2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012 Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2012, 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.
40 #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 10000
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) (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);
65 static void tipc_data_ready(struct sock *sk, int len);
66 static void tipc_write_space(struct sock *sk);
68 static const struct proto_ops packet_ops;
69 static const struct proto_ops stream_ops;
70 static const struct proto_ops msg_ops;
72 static struct proto tipc_proto;
74 static int sockets_enabled;
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
126 static void advance_rx_queue(struct sock *sk)
128 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
132 * discard_rx_queue - discard all buffers in socket receive queue
134 * Caller must hold socket lock
136 static void discard_rx_queue(struct sock *sk)
140 while ((buf = __skb_dequeue(&sk->sk_receive_queue)))
145 * reject_rx_queue - reject all buffers in socket receive queue
147 * Caller must hold socket lock
149 static void reject_rx_queue(struct sock *sk)
153 while ((buf = __skb_dequeue(&sk->sk_receive_queue)))
154 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
158 * tipc_create - create a TIPC socket
159 * @net: network namespace (must be default network)
160 * @sock: pre-allocated socket structure
161 * @protocol: protocol indicator (must be 0)
162 * @kern: caused by kernel or by userspace?
164 * This routine creates additional data structures used by the TIPC socket,
165 * initializes them, and links them together.
167 * Returns 0 on success, errno otherwise
169 static int tipc_create(struct net *net, struct socket *sock, int protocol,
172 const struct proto_ops *ops;
175 struct tipc_port *tp_ptr;
177 /* Validate arguments */
178 if (unlikely(protocol != 0))
179 return -EPROTONOSUPPORT;
181 switch (sock->type) {
184 state = SS_UNCONNECTED;
188 state = SS_UNCONNECTED;
199 /* Allocate socket's protocol area */
200 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
204 /* Allocate TIPC port for socket to use */
205 tp_ptr = tipc_createport_raw(sk, &dispatch, &wakeupdispatch,
206 TIPC_LOW_IMPORTANCE);
207 if (unlikely(!tp_ptr)) {
212 /* Finish initializing socket data structures */
216 sock_init_data(sock, sk);
217 sk->sk_backlog_rcv = backlog_rcv;
218 sk->sk_rcvbuf = TIPC_FLOW_CONTROL_WIN * 2 * TIPC_MAX_USER_MSG_SIZE * 2;
219 sk->sk_data_ready = tipc_data_ready;
220 sk->sk_write_space = tipc_write_space;
221 tipc_sk(sk)->p = tp_ptr;
222 tipc_sk(sk)->conn_timeout = CONN_TIMEOUT_DEFAULT;
224 spin_unlock_bh(tp_ptr->lock);
226 if (sock->state == SS_READY) {
227 tipc_set_portunreturnable(tp_ptr->ref, 1);
228 if (sock->type == SOCK_DGRAM)
229 tipc_set_portunreliable(tp_ptr->ref, 1);
236 * release - destroy a TIPC socket
237 * @sock: socket to destroy
239 * This routine cleans up any messages that are still queued on the socket.
240 * For DGRAM and RDM socket types, all queued messages are rejected.
241 * For SEQPACKET and STREAM socket types, the first message is rejected
242 * and any others are discarded. (If the first message on a STREAM socket
243 * is partially-read, it is discarded and the next one is rejected instead.)
245 * NOTE: Rejected messages are not necessarily returned to the sender! They
246 * are returned or discarded according to the "destination droppable" setting
247 * specified for the message by the sender.
249 * Returns 0 on success, errno otherwise
251 static int release(struct socket *sock)
253 struct sock *sk = sock->sk;
254 struct tipc_port *tport;
259 * Exit if socket isn't fully initialized (occurs when a failed accept()
260 * releases a pre-allocated child socket that was never used)
265 tport = tipc_sk_port(sk);
269 * Reject all unreceived messages, except on an active connection
270 * (which disconnects locally & sends a 'FIN+' to peer)
272 while (sock->state != SS_DISCONNECTING) {
273 buf = __skb_dequeue(&sk->sk_receive_queue);
276 if (TIPC_SKB_CB(buf)->handle != 0)
279 if ((sock->state == SS_CONNECTING) ||
280 (sock->state == SS_CONNECTED)) {
281 sock->state = SS_DISCONNECTING;
282 tipc_disconnect(tport->ref);
284 tipc_reject_msg(buf, TIPC_ERR_NO_PORT);
289 * Delete TIPC port; this ensures no more messages are queued
290 * (also disconnects an active connection & sends a 'FIN-' to peer)
292 res = tipc_deleteport(tport->ref);
294 /* Discard any remaining (connection-based) messages in receive queue */
295 discard_rx_queue(sk);
297 /* Reject any messages that accumulated in backlog queue */
298 sock->state = SS_DISCONNECTING;
308 * bind - associate or disassocate TIPC name(s) with a socket
309 * @sock: socket structure
310 * @uaddr: socket address describing name(s) and desired operation
311 * @uaddr_len: size of socket address data structure
313 * Name and name sequence binding is indicated using a positive scope value;
314 * a negative scope value unbinds the specified name. Specifying no name
315 * (i.e. a socket address length of 0) unbinds all names from the socket.
317 * Returns 0 on success, errno otherwise
319 * NOTE: This routine doesn't need to take the socket lock since it doesn't
320 * access any non-constant socket information.
322 static int bind(struct socket *sock, struct sockaddr *uaddr, int uaddr_len)
324 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
325 u32 portref = tipc_sk_port(sock->sk)->ref;
327 if (unlikely(!uaddr_len))
328 return tipc_withdraw(portref, 0, NULL);
330 if (uaddr_len < sizeof(struct sockaddr_tipc))
332 if (addr->family != AF_TIPC)
333 return -EAFNOSUPPORT;
335 if (addr->addrtype == TIPC_ADDR_NAME)
336 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
337 else if (addr->addrtype != TIPC_ADDR_NAMESEQ)
338 return -EAFNOSUPPORT;
340 if (addr->addr.nameseq.type < TIPC_RESERVED_TYPES)
343 return (addr->scope > 0) ?
344 tipc_publish(portref, addr->scope, &addr->addr.nameseq) :
345 tipc_withdraw(portref, -addr->scope, &addr->addr.nameseq);
349 * get_name - get port ID of socket or peer socket
350 * @sock: socket structure
351 * @uaddr: area for returned socket address
352 * @uaddr_len: area for returned length of socket address
353 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
355 * Returns 0 on success, errno otherwise
357 * NOTE: This routine doesn't need to take the socket lock since it only
358 * accesses socket information that is unchanging (or which changes in
359 * a completely predictable manner).
361 static int get_name(struct socket *sock, struct sockaddr *uaddr,
362 int *uaddr_len, int peer)
364 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
365 struct tipc_sock *tsock = tipc_sk(sock->sk);
367 memset(addr, 0, sizeof(*addr));
369 if ((sock->state != SS_CONNECTED) &&
370 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
372 addr->addr.id.ref = tsock->peer_name.ref;
373 addr->addr.id.node = tsock->peer_name.node;
375 addr->addr.id.ref = tsock->p->ref;
376 addr->addr.id.node = tipc_own_addr;
379 *uaddr_len = sizeof(*addr);
380 addr->addrtype = TIPC_ADDR_ID;
381 addr->family = AF_TIPC;
383 addr->addr.name.domain = 0;
389 * poll - read and possibly block on pollmask
390 * @file: file structure associated with the socket
391 * @sock: socket for which to calculate the poll bits
394 * Returns pollmask value
397 * It appears that the usual socket locking mechanisms are not useful here
398 * since the pollmask info is potentially out-of-date the moment this routine
399 * exits. TCP and other protocols seem to rely on higher level poll routines
400 * to handle any preventable race conditions, so TIPC will do the same ...
402 * TIPC sets the returned events as follows:
404 * socket state flags set
405 * ------------ ---------
406 * unconnected no read flags
407 * POLLOUT if port is not congested
409 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
412 * connected POLLIN/POLLRDNORM if data in rx queue
413 * POLLOUT if port is not congested
415 * disconnecting POLLIN/POLLRDNORM/POLLHUP
418 * listening POLLIN if SYN in rx queue
421 * ready POLLIN/POLLRDNORM if data in rx queue
422 * [connectionless] POLLOUT (since port cannot be congested)
424 * IMPORTANT: The fact that a read or write operation is indicated does NOT
425 * imply that the operation will succeed, merely that it should be performed
426 * and will not block.
428 static unsigned int poll(struct file *file, struct socket *sock,
431 struct sock *sk = sock->sk;
434 sock_poll_wait(file, sk_sleep(sk), wait);
436 switch ((int)sock->state) {
438 if (!tipc_sk_port(sk)->congested)
443 if (!tipc_sk_port(sk)->congested)
448 if (!skb_queue_empty(&sk->sk_receive_queue))
449 mask |= (POLLIN | POLLRDNORM);
451 case SS_DISCONNECTING:
452 mask = (POLLIN | POLLRDNORM | POLLHUP);
460 * dest_name_check - verify user is permitted to send to specified port name
461 * @dest: destination address
462 * @m: descriptor for message to be sent
464 * Prevents restricted configuration commands from being issued by
465 * unauthorized users.
467 * Returns 0 if permission is granted, otherwise errno
469 static int dest_name_check(struct sockaddr_tipc *dest, struct msghdr *m)
471 struct tipc_cfg_msg_hdr hdr;
473 if (likely(dest->addr.name.name.type >= TIPC_RESERVED_TYPES))
475 if (likely(dest->addr.name.name.type == TIPC_TOP_SRV))
477 if (likely(dest->addr.name.name.type != TIPC_CFG_SRV))
480 if (!m->msg_iovlen || (m->msg_iov[0].iov_len < sizeof(hdr)))
482 if (copy_from_user(&hdr, m->msg_iov[0].iov_base, sizeof(hdr)))
484 if ((ntohs(hdr.tcm_type) & 0xC000) && (!capable(CAP_NET_ADMIN)))
491 * send_msg - send message in connectionless manner
492 * @iocb: if NULL, indicates that socket lock is already held
493 * @sock: socket structure
494 * @m: message to send
495 * @total_len: length of message
497 * Message must have an destination specified explicitly.
498 * Used for SOCK_RDM and SOCK_DGRAM messages,
499 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
500 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
502 * Returns the number of bytes sent on success, or errno otherwise
504 static int send_msg(struct kiocb *iocb, struct socket *sock,
505 struct msghdr *m, size_t total_len)
507 struct sock *sk = sock->sk;
508 struct tipc_port *tport = tipc_sk_port(sk);
509 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
515 return -EDESTADDRREQ;
516 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
517 (dest->family != AF_TIPC)))
519 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
520 (m->msg_iovlen > (unsigned int)INT_MAX))
526 needs_conn = (sock->state != SS_READY);
527 if (unlikely(needs_conn)) {
528 if (sock->state == SS_LISTENING) {
532 if (sock->state != SS_UNCONNECTED) {
536 if ((tport->published) ||
537 ((sock->type == SOCK_STREAM) && (total_len != 0))) {
541 if (dest->addrtype == TIPC_ADDR_NAME) {
542 tport->conn_type = dest->addr.name.name.type;
543 tport->conn_instance = dest->addr.name.name.instance;
546 /* Abort any pending connection attempts (very unlikely) */
550 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
553 if (dest->addrtype == TIPC_ADDR_NAME) {
554 res = dest_name_check(dest, m);
557 res = tipc_send2name(tport->ref,
558 &dest->addr.name.name,
559 dest->addr.name.domain,
563 } else if (dest->addrtype == TIPC_ADDR_ID) {
564 res = tipc_send2port(tport->ref,
569 } else if (dest->addrtype == TIPC_ADDR_MCAST) {
574 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;
588 if (timeout_val <= 0L) {
589 res = timeout_val ? timeout_val : -EWOULDBLOCK;
593 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
594 !tport->congested, timeout_val);
605 * send_packet - send a connection-oriented message
606 * @iocb: if NULL, indicates that socket lock is already held
607 * @sock: socket structure
608 * @m: message to send
609 * @total_len: length of message
611 * Used for SOCK_SEQPACKET messages and SOCK_STREAM data.
613 * Returns the number of bytes sent on success, or errno otherwise
615 static int send_packet(struct kiocb *iocb, struct socket *sock,
616 struct msghdr *m, size_t total_len)
618 struct sock *sk = sock->sk;
619 struct tipc_port *tport = tipc_sk_port(sk);
620 struct sockaddr_tipc *dest = (struct sockaddr_tipc *)m->msg_name;
624 /* Handle implied connection establishment */
626 return send_msg(iocb, sock, m, total_len);
628 if ((total_len > TIPC_MAX_USER_MSG_SIZE) ||
629 (m->msg_iovlen > (unsigned int)INT_MAX))
635 timeout_val = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
638 if (unlikely(sock->state != SS_CONNECTED)) {
639 if (sock->state == SS_DISCONNECTING)
646 res = tipc_send(tport->ref, m->msg_iovlen, m->msg_iov,
648 if (likely(res != -ELINKCONG))
650 if (timeout_val <= 0L) {
651 res = timeout_val ? timeout_val : -EWOULDBLOCK;
655 timeout_val = wait_event_interruptible_timeout(*sk_sleep(sk),
656 (!tport->congested || !tport->connected), timeout_val);
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
677 static int send_stream(struct kiocb *iocb, struct socket *sock,
678 struct msghdr *m, size_t total_len)
680 struct sock *sk = sock->sk;
681 struct tipc_port *tport = tipc_sk_port(sk);
682 struct msghdr my_msg;
684 struct iovec *curr_iov;
686 char __user *curr_start;
695 /* Handle special cases where there is no connection */
696 if (unlikely(sock->state != SS_CONNECTED)) {
697 if (sock->state == SS_UNCONNECTED) {
698 res = send_packet(NULL, sock, m, total_len);
700 } else if (sock->state == SS_DISCONNECTING) {
709 if (unlikely(m->msg_name)) {
714 if ((total_len > (unsigned int)INT_MAX) ||
715 (m->msg_iovlen > (unsigned int)INT_MAX)) {
721 * Send each iovec entry using one or more messages
723 * Note: This algorithm is good for the most likely case
724 * (i.e. one large iovec entry), but could be improved to pass sets
725 * of small iovec entries into send_packet().
727 curr_iov = m->msg_iov;
728 curr_iovlen = m->msg_iovlen;
729 my_msg.msg_iov = &my_iov;
730 my_msg.msg_iovlen = 1;
731 my_msg.msg_flags = m->msg_flags;
732 my_msg.msg_name = NULL;
735 hdr_size = msg_hdr_sz(&tport->phdr);
737 while (curr_iovlen--) {
738 curr_start = curr_iov->iov_base;
739 curr_left = curr_iov->iov_len;
742 bytes_to_send = tport->max_pkt - hdr_size;
743 if (bytes_to_send > TIPC_MAX_USER_MSG_SIZE)
744 bytes_to_send = TIPC_MAX_USER_MSG_SIZE;
745 if (curr_left < bytes_to_send)
746 bytes_to_send = curr_left;
747 my_iov.iov_base = curr_start;
748 my_iov.iov_len = bytes_to_send;
749 res = send_packet(NULL, sock, &my_msg, bytes_to_send);
755 curr_left -= bytes_to_send;
756 curr_start += bytes_to_send;
757 bytes_sent += bytes_to_send;
769 * auto_connect - complete connection setup to a remote port
770 * @sock: socket structure
771 * @msg: peer's response message
773 * Returns 0 on success, errno otherwise
775 static int auto_connect(struct socket *sock, struct tipc_msg *msg)
777 struct tipc_sock *tsock = tipc_sk(sock->sk);
778 struct tipc_port *p_ptr;
780 tsock->peer_name.ref = msg_origport(msg);
781 tsock->peer_name.node = msg_orignode(msg);
782 p_ptr = tipc_port_deref(tsock->p->ref);
786 __tipc_connect(tsock->p->ref, p_ptr, &tsock->peer_name);
788 if (msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)
790 msg_set_importance(&p_ptr->phdr, (u32)msg_importance(msg));
791 sock->state = SS_CONNECTED;
796 * set_orig_addr - capture sender's address for received message
797 * @m: descriptor for message info
798 * @msg: received message header
800 * Note: Address is not captured if not requested by receiver.
802 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
804 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)m->msg_name;
807 addr->family = AF_TIPC;
808 addr->addrtype = TIPC_ADDR_ID;
809 addr->addr.id.ref = msg_origport(msg);
810 addr->addr.id.node = msg_orignode(msg);
811 addr->addr.name.domain = 0; /* could leave uninitialized */
812 addr->scope = 0; /* could leave uninitialized */
813 m->msg_namelen = sizeof(struct sockaddr_tipc);
818 * anc_data_recv - optionally capture ancillary data for received message
819 * @m: descriptor for message info
820 * @msg: received message header
821 * @tport: TIPC port associated with message
823 * Note: Ancillary data is not captured if not requested by receiver.
825 * Returns 0 if successful, otherwise errno
827 static int anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
828 struct tipc_port *tport)
836 if (likely(m->msg_controllen == 0))
839 /* Optionally capture errored message object(s) */
840 err = msg ? msg_errcode(msg) : 0;
843 anc_data[1] = msg_data_sz(msg);
844 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
848 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
855 /* Optionally capture message destination object */
856 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
860 anc_data[0] = msg_nametype(msg);
861 anc_data[1] = msg_namelower(msg);
862 anc_data[2] = msg_namelower(msg);
866 anc_data[0] = msg_nametype(msg);
867 anc_data[1] = msg_namelower(msg);
868 anc_data[2] = msg_nameupper(msg);
871 has_name = (tport->conn_type != 0);
872 anc_data[0] = tport->conn_type;
873 anc_data[1] = tport->conn_instance;
874 anc_data[2] = tport->conn_instance;
880 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
889 * recv_msg - receive packet-oriented message
891 * @m: descriptor for message info
892 * @buf_len: total size of user buffer area
893 * @flags: receive flags
895 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
896 * If the complete message doesn't fit in user area, truncate it.
898 * Returns size of returned message data, errno otherwise
900 static int recv_msg(struct kiocb *iocb, struct socket *sock,
901 struct msghdr *m, size_t buf_len, int flags)
903 struct sock *sk = sock->sk;
904 struct tipc_port *tport = tipc_sk_port(sk);
906 struct tipc_msg *msg;
912 /* Catch invalid receive requests */
913 if (unlikely(!buf_len))
918 if (unlikely(sock->state == SS_UNCONNECTED)) {
923 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
926 /* Look for a message in receive queue; wait if necessary */
927 while (skb_queue_empty(&sk->sk_receive_queue)) {
928 if (sock->state == SS_DISCONNECTING) {
933 res = timeout ? timeout : -EWOULDBLOCK;
937 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
943 /* 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 /* Discard an empty non-errored message & try again */
950 if ((!sz) && (!err)) {
951 advance_rx_queue(sk);
955 /* Capture sender's address (optional) */
956 set_orig_addr(m, msg);
958 /* Capture ancillary data (optional) */
959 res = anc_data_recv(m, msg, tport);
963 /* Capture message data (if valid) & compute return value (always) */
965 if (unlikely(buf_len < sz)) {
967 m->msg_flags |= MSG_TRUNC;
969 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg),
975 if ((sock->state == SS_READY) ||
976 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
982 /* Consume received message (optional) */
983 if (likely(!(flags & MSG_PEEK))) {
984 if ((sock->state != SS_READY) &&
985 (++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
986 tipc_acknowledge(tport->ref, tport->conn_unacked);
987 advance_rx_queue(sk);
995 * recv_stream - receive stream-oriented data
997 * @m: descriptor for message info
998 * @buf_len: total size of user buffer area
999 * @flags: receive flags
1001 * Used for SOCK_STREAM messages only. If not enough data is available
1002 * will optionally wait for more; never truncates data.
1004 * Returns size of returned message data, errno otherwise
1006 static int recv_stream(struct kiocb *iocb, struct socket *sock,
1007 struct msghdr *m, size_t buf_len, int flags)
1009 struct sock *sk = sock->sk;
1010 struct tipc_port *tport = tipc_sk_port(sk);
1011 struct sk_buff *buf;
1012 struct tipc_msg *msg;
1015 int sz_to_copy, target, needed;
1020 /* Catch invalid receive attempts */
1021 if (unlikely(!buf_len))
1026 if (unlikely((sock->state == SS_UNCONNECTED) ||
1027 (sock->state == SS_CONNECTING))) {
1032 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1033 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1036 /* Look for a message in receive queue; wait if necessary */
1037 while (skb_queue_empty(&sk->sk_receive_queue)) {
1038 if (sock->state == SS_DISCONNECTING) {
1042 if (timeout <= 0L) {
1043 res = timeout ? timeout : -EWOULDBLOCK;
1047 timeout = wait_event_interruptible_timeout(*sk_sleep(sk),
1048 tipc_rx_ready(sock),
1053 /* Look at first message in receive queue */
1054 buf = skb_peek(&sk->sk_receive_queue);
1056 sz = msg_data_sz(msg);
1057 err = msg_errcode(msg);
1059 /* Discard an empty non-errored message & try again */
1060 if ((!sz) && (!err)) {
1061 advance_rx_queue(sk);
1065 /* Optionally capture sender's address & ancillary data of first msg */
1066 if (sz_copied == 0) {
1067 set_orig_addr(m, msg);
1068 res = anc_data_recv(m, msg, tport);
1073 /* Capture message data (if valid) & compute return value (always) */
1075 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1078 needed = (buf_len - sz_copied);
1079 sz_to_copy = (sz <= needed) ? sz : needed;
1081 res = skb_copy_datagram_iovec(buf, msg_hdr_sz(msg) + offset,
1082 m->msg_iov, sz_to_copy);
1086 sz_copied += sz_to_copy;
1088 if (sz_to_copy < sz) {
1089 if (!(flags & MSG_PEEK))
1090 TIPC_SKB_CB(buf)->handle =
1091 (void *)(unsigned long)(offset + sz_to_copy);
1096 goto exit; /* can't add error msg to valid data */
1098 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1104 /* Consume received message (optional) */
1105 if (likely(!(flags & MSG_PEEK))) {
1106 if (unlikely(++tport->conn_unacked >= TIPC_FLOW_CONTROL_WIN))
1107 tipc_acknowledge(tport->ref, tport->conn_unacked);
1108 advance_rx_queue(sk);
1111 /* Loop around if more data is required */
1112 if ((sz_copied < buf_len) && /* didn't get all requested data */
1113 (!skb_queue_empty(&sk->sk_receive_queue) ||
1114 (sz_copied < target)) && /* and more is ready or required */
1115 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1116 (!err)) /* and haven't reached a FIN */
1121 return sz_copied ? sz_copied : res;
1125 * tipc_write_space - wake up thread if port congestion is released
1128 static void tipc_write_space(struct sock *sk)
1130 struct socket_wq *wq;
1133 wq = rcu_dereference(sk->sk_wq);
1134 if (wq_has_sleeper(wq))
1135 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1136 POLLWRNORM | POLLWRBAND);
1141 * tipc_data_ready - wake up threads to indicate messages have been received
1143 * @len: the length of messages
1145 static void tipc_data_ready(struct sock *sk, int len)
1147 struct socket_wq *wq;
1150 wq = rcu_dereference(sk->sk_wq);
1151 if (wq_has_sleeper(wq))
1152 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1153 POLLRDNORM | POLLRDBAND);
1158 * rx_queue_full - determine if receive queue can accept another message
1159 * @msg: message to be added to queue
1160 * @queue_size: current size of queue
1161 * @base: nominal maximum size of queue
1163 * Returns 1 if queue is unable to accept message, 0 otherwise
1165 static int rx_queue_full(struct tipc_msg *msg, u32 queue_size, u32 base)
1168 u32 imp = msg_importance(msg);
1170 if (imp == TIPC_LOW_IMPORTANCE)
1172 else if (imp == TIPC_MEDIUM_IMPORTANCE)
1173 threshold = base * 2;
1174 else if (imp == TIPC_HIGH_IMPORTANCE)
1175 threshold = base * 100;
1179 if (msg_connected(msg))
1182 return queue_size >= threshold;
1186 * filter_connect - Handle all incoming messages for a connection-based socket
1187 * @tsock: TIPC socket
1190 * Returns TIPC error status code and socket error status code
1191 * once it encounters some errors
1193 static u32 filter_connect(struct tipc_sock *tsock, struct sk_buff **buf)
1195 struct socket *sock = tsock->sk.sk_socket;
1196 struct tipc_msg *msg = buf_msg(*buf);
1197 struct sock *sk = &tsock->sk;
1198 u32 retval = TIPC_ERR_NO_PORT;
1204 switch ((int)sock->state) {
1206 /* Accept only connection-based messages sent by peer */
1207 if (msg_connected(msg) && tipc_port_peer_msg(tsock->p, msg)) {
1208 if (unlikely(msg_errcode(msg))) {
1209 sock->state = SS_DISCONNECTING;
1210 __tipc_disconnect(tsock->p);
1216 /* Accept only ACK or NACK message */
1217 if (unlikely(msg_errcode(msg))) {
1218 sock->state = SS_DISCONNECTING;
1219 sk->sk_err = -ECONNREFUSED;
1224 if (unlikely(!msg_connected(msg)))
1227 res = auto_connect(sock, msg);
1229 sock->state = SS_DISCONNECTING;
1235 /* If an incoming message is an 'ACK-', it should be
1236 * discarded here because it doesn't contain useful
1237 * data. In addition, we should try to wake up
1238 * connect() routine if sleeping.
1240 if (msg_data_sz(msg) == 0) {
1243 if (waitqueue_active(sk_sleep(sk)))
1244 wake_up_interruptible(sk_sleep(sk));
1249 case SS_UNCONNECTED:
1250 /* Accept only SYN message */
1251 if (!msg_connected(msg) && !(msg_errcode(msg)))
1254 case SS_DISCONNECTING:
1257 pr_err("Unknown socket state %u\n", sock->state);
1263 * filter_rcv - validate incoming message
1267 * Enqueues message on receive queue if acceptable; optionally handles
1268 * disconnect indication for a connected socket.
1270 * Called with socket lock already taken; port lock may also be taken.
1272 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1274 static u32 filter_rcv(struct sock *sk, struct sk_buff *buf)
1276 struct socket *sock = sk->sk_socket;
1277 struct tipc_msg *msg = buf_msg(buf);
1281 /* Reject message if it is wrong sort of message for socket */
1282 if (msg_type(msg) > TIPC_DIRECT_MSG)
1283 return TIPC_ERR_NO_PORT;
1285 if (sock->state == SS_READY) {
1286 if (msg_connected(msg))
1287 return TIPC_ERR_NO_PORT;
1289 res = filter_connect(tipc_sk(sk), &buf);
1290 if (res != TIPC_OK || buf == NULL)
1294 /* Reject message if there isn't room to queue it */
1295 recv_q_len = skb_queue_len(&sk->sk_receive_queue);
1296 if (unlikely(recv_q_len >= (OVERLOAD_LIMIT_BASE / 2))) {
1297 if (rx_queue_full(msg, recv_q_len, OVERLOAD_LIMIT_BASE / 2))
1298 return TIPC_ERR_OVERLOAD;
1301 /* Enqueue message (finally!) */
1302 TIPC_SKB_CB(buf)->handle = 0;
1303 __skb_queue_tail(&sk->sk_receive_queue, buf);
1305 sk->sk_data_ready(sk, 0);
1310 * backlog_rcv - handle incoming message from backlog queue
1314 * Caller must hold socket lock, but not port lock.
1318 static int backlog_rcv(struct sock *sk, struct sk_buff *buf)
1322 res = filter_rcv(sk, buf);
1324 tipc_reject_msg(buf, res);
1329 * dispatch - handle incoming message
1330 * @tport: TIPC port that received message
1333 * Called with port lock already taken.
1335 * Returns TIPC error status code (TIPC_OK if message is not to be rejected)
1337 static u32 dispatch(struct tipc_port *tport, struct sk_buff *buf)
1339 struct sock *sk = (struct sock *)tport->usr_handle;
1343 * Process message if socket is unlocked; otherwise add to backlog queue
1345 * This code is based on sk_receive_skb(), but must be distinct from it
1346 * since a TIPC-specific filter/reject mechanism is utilized
1349 if (!sock_owned_by_user(sk)) {
1350 res = filter_rcv(sk, buf);
1352 if (sk_add_backlog(sk, buf, sk->sk_rcvbuf))
1353 res = TIPC_ERR_OVERLOAD;
1363 * wakeupdispatch - wake up port after congestion
1364 * @tport: port to wakeup
1366 * Called with port lock already taken.
1368 static void wakeupdispatch(struct tipc_port *tport)
1370 struct sock *sk = (struct sock *)tport->usr_handle;
1372 sk->sk_write_space(sk);
1376 * connect - establish a connection to another TIPC port
1377 * @sock: socket structure
1378 * @dest: socket address for destination port
1379 * @destlen: size of socket address data structure
1380 * @flags: file-related flags associated with socket
1382 * Returns 0 on success, errno otherwise
1384 static int connect(struct socket *sock, struct sockaddr *dest, int destlen,
1387 struct sock *sk = sock->sk;
1388 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1389 struct msghdr m = {NULL,};
1390 unsigned int timeout;
1395 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1396 if (sock->state == SS_READY) {
1402 * Reject connection attempt using multicast address
1404 * Note: send_msg() validates the rest of the address fields,
1405 * so there's no need to do it here
1407 if (dst->addrtype == TIPC_ADDR_MCAST) {
1412 timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1414 switch (sock->state) {
1415 case SS_UNCONNECTED:
1416 /* Send a 'SYN-' to destination */
1418 m.msg_namelen = destlen;
1420 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1421 * indicate send_msg() is never blocked.
1424 m.msg_flags = MSG_DONTWAIT;
1426 res = send_msg(NULL, sock, &m, 0);
1427 if ((res < 0) && (res != -EWOULDBLOCK))
1430 /* Just entered SS_CONNECTING state; the only
1431 * difference is that return value in non-blocking
1432 * case is EINPROGRESS, rather than EALREADY.
1447 if (sock->state == SS_CONNECTING) {
1451 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1453 res = wait_event_interruptible_timeout(*sk_sleep(sk),
1454 sock->state != SS_CONNECTING,
1455 timeout ? (long)msecs_to_jiffies(timeout)
1456 : MAX_SCHEDULE_TIMEOUT);
1462 ; /* leave "res" unchanged */
1467 if (unlikely(sock->state == SS_DISCONNECTING))
1468 res = sock_error(sk);
1478 * listen - allow socket to listen for incoming connections
1479 * @sock: socket structure
1482 * Returns 0 on success, errno otherwise
1484 static int listen(struct socket *sock, int len)
1486 struct sock *sk = sock->sk;
1491 if (sock->state != SS_UNCONNECTED)
1494 sock->state = SS_LISTENING;
1503 * accept - wait for connection request
1504 * @sock: listening socket
1505 * @newsock: new socket that is to be connected
1506 * @flags: file-related flags associated with socket
1508 * Returns 0 on success, errno otherwise
1510 static int accept(struct socket *sock, struct socket *new_sock, int flags)
1512 struct sock *new_sk, *sk = sock->sk;
1513 struct sk_buff *buf;
1514 struct tipc_sock *new_tsock;
1515 struct tipc_port *new_tport;
1516 struct tipc_msg *msg;
1523 if (sock->state != SS_LISTENING) {
1528 while (skb_queue_empty(&sk->sk_receive_queue)) {
1529 if (flags & O_NONBLOCK) {
1534 res = wait_event_interruptible(*sk_sleep(sk),
1535 (!skb_queue_empty(&sk->sk_receive_queue)));
1541 buf = skb_peek(&sk->sk_receive_queue);
1543 res = tipc_create(sock_net(sock->sk), new_sock, 0, 0);
1547 new_sk = new_sock->sk;
1548 new_tsock = tipc_sk(new_sk);
1549 new_tport = new_tsock->p;
1550 new_ref = new_tport->ref;
1553 /* we lock on new_sk; but lockdep sees the lock on sk */
1554 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1557 * Reject any stray messages received by new socket
1558 * before the socket lock was taken (very, very unlikely)
1560 reject_rx_queue(new_sk);
1562 /* Connect new socket to it's peer */
1563 new_tsock->peer_name.ref = msg_origport(msg);
1564 new_tsock->peer_name.node = msg_orignode(msg);
1565 tipc_connect(new_ref, &new_tsock->peer_name);
1566 new_sock->state = SS_CONNECTED;
1568 tipc_set_portimportance(new_ref, msg_importance(msg));
1569 if (msg_named(msg)) {
1570 new_tport->conn_type = msg_nametype(msg);
1571 new_tport->conn_instance = msg_nameinst(msg);
1575 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
1576 * Respond to 'SYN+' by queuing it on new socket.
1578 if (!msg_data_sz(msg)) {
1579 struct msghdr m = {NULL,};
1581 advance_rx_queue(sk);
1582 send_packet(NULL, new_sock, &m, 0);
1584 __skb_dequeue(&sk->sk_receive_queue);
1585 __skb_queue_head(&new_sk->sk_receive_queue, buf);
1587 release_sock(new_sk);
1595 * shutdown - shutdown socket connection
1596 * @sock: socket structure
1597 * @how: direction to close (must be SHUT_RDWR)
1599 * Terminates connection (if necessary), then purges socket's receive queue.
1601 * 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) {
1620 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
1621 buf = __skb_dequeue(&sk->sk_receive_queue);
1623 if (TIPC_SKB_CB(buf)->handle != 0) {
1627 tipc_disconnect(tport->ref);
1628 tipc_reject_msg(buf, TIPC_CONN_SHUTDOWN);
1630 tipc_shutdown(tport->ref);
1633 sock->state = SS_DISCONNECTING;
1637 case SS_DISCONNECTING:
1639 /* Discard any unreceived messages */
1640 discard_rx_queue(sk);
1642 /* Wake up anyone sleeping in poll */
1643 sk->sk_state_change(sk);
1656 * setsockopt - set socket option
1657 * @sock: socket structure
1658 * @lvl: option level
1659 * @opt: option identifier
1660 * @ov: pointer to new option value
1661 * @ol: length of option value
1663 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
1664 * (to ease compatibility).
1666 * Returns 0 on success, errno otherwise
1668 static int setsockopt(struct socket *sock,
1669 int lvl, int opt, char __user *ov, unsigned int ol)
1671 struct sock *sk = sock->sk;
1672 struct tipc_port *tport = tipc_sk_port(sk);
1676 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1678 if (lvl != SOL_TIPC)
1679 return -ENOPROTOOPT;
1680 if (ol < sizeof(value))
1682 res = get_user(value, (u32 __user *)ov);
1689 case TIPC_IMPORTANCE:
1690 res = tipc_set_portimportance(tport->ref, value);
1692 case TIPC_SRC_DROPPABLE:
1693 if (sock->type != SOCK_STREAM)
1694 res = tipc_set_portunreliable(tport->ref, value);
1698 case TIPC_DEST_DROPPABLE:
1699 res = tipc_set_portunreturnable(tport->ref, value);
1701 case TIPC_CONN_TIMEOUT:
1702 tipc_sk(sk)->conn_timeout = value;
1703 /* no need to set "res", since already 0 at this point */
1715 * getsockopt - get socket option
1716 * @sock: socket structure
1717 * @lvl: option level
1718 * @opt: option identifier
1719 * @ov: receptacle for option value
1720 * @ol: receptacle for length of option value
1722 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
1723 * (to ease compatibility).
1725 * Returns 0 on success, errno otherwise
1727 static int getsockopt(struct socket *sock,
1728 int lvl, int opt, char __user *ov, int __user *ol)
1730 struct sock *sk = sock->sk;
1731 struct tipc_port *tport = tipc_sk_port(sk);
1736 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
1737 return put_user(0, ol);
1738 if (lvl != SOL_TIPC)
1739 return -ENOPROTOOPT;
1740 res = get_user(len, ol);
1747 case TIPC_IMPORTANCE:
1748 res = tipc_portimportance(tport->ref, &value);
1750 case TIPC_SRC_DROPPABLE:
1751 res = tipc_portunreliable(tport->ref, &value);
1753 case TIPC_DEST_DROPPABLE:
1754 res = tipc_portunreturnable(tport->ref, &value);
1756 case TIPC_CONN_TIMEOUT:
1757 value = tipc_sk(sk)->conn_timeout;
1758 /* no need to set "res", since already 0 at this point */
1760 case TIPC_NODE_RECVQ_DEPTH:
1761 value = 0; /* was tipc_queue_size, now obsolete */
1763 case TIPC_SOCK_RECVQ_DEPTH:
1764 value = skb_queue_len(&sk->sk_receive_queue);
1773 return res; /* "get" failed */
1775 if (len < sizeof(value))
1778 if (copy_to_user(ov, &value, sizeof(value)))
1781 return put_user(sizeof(value), ol);
1784 /* 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,
1793 .accept = sock_no_accept,
1794 .getname = get_name,
1796 .ioctl = sock_no_ioctl,
1797 .listen = sock_no_listen,
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 pr_err("Failed to register TIPC protocol type\n");
1876 res = sock_register(&tipc_family_ops);
1878 pr_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
1891 void tipc_socket_stop(void)
1893 if (!sockets_enabled)
1896 sockets_enabled = 0;
1897 sock_unregister(tipc_family_ops.family);
1898 proto_unregister(&tipc_proto);