->ref, tport->conn_unacked);
1148 advance_rx_queue(sk);
1151 /* Loop around if more data is required */
1153 if ((sz_copied < buf_len) && /* didn't get all requested data */
1154 (!skb_queue_empty(&sk->sk_receive_queue) ||
1155 (sz_copied < target)) && /* and more is ready or required */
1156 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1157 (!err)) /* and haven't reached a FIN */
1162 return sz_copied ? sz_copied : res;
1166 * rx_queue_full - determine if receive queue can accept another message
1167 * @msg: message to be added to queue
1168 * @queue_size: current size of queue
1169 * @base: nominal maximum size of queue
1171 * Returns 1 if queue is unable to accept message, 0 otherwise
1174 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1177 u32 imp = msg_importance(msg);
1179 if (imp == TIPC_LOW_IMPORTANCE)
1181 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1182 threshold = base * 2;
1183 else if (imp == TIPC_HIGH_IMPORTANCE)
1184 threshold = base * 100;
1188 if (msg_connected(msg))
1191 return queue_size >= threshold;
1195 * filter_rcv - validate incoming message
1199 * Enqueues message on receive queue if acceptable; optionally handles
1200 * disconnect indication for a connected socket.
1202 * Called with socket lock already taken; port lock may also be taken.
1204 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1207 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1209 struct socket *sock = sk->sk_socket;
1210 struct tipc_msg *msg = buf_msg(buf);
1213 /* Reject message if it is wrong sort of message for socket */
1216 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1217 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1218 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1221 if (sock->state == SS_READY) {
1222 if (msg_connected(msg))
1223 return TIPC_ERR_NO_PORT;
1226 return TIPC_ERR_NO_PORT;
1227 if (sock->state == SS_CONNECTED) {
1228 if (!msg_connected(msg))
1229 return TIPC_ERR_NO_PORT;
1230 } else if (sock->state == SS_CONNECTING) {
1231 if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1232 return TIPC_ERR_NO_PORT;
1233 } else if (sock->state == SS_LISTENING) {
1234 if (msg_connected(msg) || msg_errcode(msg))
1235 return TIPC_ERR_NO_PORT;
1236 } else if (sock->state == SS_DISCONNECTING) {
1237 return TIPC_ERR_NO_PORT;
1238 } else /* (sock->state == SS_UNCONNECTED) */ {
1239 if (msg_connected(msg) || msg_errcode(msg))
1240 return TIPC_ERR_NO_PORT;
1244 /* Reject message if there isn't room to queue it */
1246 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1247 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1248 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1249 return TIPC_ERR_OVERLOAD;
1251 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1252 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1253 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1254 return TIPC_ERR_OVERLOAD;
1257 /* Enqueue message (finally!) */
1259 TIPC_SKB_CB(buf)->handle = msg_data(msg);
1260 atomic_inc(&tipc_queue_size);
1261 __skb_queue_tail(&sk->sk_receive_queue, buf);
1263 /* Initiate connection termination for an incoming 'FIN' */
1265 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1266 sock->state = SS_DISCONNECTING;
1267 tipc_disconnect_port(tipc_sk_port(sk));
1270 if (waitqueue_active(sk_sleep(sk)))
1271 wake_up_interruptible(sk_sleep(sk));
1276 * backlog_rcv - handle incoming message from backlog queue
1280 * Caller must hold socket lock, but not port lock.
1285 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1289 res = filter_rcv(sk, buf);
1291 tipc_reject_msg(buf, res);
1296 * dispatch - handle incoming message
1297 * @tport: TIPC port that received message
1300 * Called with port lock already taken.
1302 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1305 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1307 struct sock *sk = (struct sock *)tport->usr_handle;
1311 * Process message if socket is unlocked; otherwise add to backlog queue
1313 * This code is based on sk_receive_skb(), but must be distinct from it
1314 * since a TIPC-specific filter/reject mechanism is utilized
1318 if (!sock_owned_by_user(sk)) {
1319 res = filter_rcv(sk, buf);
1321 if (sk_add_backlog(sk, buf))
1322 res = TIPC_ERR_OVERLOAD;
1332 * wakeupdispatch - wake up port after congestion
1333 * @tport: port to wakeup
1335 * Called with port lock already taken.
1338 static void wakeupdispatch(struct tipc_port *tport)
1340 struct sock *sk = (struct sock *)tport->usr_handle;
1342 if (waitqueue_active(sk_sleep(sk)))
1343 wake_up_interruptible(sk_sleep(sk));
1347 * connect - establish a connection to another TIPC port
1348 * @sock: socket structure
1349 * @dest: socket address for destination port
1350 * @destlen: size of socket address data structure
1351 * @flags: file-related flags associated with socket
1353 * Returns 0 on success, errno otherwise
1356 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1359 struct sock *sk = sock->sk;
1360 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1361 struct msghdr m = {NULL,};
1362 struct sk_buff *buf;
1363 struct tipc_msg *msg;
1369 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1371 if (sock->state == SS_READY) {
1376 /* For now, TIPC does not support the non-blocking form of connect() */
1378 if (flags & O_NONBLOCK) {
1383 /* Issue Posix-compliant error code if socket is in the wrong state */
1385 if (sock->state == SS_LISTENING) {
1389 if (sock->state == SS_CONNECTING) {
1393 if (sock->state != SS_UNCONNECTED) {
1399 * Reject connection attempt using multicast address
1401 * Note: send_msg() validates the rest of the address fields,
1402 * so there's no need to do it here
1405 if (dst->addrtype == TIPC_ADDR_MCAST) {
1410 /* Reject any messages already in receive queue (very unlikely) */
1412 reject_rx_queue(sk);
1414 /* Send a 'SYN-' to destination */
1417 m.msg_namelen = destlen;
1418 res = send_msg(NULL, sock, &m, 0);
1422 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1424 timeout = tipc_sk(sk)->conn_timeout;
1426 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1427 (!skb_queue_empty(&sk->sk_receive_queue) ||
1428 (sock->state != SS_CONNECTING)),
1429 timeout ? timeout : MAX_SCHEDULE_TIMEOUT);
1433 buf = skb_peek(&sk->sk_receive_queue);
1436 res = auto_connect(sock, msg);
1438 if (!msg_data_sz(msg))
1439 advance_rx_queue(sk);
1442 if (sock->state == SS_CONNECTED)
1445 res = -ECONNREFUSED;
1451 ; /* leave "res" unchanged */
1452 sock->state = SS_DISCONNECTING;
1461 * listen - allow socket to listen for incoming connections
1462 * @sock: socket structure
1465 * Returns 0 on success, errno otherwise
1468 static int listen(struct socket *sock, int len)
1470 struct sock *sk = sock->sk;
1475 if (sock->state == SS_READY)
1477 else if (sock->state != SS_UNCONNECTED)
1480 sock->state = SS_LISTENING;
1489 * accept - wait for connection request
1490 * @sock: listening socket
1491 * @newsock: new socket that is to be connected
1492 * @flags: file-related flags associated with socket
1494 * Returns 0 on success, errno otherwise
1497 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1499 struct sock *sk = sock->sk;
1500 struct sk_buff *buf;
1505 if (sock->state == SS_READY) {
1509 if (sock->state != SS_LISTENING) {
1514 while (skb_queue_empty(&sk->sk_receive_queue)) {
1515 if (flags & O_NONBLOCK) {
1520 res = wait_event_interruptible(*sk_sleep(sk),
1521 (!skb_queue_empty(&sk->sk_receive_queue)));
1527 buf = skb_peek(&sk->sk_receive_queue);
1529 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1531 struct sock *new_sk = new_sock->sk;
1532 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1533 struct tipc_port *new_tport = new_tsock->p;
1534 u32 new_ref = new_tport->ref;
1535 struct tipc_msg *msg = buf_msg(buf);
1540 * Reject any stray messages received by new socket
1541 * before the socket lock was taken (very, very unlikely)
1544 reject_rx_queue(new_sk);
1546 /* Connect new socket to it's peer */
1548 new_tsock->peer_name.ref = msg_origport(msg);
1549 new_tsock->peer_name.node = msg_orignode(msg);
1550 tipc_connect2port(new_ref, &new_tsock->peer_name);
1551 new_sock->state = SS_CONNECTED;
1553 tipc_set_portimportance(new_ref, msg_importance(msg));
1554 if (msg_named(msg)) {
1555 new_tport->conn_type = msg_nametype(msg);
1556 new_tport->conn_instance = msg_nameinst(msg);
1560 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1561 * Respond to 'SYN+' by queuing it on new socket.
1564 if (!msg_data_sz(msg)) {
1565 struct msghdr m = {NULL,};
1567 advance_rx_queue(sk);
1568 send_packet(NULL, new_sock, &m, 0);
1570 __skb_dequeue(&sk->sk_receive_queue);
1571 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1573 release_sock(new_sk);
1581 * shutdown - shutdown socket connection
1582 * @sock: socket structure
1583 * @how: direction to close (must be SHUT_RDWR)
1585 * Terminates connection (if necessary), then purges socket's receive queue.
1587 * Returns 0 on success, errno otherwise
1590 static int shutdown(struct socket *sock, int how)
1592 struct sock *sk = sock->sk;
1593 struct tipc_port *tport = tipc_sk_port(sk);
1594 struct sk_buff *buf;
1597 if (how != SHUT_RDWR)
1602 switch (sock->state) {
1606 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1608 buf = __skb_dequeue(&sk->sk_receive_queue);
1610 atomic_dec(&tipc_queue_size);
1611 if (TIPC_SKB_CB(buf)->handle != msg_data(buf_msg(buf))) {
1615 tipc_disconnect(tport->ref);
1616 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1618 tipc_shutdown(tport->ref);
1621 sock->state = SS_DISCONNECTING;
1625 case SS_DISCONNECTING:
1627 /* Discard any unreceived messages; wake up sleeping tasks */
1629 discard_rx_queue(sk);
1630 if (waitqueue_active(sk_sleep(sk)))
1631 wake_up_interruptible(sk_sleep(sk));
1644 * setsockopt - set socket option
1645 * @sock: socket structure
1646 * @lvl: option level
1647 * @opt: option identifier
1648 * @ov: pointer to new option value
1649 * @ol: length of option value
1651 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1652 * (to ease compatibility).
1654 * Returns 0 on success, errno otherwise
1657 static int setsockopt(struct socket *sock,
1658 int lvl, int opt, char __user *ov, unsigned int ol)
1660 struct sock *sk = sock->sk;
1661 struct tipc_port *tport = tipc_sk_port(sk);
1665 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1667 if (lvl != SOL_TIPC)
1668 return -ENOPROTOOPT;
1669 if (ol < sizeof(value))
1671 res = get_user(value, (u32 __user *)ov);
1678 case TIPC_IMPORTANCE:
1679 res = tipc_set_portimportance(tport->ref, value);
1681 case TIPC_SRC_DROPPABLE:
1682 if (sock->type != SOCK_STREAM)
1683 res = tipc_set_portunreliable(tport->ref, value);
1687 case TIPC_DEST_DROPPABLE:
1688 res = tipc_set_portunreturnable(tport->ref, value);
1690 case TIPC_CONN_TIMEOUT:
1691 tipc_sk(sk)->conn_timeout = msecs_to_jiffies(value);
1692 /* no need to set "res", since already 0 at this point */
1704 * getsockopt - get socket option
1705 * @sock: socket structure
1706 * @lvl: option level
1707 * @opt: option identifier
1708 * @ov: receptacle for option value
1709 * @ol: receptacle for length of option value
1711 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1712 * (to ease compatibility).
1714 * Returns 0 on success, errno otherwise
1717 static int getsockopt(struct socket *sock,
1718 int lvl, int opt, char __user *ov, int __user *ol)
1720 struct sock *sk = sock->sk;
1721 struct tipc_port *tport = tipc_sk_port(sk);
1726 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1727 return put_user(0, ol);
1728 if (lvl != SOL_TIPC)
1729 return -ENOPROTOOPT;
1730 res = get_user(len, ol);
1737 case TIPC_IMPORTANCE:
1738 res = tipc_portimportance(tport->ref, &value);
1740 case TIPC_SRC_DROPPABLE:
1741 res = tipc_portunreliable(tport->ref, &value);
1743 case TIPC_DEST_DROPPABLE:
1744 res = tipc_portunreturnable(tport->ref, &value);
1746 case TIPC_CONN_TIMEOUT:
1747 value = jiffies_to_msecs(tipc_sk(sk)->conn_timeout);
1748 /* no need to set "res", since already 0 at this point */
1750 case TIPC_NODE_RECVQ_DEPTH:
1751 value = (u32)atomic_read(&tipc_queue_size);
1753 case TIPC_SOCK_RECVQ_DEPTH:
1754 value = skb_queue_len(&sk->sk_receive_queue);
1763 return res; /* "get" failed */
1765 if (len < sizeof(value))
1768 if (copy_to_user(ov, &value, sizeof(value)))
1771 return put_user(sizeof(value), ol);
1775 * Protocol switches for the various types of TIPC sockets
1778 static const struct proto_ops msg_ops = {
1779 .owner = THIS_MODULE,
1784 .socketpair = sock_no_socketpair,
1786 .getname = get_name,
1788 .ioctl = sock_no_ioctl,
1790 .shutdown = shutdown,
1791 .setsockopt = setsockopt,
1792 .getsockopt = getsockopt,
1793 .sendmsg = send_msg,
1794 .recvmsg = recv_msg,
1795 .mmap = sock_no_mmap,
1796 .sendpage = sock_no_sendpage
1799 static const struct proto_ops packet_ops = {
1800 .owner = THIS_MODULE,
1805 .socketpair = sock_no_socketpair,
1807 .getname = get_name,
1809 .ioctl = sock_no_ioctl,
1811 .shutdown = shutdown,
1812 .setsockopt = setsockopt,
1813 .getsockopt = getsockopt,
1814 .sendmsg = send_packet,
1815 .recvmsg = recv_msg,
1816 .mmap = sock_no_mmap,
1817 .sendpage = sock_no_sendpage
1820 static const struct proto_ops stream_ops = {
1821 .owner = THIS_MODULE,
1826 .socketpair = sock_no_socketpair,
1828 .getname = get_name,
1830 .ioctl = sock_no_ioctl,
1832 .shutdown = shutdown,
1833 .setsockopt = setsockopt,
1834 .getsockopt = getsockopt,
1835 .sendmsg = send_stream,
1836 .recvmsg = recv_stream,
1837 .mmap = sock_no_mmap,
1838 .sendpage = sock_no_sendpage
1841 static const struct net_proto_family tipc_family_ops = {
1842 .owner = THIS_MODULE,
1844 .create = tipc_create
1847 static struct proto tipc_proto = {
1849 .owner = THIS_MODULE,
1850 .obj_size = sizeof(struct tipc_sock)
1854 * tipc_socket_init - initialize TIPC socket interface
1856 * Returns 0 on success, errno otherwise
1858 int tipc_socket_init(void)
1862 res = proto_register(&tipc_proto, 1);
1864 err("Failed to register TIPC protocol type\n");
1868 res = sock_register(&tipc_family_ops);
1870 err("Failed to register TIPC socket type\n");
1871 proto_unregister(&tipc_proto);
1875 sockets_enabled = 1;
1881 * tipc_socket_stop - stop TIPC socket interface
1884 void tipc_socket_stop(void)
1886 if (!sockets_enabled)
1889 sockets_enabled = 0;
1890 sock_unregister(tipc_family_ops.family);
1891 proto_unregister(&tipc_proto);