2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2014, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, 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/rhashtable.h>
38 #include <linux/jhash.h>
40 #include "name_table.h"
46 #define SS_LISTENING -1 /* socket is listening */
47 #define SS_READY -2 /* socket is connectionless */
49 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
50 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
51 #define TIPC_FWD_MSG 1
52 #define TIPC_CONN_OK 0
53 #define TIPC_CONN_PROBING 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
58 * struct tipc_sock - TIPC socket structure
59 * @sk: socket - interacts with 'port' and with user via the socket API
60 * @connected: non-zero if port is currently connected to a peer port
61 * @conn_type: TIPC type used when connection was established
62 * @conn_instance: TIPC instance used when connection was established
63 * @published: non-zero if port has one or more associated names
64 * @max_pkt: maximum packet size "hint" used when building messages sent by port
65 * @portid: unique port identity in TIPC socket hash table
66 * @phdr: preformatted message header used when sending messages
67 * @port_list: adjacent ports in TIPC's global list of ports
68 * @publications: list of publications for port
69 * @pub_count: total # of publications port has made during its lifetime
72 * @port: port - interacts with 'sk' and with the rest of the TIPC stack
73 * @peer_name: the peer of the connection, if any
74 * @conn_timeout: the time we can wait for an unresponded setup request
75 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
76 * @link_cong: non-zero if owner must sleep because of link congestion
77 * @sent_unacked: # messages sent by socket, and not yet acked by peer
78 * @rcv_unacked: # messages read by user, but not yet acked back to peer
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
91 struct list_head sock_list;
92 struct list_head publications;
95 unsigned long probing_intv;
101 struct rhash_head node;
105 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
106 static void tipc_data_ready(struct sock *sk);
107 static void tipc_write_space(struct sock *sk);
108 static int tipc_release(struct socket *sock);
109 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
110 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
111 static void tipc_sk_timeout(unsigned long data);
112 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
113 struct tipc_name_seq const *seq);
114 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
115 struct tipc_name_seq const *seq);
116 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
117 static int tipc_sk_insert(struct tipc_sock *tsk);
118 static void tipc_sk_remove(struct tipc_sock *tsk);
120 static const struct proto_ops packet_ops;
121 static const struct proto_ops stream_ops;
122 static const struct proto_ops msg_ops;
124 static struct proto tipc_proto;
125 static struct proto tipc_proto_kern;
127 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
128 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
129 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
130 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
131 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
132 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
136 * Revised TIPC socket locking policy:
138 * Most socket operations take the standard socket lock when they start
139 * and hold it until they finish (or until they need to sleep). Acquiring
140 * this lock grants the owner exclusive access to the fields of the socket
141 * data structures, with the exception of the backlog queue. A few socket
142 * operations can be done without taking the socket lock because they only
143 * read socket information that never changes during the life of the socket.
145 * Socket operations may acquire the lock for the associated TIPC port if they
146 * need to perform an operation on the port. If any routine needs to acquire
147 * both the socket lock and the port lock it must take the socket lock first
148 * to avoid the risk of deadlock.
150 * The dispatcher handling incoming messages cannot grab the socket lock in
151 * the standard fashion, since invoked it runs at the BH level and cannot block.
152 * Instead, it checks to see if the socket lock is currently owned by someone,
153 * and either handles the message itself or adds it to the socket's backlog
154 * queue; in the latter case the queued message is processed once the process
155 * owning the socket lock releases it.
157 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
158 * the problem of a blocked socket operation preventing any other operations
159 * from occurring. However, applications must be careful if they have
160 * multiple threads trying to send (or receive) on the same socket, as these
161 * operations might interfere with each other. For example, doing a connect
162 * and a receive at the same time might allow the receive to consume the
163 * ACK message meant for the connect. While additional work could be done
164 * to try and overcome this, it doesn't seem to be worthwhile at the present.
166 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
167 * that another operation that must be performed in a non-blocking manner is
168 * not delayed for very long because the lock has already been taken.
170 * NOTE: This code assumes that certain fields of a port/socket pair are
171 * constant over its lifetime; such fields can be examined without taking
172 * the socket lock and/or port lock, and do not need to be re-read even
173 * after resuming processing after waiting. These fields include:
175 * - pointer to socket sk structure (aka tipc_sock structure)
176 * - pointer to port structure
180 static u32 tsk_peer_node(struct tipc_sock *tsk)
182 return msg_destnode(&tsk->phdr);
185 static u32 tsk_peer_port(struct tipc_sock *tsk)
187 return msg_destport(&tsk->phdr);
190 static bool tsk_unreliable(struct tipc_sock *tsk)
192 return msg_src_droppable(&tsk->phdr) != 0;
195 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
197 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
200 static bool tsk_unreturnable(struct tipc_sock *tsk)
202 return msg_dest_droppable(&tsk->phdr) != 0;
205 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
207 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
210 static int tsk_importance(struct tipc_sock *tsk)
212 return msg_importance(&tsk->phdr);
215 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
217 if (imp > TIPC_CRITICAL_IMPORTANCE)
219 msg_set_importance(&tsk->phdr, (u32)imp);
223 static struct tipc_sock *tipc_sk(const struct sock *sk)
225 return container_of(sk, struct tipc_sock, sk);
228 static int tsk_conn_cong(struct tipc_sock *tsk)
230 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
234 * tsk_advance_rx_queue - discard first buffer in socket receive queue
236 * Caller must hold socket lock
238 static void tsk_advance_rx_queue(struct sock *sk)
240 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
244 * tsk_rej_rx_queue - reject all buffers in socket receive queue
246 * Caller must hold socket lock
248 static void tsk_rej_rx_queue(struct sock *sk)
252 struct net *net = sock_net(sk);
254 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
255 if (tipc_msg_reverse(net, skb, &dnode, TIPC_ERR_NO_PORT))
256 tipc_link_xmit_skb(net, skb, dnode, 0);
260 /* tsk_peer_msg - verify if message was sent by connected port's peer
262 * Handles cases where the node's network address has changed from
263 * the default of <0.0.0> to its configured setting.
265 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
267 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
268 u32 peer_port = tsk_peer_port(tsk);
272 if (unlikely(!tsk->connected))
275 if (unlikely(msg_origport(msg) != peer_port))
278 orig_node = msg_orignode(msg);
279 peer_node = tsk_peer_node(tsk);
281 if (likely(orig_node == peer_node))
284 if (!orig_node && (peer_node == tn->own_addr))
287 if (!peer_node && (orig_node == tn->own_addr))
294 * tipc_sk_create - create a TIPC socket
295 * @net: network namespace (must be default network)
296 * @sock: pre-allocated socket structure
297 * @protocol: protocol indicator (must be 0)
298 * @kern: caused by kernel or by userspace?
300 * This routine creates additional data structures used by the TIPC socket,
301 * initializes them, and links them together.
303 * Returns 0 on success, errno otherwise
305 static int tipc_sk_create(struct net *net, struct socket *sock,
306 int protocol, int kern)
308 const struct proto_ops *ops;
311 struct tipc_sock *tsk;
312 struct tipc_msg *msg;
314 /* Validate arguments */
315 if (unlikely(protocol != 0))
316 return -EPROTONOSUPPORT;
318 switch (sock->type) {
321 state = SS_UNCONNECTED;
325 state = SS_UNCONNECTED;
336 /* Allocate socket's protocol area */
338 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
340 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
346 tsk->max_pkt = MAX_PKT_DEFAULT;
347 INIT_LIST_HEAD(&tsk->publications);
349 tipc_msg_init(net, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
352 /* Finish initializing socket data structures */
355 sock_init_data(sock, sk);
356 if (tipc_sk_insert(tsk)) {
357 pr_warn("Socket create failed; port numbrer exhausted\n");
360 msg_set_origport(msg, tsk->portid);
361 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
362 sk->sk_backlog_rcv = tipc_backlog_rcv;
363 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
364 sk->sk_data_ready = tipc_data_ready;
365 sk->sk_write_space = tipc_write_space;
366 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
367 tsk->sent_unacked = 0;
368 atomic_set(&tsk->dupl_rcvcnt, 0);
370 if (sock->state == SS_READY) {
371 tsk_set_unreturnable(tsk, true);
372 if (sock->type == SOCK_DGRAM)
373 tsk_set_unreliable(tsk, true);
379 * tipc_sock_create_local - create TIPC socket from inside TIPC module
380 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
382 * We cannot use sock_creat_kern here because it bumps module user count.
383 * Since socket owner and creator is the same module we must make sure
384 * that module count remains zero for module local sockets, otherwise
385 * we cannot do rmmod.
387 * Returns 0 on success, errno otherwise
389 int tipc_sock_create_local(struct net *net, int type, struct socket **res)
393 rc = sock_create_lite(AF_TIPC, type, 0, res);
395 pr_err("Failed to create kernel socket\n");
398 tipc_sk_create(net, *res, 0, 1);
404 * tipc_sock_release_local - release socket created by tipc_sock_create_local
405 * @sock: the socket to be released.
407 * Module reference count is not incremented when such sockets are created,
408 * so we must keep it from being decremented when they are released.
410 void tipc_sock_release_local(struct socket *sock)
418 * tipc_sock_accept_local - accept a connection on a socket created
419 * with tipc_sock_create_local. Use this function to avoid that
420 * module reference count is inadvertently incremented.
422 * @sock: the accepting socket
423 * @newsock: reference to the new socket to be created
424 * @flags: socket flags
427 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
430 struct sock *sk = sock->sk;
433 ret = sock_create_lite(sk->sk_family, sk->sk_type,
434 sk->sk_protocol, newsock);
438 ret = tipc_accept(sock, *newsock, flags);
440 sock_release(*newsock);
443 (*newsock)->ops = sock->ops;
447 static void tipc_sk_callback(struct rcu_head *head)
449 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
455 * tipc_release - destroy a TIPC socket
456 * @sock: socket to destroy
458 * This routine cleans up any messages that are still queued on the socket.
459 * For DGRAM and RDM socket types, all queued messages are rejected.
460 * For SEQPACKET and STREAM socket types, the first message is rejected
461 * and any others are discarded. (If the first message on a STREAM socket
462 * is partially-read, it is discarded and the next one is rejected instead.)
464 * NOTE: Rejected messages are not necessarily returned to the sender! They
465 * are returned or discarded according to the "destination droppable" setting
466 * specified for the message by the sender.
468 * Returns 0 on success, errno otherwise
470 static int tipc_release(struct socket *sock)
472 struct sock *sk = sock->sk;
475 struct tipc_sock *tsk;
477 u32 dnode, probing_state;
480 * Exit if socket isn't fully initialized (occurs when a failed accept()
481 * releases a pre-allocated child socket that was never used)
487 tn = net_generic(net, tipc_net_id);
493 * Reject all unreceived messages, except on an active connection
494 * (which disconnects locally & sends a 'FIN+' to peer)
496 dnode = tsk_peer_node(tsk);
497 while (sock->state != SS_DISCONNECTING) {
498 skb = __skb_dequeue(&sk->sk_receive_queue);
501 if (TIPC_SKB_CB(skb)->handle != NULL)
504 if ((sock->state == SS_CONNECTING) ||
505 (sock->state == SS_CONNECTED)) {
506 sock->state = SS_DISCONNECTING;
508 tipc_node_remove_conn(net, dnode, tsk->portid);
510 if (tipc_msg_reverse(net, skb, &dnode,
512 tipc_link_xmit_skb(net, skb, dnode, 0);
516 tipc_sk_withdraw(tsk, 0, NULL);
517 probing_state = tsk->probing_state;
518 if (del_timer_sync(&sk->sk_timer) &&
519 probing_state != TIPC_CONN_PROBING)
522 if (tsk->connected) {
523 skb = tipc_msg_create(net, TIPC_CRITICAL_IMPORTANCE,
524 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
525 tn->own_addr, tsk_peer_port(tsk),
526 tsk->portid, TIPC_ERR_NO_PORT);
528 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
529 tipc_node_remove_conn(net, dnode, tsk->portid);
532 /* Discard any remaining (connection-based) messages in receive queue */
533 __skb_queue_purge(&sk->sk_receive_queue);
535 /* Reject any messages that accumulated in backlog queue */
536 sock->state = SS_DISCONNECTING;
539 call_rcu(&tsk->rcu, tipc_sk_callback);
546 * tipc_bind - associate or disassocate TIPC name(s) with a socket
547 * @sock: socket structure
548 * @uaddr: socket address describing name(s) and desired operation
549 * @uaddr_len: size of socket address data structure
551 * Name and name sequence binding is indicated using a positive scope value;
552 * a negative scope value unbinds the specified name. Specifying no name
553 * (i.e. a socket address length of 0) unbinds all names from the socket.
555 * Returns 0 on success, errno otherwise
557 * NOTE: This routine doesn't need to take the socket lock since it doesn't
558 * access any non-constant socket information.
560 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
563 struct sock *sk = sock->sk;
564 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
565 struct tipc_sock *tsk = tipc_sk(sk);
569 if (unlikely(!uaddr_len)) {
570 res = tipc_sk_withdraw(tsk, 0, NULL);
574 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
578 if (addr->family != AF_TIPC) {
583 if (addr->addrtype == TIPC_ADDR_NAME)
584 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
585 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
590 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
591 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
592 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
597 res = (addr->scope > 0) ?
598 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
599 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
606 * tipc_getname - get port ID of socket or peer socket
607 * @sock: socket structure
608 * @uaddr: area for returned socket address
609 * @uaddr_len: area for returned length of socket address
610 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
612 * Returns 0 on success, errno otherwise
614 * NOTE: This routine doesn't need to take the socket lock since it only
615 * accesses socket information that is unchanging (or which changes in
616 * a completely predictable manner).
618 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
619 int *uaddr_len, int peer)
621 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
622 struct tipc_sock *tsk = tipc_sk(sock->sk);
623 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
625 memset(addr, 0, sizeof(*addr));
627 if ((sock->state != SS_CONNECTED) &&
628 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
630 addr->addr.id.ref = tsk_peer_port(tsk);
631 addr->addr.id.node = tsk_peer_node(tsk);
633 addr->addr.id.ref = tsk->portid;
634 addr->addr.id.node = tn->own_addr;
637 *uaddr_len = sizeof(*addr);
638 addr->addrtype = TIPC_ADDR_ID;
639 addr->family = AF_TIPC;
641 addr->addr.name.domain = 0;
647 * tipc_poll - read and possibly block on pollmask
648 * @file: file structure associated with the socket
649 * @sock: socket for which to calculate the poll bits
652 * Returns pollmask value
655 * It appears that the usual socket locking mechanisms are not useful here
656 * since the pollmask info is potentially out-of-date the moment this routine
657 * exits. TCP and other protocols seem to rely on higher level poll routines
658 * to handle any preventable race conditions, so TIPC will do the same ...
660 * TIPC sets the returned events as follows:
662 * socket state flags set
663 * ------------ ---------
664 * unconnected no read flags
665 * POLLOUT if port is not congested
667 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
670 * connected POLLIN/POLLRDNORM if data in rx queue
671 * POLLOUT if port is not congested
673 * disconnecting POLLIN/POLLRDNORM/POLLHUP
676 * listening POLLIN if SYN in rx queue
679 * ready POLLIN/POLLRDNORM if data in rx queue
680 * [connectionless] POLLOUT (since port cannot be congested)
682 * IMPORTANT: The fact that a read or write operation is indicated does NOT
683 * imply that the operation will succeed, merely that it should be performed
684 * and will not block.
686 static unsigned int tipc_poll(struct file *file, struct socket *sock,
689 struct sock *sk = sock->sk;
690 struct tipc_sock *tsk = tipc_sk(sk);
693 sock_poll_wait(file, sk_sleep(sk), wait);
695 switch ((int)sock->state) {
702 if (!tsk->link_cong && !tsk_conn_cong(tsk))
707 if (!skb_queue_empty(&sk->sk_receive_queue))
708 mask |= (POLLIN | POLLRDNORM);
710 case SS_DISCONNECTING:
711 mask = (POLLIN | POLLRDNORM | POLLHUP);
719 * tipc_sendmcast - send multicast message
720 * @sock: socket structure
721 * @seq: destination address
722 * @msg: message to send
723 * @dsz: total length of message data
724 * @timeo: timeout to wait for wakeup
726 * Called from function tipc_sendmsg(), which has done all sanity checks
727 * Returns the number of bytes sent on success, or errno
729 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
730 struct msghdr *msg, size_t dsz, long timeo)
732 struct sock *sk = sock->sk;
733 struct net *net = sock_net(sk);
734 struct tipc_msg *mhdr = &tipc_sk(sk)->phdr;
735 struct sk_buff_head head;
736 struct iov_iter save = msg->msg_iter;
740 msg_set_type(mhdr, TIPC_MCAST_MSG);
741 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
742 msg_set_destport(mhdr, 0);
743 msg_set_destnode(mhdr, 0);
744 msg_set_nametype(mhdr, seq->type);
745 msg_set_namelower(mhdr, seq->lower);
746 msg_set_nameupper(mhdr, seq->upper);
747 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
750 mtu = tipc_bclink_get_mtu();
751 __skb_queue_head_init(&head);
752 rc = tipc_msg_build(net, mhdr, msg, 0, dsz, mtu, &head);
753 if (unlikely(rc < 0))
757 rc = tipc_bclink_xmit(net, &head);
758 if (likely(rc >= 0)) {
762 if (rc == -EMSGSIZE) {
763 msg->msg_iter = save;
766 if (rc != -ELINKCONG)
768 tipc_sk(sk)->link_cong = 1;
769 rc = tipc_wait_for_sndmsg(sock, &timeo);
771 __skb_queue_purge(&head);
776 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
778 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff *buf)
780 struct tipc_msg *msg = buf_msg(buf);
781 struct tipc_port_list dports = {0, NULL, };
782 struct tipc_port_list *item;
784 uint i, last, dst = 0;
785 u32 scope = TIPC_CLUSTER_SCOPE;
787 if (in_own_node(net, msg_orignode(msg)))
788 scope = TIPC_NODE_SCOPE;
790 /* Create destination port list: */
791 tipc_nametbl_mc_translate(net, msg_nametype(msg), msg_namelower(msg),
792 msg_nameupper(msg), scope, &dports);
799 for (item = &dports; item; item = item->next) {
800 for (i = 0; i < PLSIZE && ++dst <= last; i++) {
801 b = (dst != last) ? skb_clone(buf, GFP_ATOMIC) : buf;
803 pr_warn("Failed do clone mcast rcv buffer\n");
806 msg_set_destport(msg, item->ports[i]);
810 tipc_port_list_free(&dports);
814 * tipc_sk_proto_rcv - receive a connection mng protocol message
815 * @tsk: receiving socket
816 * @dnode: node to send response message to, if any
817 * @buf: buffer containing protocol message
818 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
819 * (CONN_PROBE_REPLY) message should be forwarded.
821 static int tipc_sk_proto_rcv(struct tipc_sock *tsk, u32 *dnode,
824 struct tipc_msg *msg = buf_msg(buf);
827 /* Ignore if connection cannot be validated: */
828 if (!tsk_peer_msg(tsk, msg))
831 tsk->probing_state = TIPC_CONN_OK;
833 if (msg_type(msg) == CONN_ACK) {
834 conn_cong = tsk_conn_cong(tsk);
835 tsk->sent_unacked -= msg_msgcnt(msg);
837 tsk->sk.sk_write_space(&tsk->sk);
838 } else if (msg_type(msg) == CONN_PROBE) {
839 if (!tipc_msg_reverse(sock_net(&tsk->sk), buf, dnode, TIPC_OK))
841 msg_set_type(msg, CONN_PROBE_REPLY);
844 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
850 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
852 struct sock *sk = sock->sk;
853 struct tipc_sock *tsk = tipc_sk(sk);
858 int err = sock_error(sk);
861 if (sock->state == SS_DISCONNECTING)
865 if (signal_pending(current))
866 return sock_intr_errno(*timeo_p);
868 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
869 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
870 finish_wait(sk_sleep(sk), &wait);
876 * tipc_sendmsg - send message in connectionless manner
877 * @iocb: if NULL, indicates that socket lock is already held
878 * @sock: socket structure
879 * @m: message to send
880 * @dsz: amount of user data to be sent
882 * Message must have an destination specified explicitly.
883 * Used for SOCK_RDM and SOCK_DGRAM messages,
884 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
885 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
887 * Returns the number of bytes sent on success, or errno otherwise
889 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
890 struct msghdr *m, size_t dsz)
892 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
893 struct sock *sk = sock->sk;
894 struct tipc_sock *tsk = tipc_sk(sk);
895 struct net *net = sock_net(sk);
896 struct tipc_msg *mhdr = &tsk->phdr;
898 struct sk_buff_head head;
900 struct tipc_name_seq *seq = &dest->addr.nameseq;
901 struct iov_iter save;
907 return -EDESTADDRREQ;
909 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
910 (dest->family != AF_TIPC)))
913 if (dsz > TIPC_MAX_USER_MSG_SIZE)
919 if (unlikely(sock->state != SS_READY)) {
920 if (sock->state == SS_LISTENING) {
924 if (sock->state != SS_UNCONNECTED) {
928 if (tsk->published) {
932 if (dest->addrtype == TIPC_ADDR_NAME) {
933 tsk->conn_type = dest->addr.name.name.type;
934 tsk->conn_instance = dest->addr.name.name.instance;
938 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
940 if (dest->addrtype == TIPC_ADDR_MCAST) {
941 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
943 } else if (dest->addrtype == TIPC_ADDR_NAME) {
944 u32 type = dest->addr.name.name.type;
945 u32 inst = dest->addr.name.name.instance;
946 u32 domain = dest->addr.name.domain;
949 msg_set_type(mhdr, TIPC_NAMED_MSG);
950 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
951 msg_set_nametype(mhdr, type);
952 msg_set_nameinst(mhdr, inst);
953 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
954 dport = tipc_nametbl_translate(net, type, inst, &dnode);
955 msg_set_destnode(mhdr, dnode);
956 msg_set_destport(mhdr, dport);
957 if (unlikely(!dport && !dnode)) {
961 } else if (dest->addrtype == TIPC_ADDR_ID) {
962 dnode = dest->addr.id.node;
963 msg_set_type(mhdr, TIPC_DIRECT_MSG);
964 msg_set_lookup_scope(mhdr, 0);
965 msg_set_destnode(mhdr, dnode);
966 msg_set_destport(mhdr, dest->addr.id.ref);
967 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
972 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
973 __skb_queue_head_init(&head);
974 rc = tipc_msg_build(net, mhdr, m, 0, dsz, mtu, &head);
979 skb = skb_peek(&head);
980 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
981 rc = tipc_link_xmit(net, &head, dnode, tsk->portid);
982 if (likely(rc >= 0)) {
983 if (sock->state != SS_READY)
984 sock->state = SS_CONNECTING;
988 if (rc == -EMSGSIZE) {
992 if (rc != -ELINKCONG)
995 rc = tipc_wait_for_sndmsg(sock, &timeo);
997 __skb_queue_purge(&head);
1006 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
1008 struct sock *sk = sock->sk;
1009 struct tipc_sock *tsk = tipc_sk(sk);
1014 int err = sock_error(sk);
1017 if (sock->state == SS_DISCONNECTING)
1019 else if (sock->state != SS_CONNECTED)
1023 if (signal_pending(current))
1024 return sock_intr_errno(*timeo_p);
1026 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1027 done = sk_wait_event(sk, timeo_p,
1029 !tsk_conn_cong(tsk)) ||
1031 finish_wait(sk_sleep(sk), &wait);
1037 * tipc_send_stream - send stream-oriented data
1039 * @sock: socket structure
1041 * @dsz: total length of data to be transmitted
1043 * Used for SOCK_STREAM data.
1045 * Returns the number of bytes sent on success (or partial success),
1046 * or errno if no data sent
1048 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1049 struct msghdr *m, size_t dsz)
1051 struct sock *sk = sock->sk;
1052 struct net *net = sock_net(sk);
1053 struct tipc_sock *tsk = tipc_sk(sk);
1054 struct tipc_msg *mhdr = &tsk->phdr;
1055 struct sk_buff_head head;
1056 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1057 u32 portid = tsk->portid;
1061 uint mtu, send, sent = 0;
1062 struct iov_iter save;
1064 /* Handle implied connection establishment */
1065 if (unlikely(dest)) {
1066 rc = tipc_sendmsg(iocb, sock, m, dsz);
1067 if (dsz && (dsz == rc))
1068 tsk->sent_unacked = 1;
1071 if (dsz > (uint)INT_MAX)
1077 if (unlikely(sock->state != SS_CONNECTED)) {
1078 if (sock->state == SS_DISCONNECTING)
1085 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1086 dnode = tsk_peer_node(tsk);
1091 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1092 __skb_queue_head_init(&head);
1093 rc = tipc_msg_build(net, mhdr, m, sent, send, mtu, &head);
1094 if (unlikely(rc < 0))
1097 if (likely(!tsk_conn_cong(tsk))) {
1098 rc = tipc_link_xmit(net, &head, dnode, portid);
1100 tsk->sent_unacked++;
1106 if (rc == -EMSGSIZE) {
1107 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1112 if (rc != -ELINKCONG)
1116 rc = tipc_wait_for_sndpkt(sock, &timeo);
1118 __skb_queue_purge(&head);
1123 return sent ? sent : rc;
1127 * tipc_send_packet - send a connection-oriented message
1128 * @iocb: if NULL, indicates that socket lock is already held
1129 * @sock: socket structure
1130 * @m: message to send
1131 * @dsz: length of data to be transmitted
1133 * Used for SOCK_SEQPACKET messages.
1135 * Returns the number of bytes sent on success, or errno otherwise
1137 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1138 struct msghdr *m, size_t dsz)
1140 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1143 return tipc_send_stream(iocb, sock, m, dsz);
1146 /* tipc_sk_finish_conn - complete the setup of a connection
1148 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1151 struct sock *sk = &tsk->sk;
1152 struct net *net = sock_net(sk);
1153 struct tipc_msg *msg = &tsk->phdr;
1155 msg_set_destnode(msg, peer_node);
1156 msg_set_destport(msg, peer_port);
1157 msg_set_type(msg, TIPC_CONN_MSG);
1158 msg_set_lookup_scope(msg, 0);
1159 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1161 tsk->probing_intv = CONN_PROBING_INTERVAL;
1162 tsk->probing_state = TIPC_CONN_OK;
1164 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1165 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1166 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1170 * set_orig_addr - capture sender's address for received message
1171 * @m: descriptor for message info
1172 * @msg: received message header
1174 * Note: Address is not captured if not requested by receiver.
1176 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1178 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1181 addr->family = AF_TIPC;
1182 addr->addrtype = TIPC_ADDR_ID;
1183 memset(&addr->addr, 0, sizeof(addr->addr));
1184 addr->addr.id.ref = msg_origport(msg);
1185 addr->addr.id.node = msg_orignode(msg);
1186 addr->addr.name.domain = 0; /* could leave uninitialized */
1187 addr->scope = 0; /* could leave uninitialized */
1188 m->msg_namelen = sizeof(struct sockaddr_tipc);
1193 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1194 * @m: descriptor for message info
1195 * @msg: received message header
1196 * @tsk: TIPC port associated with message
1198 * Note: Ancillary data is not captured if not requested by receiver.
1200 * Returns 0 if successful, otherwise errno
1202 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1203 struct tipc_sock *tsk)
1211 if (likely(m->msg_controllen == 0))
1214 /* Optionally capture errored message object(s) */
1215 err = msg ? msg_errcode(msg) : 0;
1216 if (unlikely(err)) {
1218 anc_data[1] = msg_data_sz(msg);
1219 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1223 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1230 /* Optionally capture message destination object */
1231 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1232 switch (dest_type) {
1233 case TIPC_NAMED_MSG:
1235 anc_data[0] = msg_nametype(msg);
1236 anc_data[1] = msg_namelower(msg);
1237 anc_data[2] = msg_namelower(msg);
1239 case TIPC_MCAST_MSG:
1241 anc_data[0] = msg_nametype(msg);
1242 anc_data[1] = msg_namelower(msg);
1243 anc_data[2] = msg_nameupper(msg);
1246 has_name = (tsk->conn_type != 0);
1247 anc_data[0] = tsk->conn_type;
1248 anc_data[1] = tsk->conn_instance;
1249 anc_data[2] = tsk->conn_instance;
1255 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1263 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1265 struct net *net = sock_net(&tsk->sk);
1266 struct tipc_net *tn = net_generic(net, tipc_net_id);
1267 struct sk_buff *skb = NULL;
1268 struct tipc_msg *msg;
1269 u32 peer_port = tsk_peer_port(tsk);
1270 u32 dnode = tsk_peer_node(tsk);
1272 if (!tsk->connected)
1274 skb = tipc_msg_create(net, CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1275 dnode, tn->own_addr, peer_port, tsk->portid,
1280 msg_set_msgcnt(msg, ack);
1281 tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1284 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1286 struct sock *sk = sock->sk;
1288 long timeo = *timeop;
1292 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1293 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1294 if (sock->state == SS_DISCONNECTING) {
1299 timeo = schedule_timeout(timeo);
1303 if (!skb_queue_empty(&sk->sk_receive_queue))
1305 err = sock_intr_errno(timeo);
1306 if (signal_pending(current))
1312 finish_wait(sk_sleep(sk), &wait);
1318 * tipc_recvmsg - receive packet-oriented message
1320 * @m: descriptor for message info
1321 * @buf_len: total size of user buffer area
1322 * @flags: receive flags
1324 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1325 * If the complete message doesn't fit in user area, truncate it.
1327 * Returns size of returned message data, errno otherwise
1329 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1330 struct msghdr *m, size_t buf_len, int flags)
1332 struct sock *sk = sock->sk;
1333 struct tipc_sock *tsk = tipc_sk(sk);
1334 struct sk_buff *buf;
1335 struct tipc_msg *msg;
1341 /* Catch invalid receive requests */
1342 if (unlikely(!buf_len))
1347 if (unlikely(sock->state == SS_UNCONNECTED)) {
1352 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1355 /* Look for a message in receive queue; wait if necessary */
1356 res = tipc_wait_for_rcvmsg(sock, &timeo);
1360 /* Look at first message in receive queue */
1361 buf = skb_peek(&sk->sk_receive_queue);
1363 sz = msg_data_sz(msg);
1364 err = msg_errcode(msg);
1366 /* Discard an empty non-errored message & try again */
1367 if ((!sz) && (!err)) {
1368 tsk_advance_rx_queue(sk);
1372 /* Capture sender's address (optional) */
1373 set_orig_addr(m, msg);
1375 /* Capture ancillary data (optional) */
1376 res = tipc_sk_anc_data_recv(m, msg, tsk);
1380 /* Capture message data (if valid) & compute return value (always) */
1382 if (unlikely(buf_len < sz)) {
1384 m->msg_flags |= MSG_TRUNC;
1386 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1391 if ((sock->state == SS_READY) ||
1392 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1398 /* Consume received message (optional) */
1399 if (likely(!(flags & MSG_PEEK))) {
1400 if ((sock->state != SS_READY) &&
1401 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1402 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1403 tsk->rcv_unacked = 0;
1405 tsk_advance_rx_queue(sk);
1413 * tipc_recv_stream - receive stream-oriented data
1415 * @m: descriptor for message info
1416 * @buf_len: total size of user buffer area
1417 * @flags: receive flags
1419 * Used for SOCK_STREAM messages only. If not enough data is available
1420 * will optionally wait for more; never truncates data.
1422 * Returns size of returned message data, errno otherwise
1424 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1425 struct msghdr *m, size_t buf_len, int flags)
1427 struct sock *sk = sock->sk;
1428 struct tipc_sock *tsk = tipc_sk(sk);
1429 struct sk_buff *buf;
1430 struct tipc_msg *msg;
1433 int sz_to_copy, target, needed;
1438 /* Catch invalid receive attempts */
1439 if (unlikely(!buf_len))
1444 if (unlikely(sock->state == SS_UNCONNECTED)) {
1449 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1450 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1453 /* Look for a message in receive queue; wait if necessary */
1454 res = tipc_wait_for_rcvmsg(sock, &timeo);
1458 /* Look at first message in receive queue */
1459 buf = skb_peek(&sk->sk_receive_queue);
1461 sz = msg_data_sz(msg);
1462 err = msg_errcode(msg);
1464 /* Discard an empty non-errored message & try again */
1465 if ((!sz) && (!err)) {
1466 tsk_advance_rx_queue(sk);
1470 /* Optionally capture sender's address & ancillary data of first msg */
1471 if (sz_copied == 0) {
1472 set_orig_addr(m, msg);
1473 res = tipc_sk_anc_data_recv(m, msg, tsk);
1478 /* Capture message data (if valid) & compute return value (always) */
1480 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1483 needed = (buf_len - sz_copied);
1484 sz_to_copy = (sz <= needed) ? sz : needed;
1486 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1491 sz_copied += sz_to_copy;
1493 if (sz_to_copy < sz) {
1494 if (!(flags & MSG_PEEK))
1495 TIPC_SKB_CB(buf)->handle =
1496 (void *)(unsigned long)(offset + sz_to_copy);
1501 goto exit; /* can't add error msg to valid data */
1503 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1509 /* Consume received message (optional) */
1510 if (likely(!(flags & MSG_PEEK))) {
1511 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1512 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1513 tsk->rcv_unacked = 0;
1515 tsk_advance_rx_queue(sk);
1518 /* Loop around if more data is required */
1519 if ((sz_copied < buf_len) && /* didn't get all requested data */
1520 (!skb_queue_empty(&sk->sk_receive_queue) ||
1521 (sz_copied < target)) && /* and more is ready or required */
1522 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1523 (!err)) /* and haven't reached a FIN */
1528 return sz_copied ? sz_copied : res;
1532 * tipc_write_space - wake up thread if port congestion is released
1535 static void tipc_write_space(struct sock *sk)
1537 struct socket_wq *wq;
1540 wq = rcu_dereference(sk->sk_wq);
1541 if (wq_has_sleeper(wq))
1542 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1543 POLLWRNORM | POLLWRBAND);
1548 * tipc_data_ready - wake up threads to indicate messages have been received
1550 * @len: the length of messages
1552 static void tipc_data_ready(struct sock *sk)
1554 struct socket_wq *wq;
1557 wq = rcu_dereference(sk->sk_wq);
1558 if (wq_has_sleeper(wq))
1559 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1560 POLLRDNORM | POLLRDBAND);
1565 * filter_connect - Handle all incoming messages for a connection-based socket
1569 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1571 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1573 struct sock *sk = &tsk->sk;
1574 struct net *net = sock_net(sk);
1575 struct socket *sock = sk->sk_socket;
1576 struct tipc_msg *msg = buf_msg(*buf);
1577 int retval = -TIPC_ERR_NO_PORT;
1582 switch ((int)sock->state) {
1584 /* Accept only connection-based messages sent by peer */
1585 if (tsk_peer_msg(tsk, msg)) {
1586 if (unlikely(msg_errcode(msg))) {
1587 sock->state = SS_DISCONNECTING;
1589 /* let timer expire on it's own */
1590 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1597 /* Accept only ACK or NACK message */
1599 if (unlikely(!msg_connected(msg)))
1602 if (unlikely(msg_errcode(msg))) {
1603 sock->state = SS_DISCONNECTING;
1604 sk->sk_err = ECONNREFUSED;
1609 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1610 sock->state = SS_DISCONNECTING;
1611 sk->sk_err = EINVAL;
1616 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1617 msg_set_importance(&tsk->phdr, msg_importance(msg));
1618 sock->state = SS_CONNECTED;
1620 /* If an incoming message is an 'ACK-', it should be
1621 * discarded here because it doesn't contain useful
1622 * data. In addition, we should try to wake up
1623 * connect() routine if sleeping.
1625 if (msg_data_sz(msg) == 0) {
1628 if (waitqueue_active(sk_sleep(sk)))
1629 wake_up_interruptible(sk_sleep(sk));
1634 case SS_UNCONNECTED:
1635 /* Accept only SYN message */
1636 if (!msg_connected(msg) && !(msg_errcode(msg)))
1639 case SS_DISCONNECTING:
1642 pr_err("Unknown socket state %u\n", sock->state);
1648 * rcvbuf_limit - get proper overload limit of socket receive queue
1652 * For all connection oriented messages, irrespective of importance,
1653 * the default overload value (i.e. 67MB) is set as limit.
1655 * For all connectionless messages, by default new queue limits are
1658 * TIPC_LOW_IMPORTANCE (4 MB)
1659 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1660 * TIPC_HIGH_IMPORTANCE (16 MB)
1661 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1663 * Returns overload limit according to corresponding message importance
1665 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1667 struct tipc_msg *msg = buf_msg(buf);
1669 if (msg_connected(msg))
1670 return sysctl_tipc_rmem[2];
1672 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1673 msg_importance(msg);
1677 * filter_rcv - validate incoming message
1681 * Enqueues message on receive queue if acceptable; optionally handles
1682 * disconnect indication for a connected socket.
1684 * Called with socket lock already taken; port lock may also be taken.
1686 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1687 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1689 static int filter_rcv(struct sock *sk, struct sk_buff *buf)
1691 struct socket *sock = sk->sk_socket;
1692 struct tipc_sock *tsk = tipc_sk(sk);
1693 struct tipc_msg *msg = buf_msg(buf);
1694 unsigned int limit = rcvbuf_limit(sk, buf);
1698 if (unlikely(msg_user(msg) == CONN_MANAGER))
1699 return tipc_sk_proto_rcv(tsk, &onode, buf);
1701 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1704 sk->sk_write_space(sk);
1708 /* Reject message if it is wrong sort of message for socket */
1709 if (msg_type(msg) > TIPC_DIRECT_MSG)
1710 return -TIPC_ERR_NO_PORT;
1712 if (sock->state == SS_READY) {
1713 if (msg_connected(msg))
1714 return -TIPC_ERR_NO_PORT;
1716 rc = filter_connect(tsk, &buf);
1717 if (rc != TIPC_OK || buf == NULL)
1721 /* Reject message if there isn't room to queue it */
1722 if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1723 return -TIPC_ERR_OVERLOAD;
1725 /* Enqueue message */
1726 TIPC_SKB_CB(buf)->handle = NULL;
1727 __skb_queue_tail(&sk->sk_receive_queue, buf);
1728 skb_set_owner_r(buf, sk);
1730 sk->sk_data_ready(sk);
1735 * tipc_backlog_rcv - handle incoming message from backlog queue
1739 * Caller must hold socket lock, but not port lock.
1743 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1747 struct tipc_sock *tsk = tipc_sk(sk);
1748 struct net *net = sock_net(sk);
1749 uint truesize = skb->truesize;
1751 rc = filter_rcv(sk, skb);
1754 if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1755 atomic_add(truesize, &tsk->dupl_rcvcnt);
1759 if ((rc < 0) && !tipc_msg_reverse(net, skb, &onode, -rc))
1762 tipc_link_xmit_skb(net, skb, onode, 0);
1768 * tipc_sk_rcv - handle incoming message
1769 * @skb: buffer containing arriving message
1771 * Returns 0 if success, or errno: -EHOSTUNREACH
1773 int tipc_sk_rcv(struct net *net, struct sk_buff *skb)
1775 struct tipc_sock *tsk;
1777 u32 dport = msg_destport(buf_msg(skb));
1782 /* Validate destination and message */
1783 tsk = tipc_sk_lookup(net, dport);
1784 if (unlikely(!tsk)) {
1785 rc = tipc_msg_eval(net, skb, &dnode);
1791 spin_lock_bh(&sk->sk_lock.slock);
1793 if (!sock_owned_by_user(sk)) {
1794 rc = filter_rcv(sk, skb);
1796 if (sk->sk_backlog.len == 0)
1797 atomic_set(&tsk->dupl_rcvcnt, 0);
1798 limit = rcvbuf_limit(sk, skb) + atomic_read(&tsk->dupl_rcvcnt);
1799 if (sk_add_backlog(sk, skb, limit))
1800 rc = -TIPC_ERR_OVERLOAD;
1802 spin_unlock_bh(&sk->sk_lock.slock);
1807 if ((rc < 0) && !tipc_msg_reverse(net, skb, &dnode, -rc))
1808 return -EHOSTUNREACH;
1810 tipc_link_xmit_skb(net, skb, dnode, 0);
1811 return (rc < 0) ? -EHOSTUNREACH : 0;
1814 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1816 struct sock *sk = sock->sk;
1821 int err = sock_error(sk);
1826 if (signal_pending(current))
1827 return sock_intr_errno(*timeo_p);
1829 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1830 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1831 finish_wait(sk_sleep(sk), &wait);
1837 * tipc_connect - establish a connection to another TIPC port
1838 * @sock: socket structure
1839 * @dest: socket address for destination port
1840 * @destlen: size of socket address data structure
1841 * @flags: file-related flags associated with socket
1843 * Returns 0 on success, errno otherwise
1845 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1846 int destlen, int flags)
1848 struct sock *sk = sock->sk;
1849 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1850 struct msghdr m = {NULL,};
1851 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1852 socket_state previous;
1857 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1858 if (sock->state == SS_READY) {
1864 * Reject connection attempt using multicast address
1866 * Note: send_msg() validates the rest of the address fields,
1867 * so there's no need to do it here
1869 if (dst->addrtype == TIPC_ADDR_MCAST) {
1874 previous = sock->state;
1875 switch (sock->state) {
1876 case SS_UNCONNECTED:
1877 /* Send a 'SYN-' to destination */
1879 m.msg_namelen = destlen;
1881 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1882 * indicate send_msg() is never blocked.
1885 m.msg_flags = MSG_DONTWAIT;
1887 res = tipc_sendmsg(NULL, sock, &m, 0);
1888 if ((res < 0) && (res != -EWOULDBLOCK))
1891 /* Just entered SS_CONNECTING state; the only
1892 * difference is that return value in non-blocking
1893 * case is EINPROGRESS, rather than EALREADY.
1897 if (previous == SS_CONNECTING)
1901 timeout = msecs_to_jiffies(timeout);
1902 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1903 res = tipc_wait_for_connect(sock, &timeout);
1918 * tipc_listen - allow socket to listen for incoming connections
1919 * @sock: socket structure
1922 * Returns 0 on success, errno otherwise
1924 static int tipc_listen(struct socket *sock, int len)
1926 struct sock *sk = sock->sk;
1931 if (sock->state != SS_UNCONNECTED)
1934 sock->state = SS_LISTENING;
1942 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1944 struct sock *sk = sock->sk;
1948 /* True wake-one mechanism for incoming connections: only
1949 * one process gets woken up, not the 'whole herd'.
1950 * Since we do not 'race & poll' for established sockets
1951 * anymore, the common case will execute the loop only once.
1954 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1955 TASK_INTERRUPTIBLE);
1956 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1958 timeo = schedule_timeout(timeo);
1962 if (!skb_queue_empty(&sk->sk_receive_queue))
1965 if (sock->state != SS_LISTENING)
1967 err = sock_intr_errno(timeo);
1968 if (signal_pending(current))
1974 finish_wait(sk_sleep(sk), &wait);
1979 * tipc_accept - wait for connection request
1980 * @sock: listening socket
1981 * @newsock: new socket that is to be connected
1982 * @flags: file-related flags associated with socket
1984 * Returns 0 on success, errno otherwise
1986 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1988 struct sock *new_sk, *sk = sock->sk;
1989 struct sk_buff *buf;
1990 struct tipc_sock *new_tsock;
1991 struct tipc_msg *msg;
1997 if (sock->state != SS_LISTENING) {
2001 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2002 res = tipc_wait_for_accept(sock, timeo);
2006 buf = skb_peek(&sk->sk_receive_queue);
2008 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2012 new_sk = new_sock->sk;
2013 new_tsock = tipc_sk(new_sk);
2016 /* we lock on new_sk; but lockdep sees the lock on sk */
2017 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2020 * Reject any stray messages received by new socket
2021 * before the socket lock was taken (very, very unlikely)
2023 tsk_rej_rx_queue(new_sk);
2025 /* Connect new socket to it's peer */
2026 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2027 new_sock->state = SS_CONNECTED;
2029 tsk_set_importance(new_tsock, msg_importance(msg));
2030 if (msg_named(msg)) {
2031 new_tsock->conn_type = msg_nametype(msg);
2032 new_tsock->conn_instance = msg_nameinst(msg);
2036 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2037 * Respond to 'SYN+' by queuing it on new socket.
2039 if (!msg_data_sz(msg)) {
2040 struct msghdr m = {NULL,};
2042 tsk_advance_rx_queue(sk);
2043 tipc_send_packet(NULL, new_sock, &m, 0);
2045 __skb_dequeue(&sk->sk_receive_queue);
2046 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2047 skb_set_owner_r(buf, new_sk);
2049 release_sock(new_sk);
2056 * tipc_shutdown - shutdown socket connection
2057 * @sock: socket structure
2058 * @how: direction to close (must be SHUT_RDWR)
2060 * Terminates connection (if necessary), then purges socket's receive queue.
2062 * Returns 0 on success, errno otherwise
2064 static int tipc_shutdown(struct socket *sock, int how)
2066 struct sock *sk = sock->sk;
2067 struct net *net = sock_net(sk);
2068 struct tipc_net *tn = net_generic(net, tipc_net_id);
2069 struct tipc_sock *tsk = tipc_sk(sk);
2070 struct sk_buff *skb;
2074 if (how != SHUT_RDWR)
2079 switch (sock->state) {
2084 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2085 skb = __skb_dequeue(&sk->sk_receive_queue);
2087 if (TIPC_SKB_CB(skb)->handle != NULL) {
2091 if (tipc_msg_reverse(net, skb, &dnode,
2092 TIPC_CONN_SHUTDOWN))
2093 tipc_link_xmit_skb(net, skb, dnode,
2095 tipc_node_remove_conn(net, dnode, tsk->portid);
2097 dnode = tsk_peer_node(tsk);
2098 skb = tipc_msg_create(net, TIPC_CRITICAL_IMPORTANCE,
2099 TIPC_CONN_MSG, SHORT_H_SIZE,
2100 0, dnode, tn->own_addr,
2102 tsk->portid, TIPC_CONN_SHUTDOWN);
2103 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2106 sock->state = SS_DISCONNECTING;
2107 tipc_node_remove_conn(net, dnode, tsk->portid);
2110 case SS_DISCONNECTING:
2112 /* Discard any unreceived messages */
2113 __skb_queue_purge(&sk->sk_receive_queue);
2115 /* Wake up anyone sleeping in poll */
2116 sk->sk_state_change(sk);
2128 static void tipc_sk_timeout(unsigned long data)
2130 struct tipc_sock *tsk = (struct tipc_sock *)data;
2131 struct sock *sk = &tsk->sk;
2132 struct net *net = sock_net(sk);
2133 struct tipc_net *tn = net_generic(net, tipc_net_id);
2134 struct sk_buff *skb = NULL;
2135 u32 peer_port, peer_node;
2138 if (!tsk->connected) {
2142 peer_port = tsk_peer_port(tsk);
2143 peer_node = tsk_peer_node(tsk);
2145 if (tsk->probing_state == TIPC_CONN_PROBING) {
2146 /* Previous probe not answered -> self abort */
2147 skb = tipc_msg_create(net, TIPC_CRITICAL_IMPORTANCE,
2148 TIPC_CONN_MSG, SHORT_H_SIZE, 0,
2149 tn->own_addr, peer_node, tsk->portid,
2150 peer_port, TIPC_ERR_NO_PORT);
2152 skb = tipc_msg_create(net, CONN_MANAGER, CONN_PROBE, INT_H_SIZE,
2153 0, peer_node, tn->own_addr,
2154 peer_port, tsk->portid, TIPC_OK);
2155 tsk->probing_state = TIPC_CONN_PROBING;
2156 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2160 tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2165 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2166 struct tipc_name_seq const *seq)
2168 struct net *net = sock_net(&tsk->sk);
2169 struct publication *publ;
2174 key = tsk->portid + tsk->pub_count + 1;
2175 if (key == tsk->portid)
2178 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2179 scope, tsk->portid, key);
2180 if (unlikely(!publ))
2183 list_add(&publ->pport_list, &tsk->publications);
2189 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2190 struct tipc_name_seq const *seq)
2192 struct net *net = sock_net(&tsk->sk);
2193 struct publication *publ;
2194 struct publication *safe;
2197 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2199 if (publ->scope != scope)
2201 if (publ->type != seq->type)
2203 if (publ->lower != seq->lower)
2205 if (publ->upper != seq->upper)
2207 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2208 publ->ref, publ->key);
2212 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2213 publ->ref, publ->key);
2216 if (list_empty(&tsk->publications))
2221 static int tipc_sk_show(struct tipc_sock *tsk, char *buf,
2222 int len, int full_id)
2224 struct net *net = sock_net(&tsk->sk);
2225 struct tipc_net *tn = net_generic(net, tipc_net_id);
2226 struct publication *publ;
2230 ret = tipc_snprintf(buf, len, "<%u.%u.%u:%u>:",
2231 tipc_zone(tn->own_addr),
2232 tipc_cluster(tn->own_addr),
2233 tipc_node(tn->own_addr), tsk->portid);
2235 ret = tipc_snprintf(buf, len, "%-10u:", tsk->portid);
2237 if (tsk->connected) {
2238 u32 dport = tsk_peer_port(tsk);
2239 u32 destnode = tsk_peer_node(tsk);
2241 ret += tipc_snprintf(buf + ret, len - ret,
2242 " connected to <%u.%u.%u:%u>",
2243 tipc_zone(destnode),
2244 tipc_cluster(destnode),
2245 tipc_node(destnode), dport);
2246 if (tsk->conn_type != 0)
2247 ret += tipc_snprintf(buf + ret, len - ret,
2248 " via {%u,%u}", tsk->conn_type,
2249 tsk->conn_instance);
2250 } else if (tsk->published) {
2251 ret += tipc_snprintf(buf + ret, len - ret, " bound to");
2252 list_for_each_entry(publ, &tsk->publications, pport_list) {
2253 if (publ->lower == publ->upper)
2254 ret += tipc_snprintf(buf + ret, len - ret,
2255 " {%u,%u}", publ->type,
2258 ret += tipc_snprintf(buf + ret, len - ret,
2259 " {%u,%u,%u}", publ->type,
2260 publ->lower, publ->upper);
2263 ret += tipc_snprintf(buf + ret, len - ret, "\n");
2267 struct sk_buff *tipc_sk_socks_show(struct net *net)
2269 struct tipc_net *tn = net_generic(net, tipc_net_id);
2270 const struct bucket_table *tbl;
2271 struct rhash_head *pos;
2272 struct sk_buff *buf;
2273 struct tlv_desc *rep_tlv;
2276 struct tipc_sock *tsk;
2280 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2283 rep_tlv = (struct tlv_desc *)buf->data;
2284 pb = TLV_DATA(rep_tlv);
2285 pb_len = ULTRA_STRING_MAX_LEN;
2288 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2289 for (i = 0; i < tbl->size; i++) {
2290 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2291 spin_lock_bh(&tsk->sk.sk_lock.slock);
2292 str_len += tipc_sk_show(tsk, pb + str_len,
2293 pb_len - str_len, 0);
2294 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2299 str_len += 1; /* for "\0" */
2300 skb_put(buf, TLV_SPACE(str_len));
2301 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2306 /* tipc_sk_reinit: set non-zero address in all existing sockets
2307 * when we go from standalone to network mode.
2309 void tipc_sk_reinit(struct net *net)
2311 struct tipc_net *tn = net_generic(net, tipc_net_id);
2312 const struct bucket_table *tbl;
2313 struct rhash_head *pos;
2314 struct tipc_sock *tsk;
2315 struct tipc_msg *msg;
2319 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2320 for (i = 0; i < tbl->size; i++) {
2321 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2322 spin_lock_bh(&tsk->sk.sk_lock.slock);
2324 msg_set_prevnode(msg, tn->own_addr);
2325 msg_set_orignode(msg, tn->own_addr);
2326 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2332 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2334 struct tipc_net *tn = net_generic(net, tipc_net_id);
2335 struct tipc_sock *tsk;
2338 tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2340 sock_hold(&tsk->sk);
2346 static int tipc_sk_insert(struct tipc_sock *tsk)
2348 struct sock *sk = &tsk->sk;
2349 struct net *net = sock_net(sk);
2350 struct tipc_net *tn = net_generic(net, tipc_net_id);
2351 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2352 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2354 while (remaining--) {
2356 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2357 portid = TIPC_MIN_PORT;
2358 tsk->portid = portid;
2359 sock_hold(&tsk->sk);
2360 if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2368 static void tipc_sk_remove(struct tipc_sock *tsk)
2370 struct sock *sk = &tsk->sk;
2371 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2373 if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2374 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2379 int tipc_sk_rht_init(struct net *net)
2381 struct tipc_net *tn = net_generic(net, tipc_net_id);
2382 struct rhashtable_params rht_params = {
2384 .head_offset = offsetof(struct tipc_sock, node),
2385 .key_offset = offsetof(struct tipc_sock, portid),
2386 .key_len = sizeof(u32), /* portid */
2388 .max_shift = 20, /* 1M */
2389 .min_shift = 8, /* 256 */
2390 .grow_decision = rht_grow_above_75,
2391 .shrink_decision = rht_shrink_below_30,
2394 return rhashtable_init(&tn->sk_rht, &rht_params);
2397 void tipc_sk_rht_destroy(struct net *net)
2399 struct tipc_net *tn = net_generic(net, tipc_net_id);
2401 /* Wait for socket readers to complete */
2404 rhashtable_destroy(&tn->sk_rht);
2408 * tipc_setsockopt - set socket option
2409 * @sock: socket structure
2410 * @lvl: option level
2411 * @opt: option identifier
2412 * @ov: pointer to new option value
2413 * @ol: length of option value
2415 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2416 * (to ease compatibility).
2418 * Returns 0 on success, errno otherwise
2420 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2421 char __user *ov, unsigned int ol)
2423 struct sock *sk = sock->sk;
2424 struct tipc_sock *tsk = tipc_sk(sk);
2428 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2430 if (lvl != SOL_TIPC)
2431 return -ENOPROTOOPT;
2432 if (ol < sizeof(value))
2434 res = get_user(value, (u32 __user *)ov);
2441 case TIPC_IMPORTANCE:
2442 res = tsk_set_importance(tsk, value);
2444 case TIPC_SRC_DROPPABLE:
2445 if (sock->type != SOCK_STREAM)
2446 tsk_set_unreliable(tsk, value);
2450 case TIPC_DEST_DROPPABLE:
2451 tsk_set_unreturnable(tsk, value);
2453 case TIPC_CONN_TIMEOUT:
2454 tipc_sk(sk)->conn_timeout = value;
2455 /* no need to set "res", since already 0 at this point */
2467 * tipc_getsockopt - get socket option
2468 * @sock: socket structure
2469 * @lvl: option level
2470 * @opt: option identifier
2471 * @ov: receptacle for option value
2472 * @ol: receptacle for length of option value
2474 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2475 * (to ease compatibility).
2477 * Returns 0 on success, errno otherwise
2479 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2480 char __user *ov, int __user *ol)
2482 struct sock *sk = sock->sk;
2483 struct tipc_sock *tsk = tipc_sk(sk);
2488 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2489 return put_user(0, ol);
2490 if (lvl != SOL_TIPC)
2491 return -ENOPROTOOPT;
2492 res = get_user(len, ol);
2499 case TIPC_IMPORTANCE:
2500 value = tsk_importance(tsk);
2502 case TIPC_SRC_DROPPABLE:
2503 value = tsk_unreliable(tsk);
2505 case TIPC_DEST_DROPPABLE:
2506 value = tsk_unreturnable(tsk);
2508 case TIPC_CONN_TIMEOUT:
2509 value = tsk->conn_timeout;
2510 /* no need to set "res", since already 0 at this point */
2512 case TIPC_NODE_RECVQ_DEPTH:
2513 value = 0; /* was tipc_queue_size, now obsolete */
2515 case TIPC_SOCK_RECVQ_DEPTH:
2516 value = skb_queue_len(&sk->sk_receive_queue);
2525 return res; /* "get" failed */
2527 if (len < sizeof(value))
2530 if (copy_to_user(ov, &value, sizeof(value)))
2533 return put_user(sizeof(value), ol);
2536 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2538 struct sock *sk = sock->sk;
2539 struct tipc_sioc_ln_req lnr;
2540 void __user *argp = (void __user *)arg;
2543 case SIOCGETLINKNAME:
2544 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2546 if (!tipc_node_get_linkname(sock_net(sk),
2547 lnr.bearer_id & 0xffff, lnr.peer,
2548 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2549 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2553 return -EADDRNOTAVAIL;
2555 return -ENOIOCTLCMD;
2559 /* Protocol switches for the various types of TIPC sockets */
2561 static const struct proto_ops msg_ops = {
2562 .owner = THIS_MODULE,
2564 .release = tipc_release,
2566 .connect = tipc_connect,
2567 .socketpair = sock_no_socketpair,
2568 .accept = sock_no_accept,
2569 .getname = tipc_getname,
2571 .ioctl = tipc_ioctl,
2572 .listen = sock_no_listen,
2573 .shutdown = tipc_shutdown,
2574 .setsockopt = tipc_setsockopt,
2575 .getsockopt = tipc_getsockopt,
2576 .sendmsg = tipc_sendmsg,
2577 .recvmsg = tipc_recvmsg,
2578 .mmap = sock_no_mmap,
2579 .sendpage = sock_no_sendpage
2582 static const struct proto_ops packet_ops = {
2583 .owner = THIS_MODULE,
2585 .release = tipc_release,
2587 .connect = tipc_connect,
2588 .socketpair = sock_no_socketpair,
2589 .accept = tipc_accept,
2590 .getname = tipc_getname,
2592 .ioctl = tipc_ioctl,
2593 .listen = tipc_listen,
2594 .shutdown = tipc_shutdown,
2595 .setsockopt = tipc_setsockopt,
2596 .getsockopt = tipc_getsockopt,
2597 .sendmsg = tipc_send_packet,
2598 .recvmsg = tipc_recvmsg,
2599 .mmap = sock_no_mmap,
2600 .sendpage = sock_no_sendpage
2603 static const struct proto_ops stream_ops = {
2604 .owner = THIS_MODULE,
2606 .release = tipc_release,
2608 .connect = tipc_connect,
2609 .socketpair = sock_no_socketpair,
2610 .accept = tipc_accept,
2611 .getname = tipc_getname,
2613 .ioctl = tipc_ioctl,
2614 .listen = tipc_listen,
2615 .shutdown = tipc_shutdown,
2616 .setsockopt = tipc_setsockopt,
2617 .getsockopt = tipc_getsockopt,
2618 .sendmsg = tipc_send_stream,
2619 .recvmsg = tipc_recv_stream,
2620 .mmap = sock_no_mmap,
2621 .sendpage = sock_no_sendpage
2624 static const struct net_proto_family tipc_family_ops = {
2625 .owner = THIS_MODULE,
2627 .create = tipc_sk_create
2630 static struct proto tipc_proto = {
2632 .owner = THIS_MODULE,
2633 .obj_size = sizeof(struct tipc_sock),
2634 .sysctl_rmem = sysctl_tipc_rmem
2637 static struct proto tipc_proto_kern = {
2639 .obj_size = sizeof(struct tipc_sock),
2640 .sysctl_rmem = sysctl_tipc_rmem
2644 * tipc_socket_init - initialize TIPC socket interface
2646 * Returns 0 on success, errno otherwise
2648 int tipc_socket_init(void)
2652 res = proto_register(&tipc_proto, 1);
2654 pr_err("Failed to register TIPC protocol type\n");
2658 res = sock_register(&tipc_family_ops);
2660 pr_err("Failed to register TIPC socket type\n");
2661 proto_unregister(&tipc_proto);
2669 * tipc_socket_stop - stop TIPC socket interface
2671 void tipc_socket_stop(void)
2673 sock_unregister(tipc_family_ops.family);
2674 proto_unregister(&tipc_proto);
2677 /* Caller should hold socket lock for the passed tipc socket. */
2678 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2682 struct nlattr *nest;
2684 peer_node = tsk_peer_node(tsk);
2685 peer_port = tsk_peer_port(tsk);
2687 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2689 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2691 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2694 if (tsk->conn_type != 0) {
2695 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2697 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2699 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2702 nla_nest_end(skb, nest);
2707 nla_nest_cancel(skb, nest);
2712 /* Caller should hold socket lock for the passed tipc socket. */
2713 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2714 struct tipc_sock *tsk)
2718 struct nlattr *attrs;
2719 struct net *net = sock_net(skb->sk);
2720 struct tipc_net *tn = net_generic(net, tipc_net_id);
2722 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2723 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2727 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2729 goto genlmsg_cancel;
2730 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2731 goto attr_msg_cancel;
2732 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2733 goto attr_msg_cancel;
2735 if (tsk->connected) {
2736 err = __tipc_nl_add_sk_con(skb, tsk);
2738 goto attr_msg_cancel;
2739 } else if (!list_empty(&tsk->publications)) {
2740 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2741 goto attr_msg_cancel;
2743 nla_nest_end(skb, attrs);
2744 genlmsg_end(skb, hdr);
2749 nla_nest_cancel(skb, attrs);
2751 genlmsg_cancel(skb, hdr);
2756 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2759 struct tipc_sock *tsk;
2760 const struct bucket_table *tbl;
2761 struct rhash_head *pos;
2762 struct net *net = sock_net(skb->sk);
2763 struct tipc_net *tn = net_generic(net, tipc_net_id);
2764 u32 tbl_id = cb->args[0];
2765 u32 prev_portid = cb->args[1];
2768 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2769 for (; tbl_id < tbl->size; tbl_id++) {
2770 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2771 spin_lock_bh(&tsk->sk.sk_lock.slock);
2772 if (prev_portid && prev_portid != tsk->portid) {
2773 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2777 err = __tipc_nl_add_sk(skb, cb, tsk);
2779 prev_portid = tsk->portid;
2780 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2784 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2789 cb->args[0] = tbl_id;
2790 cb->args[1] = prev_portid;
2795 /* Caller should hold socket lock for the passed tipc socket. */
2796 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2797 struct netlink_callback *cb,
2798 struct publication *publ)
2801 struct nlattr *attrs;
2803 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2804 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2808 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2810 goto genlmsg_cancel;
2812 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2813 goto attr_msg_cancel;
2814 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2815 goto attr_msg_cancel;
2816 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2817 goto attr_msg_cancel;
2818 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2819 goto attr_msg_cancel;
2821 nla_nest_end(skb, attrs);
2822 genlmsg_end(skb, hdr);
2827 nla_nest_cancel(skb, attrs);
2829 genlmsg_cancel(skb, hdr);
2834 /* Caller should hold socket lock for the passed tipc socket. */
2835 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2836 struct netlink_callback *cb,
2837 struct tipc_sock *tsk, u32 *last_publ)
2840 struct publication *p;
2843 list_for_each_entry(p, &tsk->publications, pport_list) {
2844 if (p->key == *last_publ)
2847 if (p->key != *last_publ) {
2848 /* We never set seq or call nl_dump_check_consistent()
2849 * this means that setting prev_seq here will cause the
2850 * consistence check to fail in the netlink callback
2851 * handler. Resulting in the last NLMSG_DONE message
2852 * having the NLM_F_DUMP_INTR flag set.
2859 p = list_first_entry(&tsk->publications, struct publication,
2863 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2864 err = __tipc_nl_add_sk_publ(skb, cb, p);
2866 *last_publ = p->key;
2875 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2878 u32 tsk_portid = cb->args[0];
2879 u32 last_publ = cb->args[1];
2880 u32 done = cb->args[2];
2881 struct net *net = sock_net(skb->sk);
2882 struct tipc_sock *tsk;
2885 struct nlattr **attrs;
2886 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2888 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2892 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2893 attrs[TIPC_NLA_SOCK],
2894 tipc_nl_sock_policy);
2898 if (!sock[TIPC_NLA_SOCK_REF])
2901 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2907 tsk = tipc_sk_lookup(net, tsk_portid);
2911 lock_sock(&tsk->sk);
2912 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2915 release_sock(&tsk->sk);
2918 cb->args[0] = tsk_portid;
2919 cb->args[1] = last_publ;