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
73 * @port: port - interacts with 'sk' and with the rest of the TIPC stack
74 * @peer_name: the peer of the connection, if any
75 * @conn_timeout: the time we can wait for an unresponded setup request
76 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
77 * @link_cong: non-zero if owner must sleep because of link congestion
78 * @sent_unacked: # messages sent by socket, and not yet acked by peer
79 * @rcv_unacked: # messages read by user, but not yet acked back to peer
80 * @node: hash table node
81 * @rcu: rcu struct for tipc_sock
92 struct list_head sock_list;
93 struct list_head publications;
96 unsigned long probing_intv;
97 struct timer_list timer;
103 struct rhash_head node;
107 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
108 static void tipc_data_ready(struct sock *sk);
109 static void tipc_write_space(struct sock *sk);
110 static int tipc_release(struct socket *sock);
111 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
112 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
113 static void tipc_sk_timeout(unsigned long portid);
114 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
115 struct tipc_name_seq const *seq);
116 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
117 struct tipc_name_seq const *seq);
118 static struct tipc_sock *tipc_sk_lookup(u32 portid);
119 static int tipc_sk_insert(struct tipc_sock *tsk);
120 static void tipc_sk_remove(struct tipc_sock *tsk);
122 static const struct proto_ops packet_ops;
123 static const struct proto_ops stream_ops;
124 static const struct proto_ops msg_ops;
126 static struct proto tipc_proto;
127 static struct proto tipc_proto_kern;
129 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
130 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
131 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
132 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
133 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
134 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
138 * Revised TIPC socket locking policy:
140 * Most socket operations take the standard socket lock when they start
141 * and hold it until they finish (or until they need to sleep). Acquiring
142 * this lock grants the owner exclusive access to the fields of the socket
143 * data structures, with the exception of the backlog queue. A few socket
144 * operations can be done without taking the socket lock because they only
145 * read socket information that never changes during the life of the socket.
147 * Socket operations may acquire the lock for the associated TIPC port if they
148 * need to perform an operation on the port. If any routine needs to acquire
149 * both the socket lock and the port lock it must take the socket lock first
150 * to avoid the risk of deadlock.
152 * The dispatcher handling incoming messages cannot grab the socket lock in
153 * the standard fashion, since invoked it runs at the BH level and cannot block.
154 * Instead, it checks to see if the socket lock is currently owned by someone,
155 * and either handles the message itself or adds it to the socket's backlog
156 * queue; in the latter case the queued message is processed once the process
157 * owning the socket lock releases it.
159 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
160 * the problem of a blocked socket operation preventing any other operations
161 * from occurring. However, applications must be careful if they have
162 * multiple threads trying to send (or receive) on the same socket, as these
163 * operations might interfere with each other. For example, doing a connect
164 * and a receive at the same time might allow the receive to consume the
165 * ACK message meant for the connect. While additional work could be done
166 * to try and overcome this, it doesn't seem to be worthwhile at the present.
168 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
169 * that another operation that must be performed in a non-blocking manner is
170 * not delayed for very long because the lock has already been taken.
172 * NOTE: This code assumes that certain fields of a port/socket pair are
173 * constant over its lifetime; such fields can be examined without taking
174 * the socket lock and/or port lock, and do not need to be re-read even
175 * after resuming processing after waiting. These fields include:
177 * - pointer to socket sk structure (aka tipc_sock structure)
178 * - pointer to port structure
182 /* Protects tipc socket hash table mutations */
183 static struct rhashtable tipc_sk_rht;
185 static u32 tsk_peer_node(struct tipc_sock *tsk)
187 return msg_destnode(&tsk->phdr);
190 static u32 tsk_peer_port(struct tipc_sock *tsk)
192 return msg_destport(&tsk->phdr);
195 static bool tsk_unreliable(struct tipc_sock *tsk)
197 return msg_src_droppable(&tsk->phdr) != 0;
200 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
202 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
205 static bool tsk_unreturnable(struct tipc_sock *tsk)
207 return msg_dest_droppable(&tsk->phdr) != 0;
210 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
212 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
215 static int tsk_importance(struct tipc_sock *tsk)
217 return msg_importance(&tsk->phdr);
220 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
222 if (imp > TIPC_CRITICAL_IMPORTANCE)
224 msg_set_importance(&tsk->phdr, (u32)imp);
228 static struct tipc_sock *tipc_sk(const struct sock *sk)
230 return container_of(sk, struct tipc_sock, sk);
233 static int tsk_conn_cong(struct tipc_sock *tsk)
235 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
239 * tsk_advance_rx_queue - discard first buffer in socket receive queue
241 * Caller must hold socket lock
243 static void tsk_advance_rx_queue(struct sock *sk)
245 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
249 * tsk_rej_rx_queue - reject all buffers in socket receive queue
251 * Caller must hold socket lock
253 static void tsk_rej_rx_queue(struct sock *sk)
258 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
259 if (tipc_msg_reverse(skb, &dnode, TIPC_ERR_NO_PORT))
260 tipc_link_xmit_skb(skb, dnode, 0);
264 /* tsk_peer_msg - verify if message was sent by connected port's peer
266 * Handles cases where the node's network address has changed from
267 * the default of <0.0.0> to its configured setting.
269 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
271 u32 peer_port = tsk_peer_port(tsk);
275 if (unlikely(!tsk->connected))
278 if (unlikely(msg_origport(msg) != peer_port))
281 orig_node = msg_orignode(msg);
282 peer_node = tsk_peer_node(tsk);
284 if (likely(orig_node == peer_node))
287 if (!orig_node && (peer_node == tipc_own_addr))
290 if (!peer_node && (orig_node == tipc_own_addr))
297 * tipc_sk_create - create a TIPC socket
298 * @net: network namespace (must be default network)
299 * @sock: pre-allocated socket structure
300 * @protocol: protocol indicator (must be 0)
301 * @kern: caused by kernel or by userspace?
303 * This routine creates additional data structures used by the TIPC socket,
304 * initializes them, and links them together.
306 * Returns 0 on success, errno otherwise
308 static int tipc_sk_create(struct net *net, struct socket *sock,
309 int protocol, int kern)
311 const struct proto_ops *ops;
314 struct tipc_sock *tsk;
315 struct tipc_msg *msg;
317 /* Validate arguments */
318 if (unlikely(protocol != 0))
319 return -EPROTONOSUPPORT;
321 switch (sock->type) {
324 state = SS_UNCONNECTED;
328 state = SS_UNCONNECTED;
339 /* Allocate socket's protocol area */
341 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
343 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
349 tsk->max_pkt = MAX_PKT_DEFAULT;
350 INIT_LIST_HEAD(&tsk->publications);
352 tipc_msg_init(msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
355 /* Finish initializing socket data structures */
358 sock_init_data(sock, sk);
359 if (tipc_sk_insert(tsk)) {
360 pr_warn("Socket create failed; port numbrer exhausted\n");
363 msg_set_origport(msg, tsk->portid);
364 setup_timer(&tsk->timer, tipc_sk_timeout, tsk->portid);
365 sk->sk_backlog_rcv = tipc_backlog_rcv;
366 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
367 sk->sk_data_ready = tipc_data_ready;
368 sk->sk_write_space = tipc_write_space;
369 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
370 tsk->sent_unacked = 0;
371 atomic_set(&tsk->dupl_rcvcnt, 0);
373 if (sock->state == SS_READY) {
374 tsk_set_unreturnable(tsk, true);
375 if (sock->type == SOCK_DGRAM)
376 tsk_set_unreliable(tsk, true);
382 * tipc_sock_create_local - create TIPC socket from inside TIPC module
383 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
385 * We cannot use sock_creat_kern here because it bumps module user count.
386 * Since socket owner and creator is the same module we must make sure
387 * that module count remains zero for module local sockets, otherwise
388 * we cannot do rmmod.
390 * Returns 0 on success, errno otherwise
392 int tipc_sock_create_local(int type, struct socket **res)
396 rc = sock_create_lite(AF_TIPC, type, 0, res);
398 pr_err("Failed to create kernel socket\n");
401 tipc_sk_create(&init_net, *res, 0, 1);
407 * tipc_sock_release_local - release socket created by tipc_sock_create_local
408 * @sock: the socket to be released.
410 * Module reference count is not incremented when such sockets are created,
411 * so we must keep it from being decremented when they are released.
413 void tipc_sock_release_local(struct socket *sock)
421 * tipc_sock_accept_local - accept a connection on a socket created
422 * with tipc_sock_create_local. Use this function to avoid that
423 * module reference count is inadvertently incremented.
425 * @sock: the accepting socket
426 * @newsock: reference to the new socket to be created
427 * @flags: socket flags
430 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
433 struct sock *sk = sock->sk;
436 ret = sock_create_lite(sk->sk_family, sk->sk_type,
437 sk->sk_protocol, newsock);
441 ret = tipc_accept(sock, *newsock, flags);
443 sock_release(*newsock);
446 (*newsock)->ops = sock->ops;
450 static void tipc_sk_callback(struct rcu_head *head)
452 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
458 * tipc_release - destroy a TIPC socket
459 * @sock: socket to destroy
461 * This routine cleans up any messages that are still queued on the socket.
462 * For DGRAM and RDM socket types, all queued messages are rejected.
463 * For SEQPACKET and STREAM socket types, the first message is rejected
464 * and any others are discarded. (If the first message on a STREAM socket
465 * is partially-read, it is discarded and the next one is rejected instead.)
467 * NOTE: Rejected messages are not necessarily returned to the sender! They
468 * are returned or discarded according to the "destination droppable" setting
469 * specified for the message by the sender.
471 * Returns 0 on success, errno otherwise
473 static int tipc_release(struct socket *sock)
475 struct sock *sk = sock->sk;
476 struct tipc_sock *tsk;
481 * Exit if socket isn't fully initialized (occurs when a failed accept()
482 * releases a pre-allocated child socket that was never used)
491 * Reject all unreceived messages, except on an active connection
492 * (which disconnects locally & sends a 'FIN+' to peer)
494 dnode = tsk_peer_node(tsk);
495 while (sock->state != SS_DISCONNECTING) {
496 skb = __skb_dequeue(&sk->sk_receive_queue);
499 if (TIPC_SKB_CB(skb)->handle != NULL)
502 if ((sock->state == SS_CONNECTING) ||
503 (sock->state == SS_CONNECTED)) {
504 sock->state = SS_DISCONNECTING;
506 tipc_node_remove_conn(dnode, tsk->portid);
508 if (tipc_msg_reverse(skb, &dnode, TIPC_ERR_NO_PORT))
509 tipc_link_xmit_skb(skb, dnode, 0);
513 tipc_sk_withdraw(tsk, 0, NULL);
514 del_timer_sync(&tsk->timer);
516 if (tsk->connected) {
517 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
518 SHORT_H_SIZE, 0, dnode, tipc_own_addr,
520 tsk->portid, TIPC_ERR_NO_PORT);
522 tipc_link_xmit_skb(skb, dnode, tsk->portid);
523 tipc_node_remove_conn(dnode, tsk->portid);
526 /* Discard any remaining (connection-based) messages in receive queue */
527 __skb_queue_purge(&sk->sk_receive_queue);
529 /* Reject any messages that accumulated in backlog queue */
530 sock->state = SS_DISCONNECTING;
533 call_rcu(&tsk->rcu, tipc_sk_callback);
540 * tipc_bind - associate or disassocate TIPC name(s) with a socket
541 * @sock: socket structure
542 * @uaddr: socket address describing name(s) and desired operation
543 * @uaddr_len: size of socket address data structure
545 * Name and name sequence binding is indicated using a positive scope value;
546 * a negative scope value unbinds the specified name. Specifying no name
547 * (i.e. a socket address length of 0) unbinds all names from the socket.
549 * Returns 0 on success, errno otherwise
551 * NOTE: This routine doesn't need to take the socket lock since it doesn't
552 * access any non-constant socket information.
554 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
557 struct sock *sk = sock->sk;
558 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
559 struct tipc_sock *tsk = tipc_sk(sk);
563 if (unlikely(!uaddr_len)) {
564 res = tipc_sk_withdraw(tsk, 0, NULL);
568 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
572 if (addr->family != AF_TIPC) {
577 if (addr->addrtype == TIPC_ADDR_NAME)
578 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
579 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
584 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
585 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
586 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
591 res = (addr->scope > 0) ?
592 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
593 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
600 * tipc_getname - get port ID of socket or peer socket
601 * @sock: socket structure
602 * @uaddr: area for returned socket address
603 * @uaddr_len: area for returned length of socket address
604 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
606 * Returns 0 on success, errno otherwise
608 * NOTE: This routine doesn't need to take the socket lock since it only
609 * accesses socket information that is unchanging (or which changes in
610 * a completely predictable manner).
612 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
613 int *uaddr_len, int peer)
615 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
616 struct tipc_sock *tsk = tipc_sk(sock->sk);
618 memset(addr, 0, sizeof(*addr));
620 if ((sock->state != SS_CONNECTED) &&
621 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
623 addr->addr.id.ref = tsk_peer_port(tsk);
624 addr->addr.id.node = tsk_peer_node(tsk);
626 addr->addr.id.ref = tsk->portid;
627 addr->addr.id.node = tipc_own_addr;
630 *uaddr_len = sizeof(*addr);
631 addr->addrtype = TIPC_ADDR_ID;
632 addr->family = AF_TIPC;
634 addr->addr.name.domain = 0;
640 * tipc_poll - read and possibly block on pollmask
641 * @file: file structure associated with the socket
642 * @sock: socket for which to calculate the poll bits
645 * Returns pollmask value
648 * It appears that the usual socket locking mechanisms are not useful here
649 * since the pollmask info is potentially out-of-date the moment this routine
650 * exits. TCP and other protocols seem to rely on higher level poll routines
651 * to handle any preventable race conditions, so TIPC will do the same ...
653 * TIPC sets the returned events as follows:
655 * socket state flags set
656 * ------------ ---------
657 * unconnected no read flags
658 * POLLOUT if port is not congested
660 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
663 * connected POLLIN/POLLRDNORM if data in rx queue
664 * POLLOUT if port is not congested
666 * disconnecting POLLIN/POLLRDNORM/POLLHUP
669 * listening POLLIN if SYN in rx queue
672 * ready POLLIN/POLLRDNORM if data in rx queue
673 * [connectionless] POLLOUT (since port cannot be congested)
675 * IMPORTANT: The fact that a read or write operation is indicated does NOT
676 * imply that the operation will succeed, merely that it should be performed
677 * and will not block.
679 static unsigned int tipc_poll(struct file *file, struct socket *sock,
682 struct sock *sk = sock->sk;
683 struct tipc_sock *tsk = tipc_sk(sk);
686 sock_poll_wait(file, sk_sleep(sk), wait);
688 switch ((int)sock->state) {
695 if (!tsk->link_cong && !tsk_conn_cong(tsk))
700 if (!skb_queue_empty(&sk->sk_receive_queue))
701 mask |= (POLLIN | POLLRDNORM);
703 case SS_DISCONNECTING:
704 mask = (POLLIN | POLLRDNORM | POLLHUP);
712 * tipc_sendmcast - send multicast message
713 * @sock: socket structure
714 * @seq: destination address
715 * @msg: message to send
716 * @dsz: total length of message data
717 * @timeo: timeout to wait for wakeup
719 * Called from function tipc_sendmsg(), which has done all sanity checks
720 * Returns the number of bytes sent on success, or errno
722 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
723 struct msghdr *msg, size_t dsz, long timeo)
725 struct sock *sk = sock->sk;
726 struct tipc_msg *mhdr = &tipc_sk(sk)->phdr;
727 struct sk_buff_head head;
731 msg_set_type(mhdr, TIPC_MCAST_MSG);
732 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
733 msg_set_destport(mhdr, 0);
734 msg_set_destnode(mhdr, 0);
735 msg_set_nametype(mhdr, seq->type);
736 msg_set_namelower(mhdr, seq->lower);
737 msg_set_nameupper(mhdr, seq->upper);
738 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
741 mtu = tipc_bclink_get_mtu();
742 __skb_queue_head_init(&head);
743 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &head);
744 if (unlikely(rc < 0))
748 rc = tipc_bclink_xmit(&head);
749 if (likely(rc >= 0)) {
755 if (rc != -ELINKCONG)
757 tipc_sk(sk)->link_cong = 1;
758 rc = tipc_wait_for_sndmsg(sock, &timeo);
760 __skb_queue_purge(&head);
765 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
767 void tipc_sk_mcast_rcv(struct sk_buff *buf)
769 struct tipc_msg *msg = buf_msg(buf);
770 struct tipc_port_list dports = {0, NULL, };
771 struct tipc_port_list *item;
773 uint i, last, dst = 0;
774 u32 scope = TIPC_CLUSTER_SCOPE;
776 if (in_own_node(msg_orignode(msg)))
777 scope = TIPC_NODE_SCOPE;
779 /* Create destination port list: */
780 tipc_nametbl_mc_translate(msg_nametype(msg),
791 for (item = &dports; item; item = item->next) {
792 for (i = 0; i < PLSIZE && ++dst <= last; i++) {
793 b = (dst != last) ? skb_clone(buf, GFP_ATOMIC) : buf;
795 pr_warn("Failed do clone mcast rcv buffer\n");
798 msg_set_destport(msg, item->ports[i]);
802 tipc_port_list_free(&dports);
806 * tipc_sk_proto_rcv - receive a connection mng protocol message
807 * @tsk: receiving socket
808 * @dnode: node to send response message to, if any
809 * @buf: buffer containing protocol message
810 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
811 * (CONN_PROBE_REPLY) message should be forwarded.
813 static int tipc_sk_proto_rcv(struct tipc_sock *tsk, u32 *dnode,
816 struct tipc_msg *msg = buf_msg(buf);
819 /* Ignore if connection cannot be validated: */
820 if (!tsk_peer_msg(tsk, msg))
823 tsk->probing_state = TIPC_CONN_OK;
825 if (msg_type(msg) == CONN_ACK) {
826 conn_cong = tsk_conn_cong(tsk);
827 tsk->sent_unacked -= msg_msgcnt(msg);
829 tsk->sk.sk_write_space(&tsk->sk);
830 } else if (msg_type(msg) == CONN_PROBE) {
831 if (!tipc_msg_reverse(buf, dnode, TIPC_OK))
833 msg_set_type(msg, CONN_PROBE_REPLY);
836 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
842 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
844 struct sock *sk = sock->sk;
845 struct tipc_sock *tsk = tipc_sk(sk);
850 int err = sock_error(sk);
853 if (sock->state == SS_DISCONNECTING)
857 if (signal_pending(current))
858 return sock_intr_errno(*timeo_p);
860 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
861 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
862 finish_wait(sk_sleep(sk), &wait);
868 * tipc_sendmsg - send message in connectionless manner
869 * @iocb: if NULL, indicates that socket lock is already held
870 * @sock: socket structure
871 * @m: message to send
872 * @dsz: amount of user data to be sent
874 * Message must have an destination specified explicitly.
875 * Used for SOCK_RDM and SOCK_DGRAM messages,
876 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
877 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
879 * Returns the number of bytes sent on success, or errno otherwise
881 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
882 struct msghdr *m, size_t dsz)
884 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
885 struct sock *sk = sock->sk;
886 struct tipc_sock *tsk = tipc_sk(sk);
887 struct tipc_msg *mhdr = &tsk->phdr;
889 struct sk_buff_head head;
891 struct tipc_name_seq *seq = &dest->addr.nameseq;
897 return -EDESTADDRREQ;
899 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
900 (dest->family != AF_TIPC)))
903 if (dsz > TIPC_MAX_USER_MSG_SIZE)
909 if (unlikely(sock->state != SS_READY)) {
910 if (sock->state == SS_LISTENING) {
914 if (sock->state != SS_UNCONNECTED) {
918 if (tsk->published) {
922 if (dest->addrtype == TIPC_ADDR_NAME) {
923 tsk->conn_type = dest->addr.name.name.type;
924 tsk->conn_instance = dest->addr.name.name.instance;
928 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
930 if (dest->addrtype == TIPC_ADDR_MCAST) {
931 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
933 } else if (dest->addrtype == TIPC_ADDR_NAME) {
934 u32 type = dest->addr.name.name.type;
935 u32 inst = dest->addr.name.name.instance;
936 u32 domain = dest->addr.name.domain;
939 msg_set_type(mhdr, TIPC_NAMED_MSG);
940 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
941 msg_set_nametype(mhdr, type);
942 msg_set_nameinst(mhdr, inst);
943 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
944 dport = tipc_nametbl_translate(type, inst, &dnode);
945 msg_set_destnode(mhdr, dnode);
946 msg_set_destport(mhdr, dport);
947 if (unlikely(!dport && !dnode)) {
951 } else if (dest->addrtype == TIPC_ADDR_ID) {
952 dnode = dest->addr.id.node;
953 msg_set_type(mhdr, TIPC_DIRECT_MSG);
954 msg_set_lookup_scope(mhdr, 0);
955 msg_set_destnode(mhdr, dnode);
956 msg_set_destport(mhdr, dest->addr.id.ref);
957 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
961 mtu = tipc_node_get_mtu(dnode, tsk->portid);
962 __skb_queue_head_init(&head);
963 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &head);
968 skb = skb_peek(&head);
969 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
970 rc = tipc_link_xmit(&head, dnode, tsk->portid);
971 if (likely(rc >= 0)) {
972 if (sock->state != SS_READY)
973 sock->state = SS_CONNECTING;
979 if (rc != -ELINKCONG)
982 rc = tipc_wait_for_sndmsg(sock, &timeo);
984 __skb_queue_purge(&head);
993 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
995 struct sock *sk = sock->sk;
996 struct tipc_sock *tsk = tipc_sk(sk);
1001 int err = sock_error(sk);
1004 if (sock->state == SS_DISCONNECTING)
1006 else if (sock->state != SS_CONNECTED)
1010 if (signal_pending(current))
1011 return sock_intr_errno(*timeo_p);
1013 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1014 done = sk_wait_event(sk, timeo_p,
1016 !tsk_conn_cong(tsk)) ||
1018 finish_wait(sk_sleep(sk), &wait);
1024 * tipc_send_stream - send stream-oriented data
1026 * @sock: socket structure
1028 * @dsz: total length of data to be transmitted
1030 * Used for SOCK_STREAM data.
1032 * Returns the number of bytes sent on success (or partial success),
1033 * or errno if no data sent
1035 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1036 struct msghdr *m, size_t dsz)
1038 struct sock *sk = sock->sk;
1039 struct tipc_sock *tsk = tipc_sk(sk);
1040 struct tipc_msg *mhdr = &tsk->phdr;
1041 struct sk_buff_head head;
1042 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1043 u32 portid = tsk->portid;
1047 uint mtu, send, sent = 0;
1049 /* Handle implied connection establishment */
1050 if (unlikely(dest)) {
1051 rc = tipc_sendmsg(iocb, sock, m, dsz);
1052 if (dsz && (dsz == rc))
1053 tsk->sent_unacked = 1;
1056 if (dsz > (uint)INT_MAX)
1062 if (unlikely(sock->state != SS_CONNECTED)) {
1063 if (sock->state == SS_DISCONNECTING)
1070 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1071 dnode = tsk_peer_node(tsk);
1075 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1076 __skb_queue_head_init(&head);
1077 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &head);
1078 if (unlikely(rc < 0))
1081 if (likely(!tsk_conn_cong(tsk))) {
1082 rc = tipc_link_xmit(&head, dnode, portid);
1084 tsk->sent_unacked++;
1090 if (rc == -EMSGSIZE) {
1091 tsk->max_pkt = tipc_node_get_mtu(dnode, portid);
1094 if (rc != -ELINKCONG)
1098 rc = tipc_wait_for_sndpkt(sock, &timeo);
1100 __skb_queue_purge(&head);
1105 return sent ? sent : rc;
1109 * tipc_send_packet - send a connection-oriented message
1110 * @iocb: if NULL, indicates that socket lock is already held
1111 * @sock: socket structure
1112 * @m: message to send
1113 * @dsz: length of data to be transmitted
1115 * Used for SOCK_SEQPACKET messages.
1117 * Returns the number of bytes sent on success, or errno otherwise
1119 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1120 struct msghdr *m, size_t dsz)
1122 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1125 return tipc_send_stream(iocb, sock, m, dsz);
1128 /* tipc_sk_finish_conn - complete the setup of a connection
1130 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1133 struct tipc_msg *msg = &tsk->phdr;
1135 msg_set_destnode(msg, peer_node);
1136 msg_set_destport(msg, peer_port);
1137 msg_set_type(msg, TIPC_CONN_MSG);
1138 msg_set_lookup_scope(msg, 0);
1139 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1141 tsk->probing_intv = CONN_PROBING_INTERVAL;
1142 tsk->probing_state = TIPC_CONN_OK;
1144 mod_timer(&tsk->timer, jiffies + tsk->probing_intv);
1145 tipc_node_add_conn(peer_node, tsk->portid, peer_port);
1146 tsk->max_pkt = tipc_node_get_mtu(peer_node, tsk->portid);
1150 * set_orig_addr - capture sender's address for received message
1151 * @m: descriptor for message info
1152 * @msg: received message header
1154 * Note: Address is not captured if not requested by receiver.
1156 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1158 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1161 addr->family = AF_TIPC;
1162 addr->addrtype = TIPC_ADDR_ID;
1163 memset(&addr->addr, 0, sizeof(addr->addr));
1164 addr->addr.id.ref = msg_origport(msg);
1165 addr->addr.id.node = msg_orignode(msg);
1166 addr->addr.name.domain = 0; /* could leave uninitialized */
1167 addr->scope = 0; /* could leave uninitialized */
1168 m->msg_namelen = sizeof(struct sockaddr_tipc);
1173 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1174 * @m: descriptor for message info
1175 * @msg: received message header
1176 * @tsk: TIPC port associated with message
1178 * Note: Ancillary data is not captured if not requested by receiver.
1180 * Returns 0 if successful, otherwise errno
1182 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1183 struct tipc_sock *tsk)
1191 if (likely(m->msg_controllen == 0))
1194 /* Optionally capture errored message object(s) */
1195 err = msg ? msg_errcode(msg) : 0;
1196 if (unlikely(err)) {
1198 anc_data[1] = msg_data_sz(msg);
1199 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1203 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1210 /* Optionally capture message destination object */
1211 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1212 switch (dest_type) {
1213 case TIPC_NAMED_MSG:
1215 anc_data[0] = msg_nametype(msg);
1216 anc_data[1] = msg_namelower(msg);
1217 anc_data[2] = msg_namelower(msg);
1219 case TIPC_MCAST_MSG:
1221 anc_data[0] = msg_nametype(msg);
1222 anc_data[1] = msg_namelower(msg);
1223 anc_data[2] = msg_nameupper(msg);
1226 has_name = (tsk->conn_type != 0);
1227 anc_data[0] = tsk->conn_type;
1228 anc_data[1] = tsk->conn_instance;
1229 anc_data[2] = tsk->conn_instance;
1235 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1243 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1245 struct sk_buff *skb = NULL;
1246 struct tipc_msg *msg;
1247 u32 peer_port = tsk_peer_port(tsk);
1248 u32 dnode = tsk_peer_node(tsk);
1250 if (!tsk->connected)
1252 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, dnode,
1253 tipc_own_addr, peer_port, tsk->portid, TIPC_OK);
1257 msg_set_msgcnt(msg, ack);
1258 tipc_link_xmit_skb(skb, dnode, msg_link_selector(msg));
1261 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1263 struct sock *sk = sock->sk;
1265 long timeo = *timeop;
1269 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1270 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1271 if (sock->state == SS_DISCONNECTING) {
1276 timeo = schedule_timeout(timeo);
1280 if (!skb_queue_empty(&sk->sk_receive_queue))
1282 err = sock_intr_errno(timeo);
1283 if (signal_pending(current))
1289 finish_wait(sk_sleep(sk), &wait);
1295 * tipc_recvmsg - receive packet-oriented message
1297 * @m: descriptor for message info
1298 * @buf_len: total size of user buffer area
1299 * @flags: receive flags
1301 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1302 * If the complete message doesn't fit in user area, truncate it.
1304 * Returns size of returned message data, errno otherwise
1306 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1307 struct msghdr *m, size_t buf_len, int flags)
1309 struct sock *sk = sock->sk;
1310 struct tipc_sock *tsk = tipc_sk(sk);
1311 struct sk_buff *buf;
1312 struct tipc_msg *msg;
1318 /* Catch invalid receive requests */
1319 if (unlikely(!buf_len))
1324 if (unlikely(sock->state == SS_UNCONNECTED)) {
1329 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1332 /* Look for a message in receive queue; wait if necessary */
1333 res = tipc_wait_for_rcvmsg(sock, &timeo);
1337 /* Look at first message in receive queue */
1338 buf = skb_peek(&sk->sk_receive_queue);
1340 sz = msg_data_sz(msg);
1341 err = msg_errcode(msg);
1343 /* Discard an empty non-errored message & try again */
1344 if ((!sz) && (!err)) {
1345 tsk_advance_rx_queue(sk);
1349 /* Capture sender's address (optional) */
1350 set_orig_addr(m, msg);
1352 /* Capture ancillary data (optional) */
1353 res = tipc_sk_anc_data_recv(m, msg, tsk);
1357 /* Capture message data (if valid) & compute return value (always) */
1359 if (unlikely(buf_len < sz)) {
1361 m->msg_flags |= MSG_TRUNC;
1363 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1368 if ((sock->state == SS_READY) ||
1369 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1375 /* Consume received message (optional) */
1376 if (likely(!(flags & MSG_PEEK))) {
1377 if ((sock->state != SS_READY) &&
1378 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1379 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1380 tsk->rcv_unacked = 0;
1382 tsk_advance_rx_queue(sk);
1390 * tipc_recv_stream - receive stream-oriented data
1392 * @m: descriptor for message info
1393 * @buf_len: total size of user buffer area
1394 * @flags: receive flags
1396 * Used for SOCK_STREAM messages only. If not enough data is available
1397 * will optionally wait for more; never truncates data.
1399 * Returns size of returned message data, errno otherwise
1401 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1402 struct msghdr *m, size_t buf_len, int flags)
1404 struct sock *sk = sock->sk;
1405 struct tipc_sock *tsk = tipc_sk(sk);
1406 struct sk_buff *buf;
1407 struct tipc_msg *msg;
1410 int sz_to_copy, target, needed;
1415 /* Catch invalid receive attempts */
1416 if (unlikely(!buf_len))
1421 if (unlikely(sock->state == SS_UNCONNECTED)) {
1426 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1427 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1430 /* Look for a message in receive queue; wait if necessary */
1431 res = tipc_wait_for_rcvmsg(sock, &timeo);
1435 /* Look at first message in receive queue */
1436 buf = skb_peek(&sk->sk_receive_queue);
1438 sz = msg_data_sz(msg);
1439 err = msg_errcode(msg);
1441 /* Discard an empty non-errored message & try again */
1442 if ((!sz) && (!err)) {
1443 tsk_advance_rx_queue(sk);
1447 /* Optionally capture sender's address & ancillary data of first msg */
1448 if (sz_copied == 0) {
1449 set_orig_addr(m, msg);
1450 res = tipc_sk_anc_data_recv(m, msg, tsk);
1455 /* Capture message data (if valid) & compute return value (always) */
1457 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1460 needed = (buf_len - sz_copied);
1461 sz_to_copy = (sz <= needed) ? sz : needed;
1463 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1468 sz_copied += sz_to_copy;
1470 if (sz_to_copy < sz) {
1471 if (!(flags & MSG_PEEK))
1472 TIPC_SKB_CB(buf)->handle =
1473 (void *)(unsigned long)(offset + sz_to_copy);
1478 goto exit; /* can't add error msg to valid data */
1480 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1486 /* Consume received message (optional) */
1487 if (likely(!(flags & MSG_PEEK))) {
1488 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1489 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1490 tsk->rcv_unacked = 0;
1492 tsk_advance_rx_queue(sk);
1495 /* Loop around if more data is required */
1496 if ((sz_copied < buf_len) && /* didn't get all requested data */
1497 (!skb_queue_empty(&sk->sk_receive_queue) ||
1498 (sz_copied < target)) && /* and more is ready or required */
1499 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1500 (!err)) /* and haven't reached a FIN */
1505 return sz_copied ? sz_copied : res;
1509 * tipc_write_space - wake up thread if port congestion is released
1512 static void tipc_write_space(struct sock *sk)
1514 struct socket_wq *wq;
1517 wq = rcu_dereference(sk->sk_wq);
1518 if (wq_has_sleeper(wq))
1519 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1520 POLLWRNORM | POLLWRBAND);
1525 * tipc_data_ready - wake up threads to indicate messages have been received
1527 * @len: the length of messages
1529 static void tipc_data_ready(struct sock *sk)
1531 struct socket_wq *wq;
1534 wq = rcu_dereference(sk->sk_wq);
1535 if (wq_has_sleeper(wq))
1536 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1537 POLLRDNORM | POLLRDBAND);
1542 * filter_connect - Handle all incoming messages for a connection-based socket
1546 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1548 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1550 struct sock *sk = &tsk->sk;
1551 struct socket *sock = sk->sk_socket;
1552 struct tipc_msg *msg = buf_msg(*buf);
1553 int retval = -TIPC_ERR_NO_PORT;
1558 switch ((int)sock->state) {
1560 /* Accept only connection-based messages sent by peer */
1561 if (tsk_peer_msg(tsk, msg)) {
1562 if (unlikely(msg_errcode(msg))) {
1563 sock->state = SS_DISCONNECTING;
1565 /* let timer expire on it's own */
1566 tipc_node_remove_conn(tsk_peer_node(tsk),
1573 /* Accept only ACK or NACK message */
1575 if (unlikely(!msg_connected(msg)))
1578 if (unlikely(msg_errcode(msg))) {
1579 sock->state = SS_DISCONNECTING;
1580 sk->sk_err = ECONNREFUSED;
1585 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1586 sock->state = SS_DISCONNECTING;
1587 sk->sk_err = EINVAL;
1592 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1593 msg_set_importance(&tsk->phdr, msg_importance(msg));
1594 sock->state = SS_CONNECTED;
1596 /* If an incoming message is an 'ACK-', it should be
1597 * discarded here because it doesn't contain useful
1598 * data. In addition, we should try to wake up
1599 * connect() routine if sleeping.
1601 if (msg_data_sz(msg) == 0) {
1604 if (waitqueue_active(sk_sleep(sk)))
1605 wake_up_interruptible(sk_sleep(sk));
1610 case SS_UNCONNECTED:
1611 /* Accept only SYN message */
1612 if (!msg_connected(msg) && !(msg_errcode(msg)))
1615 case SS_DISCONNECTING:
1618 pr_err("Unknown socket state %u\n", sock->state);
1624 * rcvbuf_limit - get proper overload limit of socket receive queue
1628 * For all connection oriented messages, irrespective of importance,
1629 * the default overload value (i.e. 67MB) is set as limit.
1631 * For all connectionless messages, by default new queue limits are
1634 * TIPC_LOW_IMPORTANCE (4 MB)
1635 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1636 * TIPC_HIGH_IMPORTANCE (16 MB)
1637 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1639 * Returns overload limit according to corresponding message importance
1641 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1643 struct tipc_msg *msg = buf_msg(buf);
1645 if (msg_connected(msg))
1646 return sysctl_tipc_rmem[2];
1648 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1649 msg_importance(msg);
1653 * filter_rcv - validate incoming message
1657 * Enqueues message on receive queue if acceptable; optionally handles
1658 * disconnect indication for a connected socket.
1660 * Called with socket lock already taken; port lock may also be taken.
1662 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1663 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1665 static int filter_rcv(struct sock *sk, struct sk_buff *buf)
1667 struct socket *sock = sk->sk_socket;
1668 struct tipc_sock *tsk = tipc_sk(sk);
1669 struct tipc_msg *msg = buf_msg(buf);
1670 unsigned int limit = rcvbuf_limit(sk, buf);
1674 if (unlikely(msg_user(msg) == CONN_MANAGER))
1675 return tipc_sk_proto_rcv(tsk, &onode, buf);
1677 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1680 sk->sk_write_space(sk);
1684 /* Reject message if it is wrong sort of message for socket */
1685 if (msg_type(msg) > TIPC_DIRECT_MSG)
1686 return -TIPC_ERR_NO_PORT;
1688 if (sock->state == SS_READY) {
1689 if (msg_connected(msg))
1690 return -TIPC_ERR_NO_PORT;
1692 rc = filter_connect(tsk, &buf);
1693 if (rc != TIPC_OK || buf == NULL)
1697 /* Reject message if there isn't room to queue it */
1698 if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1699 return -TIPC_ERR_OVERLOAD;
1701 /* Enqueue message */
1702 TIPC_SKB_CB(buf)->handle = NULL;
1703 __skb_queue_tail(&sk->sk_receive_queue, buf);
1704 skb_set_owner_r(buf, sk);
1706 sk->sk_data_ready(sk);
1711 * tipc_backlog_rcv - handle incoming message from backlog queue
1715 * Caller must hold socket lock, but not port lock.
1719 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1723 struct tipc_sock *tsk = tipc_sk(sk);
1724 uint truesize = skb->truesize;
1726 rc = filter_rcv(sk, skb);
1729 if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1730 atomic_add(truesize, &tsk->dupl_rcvcnt);
1734 if ((rc < 0) && !tipc_msg_reverse(skb, &onode, -rc))
1737 tipc_link_xmit_skb(skb, onode, 0);
1743 * tipc_sk_rcv - handle incoming message
1744 * @skb: buffer containing arriving message
1746 * Returns 0 if success, or errno: -EHOSTUNREACH
1748 int tipc_sk_rcv(struct sk_buff *skb)
1750 struct tipc_sock *tsk;
1752 u32 dport = msg_destport(buf_msg(skb));
1757 /* Validate destination and message */
1758 tsk = tipc_sk_lookup(dport);
1759 if (unlikely(!tsk)) {
1760 rc = tipc_msg_eval(skb, &dnode);
1766 spin_lock_bh(&sk->sk_lock.slock);
1768 if (!sock_owned_by_user(sk)) {
1769 rc = filter_rcv(sk, skb);
1771 if (sk->sk_backlog.len == 0)
1772 atomic_set(&tsk->dupl_rcvcnt, 0);
1773 limit = rcvbuf_limit(sk, skb) + atomic_read(&tsk->dupl_rcvcnt);
1774 if (sk_add_backlog(sk, skb, limit))
1775 rc = -TIPC_ERR_OVERLOAD;
1777 spin_unlock_bh(&sk->sk_lock.slock);
1782 if ((rc < 0) && !tipc_msg_reverse(skb, &dnode, -rc))
1783 return -EHOSTUNREACH;
1785 tipc_link_xmit_skb(skb, dnode, 0);
1786 return (rc < 0) ? -EHOSTUNREACH : 0;
1789 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1791 struct sock *sk = sock->sk;
1796 int err = sock_error(sk);
1801 if (signal_pending(current))
1802 return sock_intr_errno(*timeo_p);
1804 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1805 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1806 finish_wait(sk_sleep(sk), &wait);
1812 * tipc_connect - establish a connection to another TIPC port
1813 * @sock: socket structure
1814 * @dest: socket address for destination port
1815 * @destlen: size of socket address data structure
1816 * @flags: file-related flags associated with socket
1818 * Returns 0 on success, errno otherwise
1820 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1821 int destlen, int flags)
1823 struct sock *sk = sock->sk;
1824 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1825 struct msghdr m = {NULL,};
1826 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1827 socket_state previous;
1832 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1833 if (sock->state == SS_READY) {
1839 * Reject connection attempt using multicast address
1841 * Note: send_msg() validates the rest of the address fields,
1842 * so there's no need to do it here
1844 if (dst->addrtype == TIPC_ADDR_MCAST) {
1849 previous = sock->state;
1850 switch (sock->state) {
1851 case SS_UNCONNECTED:
1852 /* Send a 'SYN-' to destination */
1854 m.msg_namelen = destlen;
1856 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1857 * indicate send_msg() is never blocked.
1860 m.msg_flags = MSG_DONTWAIT;
1862 res = tipc_sendmsg(NULL, sock, &m, 0);
1863 if ((res < 0) && (res != -EWOULDBLOCK))
1866 /* Just entered SS_CONNECTING state; the only
1867 * difference is that return value in non-blocking
1868 * case is EINPROGRESS, rather than EALREADY.
1872 if (previous == SS_CONNECTING)
1876 timeout = msecs_to_jiffies(timeout);
1877 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1878 res = tipc_wait_for_connect(sock, &timeout);
1893 * tipc_listen - allow socket to listen for incoming connections
1894 * @sock: socket structure
1897 * Returns 0 on success, errno otherwise
1899 static int tipc_listen(struct socket *sock, int len)
1901 struct sock *sk = sock->sk;
1906 if (sock->state != SS_UNCONNECTED)
1909 sock->state = SS_LISTENING;
1917 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1919 struct sock *sk = sock->sk;
1923 /* True wake-one mechanism for incoming connections: only
1924 * one process gets woken up, not the 'whole herd'.
1925 * Since we do not 'race & poll' for established sockets
1926 * anymore, the common case will execute the loop only once.
1929 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1930 TASK_INTERRUPTIBLE);
1931 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1933 timeo = schedule_timeout(timeo);
1937 if (!skb_queue_empty(&sk->sk_receive_queue))
1940 if (sock->state != SS_LISTENING)
1942 err = sock_intr_errno(timeo);
1943 if (signal_pending(current))
1949 finish_wait(sk_sleep(sk), &wait);
1954 * tipc_accept - wait for connection request
1955 * @sock: listening socket
1956 * @newsock: new socket that is to be connected
1957 * @flags: file-related flags associated with socket
1959 * Returns 0 on success, errno otherwise
1961 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1963 struct sock *new_sk, *sk = sock->sk;
1964 struct sk_buff *buf;
1965 struct tipc_sock *new_tsock;
1966 struct tipc_msg *msg;
1972 if (sock->state != SS_LISTENING) {
1976 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1977 res = tipc_wait_for_accept(sock, timeo);
1981 buf = skb_peek(&sk->sk_receive_queue);
1983 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1987 new_sk = new_sock->sk;
1988 new_tsock = tipc_sk(new_sk);
1991 /* we lock on new_sk; but lockdep sees the lock on sk */
1992 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
1995 * Reject any stray messages received by new socket
1996 * before the socket lock was taken (very, very unlikely)
1998 tsk_rej_rx_queue(new_sk);
2000 /* Connect new socket to it's peer */
2001 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2002 new_sock->state = SS_CONNECTED;
2004 tsk_set_importance(new_tsock, msg_importance(msg));
2005 if (msg_named(msg)) {
2006 new_tsock->conn_type = msg_nametype(msg);
2007 new_tsock->conn_instance = msg_nameinst(msg);
2011 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2012 * Respond to 'SYN+' by queuing it on new socket.
2014 if (!msg_data_sz(msg)) {
2015 struct msghdr m = {NULL,};
2017 tsk_advance_rx_queue(sk);
2018 tipc_send_packet(NULL, new_sock, &m, 0);
2020 __skb_dequeue(&sk->sk_receive_queue);
2021 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2022 skb_set_owner_r(buf, new_sk);
2024 release_sock(new_sk);
2031 * tipc_shutdown - shutdown socket connection
2032 * @sock: socket structure
2033 * @how: direction to close (must be SHUT_RDWR)
2035 * Terminates connection (if necessary), then purges socket's receive queue.
2037 * Returns 0 on success, errno otherwise
2039 static int tipc_shutdown(struct socket *sock, int how)
2041 struct sock *sk = sock->sk;
2042 struct tipc_sock *tsk = tipc_sk(sk);
2043 struct sk_buff *skb;
2047 if (how != SHUT_RDWR)
2052 switch (sock->state) {
2057 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2058 skb = __skb_dequeue(&sk->sk_receive_queue);
2060 if (TIPC_SKB_CB(skb)->handle != NULL) {
2064 if (tipc_msg_reverse(skb, &dnode, TIPC_CONN_SHUTDOWN))
2065 tipc_link_xmit_skb(skb, dnode, tsk->portid);
2066 tipc_node_remove_conn(dnode, tsk->portid);
2068 dnode = tsk_peer_node(tsk);
2069 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2070 TIPC_CONN_MSG, SHORT_H_SIZE,
2071 0, dnode, tipc_own_addr,
2073 tsk->portid, TIPC_CONN_SHUTDOWN);
2074 tipc_link_xmit_skb(skb, dnode, tsk->portid);
2077 sock->state = SS_DISCONNECTING;
2078 tipc_node_remove_conn(dnode, tsk->portid);
2081 case SS_DISCONNECTING:
2083 /* Discard any unreceived messages */
2084 __skb_queue_purge(&sk->sk_receive_queue);
2086 /* Wake up anyone sleeping in poll */
2087 sk->sk_state_change(sk);
2099 static void tipc_sk_timeout(unsigned long portid)
2101 struct tipc_sock *tsk;
2103 struct sk_buff *skb = NULL;
2104 u32 peer_port, peer_node;
2106 tsk = tipc_sk_lookup(portid);
2112 if (!tsk->connected) {
2116 peer_port = tsk_peer_port(tsk);
2117 peer_node = tsk_peer_node(tsk);
2119 if (tsk->probing_state == TIPC_CONN_PROBING) {
2120 /* Previous probe not answered -> self abort */
2121 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
2122 SHORT_H_SIZE, 0, tipc_own_addr,
2123 peer_node, portid, peer_port,
2126 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE,
2127 0, peer_node, tipc_own_addr,
2128 peer_port, portid, TIPC_OK);
2129 tsk->probing_state = TIPC_CONN_PROBING;
2130 mod_timer(&tsk->timer, jiffies + tsk->probing_intv);
2134 tipc_link_xmit_skb(skb, peer_node, portid);
2139 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2140 struct tipc_name_seq const *seq)
2142 struct publication *publ;
2147 key = tsk->portid + tsk->pub_count + 1;
2148 if (key == tsk->portid)
2151 publ = tipc_nametbl_publish(seq->type, seq->lower, seq->upper,
2152 scope, tsk->portid, key);
2153 if (unlikely(!publ))
2156 list_add(&publ->pport_list, &tsk->publications);
2162 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2163 struct tipc_name_seq const *seq)
2165 struct publication *publ;
2166 struct publication *safe;
2169 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2171 if (publ->scope != scope)
2173 if (publ->type != seq->type)
2175 if (publ->lower != seq->lower)
2177 if (publ->upper != seq->upper)
2179 tipc_nametbl_withdraw(publ->type, publ->lower,
2180 publ->ref, publ->key);
2184 tipc_nametbl_withdraw(publ->type, publ->lower,
2185 publ->ref, publ->key);
2188 if (list_empty(&tsk->publications))
2193 static int tipc_sk_show(struct tipc_sock *tsk, char *buf,
2194 int len, int full_id)
2196 struct publication *publ;
2200 ret = tipc_snprintf(buf, len, "<%u.%u.%u:%u>:",
2201 tipc_zone(tipc_own_addr),
2202 tipc_cluster(tipc_own_addr),
2203 tipc_node(tipc_own_addr), tsk->portid);
2205 ret = tipc_snprintf(buf, len, "%-10u:", tsk->portid);
2207 if (tsk->connected) {
2208 u32 dport = tsk_peer_port(tsk);
2209 u32 destnode = tsk_peer_node(tsk);
2211 ret += tipc_snprintf(buf + ret, len - ret,
2212 " connected to <%u.%u.%u:%u>",
2213 tipc_zone(destnode),
2214 tipc_cluster(destnode),
2215 tipc_node(destnode), dport);
2216 if (tsk->conn_type != 0)
2217 ret += tipc_snprintf(buf + ret, len - ret,
2218 " via {%u,%u}", tsk->conn_type,
2219 tsk->conn_instance);
2220 } else if (tsk->published) {
2221 ret += tipc_snprintf(buf + ret, len - ret, " bound to");
2222 list_for_each_entry(publ, &tsk->publications, pport_list) {
2223 if (publ->lower == publ->upper)
2224 ret += tipc_snprintf(buf + ret, len - ret,
2225 " {%u,%u}", publ->type,
2228 ret += tipc_snprintf(buf + ret, len - ret,
2229 " {%u,%u,%u}", publ->type,
2230 publ->lower, publ->upper);
2233 ret += tipc_snprintf(buf + ret, len - ret, "\n");
2237 struct sk_buff *tipc_sk_socks_show(void)
2239 const struct bucket_table *tbl;
2240 struct rhash_head *pos;
2241 struct sk_buff *buf;
2242 struct tlv_desc *rep_tlv;
2245 struct tipc_sock *tsk;
2249 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2252 rep_tlv = (struct tlv_desc *)buf->data;
2253 pb = TLV_DATA(rep_tlv);
2254 pb_len = ULTRA_STRING_MAX_LEN;
2257 tbl = rht_dereference_rcu((&tipc_sk_rht)->tbl, &tipc_sk_rht);
2258 for (i = 0; i < tbl->size; i++) {
2259 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2260 spin_lock_bh(&tsk->sk.sk_lock.slock);
2261 str_len += tipc_sk_show(tsk, pb + str_len,
2262 pb_len - str_len, 0);
2263 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2268 str_len += 1; /* for "\0" */
2269 skb_put(buf, TLV_SPACE(str_len));
2270 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2275 /* tipc_sk_reinit: set non-zero address in all existing sockets
2276 * when we go from standalone to network mode.
2278 void tipc_sk_reinit(void)
2280 const struct bucket_table *tbl;
2281 struct rhash_head *pos;
2282 struct tipc_sock *tsk;
2283 struct tipc_msg *msg;
2287 tbl = rht_dereference_rcu((&tipc_sk_rht)->tbl, &tipc_sk_rht);
2288 for (i = 0; i < tbl->size; i++) {
2289 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2290 spin_lock_bh(&tsk->sk.sk_lock.slock);
2292 msg_set_prevnode(msg, tipc_own_addr);
2293 msg_set_orignode(msg, tipc_own_addr);
2294 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2300 static struct tipc_sock *tipc_sk_lookup(u32 portid)
2302 struct tipc_sock *tsk;
2305 tsk = rhashtable_lookup(&tipc_sk_rht, &portid);
2307 sock_hold(&tsk->sk);
2313 static int tipc_sk_insert(struct tipc_sock *tsk)
2315 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2316 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2318 while (remaining--) {
2320 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2321 portid = TIPC_MIN_PORT;
2322 tsk->portid = portid;
2323 sock_hold(&tsk->sk);
2324 if (rhashtable_lookup_insert(&tipc_sk_rht, &tsk->node))
2332 static void tipc_sk_remove(struct tipc_sock *tsk)
2334 struct sock *sk = &tsk->sk;
2336 if (rhashtable_remove(&tipc_sk_rht, &tsk->node)) {
2337 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2342 int tipc_sk_rht_init(void)
2344 struct rhashtable_params rht_params = {
2346 .head_offset = offsetof(struct tipc_sock, node),
2347 .key_offset = offsetof(struct tipc_sock, portid),
2348 .key_len = sizeof(u32), /* portid */
2350 .max_shift = 20, /* 1M */
2351 .min_shift = 8, /* 256 */
2352 .grow_decision = rht_grow_above_75,
2353 .shrink_decision = rht_shrink_below_30,
2356 return rhashtable_init(&tipc_sk_rht, &rht_params);
2359 void tipc_sk_rht_destroy(void)
2361 /* Wait for socket readers to complete */
2364 rhashtable_destroy(&tipc_sk_rht);
2368 * tipc_setsockopt - set socket option
2369 * @sock: socket structure
2370 * @lvl: option level
2371 * @opt: option identifier
2372 * @ov: pointer to new option value
2373 * @ol: length of option value
2375 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2376 * (to ease compatibility).
2378 * Returns 0 on success, errno otherwise
2380 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2381 char __user *ov, unsigned int ol)
2383 struct sock *sk = sock->sk;
2384 struct tipc_sock *tsk = tipc_sk(sk);
2388 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2390 if (lvl != SOL_TIPC)
2391 return -ENOPROTOOPT;
2392 if (ol < sizeof(value))
2394 res = get_user(value, (u32 __user *)ov);
2401 case TIPC_IMPORTANCE:
2402 res = tsk_set_importance(tsk, value);
2404 case TIPC_SRC_DROPPABLE:
2405 if (sock->type != SOCK_STREAM)
2406 tsk_set_unreliable(tsk, value);
2410 case TIPC_DEST_DROPPABLE:
2411 tsk_set_unreturnable(tsk, value);
2413 case TIPC_CONN_TIMEOUT:
2414 tipc_sk(sk)->conn_timeout = value;
2415 /* no need to set "res", since already 0 at this point */
2427 * tipc_getsockopt - get socket option
2428 * @sock: socket structure
2429 * @lvl: option level
2430 * @opt: option identifier
2431 * @ov: receptacle for option value
2432 * @ol: receptacle for length of option value
2434 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2435 * (to ease compatibility).
2437 * Returns 0 on success, errno otherwise
2439 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2440 char __user *ov, int __user *ol)
2442 struct sock *sk = sock->sk;
2443 struct tipc_sock *tsk = tipc_sk(sk);
2448 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2449 return put_user(0, ol);
2450 if (lvl != SOL_TIPC)
2451 return -ENOPROTOOPT;
2452 res = get_user(len, ol);
2459 case TIPC_IMPORTANCE:
2460 value = tsk_importance(tsk);
2462 case TIPC_SRC_DROPPABLE:
2463 value = tsk_unreliable(tsk);
2465 case TIPC_DEST_DROPPABLE:
2466 value = tsk_unreturnable(tsk);
2468 case TIPC_CONN_TIMEOUT:
2469 value = tsk->conn_timeout;
2470 /* no need to set "res", since already 0 at this point */
2472 case TIPC_NODE_RECVQ_DEPTH:
2473 value = 0; /* was tipc_queue_size, now obsolete */
2475 case TIPC_SOCK_RECVQ_DEPTH:
2476 value = skb_queue_len(&sk->sk_receive_queue);
2485 return res; /* "get" failed */
2487 if (len < sizeof(value))
2490 if (copy_to_user(ov, &value, sizeof(value)))
2493 return put_user(sizeof(value), ol);
2496 static int tipc_ioctl(struct socket *sk, unsigned int cmd, unsigned long arg)
2498 struct tipc_sioc_ln_req lnr;
2499 void __user *argp = (void __user *)arg;
2502 case SIOCGETLINKNAME:
2503 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2505 if (!tipc_node_get_linkname(lnr.bearer_id & 0xffff, lnr.peer,
2506 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2507 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2511 return -EADDRNOTAVAIL;
2513 return -ENOIOCTLCMD;
2517 /* Protocol switches for the various types of TIPC sockets */
2519 static const struct proto_ops msg_ops = {
2520 .owner = THIS_MODULE,
2522 .release = tipc_release,
2524 .connect = tipc_connect,
2525 .socketpair = sock_no_socketpair,
2526 .accept = sock_no_accept,
2527 .getname = tipc_getname,
2529 .ioctl = tipc_ioctl,
2530 .listen = sock_no_listen,
2531 .shutdown = tipc_shutdown,
2532 .setsockopt = tipc_setsockopt,
2533 .getsockopt = tipc_getsockopt,
2534 .sendmsg = tipc_sendmsg,
2535 .recvmsg = tipc_recvmsg,
2536 .mmap = sock_no_mmap,
2537 .sendpage = sock_no_sendpage
2540 static const struct proto_ops packet_ops = {
2541 .owner = THIS_MODULE,
2543 .release = tipc_release,
2545 .connect = tipc_connect,
2546 .socketpair = sock_no_socketpair,
2547 .accept = tipc_accept,
2548 .getname = tipc_getname,
2550 .ioctl = tipc_ioctl,
2551 .listen = tipc_listen,
2552 .shutdown = tipc_shutdown,
2553 .setsockopt = tipc_setsockopt,
2554 .getsockopt = tipc_getsockopt,
2555 .sendmsg = tipc_send_packet,
2556 .recvmsg = tipc_recvmsg,
2557 .mmap = sock_no_mmap,
2558 .sendpage = sock_no_sendpage
2561 static const struct proto_ops stream_ops = {
2562 .owner = THIS_MODULE,
2564 .release = tipc_release,
2566 .connect = tipc_connect,
2567 .socketpair = sock_no_socketpair,
2568 .accept = tipc_accept,
2569 .getname = tipc_getname,
2571 .ioctl = tipc_ioctl,
2572 .listen = tipc_listen,
2573 .shutdown = tipc_shutdown,
2574 .setsockopt = tipc_setsockopt,
2575 .getsockopt = tipc_getsockopt,
2576 .sendmsg = tipc_send_stream,
2577 .recvmsg = tipc_recv_stream,
2578 .mmap = sock_no_mmap,
2579 .sendpage = sock_no_sendpage
2582 static const struct net_proto_family tipc_family_ops = {
2583 .owner = THIS_MODULE,
2585 .create = tipc_sk_create
2588 static struct proto tipc_proto = {
2590 .owner = THIS_MODULE,
2591 .obj_size = sizeof(struct tipc_sock),
2592 .sysctl_rmem = sysctl_tipc_rmem
2595 static struct proto tipc_proto_kern = {
2597 .obj_size = sizeof(struct tipc_sock),
2598 .sysctl_rmem = sysctl_tipc_rmem
2602 * tipc_socket_init - initialize TIPC socket interface
2604 * Returns 0 on success, errno otherwise
2606 int tipc_socket_init(void)
2610 res = proto_register(&tipc_proto, 1);
2612 pr_err("Failed to register TIPC protocol type\n");
2616 res = sock_register(&tipc_family_ops);
2618 pr_err("Failed to register TIPC socket type\n");
2619 proto_unregister(&tipc_proto);
2627 * tipc_socket_stop - stop TIPC socket interface
2629 void tipc_socket_stop(void)
2631 sock_unregister(tipc_family_ops.family);
2632 proto_unregister(&tipc_proto);
2635 /* Caller should hold socket lock for the passed tipc socket. */
2636 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2640 struct nlattr *nest;
2642 peer_node = tsk_peer_node(tsk);
2643 peer_port = tsk_peer_port(tsk);
2645 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2647 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2649 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2652 if (tsk->conn_type != 0) {
2653 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2655 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2657 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2660 nla_nest_end(skb, nest);
2665 nla_nest_cancel(skb, nest);
2670 /* Caller should hold socket lock for the passed tipc socket. */
2671 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2672 struct tipc_sock *tsk)
2676 struct nlattr *attrs;
2678 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2679 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2683 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2685 goto genlmsg_cancel;
2686 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2687 goto attr_msg_cancel;
2688 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr))
2689 goto attr_msg_cancel;
2691 if (tsk->connected) {
2692 err = __tipc_nl_add_sk_con(skb, tsk);
2694 goto attr_msg_cancel;
2695 } else if (!list_empty(&tsk->publications)) {
2696 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2697 goto attr_msg_cancel;
2699 nla_nest_end(skb, attrs);
2700 genlmsg_end(skb, hdr);
2705 nla_nest_cancel(skb, attrs);
2707 genlmsg_cancel(skb, hdr);
2712 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2715 struct tipc_sock *tsk;
2716 const struct bucket_table *tbl;
2717 struct rhash_head *pos;
2718 u32 prev_portid = cb->args[0];
2719 u32 portid = prev_portid;
2723 tbl = rht_dereference_rcu((&tipc_sk_rht)->tbl, &tipc_sk_rht);
2724 for (i = 0; i < tbl->size; i++) {
2725 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2726 spin_lock_bh(&tsk->sk.sk_lock.slock);
2727 portid = tsk->portid;
2728 err = __tipc_nl_add_sk(skb, cb, tsk);
2729 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2733 prev_portid = portid;
2738 cb->args[0] = prev_portid;
2743 /* Caller should hold socket lock for the passed tipc socket. */
2744 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2745 struct netlink_callback *cb,
2746 struct publication *publ)
2749 struct nlattr *attrs;
2751 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2752 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2756 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2758 goto genlmsg_cancel;
2760 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2761 goto attr_msg_cancel;
2762 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2763 goto attr_msg_cancel;
2764 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2765 goto attr_msg_cancel;
2766 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2767 goto attr_msg_cancel;
2769 nla_nest_end(skb, attrs);
2770 genlmsg_end(skb, hdr);
2775 nla_nest_cancel(skb, attrs);
2777 genlmsg_cancel(skb, hdr);
2782 /* Caller should hold socket lock for the passed tipc socket. */
2783 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2784 struct netlink_callback *cb,
2785 struct tipc_sock *tsk, u32 *last_publ)
2788 struct publication *p;
2791 list_for_each_entry(p, &tsk->publications, pport_list) {
2792 if (p->key == *last_publ)
2795 if (p->key != *last_publ) {
2796 /* We never set seq or call nl_dump_check_consistent()
2797 * this means that setting prev_seq here will cause the
2798 * consistence check to fail in the netlink callback
2799 * handler. Resulting in the last NLMSG_DONE message
2800 * having the NLM_F_DUMP_INTR flag set.
2807 p = list_first_entry(&tsk->publications, struct publication,
2811 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2812 err = __tipc_nl_add_sk_publ(skb, cb, p);
2814 *last_publ = p->key;
2823 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2826 u32 tsk_portid = cb->args[0];
2827 u32 last_publ = cb->args[1];
2828 u32 done = cb->args[2];
2829 struct tipc_sock *tsk;
2832 struct nlattr **attrs;
2833 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2835 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2839 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2840 attrs[TIPC_NLA_SOCK],
2841 tipc_nl_sock_policy);
2845 if (!sock[TIPC_NLA_SOCK_REF])
2848 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2854 tsk = tipc_sk_lookup(tsk_portid);
2858 lock_sock(&tsk->sk);
2859 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2862 release_sock(&tsk->sk);
2865 cb->args[0] = tsk_portid;
2866 cb->args[1] = last_publ;