2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2011, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/export.h>
43 #define SS_LISTENING -1 /* socket is listening */
44 #define SS_READY -2 /* socket is connectionless */
46 #define OVERLOAD_LIMIT_BASE 5000
47 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
52 struct tipc_portid peer_name;
53 unsigned int conn_timeout;
56 #define tipc_sk(sk) ((struct tipc_sock *)(sk))
57 #define tipc_sk_port(sk) ((struct tipc_port *)(tipc_sk(sk)->p))
59 #define tipc_rx_ready(sock) (!skb_queue_empty(&sock->sk->sk_receive_queue) || \
60 (sock->state == SS_DISCONNECTING))
62 static int backlog_rcv(struct sock *sk, struct sk_buff *skb);
63 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf);
64 static void wakeupdispatch(struct tipc_port *tport);
66 static const struct proto_ops packet_ops;
67 static const struct proto_ops stream_ops;
68 static const struct proto_ops msg_ops;
70 static struct proto tipc_proto;
72 static int sockets_enabled;
74 static atomic_t tipc_queue_size = ATOMIC_INIT(0);
77 * Revised TIPC socket locking policy:
79 * Most socket operations take the standard socket lock when they start
80 * and hold it until they finish (or until they need to sleep). Acquiring
81 * this lock grants the owner exclusive access to the fields of the socket
82 * data structures, with the exception of the backlog queue. A few socket
83 * operations can be done without taking the socket lock because they only
84 * read socket information that never changes during the life of the socket.
86 * Socket operations may acquire the lock for the associated TIPC port if they
87 * need to perform an operation on the port. If any routine needs to acquire
88 * both the socket lock and the port lock it must take the socket lock first
89 * to avoid the risk of deadlock.
91 * The dispatcher handling incoming messages cannot grab the socket lock in
92 * the standard fashion, since invoked it runs at the BH level and cannot block.
93 * Instead, it checks to see if the socket lock is currently owned by someone,
94 * and either handles the message itself or adds it to the socket's backlog
95 * queue; in the latter case the queued message is processed once the process
96 * owning the socket lock releases it.
98 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
99 * the problem of a blocked socket operation preventing any other operations
100 * from occurring. However, applications must be careful if they have
101 * multiple threads trying to send (or receive) on the same socket, as these
102 * operations might interfere with each other. For example, doing a connect
103 * and a receive at the same time might allow the receive to consume the
104 * ACK message meant for the connect. While additional work could be done
105 * to try and overcome this, it doesn't seem to be worthwhile at the present.
107 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
108 * that another operation that must be performed in a non-blocking manner is
109 * not delayed for very long because the lock has already been taken.
111 * NOTE: This code assumes that certain fields of a port/socket pair are
112 * constant over its lifetime; such fields can be examined without taking
113 * the socket lock and/or port lock, and do not need to be re-read even
114 * after resuming processing after waiting. These fields include:
116 * - pointer to socket sk structure (aka tipc_sock structure)
117 * - pointer to port structure
122 * advance_rx_queue - discard first buffer in socket receive queue
124 * Caller must hold socket lock
127 static void advance_rx_queue(struct sock *sk)
129 buf_discard(__skb_dequeue(&sk->sk_receive_queue));
130 atomic_dec(&tipc_queue_size);
134 * discard_rx_queue - discard all buffers in socket receive queue
136 * Caller must hold socket lock
139 static void discard_rx_queue(struct sock *sk)
143 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
144 atomic_dec(&tipc_queue_size);
150 * reject_rx_queue - reject all buffers in socket receive queue
152 * Caller must hold socket lock
155 static void reject_rx_queue(struct sock *sk)
159 while ((buf = __skb_dequeue(&sk->sk_receive_queue))) {
160 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
161 atomic_dec(&tipc_queue_size);
166 * tipc_create - create a TIPC socket
167 * @net: network namespace (must be default network)
168 * @sock: pre-allocated socket structure
169 * @protocol: protocol indicator (must be 0)
170 * @kern: caused by kernel or by userspace?
172 * This routine creates additional data structures used by the TIPC socket,
173 * initializes them, and links them together.
175 * Returns 0 on success, errno otherwise
178 static int tipc_create(struct net *net, struct socket *sock, int protocol,
181 const struct proto_ops *ops;
184 struct tipc_port *tp_ptr;
186 /* Validate arguments */
188 if (!net_eq(net, &init_net))
189 return -EAFNOSUPPORT;
191 if (unlikely(protocol != 0))
192 return -EPROTONOSUPPORT;
194 switch (sock->type) {
197 state = SS_UNCONNECTED;
201 state = SS_UNCONNECTED;
212 /* Allocate socket's protocol area */
214 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
218 /* Allocate TIPC port for socket to use */
220 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
221 TIPC_LOW_IMPORTANCE);
222 if (unlikely(!tp_ptr)) {
227 /* Finish initializing socket data structures */
232 sock_init_data(sock, sk);
233 sk->sk_backlog_rcv = backlog_rcv;
234 tipc_sk(sk)->p = tp_ptr;
235 tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT;
237 spin_unlock_bh(tp_ptr->lock);
239 if (sock->state == SS_READY) {
240 tipc_set_portunreturnable(tp_ptr->ref, 1);
241 if (sock->type == SOCK_DGRAM)
242 tipc_set_portunreliable(tp_ptr->ref, 1);
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 != 0)
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;
328 * bind - associate or disassocate TIPC name(s) with a socket
329 * @sock: socket structure
330 * @uaddr: socket address describing name(s) and desired operation
331 * @uaddr_len: size of socket address data structure
333 * Name and name sequence binding is indicated using a positive scope value;
334 * a negative scope value unbinds the specified name. Specifying no name
335 * (i.e. a socket address length of 0) unbinds all names from the socket.
337 * Returns 0 on success, errno otherwise
339 * NOTE: This routine doesn't need to take the socket lock since it doesn't
340 * access any non-constant socket information.
343 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
345 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
346 u32 portref = tipc_sk_port(sock->sk)->ref;
348 if (unlikely(!uaddr_len))
349 return tipc_withdraw(portref, 0, NULL);
351 if (uaddr_len < sizeof(struct sockaddr_tipc))
353 if (addr->family != AF_TIPC)
354 return -EAFNOSUPPORT;
356 if (addr->addrtype == TIPC_ADDR_NAME)
357 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
358 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
359 return -EAFNOSUPPORT;
361 return (addr->scope > 0) ?
362 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
363 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
367 * get_name - get port ID of socket or peer socket
368 * @sock: socket structure
369 * @uaddr: area for returned socket address
370 * @uaddr_len: area for returned length of socket address
371 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
373 * Returns 0 on success, errno otherwise
375 * NOTE: This routine doesn't need to take the socket lock since it only
376 * accesses socket information that is unchanging (or which changes in
377 * a completely predictable manner).
380 static int get_name(struct socket *sock, struct sockaddr *uaddr,
381 int *uaddr_len, int peer)
383 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
384 struct tipc_sock *tsock = tipc_sk(sock->sk);
386 memset(addr, 0, sizeof(*addr));
388 if ((sock->state != SS_CONNECTED) &&
389 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
391 addr->addr.id.ref = tsock->peer_name.ref;
392 addr->addr.id.node = tsock->peer_name.node;
394 addr->addr.id.ref = tsock->p->ref;
395 addr->addr.id.node = tipc_own_addr;
398 *uaddr_len = sizeof(*addr);
399 addr->addrtype = TIPC_ADDR_ID;
400 addr->family = AF_TIPC;
402 addr->addr.name.domain = 0;
408 * poll - read and possibly block on pollmask
409 * @file: file structure associated with the socket
410 * @sock: socket for which to calculate the poll bits
413 * Returns pollmask value
416 * It appears that the usual socket locking mechanisms are not useful here
417 * since the pollmask info is potentially out-of-date the moment this routine
418 * exits. TCP and other protocols seem to rely on higher level poll routines
419 * to handle any preventable race conditions, so TIPC will do the same ...
421 * TIPC sets the returned events as follows:
423 * socket state flags set
424 * ------------ ---------
425 * unconnected no read flags
428 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
431 * connected POLLIN/POLLRDNORM if data in rx queue
432 * POLLOUT if port is not congested
434 * disconnecting POLLIN/POLLRDNORM/POLLHUP
437 * listening POLLIN if SYN in rx queue
440 * ready POLLIN/POLLRDNORM if data in rx queue
441 * [connectionless] POLLOUT (since port cannot be congested)
443 * IMPORTANT: The fact that a read or write operation is indicated does NOT
444 * imply that the operation will succeed, merely that it should be performed
445 * and will not block.
448 static unsigned int poll(struct file *file, struct socket *sock,
451 struct sock *sk = sock->sk;
454 poll_wait(file, sk_sleep(sk), wait);
456 switch ((int)sock->state) {
459 if (!tipc_sk_port(sk)->congested)
464 if (!skb_queue_empty(&sk->sk_receive_queue))
465 mask |= (POLLIN | POLLRDNORM);
467 case SS_DISCONNECTING:
468 mask = (POLLIN | POLLRDNORM | POLLHUP);
476 * dest_name_check - verify user is permitted to send to specified port name
477 * @dest: destination address
478 * @m: descriptor for message to be sent
480 * Prevents restricted configuration commands from being issued by
481 * unauthorized users.
483 * Returns 0 if permission is granted, otherwise errno
486 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
488 struct tipc_cfg_msg_hdr hdr;
490 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
492 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
494 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
497 if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
499 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
501 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
508 * send_msg - send message in connectionless manner
509 * @iocb: if NULL, indicates that socket lock is already held
510 * @sock: socket structure
511 * @m: message to send
512 * @total_len: length of message
514 * Message must have an destination specified explicitly.
515 * Used for SOCK_RDM and SOCK_DGRAM messages,
516 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
517 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
519 * Returns the number of bytes sent on success, or errno otherwise
522 static int send_msg(struct kiocb *iocb, struct socket *sock,
523 struct msghdr *m, size_t total_len)
525 struct sock *sk = sock->sk;
526 struct tipc_port *tport = tipc_sk_port(sk);
527 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
533 return -EDESTADDRREQ;
534 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
535 (dest->family != AF_TIPC)))
537 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
538 (m->msg_iovlen > (unsigned)INT_MAX))
544 needs_conn = (sock->state != SS_READY);
545 if (unlikely(needs_conn)) {
546 if (sock->state == SS_LISTENING) {
550 if (sock->state != SS_UNCONNECTED) {
554 if ((tport->published) ||
555 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
559 if (dest->addrtype == TIPC_ADDR_NAME) {
560 tport->conn_type = dest->addr.name.name.type;
561 tport->conn_instance = dest->addr.name.name.instance;
564 /* Abort any pending connection attempts (very unlikely) */
569 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
572 if (dest->addrtype == TIPC_ADDR_NAME) {
573 res = dest_name_check(dest, m);
576 res = tipc_send2name(tport->ref,
577 &dest->addr.name.name,
578 dest->addr.name.domain,
582 } else if (dest->addrtype == TIPC_ADDR_ID) {
583 res = tipc_send2port(tport->ref,
588 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
593 res = dest_name_check(dest, m);
596 res = tipc_multicast(tport->ref,
602 if (likely(res != -ELINKCONG)) {
603 if (needs_conn && (res >= 0))
604 sock->state = SS_CONNECTING;
607 if (timeout_val <= 0L) {
608 res = timeout_val ? timeout_val : -EWOULDBLOCK;
612 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
613 !tport->congested, timeout_val);
624 * send_packet - send a connection-oriented message
625 * @iocb: if NULL, indicates that socket lock is already held
626 * @sock: socket structure
627 * @m: message to send
628 * @total_len: length of message
630 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
632 * Returns the number of bytes sent on success, or errno otherwise
635 static int send_packet(struct kiocb *iocb, struct socket *sock,
636 struct msghdr *m, size_t total_len)
638 struct sock *sk = sock->sk;
639 struct tipc_port *tport = tipc_sk_port(sk);
640 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
644 /* Handle implied connection establishment */
647 return send_msg(iocb, sock, m, total_len);
649 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
650 (m->msg_iovlen > (unsigned)INT_MAX))
656 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
659 if (unlikely(sock->state != SS_CONNECTED)) {
660 if (sock->state == SS_DISCONNECTING)
667 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
669 if (likely(res != -ELINKCONG))
671 if (timeout_val <= 0L) {
672 res = timeout_val ? timeout_val : -EWOULDBLOCK;
676 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
677 (!tport->congested || !tport->connected), timeout_val);
687 * send_stream - send stream-oriented data
689 * @sock: socket structure
691 * @total_len: total length of data to be sent
693 * Used for SOCK_STREAM data.
695 * Returns the number of bytes sent on success (or partial success),
696 * or errno if no data sent
699 static int send_stream(struct kiocb *iocb, struct socket *sock,
700 struct msghdr *m, size_t total_len)
702 struct sock *sk = sock->sk;
703 struct tipc_port *tport = tipc_sk_port(sk);
704 struct msghdr my_msg;
706 struct iovec *curr_iov;
708 char __user *curr_start;
717 /* Handle special cases where there is no connection */
719 if (unlikely(sock->state != SS_CONNECTED)) {
720 if (sock->state == SS_UNCONNECTED) {
721 res = send_packet(NULL, sock, m, total_len);
723 } else if (sock->state == SS_DISCONNECTING) {
732 if (unlikely(m->msg_name)) {
737 if ((total_len > (unsigned)INT_MAX) ||
738 (m->msg_iovlen > (unsigned)INT_MAX)) {
744 * Send each iovec entry using one or more messages
746 * Note: This algorithm is good for the most likely case
747 * (i.e. one large iovec entry), but could be improved to pass sets
748 * of small iovec entries into send_packet().
751 curr_iov = m->msg_iov;
752 curr_iovlen = m->msg_iovlen;
753 my_msg.msg_iov = &my_iov;
754 my_msg.msg_iovlen = 1;
755 my_msg.msg_flags = m->msg_flags;
756 my_msg.msg_name = NULL;
759 hdr_size = msg_hdr_sz(&tport->phdr);
761 while (curr_iovlen--) {
762 curr_start = curr_iov->iov_base;
763 curr_left = curr_iov->iov_len;
766 bytes_to_send = tport->max_pkt - hdr_size;
767 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
768 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
769 if (curr_left < bytes_to_send)
770 bytes_to_send = curr_left;
771 my_iov.iov_base = curr_start;
772 my_iov.iov_len = bytes_to_send;
773 res = send_packet(NULL, sock, &my_msg, bytes_to_send);
779 curr_left -= bytes_to_send;
780 curr_start += bytes_to_send;
781 bytes_sent += bytes_to_send;
793 * auto_connect - complete connection setup to a remote port
794 * @sock: socket structure
795 * @msg: peer's response message
797 * Returns 0 on success, errno otherwise
800 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
802 struct tipc_sock *tsock = tipc_sk(sock->sk);
804 if (msg_errcode(msg)) {
805 sock->state = SS_DISCONNECTING;
806 return -ECONNREFUSED;
809 tsock->peer_name.ref = msg_origport(msg);
810 tsock->peer_name.node = msg_orignode(msg);
811 tipc_connect2port(tsock->p->ref, &tsock->peer_name);
812 tipc_set_portimportance(tsock->p->ref, msg_importance(msg));
813 sock->state = SS_CONNECTED;
818 * set_orig_addr - capture sender's address for received message
819 * @m: descriptor for message info
820 * @msg: received message header
822 * Note: Address is not captured if not requested by receiver.
825 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
827 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
830 addr->family = AF_TIPC;
831 addr->addrtype = TIPC_ADDR_ID;
832 memset(&addr->addr, 0, sizeof(addr->addr));
833 addr->addr.id.ref = msg_origport(msg);
834 addr->addr.id.node = msg_orignode(msg);
835 addr->addr.name.domain = 0; /* could leave uninitialized */
836 addr->scope = 0; /* could leave uninitialized */
837 m->msg_namelen = sizeof(struct sockaddr_tipc);
842 * anc_data_recv - optionally capture ancillary data for received message
843 * @m: descriptor for message info
844 * @msg: received message header
845 * @tport: TIPC port associated with message
847 * Note: Ancillary data is not captured if not requested by receiver.
849 * Returns 0 if successful, otherwise errno
852 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
853 struct tipc_port *tport)
861 if (likely(m->msg_controllen == 0))
864 /* Optionally capture errored message object(s) */
866 err = msg ? msg_errcode(msg) : 0;
869 anc_data[1] = msg_data_sz(msg);
870 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
874 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
881 /* Optionally capture message destination object */
883 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
887 anc_data[0] = msg_nametype(msg);
888 anc_data[1] = msg_namelower(msg);
889 anc_data[2] = msg_namelower(msg);
893 anc_data[0] = msg_nametype(msg);
894 anc_data[1] = msg_namelower(msg);
895 anc_data[2] = msg_nameupper(msg);
898 has_name = (tport->conn_type != 0);
899 anc_data[0] = tport->conn_type;
900 anc_data[1] = tport->conn_instance;
901 anc_data[2] = tport->conn_instance;
907 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
916 * recv_msg - receive packet-oriented message
918 * @m: descriptor for message info
919 * @buf_len: total size of user buffer area
920 * @flags: receive flags
922 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
923 * If the complete message doesn't fit in user area, truncate it.
925 * Returns size of returned message data, errno otherwise
928 static int recv_msg(struct kiocb *iocb, struct socket *sock,
929 struct msghdr *m, size_t buf_len, int flags)
931 struct sock *sk = sock->sk;
932 struct tipc_port *tport = tipc_sk_port(sk);
934 struct tipc_msg *msg;
940 /* Catch invalid receive requests */
942 if (unlikely(!buf_len))
947 if (unlikely(sock->state == SS_UNCONNECTED)) {
952 /* will be updated in set_orig_addr() if needed */
955 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
958 /* Look for a message in receive queue; wait if necessary */
960 while (skb_queue_empty(&sk->sk_receive_queue)) {
961 if (sock->state == SS_DISCONNECTING) {
966 res = timeout ? timeout : -EWOULDBLOCK;
970 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
976 /* Look at first message in receive queue */
978 buf = skb_peek(&sk->sk_receive_queue);
980 sz = msg_data_sz(msg);
981 err = msg_errcode(msg);
983 /* Complete connection setup for an implied connect */
985 if (unlikely(sock->state == SS_CONNECTING)) {
986 res = auto_connect(sock, msg);
991 /* Discard an empty non-errored message & try again */
993 if ((!sz) && (!err)) {
994 advance_rx_queue(sk);
998 /* Capture sender's address (optional) */
1000 set_orig_addr(m, msg);
1002 /* Capture ancillary data (optional) */
1004 res = anc_data_recv(m, msg, tport);
1008 /* Capture message data (if valid) & compute return value (always) */
1011 if (unlikely(buf_len < sz)) {
1013 m->msg_flags |= MSG_TRUNC;
1015 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
1021 if ((sock->state == SS_READY) ||
1022 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1028 /* Consume received message (optional) */
1030 if (likely(!(flags & MSG_PEEK))) {
1031 if ((sock->state != SS_READY) &&
1032 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1033 tipc_acknowledge(tport->ref, tport->conn_unacked);
1034 advance_rx_queue(sk);
1042 * recv_stream - receive stream-oriented data
1044 * @m: descriptor for message info
1045 * @buf_len: total size of user buffer area
1046 * @flags: receive flags
1048 * Used for SOCK_STREAM messages only. If not enough data is available
1049 * will optionally wait for more; never truncates data.
1051 * Returns size of returned message data, errno otherwise
1054 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1055 struct msghdr *m, size_t buf_len, int flags)
1057 struct sock *sk = sock->sk;
1058 struct tipc_port *tport = tipc_sk_port(sk);
1059 struct sk_buff *buf;
1060 struct tipc_msg *msg;
1063 int sz_to_copy, target, needed;
1068 /* Catch invalid receive attempts */
1070 if (unlikely(!buf_len))
1075 if (unlikely((sock->state == SS_UNCONNECTED) ||
1076 (sock->state == SS_CONNECTING))) {
1081 /* will be updated in set_orig_addr() if needed */
1084 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1085 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1088 /* Look for a message in receive queue; wait if necessary */
1090 while (skb_queue_empty(&sk->sk_receive_queue)) {
1091 if (sock->state == SS_DISCONNECTING) {
1095 if (timeout <= 0L) {
1096 res = timeout ? timeout : -EWOULDBLOCK;
1100 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1101 tipc_rx_ready(sock),
1106 /* Look at first message in receive queue */
1108 buf = skb_peek(&sk->sk_receive_queue);
1110 sz = msg_data_sz(msg);
1111 err = msg_errcode(msg);
1113 /* Discard an empty non-errored message & try again */
1115 if ((!sz) && (!err)) {
1116 advance_rx_queue(sk);
1120 /* Optionally capture sender's address & ancillary data of first msg */
1122 if (sz_copied == 0) {
1123 set_orig_addr(m, msg);
1124 res = anc_data_recv(m, msg, tport);
1129 /* Capture message data (if valid) & compute return value (always) */
1132 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1135 needed = (buf_len - sz_copied);
1136 sz_to_copy = (sz <= needed) ? sz : needed;
1138 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1139 m->msg_iov, sz_to_copy);
1143 sz_copied += sz_to_copy;
1145 if (sz_to_copy < sz) {
1146 if (!(flags & MSG_PEEK))
1147 TIPC_SKB_CB(buf)->handle =
1148 (void *)(unsigned long)(offset + sz_to_copy);
1153 goto exit; /* can't add error msg to valid data */
1155 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1161 /* Consume received message (optional) */
1163 if (likely(!(flags & MSG_PEEK))) {
1164 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1165 tipc_acknowledge(tport->ref, tport->conn_unacked);
1166 advance_rx_queue(sk);
1169 /* Loop around if more data is required */
1171 if ((sz_copied < buf_len) && /* didn't get all requested data */
1172 (!skb_queue_empty(&sk->sk_receive_queue) ||
1173 (sz_copied < target)) && /* and more is ready or required */
1174 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1175 (!err)) /* and haven't reached a FIN */
1180 return sz_copied ? sz_copied : res;
1184 * rx_queue_full - determine if receive queue can accept another message
1185 * @msg: message to be added to queue
1186 * @queue_size: current size of queue
1187 * @base: nominal maximum size of queue
1189 * Returns 1 if queue is unable to accept message, 0 otherwise
1192 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1195 u32 imp = msg_importance(msg);
1197 if (imp == TIPC_LOW_IMPORTANCE)
1199 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1200 threshold = base * 2;
1201 else if (imp == TIPC_HIGH_IMPORTANCE)
1202 threshold = base * 100;
1206 if (msg_connected(msg))
1209 return queue_size >= threshold;
1213 * filter_rcv - validate incoming message
1217 * Enqueues message on receive queue if acceptable; optionally handles
1218 * disconnect indication for a connected socket.
1220 * Called with socket lock already taken; port lock may also be taken.
1222 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1225 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1227 struct socket *sock = sk->sk_socket;
1228 struct tipc_msg *msg = buf_msg(buf);
1231 /* Reject message if it is wrong sort of message for socket */
1234 * WOULD IT BE BETTER TO JUST DISCARD THESE MESSAGES INSTEAD?
1235 * "NO PORT" ISN'T REALLY THE RIGHT ERROR CODE, AND THERE MAY
1236 * BE SECURITY IMPLICATIONS INHERENT IN REJECTING INVALID TRAFFIC
1239 if (sock->state == SS_READY) {
1240 if (msg_connected(msg))
1241 return TIPC_ERR_NO_PORT;
1244 return TIPC_ERR_NO_PORT;
1245 if (sock->state == SS_CONNECTED) {
1246 if (!msg_connected(msg))
1247 return TIPC_ERR_NO_PORT;
1248 } else if (sock->state == SS_CONNECTING) {
1249 if (!msg_connected(msg) && (msg_errcode(msg) == 0))
1250 return TIPC_ERR_NO_PORT;
1251 } else if (sock->state == SS_LISTENING) {
1252 if (msg_connected(msg) || msg_errcode(msg))
1253 return TIPC_ERR_NO_PORT;
1254 } else if (sock->state == SS_DISCONNECTING) {
1255 return TIPC_ERR_NO_PORT;
1256 } else /* (sock->state == SS_UNCONNECTED) */ {
1257 if (msg_connected(msg) || msg_errcode(msg))
1258 return TIPC_ERR_NO_PORT;
1262 /* Reject message if there isn't room to queue it */
1264 recv_q_len = (u32)atomic_read(&tipc_queue_size);
1265 if (unlikely(recv_q_len >= OVERLOAD_LIMIT_BASE)) {
1266 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE))
1267 return TIPC_ERR_OVERLOAD;
1269 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1270 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1271 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1272 return TIPC_ERR_OVERLOAD;
1275 /* Enqueue message (finally!) */
1277 TIPC_SKB_CB(buf)->handle = 0;
1278 atomic_inc(&tipc_queue_size);
1279 __skb_queue_tail(&sk->sk_receive_queue, buf);
1281 /* Initiate connection termination for an incoming 'FIN' */
1283 if (unlikely(msg_errcode(msg) && (sock->state == SS_CONNECTED))) {
1284 sock->state = SS_DISCONNECTING;
1285 tipc_disconnect_port(tipc_sk_port(sk));
1288 if (waitqueue_active(sk_sleep(sk)))
1289 wake_up_interruptible(sk_sleep(sk));
1294 * backlog_rcv - handle incoming message from backlog queue
1298 * Caller must hold socket lock, but not port lock.
1303 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1307 res = filter_rcv(sk, buf);
1309 tipc_reject_msg(buf, res);
1314 * dispatch - handle incoming message
1315 * @tport: TIPC port that received message
1318 * Called with port lock already taken.
1320 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1323 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1325 struct sock *sk = (struct sock *)tport->usr_handle;
1329 * Process message if socket is unlocked; otherwise add to backlog queue
1331 * This code is based on sk_receive_skb(), but must be distinct from it
1332 * since a TIPC-specific filter/reject mechanism is utilized
1336 if (!sock_owned_by_user(sk)) {
1337 res = filter_rcv(sk, buf);
1339 if (sk_add_backlog(sk, buf))
1340 res = TIPC_ERR_OVERLOAD;
1350 * wakeupdispatch - wake up port after congestion
1351 * @tport: port to wakeup
1353 * Called with port lock already taken.
1356 static void wakeupdispatch(struct tipc_port *tport)
1358 struct sock *sk = (struct sock *)tport->usr_handle;
1360 if (waitqueue_active(sk_sleep(sk)))
1361 wake_up_interruptible(sk_sleep(sk));
1365 * connect - establish a connection to another TIPC port
1366 * @sock: socket structure
1367 * @dest: socket address for destination port
1368 * @destlen: size of socket address data structure
1369 * @flags: file-related flags associated with socket
1371 * Returns 0 on success, errno otherwise
1374 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1377 struct sock *sk = sock->sk;
1378 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1379 struct msghdr m = {NULL,};
1380 struct sk_buff *buf;
1381 struct tipc_msg *msg;
1382 unsigned int timeout;
1387 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1389 if (sock->state == SS_READY) {
1394 /* For now, TIPC does not support the non-blocking form of connect() */
1396 if (flags & O_NONBLOCK) {
1401 /* Issue Posix-compliant error code if socket is in the wrong state */
1403 if (sock->state == SS_LISTENING) {
1407 if (sock->state == SS_CONNECTING) {
1411 if (sock->state != SS_UNCONNECTED) {
1417 * Reject connection attempt using multicast address
1419 * Note: send_msg() validates the rest of the address fields,
1420 * so there's no need to do it here
1423 if (dst->addrtype == TIPC_ADDR_MCAST) {
1428 /* Reject any messages already in receive queue (very unlikely) */
1430 reject_rx_queue(sk);
1432 /* Send a 'SYN-' to destination */
1435 m.msg_namelen = destlen;
1436 res = send_msg(NULL, sock, &m, 0);
1440 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1442 timeout = tipc_sk(sk)->conn_timeout;
1444 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1445 (!skb_queue_empty(&sk->sk_receive_queue) ||
1446 (sock->state != SS_CONNECTING)),
1447 timeout ? (long)msecs_to_jiffies(timeout)
1448 : MAX_SCHEDULE_TIMEOUT);
1452 buf = skb_peek(&sk->sk_receive_queue);
1455 res = auto_connect(sock, msg);
1457 if (!msg_data_sz(msg))
1458 advance_rx_queue(sk);
1461 if (sock->state == SS_CONNECTED)
1464 res = -ECONNREFUSED;
1470 ; /* leave "res" unchanged */
1471 sock->state = SS_DISCONNECTING;
1480 * listen - allow socket to listen for incoming connections
1481 * @sock: socket structure
1484 * Returns 0 on success, errno otherwise
1487 static int listen(struct socket *sock, int len)
1489 struct sock *sk = sock->sk;
1494 if (sock->state != SS_UNCONNECTED)
1497 sock->state = SS_LISTENING;
1506 * accept - wait for connection request
1507 * @sock: listening socket
1508 * @newsock: new socket that is to be connected
1509 * @flags: file-related flags associated with socket
1511 * Returns 0 on success, errno otherwise
1514 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1516 struct sock *sk = sock->sk;
1517 struct sk_buff *buf;
1522 if (sock->state != SS_LISTENING) {
1527 while (skb_queue_empty(&sk->sk_receive_queue)) {
1528 if (flags & O_NONBLOCK) {
1533 res = wait_event_interruptible(*sk_sleep(sk),
1534 (!skb_queue_empty(&sk->sk_receive_queue)));
1540 buf = skb_peek(&sk->sk_receive_queue);
1542 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1544 struct sock *new_sk = new_sock->sk;
1545 struct tipc_sock *new_tsock = tipc_sk(new_sk);
1546 struct tipc_port *new_tport = new_tsock->p;
1547 u32 new_ref = new_tport->ref;
1548 struct tipc_msg *msg = buf_msg(buf);
1553 * Reject any stray messages received by new socket
1554 * before the socket lock was taken (very, very unlikely)
1557 reject_rx_queue(new_sk);
1559 /* Connect new socket to it's peer */
1561 new_tsock->peer_name.ref = msg_origport(msg);
1562 new_tsock->peer_name.node = msg_orignode(msg);
1563 tipc_connect2port(new_ref, &new_tsock->peer_name);
1564 new_sock->state = SS_CONNECTED;
1566 tipc_set_portimportance(new_ref, msg_importance(msg));
1567 if (msg_named(msg)) {
1568 new_tport->conn_type = msg_nametype(msg);
1569 new_tport->conn_instance = msg_nameinst(msg);
1573 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1574 * Respond to 'SYN+' by queuing it on new socket.
1577 if (!msg_data_sz(msg)) {
1578 struct msghdr m = {NULL,};
1580 advance_rx_queue(sk);
1581 send_packet(NULL, new_sock, &m, 0);
1583 __skb_dequeue(&sk->sk_receive_queue);
1584 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1586 release_sock(new_sk);
1594 * shutdown - shutdown socket connection
1595 * @sock: socket structure
1596 * @how: direction to close (must be SHUT_RDWR)
1598 * Terminates connection (if necessary), then purges socket's receive queue.
1600 * Returns 0 on success, errno otherwise
1603 static int shutdown(struct socket *sock, int how)
1605 struct sock *sk = sock->sk;
1606 struct tipc_port *tport = tipc_sk_port(sk);
1607 struct sk_buff *buf;
1610 if (how != SHUT_RDWR)
1615 switch (sock->state) {
1619 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1621 buf = __skb_dequeue(&sk->sk_receive_queue);
1623 atomic_dec(&tipc_queue_size);
1624 if (TIPC_SKB_CB(buf)->handle != 0) {
1628 tipc_disconnect(tport->ref);
1629 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1631 tipc_shutdown(tport->ref);
1634 sock->state = SS_DISCONNECTING;
1638 case SS_DISCONNECTING:
1640 /* Discard any unreceived messages; wake up sleeping tasks */
1642 discard_rx_queue(sk);
1643 if (waitqueue_active(sk_sleep(sk)))
1644 wake_up_interruptible(sk_sleep(sk));
1657 * setsockopt - set socket option
1658 * @sock: socket structure
1659 * @lvl: option level
1660 * @opt: option identifier
1661 * @ov: pointer to new option value
1662 * @ol: length of option value
1664 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1665 * (to ease compatibility).
1667 * Returns 0 on success, errno otherwise
1670 static int setsockopt(struct socket *sock,
1671 int lvl, int opt, char __user *ov, unsigned int ol)
1673 struct sock *sk = sock->sk;
1674 struct tipc_port *tport = tipc_sk_port(sk);
1678 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1680 if (lvl != SOL_TIPC)
1681 return -ENOPROTOOPT;
1682 if (ol < sizeof(value))
1684 res = get_user(value, (u32 __user *)ov);
1691 case TIPC_IMPORTANCE:
1692 res = tipc_set_portimportance(tport->ref, value);
1694 case TIPC_SRC_DROPPABLE:
1695 if (sock->type != SOCK_STREAM)
1696 res = tipc_set_portunreliable(tport->ref, value);
1700 case TIPC_DEST_DROPPABLE:
1701 res = tipc_set_portunreturnable(tport->ref, value);
1703 case TIPC_CONN_TIMEOUT:
1704 tipc_sk(sk)->conn_timeout = value;
1705 /* no need to set "res", since already 0 at this point */
1717 * getsockopt - get socket option
1718 * @sock: socket structure
1719 * @lvl: option level
1720 * @opt: option identifier
1721 * @ov: receptacle for option value
1722 * @ol: receptacle for length of option value
1724 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1725 * (to ease compatibility).
1727 * Returns 0 on success, errno otherwise
1730 static int getsockopt(struct socket *sock,
1731 int lvl, int opt, char __user *ov, int __user *ol)
1733 struct sock *sk = sock->sk;
1734 struct tipc_port *tport = tipc_sk_port(sk);
1739 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1740 return put_user(0, ol);
1741 if (lvl != SOL_TIPC)
1742 return -ENOPROTOOPT;
1743 res = get_user(len, ol);
1750 case TIPC_IMPORTANCE:
1751 res = tipc_portimportance(tport->ref, &value);
1753 case TIPC_SRC_DROPPABLE:
1754 res = tipc_portunreliable(tport->ref, &value);
1756 case TIPC_DEST_DROPPABLE:
1757 res = tipc_portunreturnable(tport->ref, &value);
1759 case TIPC_CONN_TIMEOUT:
1760 value = tipc_sk(sk)->conn_timeout;
1761 /* no need to set "res", since already 0 at this point */
1763 case TIPC_NODE_RECVQ_DEPTH:
1764 value = (u32)atomic_read(&tipc_queue_size);
1766 case TIPC_SOCK_RECVQ_DEPTH:
1767 value = skb_queue_len(&sk->sk_receive_queue);
1776 return res; /* "get" failed */
1778 if (len < sizeof(value))
1781 if (copy_to_user(ov, &value, sizeof(value)))
1784 return put_user(sizeof(value), ol);
1788 * Protocol switches for the various types of TIPC sockets
1791 static const struct proto_ops msg_ops = {
1792 .owner = THIS_MODULE,
1797 .socketpair = sock_no_socketpair,
1798 .accept = sock_no_accept,
1799 .getname = get_name,
1801 .ioctl = sock_no_ioctl,
1802 .listen = sock_no_listen,
1803 .shutdown = shutdown,
1804 .setsockopt = setsockopt,
1805 .getsockopt = getsockopt,
1806 .sendmsg = send_msg,
1807 .recvmsg = recv_msg,
1808 .mmap = sock_no_mmap,
1809 .sendpage = sock_no_sendpage
1812 static const struct proto_ops packet_ops = {
1813 .owner = THIS_MODULE,
1818 .socketpair = sock_no_socketpair,
1820 .getname = get_name,
1822 .ioctl = sock_no_ioctl,
1824 .shutdown = shutdown,
1825 .setsockopt = setsockopt,
1826 .getsockopt = getsockopt,
1827 .sendmsg = send_packet,
1828 .recvmsg = recv_msg,
1829 .mmap = sock_no_mmap,
1830 .sendpage = sock_no_sendpage
1833 static const struct proto_ops stream_ops = {
1834 .owner = THIS_MODULE,
1839 .socketpair = sock_no_socketpair,
1841 .getname = get_name,
1843 .ioctl = sock_no_ioctl,
1845 .shutdown = shutdown,
1846 .setsockopt = setsockopt,
1847 .getsockopt = getsockopt,
1848 .sendmsg = send_stream,
1849 .recvmsg = recv_stream,
1850 .mmap = sock_no_mmap,
1851 .sendpage = sock_no_sendpage
1854 static const struct net_proto_family tipc_family_ops = {
1855 .owner = THIS_MODULE,
1857 .create = tipc_create
1860 static struct proto tipc_proto = {
1862 .owner = THIS_MODULE,
1863 .obj_size = sizeof(struct tipc_sock)
1867 * tipc_socket_init - initialize TIPC socket interface
1869 * Returns 0 on success, errno otherwise
1871 int tipc_socket_init(void)
1875 res = proto_register(&tipc_proto, 1);
1877 err("Failed to register TIPC protocol type\n");
1881 res = sock_register(&tipc_family_ops);
1883 err("Failed to register TIPC socket type\n");
1884 proto_unregister(&tipc_proto);
1888 sockets_enabled = 1;
1894 * tipc_socket_stop - stop TIPC socket interface
1897 void tipc_socket_stop(void)
1899 if (!sockets_enabled)
1902 sockets_enabled = 0;
1903 sock_unregister(tipc_family_ops.family);
1904 proto_unregister(&tipc_proto);