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 data);
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(struct net *net, 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 static u32 tsk_peer_node(struct tipc_sock *tsk)
184 return msg_destnode(&tsk->phdr);
187 static u32 tsk_peer_port(struct tipc_sock *tsk)
189 return msg_destport(&tsk->phdr);
192 static bool tsk_unreliable(struct tipc_sock *tsk)
194 return msg_src_droppable(&tsk->phdr) != 0;
197 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
199 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
202 static bool tsk_unreturnable(struct tipc_sock *tsk)
204 return msg_dest_droppable(&tsk->phdr) != 0;
207 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
209 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
212 static int tsk_importance(struct tipc_sock *tsk)
214 return msg_importance(&tsk->phdr);
217 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
219 if (imp > TIPC_CRITICAL_IMPORTANCE)
221 msg_set_importance(&tsk->phdr, (u32)imp);
225 static struct tipc_sock *tipc_sk(const struct sock *sk)
227 return container_of(sk, struct tipc_sock, sk);
230 static int tsk_conn_cong(struct tipc_sock *tsk)
232 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
236 * tsk_advance_rx_queue - discard first buffer in socket receive queue
238 * Caller must hold socket lock
240 static void tsk_advance_rx_queue(struct sock *sk)
242 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
246 * tsk_rej_rx_queue - reject all buffers in socket receive queue
248 * Caller must hold socket lock
250 static void tsk_rej_rx_queue(struct sock *sk)
255 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
256 if (tipc_msg_reverse(skb, &dnode, TIPC_ERR_NO_PORT))
257 tipc_link_xmit_skb(sock_net(sk), skb, dnode, 0);
261 /* tsk_peer_msg - verify if message was sent by connected port's peer
263 * Handles cases where the node's network address has changed from
264 * the default of <0.0.0> to its configured setting.
266 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
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 == tipc_own_addr))
287 if (!peer_node && (orig_node == tipc_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(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(&tsk->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(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(&init_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;
473 struct net *net = sock_net(sk);
474 struct tipc_sock *tsk;
476 u32 dnode, probing_state;
479 * Exit if socket isn't fully initialized (occurs when a failed accept()
480 * releases a pre-allocated child socket that was never used)
489 * Reject all unreceived messages, except on an active connection
490 * (which disconnects locally & sends a 'FIN+' to peer)
492 dnode = tsk_peer_node(tsk);
493 while (sock->state != SS_DISCONNECTING) {
494 skb = __skb_dequeue(&sk->sk_receive_queue);
497 if (TIPC_SKB_CB(skb)->handle != NULL)
500 if ((sock->state == SS_CONNECTING) ||
501 (sock->state == SS_CONNECTED)) {
502 sock->state = SS_DISCONNECTING;
504 tipc_node_remove_conn(net, dnode, tsk->portid);
506 if (tipc_msg_reverse(skb, &dnode, TIPC_ERR_NO_PORT))
507 tipc_link_xmit_skb(net, skb, dnode, 0);
511 tipc_sk_withdraw(tsk, 0, NULL);
512 probing_state = tsk->probing_state;
513 if (del_timer_sync(&tsk->timer) && probing_state != TIPC_CONN_PROBING)
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(net, skb, dnode, tsk->portid);
523 tipc_node_remove_conn(net, 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 net *net = sock_net(sk);
727 struct tipc_msg *mhdr = &tipc_sk(sk)->phdr;
728 struct sk_buff_head head;
732 msg_set_type(mhdr, TIPC_MCAST_MSG);
733 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
734 msg_set_destport(mhdr, 0);
735 msg_set_destnode(mhdr, 0);
736 msg_set_nametype(mhdr, seq->type);
737 msg_set_namelower(mhdr, seq->lower);
738 msg_set_nameupper(mhdr, seq->upper);
739 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
742 mtu = tipc_bclink_get_mtu();
743 __skb_queue_head_init(&head);
744 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &head);
745 if (unlikely(rc < 0))
749 rc = tipc_bclink_xmit(net, &head);
750 if (likely(rc >= 0)) {
756 if (rc != -ELINKCONG)
758 tipc_sk(sk)->link_cong = 1;
759 rc = tipc_wait_for_sndmsg(sock, &timeo);
761 __skb_queue_purge(&head);
766 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
768 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff *buf)
770 struct tipc_msg *msg = buf_msg(buf);
771 struct tipc_port_list dports = {0, NULL, };
772 struct tipc_port_list *item;
774 uint i, last, dst = 0;
775 u32 scope = TIPC_CLUSTER_SCOPE;
777 if (in_own_node(msg_orignode(msg)))
778 scope = TIPC_NODE_SCOPE;
780 /* Create destination port list: */
781 tipc_nametbl_mc_translate(msg_nametype(msg),
792 for (item = &dports; item; item = item->next) {
793 for (i = 0; i < PLSIZE && ++dst <= last; i++) {
794 b = (dst != last) ? skb_clone(buf, GFP_ATOMIC) : buf;
796 pr_warn("Failed do clone mcast rcv buffer\n");
799 msg_set_destport(msg, item->ports[i]);
803 tipc_port_list_free(&dports);
807 * tipc_sk_proto_rcv - receive a connection mng protocol message
808 * @tsk: receiving socket
809 * @dnode: node to send response message to, if any
810 * @buf: buffer containing protocol message
811 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
812 * (CONN_PROBE_REPLY) message should be forwarded.
814 static int tipc_sk_proto_rcv(struct tipc_sock *tsk, u32 *dnode,
817 struct tipc_msg *msg = buf_msg(buf);
820 /* Ignore if connection cannot be validated: */
821 if (!tsk_peer_msg(tsk, msg))
824 tsk->probing_state = TIPC_CONN_OK;
826 if (msg_type(msg) == CONN_ACK) {
827 conn_cong = tsk_conn_cong(tsk);
828 tsk->sent_unacked -= msg_msgcnt(msg);
830 tsk->sk.sk_write_space(&tsk->sk);
831 } else if (msg_type(msg) == CONN_PROBE) {
832 if (!tipc_msg_reverse(buf, dnode, TIPC_OK))
834 msg_set_type(msg, CONN_PROBE_REPLY);
837 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
843 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
845 struct sock *sk = sock->sk;
846 struct tipc_sock *tsk = tipc_sk(sk);
851 int err = sock_error(sk);
854 if (sock->state == SS_DISCONNECTING)
858 if (signal_pending(current))
859 return sock_intr_errno(*timeo_p);
861 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
862 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
863 finish_wait(sk_sleep(sk), &wait);
869 * tipc_sendmsg - send message in connectionless manner
870 * @iocb: if NULL, indicates that socket lock is already held
871 * @sock: socket structure
872 * @m: message to send
873 * @dsz: amount of user data to be sent
875 * Message must have an destination specified explicitly.
876 * Used for SOCK_RDM and SOCK_DGRAM messages,
877 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
878 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
880 * Returns the number of bytes sent on success, or errno otherwise
882 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
883 struct msghdr *m, size_t dsz)
885 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
886 struct sock *sk = sock->sk;
887 struct tipc_sock *tsk = tipc_sk(sk);
888 struct net *net = sock_net(sk);
889 struct tipc_msg *mhdr = &tsk->phdr;
891 struct sk_buff_head head;
893 struct tipc_name_seq *seq = &dest->addr.nameseq;
899 return -EDESTADDRREQ;
901 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
902 (dest->family != AF_TIPC)))
905 if (dsz > TIPC_MAX_USER_MSG_SIZE)
911 if (unlikely(sock->state != SS_READY)) {
912 if (sock->state == SS_LISTENING) {
916 if (sock->state != SS_UNCONNECTED) {
920 if (tsk->published) {
924 if (dest->addrtype == TIPC_ADDR_NAME) {
925 tsk->conn_type = dest->addr.name.name.type;
926 tsk->conn_instance = dest->addr.name.name.instance;
930 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
932 if (dest->addrtype == TIPC_ADDR_MCAST) {
933 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
935 } else if (dest->addrtype == TIPC_ADDR_NAME) {
936 u32 type = dest->addr.name.name.type;
937 u32 inst = dest->addr.name.name.instance;
938 u32 domain = dest->addr.name.domain;
941 msg_set_type(mhdr, TIPC_NAMED_MSG);
942 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
943 msg_set_nametype(mhdr, type);
944 msg_set_nameinst(mhdr, inst);
945 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
946 dport = tipc_nametbl_translate(type, inst, &dnode);
947 msg_set_destnode(mhdr, dnode);
948 msg_set_destport(mhdr, dport);
949 if (unlikely(!dport && !dnode)) {
953 } else if (dest->addrtype == TIPC_ADDR_ID) {
954 dnode = dest->addr.id.node;
955 msg_set_type(mhdr, TIPC_DIRECT_MSG);
956 msg_set_lookup_scope(mhdr, 0);
957 msg_set_destnode(mhdr, dnode);
958 msg_set_destport(mhdr, dest->addr.id.ref);
959 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
963 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
964 __skb_queue_head_init(&head);
965 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &head);
970 skb = skb_peek(&head);
971 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
972 rc = tipc_link_xmit(net, &head, dnode, tsk->portid);
973 if (likely(rc >= 0)) {
974 if (sock->state != SS_READY)
975 sock->state = SS_CONNECTING;
981 if (rc != -ELINKCONG)
984 rc = tipc_wait_for_sndmsg(sock, &timeo);
986 __skb_queue_purge(&head);
995 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
997 struct sock *sk = sock->sk;
998 struct tipc_sock *tsk = tipc_sk(sk);
1003 int err = sock_error(sk);
1006 if (sock->state == SS_DISCONNECTING)
1008 else if (sock->state != SS_CONNECTED)
1012 if (signal_pending(current))
1013 return sock_intr_errno(*timeo_p);
1015 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1016 done = sk_wait_event(sk, timeo_p,
1018 !tsk_conn_cong(tsk)) ||
1020 finish_wait(sk_sleep(sk), &wait);
1026 * tipc_send_stream - send stream-oriented data
1028 * @sock: socket structure
1030 * @dsz: total length of data to be transmitted
1032 * Used for SOCK_STREAM data.
1034 * Returns the number of bytes sent on success (or partial success),
1035 * or errno if no data sent
1037 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1038 struct msghdr *m, size_t dsz)
1040 struct sock *sk = sock->sk;
1041 struct net *net = sock_net(sk);
1042 struct tipc_sock *tsk = tipc_sk(sk);
1043 struct tipc_msg *mhdr = &tsk->phdr;
1044 struct sk_buff_head head;
1045 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1046 u32 portid = tsk->portid;
1050 uint mtu, send, sent = 0;
1052 /* Handle implied connection establishment */
1053 if (unlikely(dest)) {
1054 rc = tipc_sendmsg(iocb, sock, m, dsz);
1055 if (dsz && (dsz == rc))
1056 tsk->sent_unacked = 1;
1059 if (dsz > (uint)INT_MAX)
1065 if (unlikely(sock->state != SS_CONNECTED)) {
1066 if (sock->state == SS_DISCONNECTING)
1073 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1074 dnode = tsk_peer_node(tsk);
1078 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1079 __skb_queue_head_init(&head);
1080 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &head);
1081 if (unlikely(rc < 0))
1084 if (likely(!tsk_conn_cong(tsk))) {
1085 rc = tipc_link_xmit(net, &head, dnode, portid);
1087 tsk->sent_unacked++;
1093 if (rc == -EMSGSIZE) {
1094 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1098 if (rc != -ELINKCONG)
1102 rc = tipc_wait_for_sndpkt(sock, &timeo);
1104 __skb_queue_purge(&head);
1109 return sent ? sent : rc;
1113 * tipc_send_packet - send a connection-oriented message
1114 * @iocb: if NULL, indicates that socket lock is already held
1115 * @sock: socket structure
1116 * @m: message to send
1117 * @dsz: length of data to be transmitted
1119 * Used for SOCK_SEQPACKET messages.
1121 * Returns the number of bytes sent on success, or errno otherwise
1123 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1124 struct msghdr *m, size_t dsz)
1126 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1129 return tipc_send_stream(iocb, sock, m, dsz);
1132 /* tipc_sk_finish_conn - complete the setup of a connection
1134 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1137 struct net *net = sock_net(&tsk->sk);
1138 struct tipc_msg *msg = &tsk->phdr;
1140 msg_set_destnode(msg, peer_node);
1141 msg_set_destport(msg, peer_port);
1142 msg_set_type(msg, TIPC_CONN_MSG);
1143 msg_set_lookup_scope(msg, 0);
1144 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1146 tsk->probing_intv = CONN_PROBING_INTERVAL;
1147 tsk->probing_state = TIPC_CONN_OK;
1149 if (!mod_timer(&tsk->timer, jiffies + tsk->probing_intv))
1150 sock_hold(&tsk->sk);
1151 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1152 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1156 * set_orig_addr - capture sender's address for received message
1157 * @m: descriptor for message info
1158 * @msg: received message header
1160 * Note: Address is not captured if not requested by receiver.
1162 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1164 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1167 addr->family = AF_TIPC;
1168 addr->addrtype = TIPC_ADDR_ID;
1169 memset(&addr->addr, 0, sizeof(addr->addr));
1170 addr->addr.id.ref = msg_origport(msg);
1171 addr->addr.id.node = msg_orignode(msg);
1172 addr->addr.name.domain = 0; /* could leave uninitialized */
1173 addr->scope = 0; /* could leave uninitialized */
1174 m->msg_namelen = sizeof(struct sockaddr_tipc);
1179 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1180 * @m: descriptor for message info
1181 * @msg: received message header
1182 * @tsk: TIPC port associated with message
1184 * Note: Ancillary data is not captured if not requested by receiver.
1186 * Returns 0 if successful, otherwise errno
1188 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1189 struct tipc_sock *tsk)
1197 if (likely(m->msg_controllen == 0))
1200 /* Optionally capture errored message object(s) */
1201 err = msg ? msg_errcode(msg) : 0;
1202 if (unlikely(err)) {
1204 anc_data[1] = msg_data_sz(msg);
1205 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1209 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1216 /* Optionally capture message destination object */
1217 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1218 switch (dest_type) {
1219 case TIPC_NAMED_MSG:
1221 anc_data[0] = msg_nametype(msg);
1222 anc_data[1] = msg_namelower(msg);
1223 anc_data[2] = msg_namelower(msg);
1225 case TIPC_MCAST_MSG:
1227 anc_data[0] = msg_nametype(msg);
1228 anc_data[1] = msg_namelower(msg);
1229 anc_data[2] = msg_nameupper(msg);
1232 has_name = (tsk->conn_type != 0);
1233 anc_data[0] = tsk->conn_type;
1234 anc_data[1] = tsk->conn_instance;
1235 anc_data[2] = tsk->conn_instance;
1241 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1249 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1251 struct net *net = sock_net(&tsk->sk);
1252 struct sk_buff *skb = NULL;
1253 struct tipc_msg *msg;
1254 u32 peer_port = tsk_peer_port(tsk);
1255 u32 dnode = tsk_peer_node(tsk);
1257 if (!tsk->connected)
1259 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0, dnode,
1260 tipc_own_addr, peer_port, tsk->portid, TIPC_OK);
1264 msg_set_msgcnt(msg, ack);
1265 tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1268 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1270 struct sock *sk = sock->sk;
1272 long timeo = *timeop;
1276 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1277 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1278 if (sock->state == SS_DISCONNECTING) {
1283 timeo = schedule_timeout(timeo);
1287 if (!skb_queue_empty(&sk->sk_receive_queue))
1289 err = sock_intr_errno(timeo);
1290 if (signal_pending(current))
1296 finish_wait(sk_sleep(sk), &wait);
1302 * tipc_recvmsg - receive packet-oriented message
1304 * @m: descriptor for message info
1305 * @buf_len: total size of user buffer area
1306 * @flags: receive flags
1308 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1309 * If the complete message doesn't fit in user area, truncate it.
1311 * Returns size of returned message data, errno otherwise
1313 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1314 struct msghdr *m, size_t buf_len, int flags)
1316 struct sock *sk = sock->sk;
1317 struct tipc_sock *tsk = tipc_sk(sk);
1318 struct sk_buff *buf;
1319 struct tipc_msg *msg;
1325 /* Catch invalid receive requests */
1326 if (unlikely(!buf_len))
1331 if (unlikely(sock->state == SS_UNCONNECTED)) {
1336 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1339 /* Look for a message in receive queue; wait if necessary */
1340 res = tipc_wait_for_rcvmsg(sock, &timeo);
1344 /* Look at first message in receive queue */
1345 buf = skb_peek(&sk->sk_receive_queue);
1347 sz = msg_data_sz(msg);
1348 err = msg_errcode(msg);
1350 /* Discard an empty non-errored message & try again */
1351 if ((!sz) && (!err)) {
1352 tsk_advance_rx_queue(sk);
1356 /* Capture sender's address (optional) */
1357 set_orig_addr(m, msg);
1359 /* Capture ancillary data (optional) */
1360 res = tipc_sk_anc_data_recv(m, msg, tsk);
1364 /* Capture message data (if valid) & compute return value (always) */
1366 if (unlikely(buf_len < sz)) {
1368 m->msg_flags |= MSG_TRUNC;
1370 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1375 if ((sock->state == SS_READY) ||
1376 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1382 /* Consume received message (optional) */
1383 if (likely(!(flags & MSG_PEEK))) {
1384 if ((sock->state != SS_READY) &&
1385 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1386 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1387 tsk->rcv_unacked = 0;
1389 tsk_advance_rx_queue(sk);
1397 * tipc_recv_stream - receive stream-oriented data
1399 * @m: descriptor for message info
1400 * @buf_len: total size of user buffer area
1401 * @flags: receive flags
1403 * Used for SOCK_STREAM messages only. If not enough data is available
1404 * will optionally wait for more; never truncates data.
1406 * Returns size of returned message data, errno otherwise
1408 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1409 struct msghdr *m, size_t buf_len, int flags)
1411 struct sock *sk = sock->sk;
1412 struct tipc_sock *tsk = tipc_sk(sk);
1413 struct sk_buff *buf;
1414 struct tipc_msg *msg;
1417 int sz_to_copy, target, needed;
1422 /* Catch invalid receive attempts */
1423 if (unlikely(!buf_len))
1428 if (unlikely(sock->state == SS_UNCONNECTED)) {
1433 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1434 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1437 /* Look for a message in receive queue; wait if necessary */
1438 res = tipc_wait_for_rcvmsg(sock, &timeo);
1442 /* Look at first message in receive queue */
1443 buf = skb_peek(&sk->sk_receive_queue);
1445 sz = msg_data_sz(msg);
1446 err = msg_errcode(msg);
1448 /* Discard an empty non-errored message & try again */
1449 if ((!sz) && (!err)) {
1450 tsk_advance_rx_queue(sk);
1454 /* Optionally capture sender's address & ancillary data of first msg */
1455 if (sz_copied == 0) {
1456 set_orig_addr(m, msg);
1457 res = tipc_sk_anc_data_recv(m, msg, tsk);
1462 /* Capture message data (if valid) & compute return value (always) */
1464 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1467 needed = (buf_len - sz_copied);
1468 sz_to_copy = (sz <= needed) ? sz : needed;
1470 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1475 sz_copied += sz_to_copy;
1477 if (sz_to_copy < sz) {
1478 if (!(flags & MSG_PEEK))
1479 TIPC_SKB_CB(buf)->handle =
1480 (void *)(unsigned long)(offset + sz_to_copy);
1485 goto exit; /* can't add error msg to valid data */
1487 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1493 /* Consume received message (optional) */
1494 if (likely(!(flags & MSG_PEEK))) {
1495 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1496 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1497 tsk->rcv_unacked = 0;
1499 tsk_advance_rx_queue(sk);
1502 /* Loop around if more data is required */
1503 if ((sz_copied < buf_len) && /* didn't get all requested data */
1504 (!skb_queue_empty(&sk->sk_receive_queue) ||
1505 (sz_copied < target)) && /* and more is ready or required */
1506 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1507 (!err)) /* and haven't reached a FIN */
1512 return sz_copied ? sz_copied : res;
1516 * tipc_write_space - wake up thread if port congestion is released
1519 static void tipc_write_space(struct sock *sk)
1521 struct socket_wq *wq;
1524 wq = rcu_dereference(sk->sk_wq);
1525 if (wq_has_sleeper(wq))
1526 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1527 POLLWRNORM | POLLWRBAND);
1532 * tipc_data_ready - wake up threads to indicate messages have been received
1534 * @len: the length of messages
1536 static void tipc_data_ready(struct sock *sk)
1538 struct socket_wq *wq;
1541 wq = rcu_dereference(sk->sk_wq);
1542 if (wq_has_sleeper(wq))
1543 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1544 POLLRDNORM | POLLRDBAND);
1549 * filter_connect - Handle all incoming messages for a connection-based socket
1553 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1555 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1557 struct sock *sk = &tsk->sk;
1558 struct net *net = sock_net(sk);
1559 struct socket *sock = sk->sk_socket;
1560 struct tipc_msg *msg = buf_msg(*buf);
1561 int retval = -TIPC_ERR_NO_PORT;
1566 switch ((int)sock->state) {
1568 /* Accept only connection-based messages sent by peer */
1569 if (tsk_peer_msg(tsk, msg)) {
1570 if (unlikely(msg_errcode(msg))) {
1571 sock->state = SS_DISCONNECTING;
1573 /* let timer expire on it's own */
1574 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1581 /* Accept only ACK or NACK message */
1583 if (unlikely(!msg_connected(msg)))
1586 if (unlikely(msg_errcode(msg))) {
1587 sock->state = SS_DISCONNECTING;
1588 sk->sk_err = ECONNREFUSED;
1593 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1594 sock->state = SS_DISCONNECTING;
1595 sk->sk_err = EINVAL;
1600 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1601 msg_set_importance(&tsk->phdr, msg_importance(msg));
1602 sock->state = SS_CONNECTED;
1604 /* If an incoming message is an 'ACK-', it should be
1605 * discarded here because it doesn't contain useful
1606 * data. In addition, we should try to wake up
1607 * connect() routine if sleeping.
1609 if (msg_data_sz(msg) == 0) {
1612 if (waitqueue_active(sk_sleep(sk)))
1613 wake_up_interruptible(sk_sleep(sk));
1618 case SS_UNCONNECTED:
1619 /* Accept only SYN message */
1620 if (!msg_connected(msg) && !(msg_errcode(msg)))
1623 case SS_DISCONNECTING:
1626 pr_err("Unknown socket state %u\n", sock->state);
1632 * rcvbuf_limit - get proper overload limit of socket receive queue
1636 * For all connection oriented messages, irrespective of importance,
1637 * the default overload value (i.e. 67MB) is set as limit.
1639 * For all connectionless messages, by default new queue limits are
1642 * TIPC_LOW_IMPORTANCE (4 MB)
1643 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1644 * TIPC_HIGH_IMPORTANCE (16 MB)
1645 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1647 * Returns overload limit according to corresponding message importance
1649 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1651 struct tipc_msg *msg = buf_msg(buf);
1653 if (msg_connected(msg))
1654 return sysctl_tipc_rmem[2];
1656 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1657 msg_importance(msg);
1661 * filter_rcv - validate incoming message
1665 * Enqueues message on receive queue if acceptable; optionally handles
1666 * disconnect indication for a connected socket.
1668 * Called with socket lock already taken; port lock may also be taken.
1670 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1671 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1673 static int filter_rcv(struct sock *sk, struct sk_buff *buf)
1675 struct socket *sock = sk->sk_socket;
1676 struct tipc_sock *tsk = tipc_sk(sk);
1677 struct tipc_msg *msg = buf_msg(buf);
1678 unsigned int limit = rcvbuf_limit(sk, buf);
1682 if (unlikely(msg_user(msg) == CONN_MANAGER))
1683 return tipc_sk_proto_rcv(tsk, &onode, buf);
1685 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1688 sk->sk_write_space(sk);
1692 /* Reject message if it is wrong sort of message for socket */
1693 if (msg_type(msg) > TIPC_DIRECT_MSG)
1694 return -TIPC_ERR_NO_PORT;
1696 if (sock->state == SS_READY) {
1697 if (msg_connected(msg))
1698 return -TIPC_ERR_NO_PORT;
1700 rc = filter_connect(tsk, &buf);
1701 if (rc != TIPC_OK || buf == NULL)
1705 /* Reject message if there isn't room to queue it */
1706 if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1707 return -TIPC_ERR_OVERLOAD;
1709 /* Enqueue message */
1710 TIPC_SKB_CB(buf)->handle = NULL;
1711 __skb_queue_tail(&sk->sk_receive_queue, buf);
1712 skb_set_owner_r(buf, sk);
1714 sk->sk_data_ready(sk);
1719 * tipc_backlog_rcv - handle incoming message from backlog queue
1723 * Caller must hold socket lock, but not port lock.
1727 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1731 struct tipc_sock *tsk = tipc_sk(sk);
1732 uint truesize = skb->truesize;
1734 rc = filter_rcv(sk, skb);
1737 if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1738 atomic_add(truesize, &tsk->dupl_rcvcnt);
1742 if ((rc < 0) && !tipc_msg_reverse(skb, &onode, -rc))
1745 tipc_link_xmit_skb(sock_net(sk), skb, onode, 0);
1751 * tipc_sk_rcv - handle incoming message
1752 * @skb: buffer containing arriving message
1754 * Returns 0 if success, or errno: -EHOSTUNREACH
1756 int tipc_sk_rcv(struct net *net, struct sk_buff *skb)
1758 struct tipc_sock *tsk;
1760 u32 dport = msg_destport(buf_msg(skb));
1765 /* Validate destination and message */
1766 tsk = tipc_sk_lookup(net, dport);
1767 if (unlikely(!tsk)) {
1768 rc = tipc_msg_eval(skb, &dnode);
1774 spin_lock_bh(&sk->sk_lock.slock);
1776 if (!sock_owned_by_user(sk)) {
1777 rc = filter_rcv(sk, skb);
1779 if (sk->sk_backlog.len == 0)
1780 atomic_set(&tsk->dupl_rcvcnt, 0);
1781 limit = rcvbuf_limit(sk, skb) + atomic_read(&tsk->dupl_rcvcnt);
1782 if (sk_add_backlog(sk, skb, limit))
1783 rc = -TIPC_ERR_OVERLOAD;
1785 spin_unlock_bh(&sk->sk_lock.slock);
1790 if ((rc < 0) && !tipc_msg_reverse(skb, &dnode, -rc))
1791 return -EHOSTUNREACH;
1793 tipc_link_xmit_skb(net, skb, dnode, 0);
1794 return (rc < 0) ? -EHOSTUNREACH : 0;
1797 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1799 struct sock *sk = sock->sk;
1804 int err = sock_error(sk);
1809 if (signal_pending(current))
1810 return sock_intr_errno(*timeo_p);
1812 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1813 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1814 finish_wait(sk_sleep(sk), &wait);
1820 * tipc_connect - establish a connection to another TIPC port
1821 * @sock: socket structure
1822 * @dest: socket address for destination port
1823 * @destlen: size of socket address data structure
1824 * @flags: file-related flags associated with socket
1826 * Returns 0 on success, errno otherwise
1828 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1829 int destlen, int flags)
1831 struct sock *sk = sock->sk;
1832 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1833 struct msghdr m = {NULL,};
1834 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1835 socket_state previous;
1840 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1841 if (sock->state == SS_READY) {
1847 * Reject connection attempt using multicast address
1849 * Note: send_msg() validates the rest of the address fields,
1850 * so there's no need to do it here
1852 if (dst->addrtype == TIPC_ADDR_MCAST) {
1857 previous = sock->state;
1858 switch (sock->state) {
1859 case SS_UNCONNECTED:
1860 /* Send a 'SYN-' to destination */
1862 m.msg_namelen = destlen;
1864 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1865 * indicate send_msg() is never blocked.
1868 m.msg_flags = MSG_DONTWAIT;
1870 res = tipc_sendmsg(NULL, sock, &m, 0);
1871 if ((res < 0) && (res != -EWOULDBLOCK))
1874 /* Just entered SS_CONNECTING state; the only
1875 * difference is that return value in non-blocking
1876 * case is EINPROGRESS, rather than EALREADY.
1880 if (previous == SS_CONNECTING)
1884 timeout = msecs_to_jiffies(timeout);
1885 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1886 res = tipc_wait_for_connect(sock, &timeout);
1901 * tipc_listen - allow socket to listen for incoming connections
1902 * @sock: socket structure
1905 * Returns 0 on success, errno otherwise
1907 static int tipc_listen(struct socket *sock, int len)
1909 struct sock *sk = sock->sk;
1914 if (sock->state != SS_UNCONNECTED)
1917 sock->state = SS_LISTENING;
1925 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1927 struct sock *sk = sock->sk;
1931 /* True wake-one mechanism for incoming connections: only
1932 * one process gets woken up, not the 'whole herd'.
1933 * Since we do not 'race & poll' for established sockets
1934 * anymore, the common case will execute the loop only once.
1937 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1938 TASK_INTERRUPTIBLE);
1939 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1941 timeo = schedule_timeout(timeo);
1945 if (!skb_queue_empty(&sk->sk_receive_queue))
1948 if (sock->state != SS_LISTENING)
1950 err = sock_intr_errno(timeo);
1951 if (signal_pending(current))
1957 finish_wait(sk_sleep(sk), &wait);
1962 * tipc_accept - wait for connection request
1963 * @sock: listening socket
1964 * @newsock: new socket that is to be connected
1965 * @flags: file-related flags associated with socket
1967 * Returns 0 on success, errno otherwise
1969 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1971 struct sock *new_sk, *sk = sock->sk;
1972 struct sk_buff *buf;
1973 struct tipc_sock *new_tsock;
1974 struct tipc_msg *msg;
1980 if (sock->state != SS_LISTENING) {
1984 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1985 res = tipc_wait_for_accept(sock, timeo);
1989 buf = skb_peek(&sk->sk_receive_queue);
1991 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
1995 new_sk = new_sock->sk;
1996 new_tsock = tipc_sk(new_sk);
1999 /* we lock on new_sk; but lockdep sees the lock on sk */
2000 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2003 * Reject any stray messages received by new socket
2004 * before the socket lock was taken (very, very unlikely)
2006 tsk_rej_rx_queue(new_sk);
2008 /* Connect new socket to it's peer */
2009 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2010 new_sock->state = SS_CONNECTED;
2012 tsk_set_importance(new_tsock, msg_importance(msg));
2013 if (msg_named(msg)) {
2014 new_tsock->conn_type = msg_nametype(msg);
2015 new_tsock->conn_instance = msg_nameinst(msg);
2019 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2020 * Respond to 'SYN+' by queuing it on new socket.
2022 if (!msg_data_sz(msg)) {
2023 struct msghdr m = {NULL,};
2025 tsk_advance_rx_queue(sk);
2026 tipc_send_packet(NULL, new_sock, &m, 0);
2028 __skb_dequeue(&sk->sk_receive_queue);
2029 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2030 skb_set_owner_r(buf, new_sk);
2032 release_sock(new_sk);
2039 * tipc_shutdown - shutdown socket connection
2040 * @sock: socket structure
2041 * @how: direction to close (must be SHUT_RDWR)
2043 * Terminates connection (if necessary), then purges socket's receive queue.
2045 * Returns 0 on success, errno otherwise
2047 static int tipc_shutdown(struct socket *sock, int how)
2049 struct sock *sk = sock->sk;
2050 struct net *net = sock_net(sk);
2051 struct tipc_sock *tsk = tipc_sk(sk);
2052 struct sk_buff *skb;
2056 if (how != SHUT_RDWR)
2061 switch (sock->state) {
2066 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2067 skb = __skb_dequeue(&sk->sk_receive_queue);
2069 if (TIPC_SKB_CB(skb)->handle != NULL) {
2073 if (tipc_msg_reverse(skb, &dnode, TIPC_CONN_SHUTDOWN))
2074 tipc_link_xmit_skb(net, skb, dnode,
2076 tipc_node_remove_conn(net, dnode, tsk->portid);
2078 dnode = tsk_peer_node(tsk);
2079 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2080 TIPC_CONN_MSG, SHORT_H_SIZE,
2081 0, dnode, tipc_own_addr,
2083 tsk->portid, TIPC_CONN_SHUTDOWN);
2084 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2087 sock->state = SS_DISCONNECTING;
2088 tipc_node_remove_conn(net, dnode, tsk->portid);
2091 case SS_DISCONNECTING:
2093 /* Discard any unreceived messages */
2094 __skb_queue_purge(&sk->sk_receive_queue);
2096 /* Wake up anyone sleeping in poll */
2097 sk->sk_state_change(sk);
2109 static void tipc_sk_timeout(unsigned long data)
2111 struct tipc_sock *tsk = (struct tipc_sock *)data;
2112 struct sock *sk = &tsk->sk;
2113 struct sk_buff *skb = NULL;
2114 u32 peer_port, peer_node;
2117 if (!tsk->connected) {
2121 peer_port = tsk_peer_port(tsk);
2122 peer_node = tsk_peer_node(tsk);
2124 if (tsk->probing_state == TIPC_CONN_PROBING) {
2125 /* Previous probe not answered -> self abort */
2126 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
2127 SHORT_H_SIZE, 0, tipc_own_addr,
2128 peer_node, tsk->portid, peer_port,
2131 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE,
2132 0, peer_node, tipc_own_addr,
2133 peer_port, tsk->portid, TIPC_OK);
2134 tsk->probing_state = TIPC_CONN_PROBING;
2135 if (!mod_timer(&tsk->timer, jiffies + tsk->probing_intv))
2140 tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2145 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2146 struct tipc_name_seq const *seq)
2148 struct net *net = sock_net(&tsk->sk);
2149 struct publication *publ;
2154 key = tsk->portid + tsk->pub_count + 1;
2155 if (key == tsk->portid)
2158 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2159 scope, tsk->portid, key);
2160 if (unlikely(!publ))
2163 list_add(&publ->pport_list, &tsk->publications);
2169 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2170 struct tipc_name_seq const *seq)
2172 struct net *net = sock_net(&tsk->sk);
2173 struct publication *publ;
2174 struct publication *safe;
2177 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2179 if (publ->scope != scope)
2181 if (publ->type != seq->type)
2183 if (publ->lower != seq->lower)
2185 if (publ->upper != seq->upper)
2187 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2188 publ->ref, publ->key);
2192 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2193 publ->ref, publ->key);
2196 if (list_empty(&tsk->publications))
2201 static int tipc_sk_show(struct tipc_sock *tsk, char *buf,
2202 int len, int full_id)
2204 struct publication *publ;
2208 ret = tipc_snprintf(buf, len, "<%u.%u.%u:%u>:",
2209 tipc_zone(tipc_own_addr),
2210 tipc_cluster(tipc_own_addr),
2211 tipc_node(tipc_own_addr), tsk->portid);
2213 ret = tipc_snprintf(buf, len, "%-10u:", tsk->portid);
2215 if (tsk->connected) {
2216 u32 dport = tsk_peer_port(tsk);
2217 u32 destnode = tsk_peer_node(tsk);
2219 ret += tipc_snprintf(buf + ret, len - ret,
2220 " connected to <%u.%u.%u:%u>",
2221 tipc_zone(destnode),
2222 tipc_cluster(destnode),
2223 tipc_node(destnode), dport);
2224 if (tsk->conn_type != 0)
2225 ret += tipc_snprintf(buf + ret, len - ret,
2226 " via {%u,%u}", tsk->conn_type,
2227 tsk->conn_instance);
2228 } else if (tsk->published) {
2229 ret += tipc_snprintf(buf + ret, len - ret, " bound to");
2230 list_for_each_entry(publ, &tsk->publications, pport_list) {
2231 if (publ->lower == publ->upper)
2232 ret += tipc_snprintf(buf + ret, len - ret,
2233 " {%u,%u}", publ->type,
2236 ret += tipc_snprintf(buf + ret, len - ret,
2237 " {%u,%u,%u}", publ->type,
2238 publ->lower, publ->upper);
2241 ret += tipc_snprintf(buf + ret, len - ret, "\n");
2245 struct sk_buff *tipc_sk_socks_show(struct net *net)
2247 struct tipc_net *tn = net_generic(net, tipc_net_id);
2248 const struct bucket_table *tbl;
2249 struct rhash_head *pos;
2250 struct sk_buff *buf;
2251 struct tlv_desc *rep_tlv;
2254 struct tipc_sock *tsk;
2258 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2261 rep_tlv = (struct tlv_desc *)buf->data;
2262 pb = TLV_DATA(rep_tlv);
2263 pb_len = ULTRA_STRING_MAX_LEN;
2266 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2267 for (i = 0; i < tbl->size; i++) {
2268 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2269 spin_lock_bh(&tsk->sk.sk_lock.slock);
2270 str_len += tipc_sk_show(tsk, pb + str_len,
2271 pb_len - str_len, 0);
2272 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2277 str_len += 1; /* for "\0" */
2278 skb_put(buf, TLV_SPACE(str_len));
2279 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2284 /* tipc_sk_reinit: set non-zero address in all existing sockets
2285 * when we go from standalone to network mode.
2287 void tipc_sk_reinit(struct net *net)
2289 struct tipc_net *tn = net_generic(net, tipc_net_id);
2290 const struct bucket_table *tbl;
2291 struct rhash_head *pos;
2292 struct tipc_sock *tsk;
2293 struct tipc_msg *msg;
2297 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2298 for (i = 0; i < tbl->size; i++) {
2299 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2300 spin_lock_bh(&tsk->sk.sk_lock.slock);
2302 msg_set_prevnode(msg, tipc_own_addr);
2303 msg_set_orignode(msg, tipc_own_addr);
2304 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2310 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2312 struct tipc_net *tn = net_generic(net, tipc_net_id);
2313 struct tipc_sock *tsk;
2316 tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2318 sock_hold(&tsk->sk);
2324 static int tipc_sk_insert(struct tipc_sock *tsk)
2326 struct sock *sk = &tsk->sk;
2327 struct net *net = sock_net(sk);
2328 struct tipc_net *tn = net_generic(net, tipc_net_id);
2329 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2330 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2332 while (remaining--) {
2334 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2335 portid = TIPC_MIN_PORT;
2336 tsk->portid = portid;
2337 sock_hold(&tsk->sk);
2338 if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2346 static void tipc_sk_remove(struct tipc_sock *tsk)
2348 struct sock *sk = &tsk->sk;
2349 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2351 if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2352 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2357 int tipc_sk_rht_init(struct net *net)
2359 struct tipc_net *tn = net_generic(net, tipc_net_id);
2360 struct rhashtable_params rht_params = {
2362 .head_offset = offsetof(struct tipc_sock, node),
2363 .key_offset = offsetof(struct tipc_sock, portid),
2364 .key_len = sizeof(u32), /* portid */
2366 .max_shift = 20, /* 1M */
2367 .min_shift = 8, /* 256 */
2368 .grow_decision = rht_grow_above_75,
2369 .shrink_decision = rht_shrink_below_30,
2372 return rhashtable_init(&tn->sk_rht, &rht_params);
2375 void tipc_sk_rht_destroy(struct net *net)
2377 struct tipc_net *tn = net_generic(net, tipc_net_id);
2379 /* Wait for socket readers to complete */
2382 rhashtable_destroy(&tn->sk_rht);
2386 * tipc_setsockopt - set socket option
2387 * @sock: socket structure
2388 * @lvl: option level
2389 * @opt: option identifier
2390 * @ov: pointer to new option value
2391 * @ol: length of option value
2393 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2394 * (to ease compatibility).
2396 * Returns 0 on success, errno otherwise
2398 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2399 char __user *ov, unsigned int ol)
2401 struct sock *sk = sock->sk;
2402 struct tipc_sock *tsk = tipc_sk(sk);
2406 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2408 if (lvl != SOL_TIPC)
2409 return -ENOPROTOOPT;
2410 if (ol < sizeof(value))
2412 res = get_user(value, (u32 __user *)ov);
2419 case TIPC_IMPORTANCE:
2420 res = tsk_set_importance(tsk, value);
2422 case TIPC_SRC_DROPPABLE:
2423 if (sock->type != SOCK_STREAM)
2424 tsk_set_unreliable(tsk, value);
2428 case TIPC_DEST_DROPPABLE:
2429 tsk_set_unreturnable(tsk, value);
2431 case TIPC_CONN_TIMEOUT:
2432 tipc_sk(sk)->conn_timeout = value;
2433 /* no need to set "res", since already 0 at this point */
2445 * tipc_getsockopt - get socket option
2446 * @sock: socket structure
2447 * @lvl: option level
2448 * @opt: option identifier
2449 * @ov: receptacle for option value
2450 * @ol: receptacle for length of option value
2452 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2453 * (to ease compatibility).
2455 * Returns 0 on success, errno otherwise
2457 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2458 char __user *ov, int __user *ol)
2460 struct sock *sk = sock->sk;
2461 struct tipc_sock *tsk = tipc_sk(sk);
2466 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2467 return put_user(0, ol);
2468 if (lvl != SOL_TIPC)
2469 return -ENOPROTOOPT;
2470 res = get_user(len, ol);
2477 case TIPC_IMPORTANCE:
2478 value = tsk_importance(tsk);
2480 case TIPC_SRC_DROPPABLE:
2481 value = tsk_unreliable(tsk);
2483 case TIPC_DEST_DROPPABLE:
2484 value = tsk_unreturnable(tsk);
2486 case TIPC_CONN_TIMEOUT:
2487 value = tsk->conn_timeout;
2488 /* no need to set "res", since already 0 at this point */
2490 case TIPC_NODE_RECVQ_DEPTH:
2491 value = 0; /* was tipc_queue_size, now obsolete */
2493 case TIPC_SOCK_RECVQ_DEPTH:
2494 value = skb_queue_len(&sk->sk_receive_queue);
2503 return res; /* "get" failed */
2505 if (len < sizeof(value))
2508 if (copy_to_user(ov, &value, sizeof(value)))
2511 return put_user(sizeof(value), ol);
2514 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2516 struct sock *sk = sock->sk;
2517 struct tipc_sioc_ln_req lnr;
2518 void __user *argp = (void __user *)arg;
2521 case SIOCGETLINKNAME:
2522 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2524 if (!tipc_node_get_linkname(sock_net(sk),
2525 lnr.bearer_id & 0xffff, lnr.peer,
2526 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2527 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2531 return -EADDRNOTAVAIL;
2533 return -ENOIOCTLCMD;
2537 /* Protocol switches for the various types of TIPC sockets */
2539 static const struct proto_ops msg_ops = {
2540 .owner = THIS_MODULE,
2542 .release = tipc_release,
2544 .connect = tipc_connect,
2545 .socketpair = sock_no_socketpair,
2546 .accept = sock_no_accept,
2547 .getname = tipc_getname,
2549 .ioctl = tipc_ioctl,
2550 .listen = sock_no_listen,
2551 .shutdown = tipc_shutdown,
2552 .setsockopt = tipc_setsockopt,
2553 .getsockopt = tipc_getsockopt,
2554 .sendmsg = tipc_sendmsg,
2555 .recvmsg = tipc_recvmsg,
2556 .mmap = sock_no_mmap,
2557 .sendpage = sock_no_sendpage
2560 static const struct proto_ops packet_ops = {
2561 .owner = THIS_MODULE,
2563 .release = tipc_release,
2565 .connect = tipc_connect,
2566 .socketpair = sock_no_socketpair,
2567 .accept = tipc_accept,
2568 .getname = tipc_getname,
2570 .ioctl = tipc_ioctl,
2571 .listen = tipc_listen,
2572 .shutdown = tipc_shutdown,
2573 .setsockopt = tipc_setsockopt,
2574 .getsockopt = tipc_getsockopt,
2575 .sendmsg = tipc_send_packet,
2576 .recvmsg = tipc_recvmsg,
2577 .mmap = sock_no_mmap,
2578 .sendpage = sock_no_sendpage
2581 static const struct proto_ops stream_ops = {
2582 .owner = THIS_MODULE,
2584 .release = tipc_release,
2586 .connect = tipc_connect,
2587 .socketpair = sock_no_socketpair,
2588 .accept = tipc_accept,
2589 .getname = tipc_getname,
2591 .ioctl = tipc_ioctl,
2592 .listen = tipc_listen,
2593 .shutdown = tipc_shutdown,
2594 .setsockopt = tipc_setsockopt,
2595 .getsockopt = tipc_getsockopt,
2596 .sendmsg = tipc_send_stream,
2597 .recvmsg = tipc_recv_stream,
2598 .mmap = sock_no_mmap,
2599 .sendpage = sock_no_sendpage
2602 static const struct net_proto_family tipc_family_ops = {
2603 .owner = THIS_MODULE,
2605 .create = tipc_sk_create
2608 static struct proto tipc_proto = {
2610 .owner = THIS_MODULE,
2611 .obj_size = sizeof(struct tipc_sock),
2612 .sysctl_rmem = sysctl_tipc_rmem
2615 static struct proto tipc_proto_kern = {
2617 .obj_size = sizeof(struct tipc_sock),
2618 .sysctl_rmem = sysctl_tipc_rmem
2622 * tipc_socket_init - initialize TIPC socket interface
2624 * Returns 0 on success, errno otherwise
2626 int tipc_socket_init(void)
2630 res = proto_register(&tipc_proto, 1);
2632 pr_err("Failed to register TIPC protocol type\n");
2636 res = sock_register(&tipc_family_ops);
2638 pr_err("Failed to register TIPC socket type\n");
2639 proto_unregister(&tipc_proto);
2647 * tipc_socket_stop - stop TIPC socket interface
2649 void tipc_socket_stop(void)
2651 sock_unregister(tipc_family_ops.family);
2652 proto_unregister(&tipc_proto);
2655 /* Caller should hold socket lock for the passed tipc socket. */
2656 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2660 struct nlattr *nest;
2662 peer_node = tsk_peer_node(tsk);
2663 peer_port = tsk_peer_port(tsk);
2665 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2667 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2669 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2672 if (tsk->conn_type != 0) {
2673 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2675 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2677 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2680 nla_nest_end(skb, nest);
2685 nla_nest_cancel(skb, nest);
2690 /* Caller should hold socket lock for the passed tipc socket. */
2691 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2692 struct tipc_sock *tsk)
2696 struct nlattr *attrs;
2698 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2699 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2703 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2705 goto genlmsg_cancel;
2706 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2707 goto attr_msg_cancel;
2708 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr))
2709 goto attr_msg_cancel;
2711 if (tsk->connected) {
2712 err = __tipc_nl_add_sk_con(skb, tsk);
2714 goto attr_msg_cancel;
2715 } else if (!list_empty(&tsk->publications)) {
2716 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2717 goto attr_msg_cancel;
2719 nla_nest_end(skb, attrs);
2720 genlmsg_end(skb, hdr);
2725 nla_nest_cancel(skb, attrs);
2727 genlmsg_cancel(skb, hdr);
2732 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2735 struct tipc_sock *tsk;
2736 const struct bucket_table *tbl;
2737 struct rhash_head *pos;
2738 u32 prev_portid = cb->args[0];
2739 u32 portid = prev_portid;
2740 struct net *net = sock_net(skb->sk);
2741 struct tipc_net *tn = net_generic(net, tipc_net_id);
2745 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2746 for (i = 0; i < tbl->size; i++) {
2747 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2748 spin_lock_bh(&tsk->sk.sk_lock.slock);
2749 portid = tsk->portid;
2750 err = __tipc_nl_add_sk(skb, cb, tsk);
2751 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2755 prev_portid = portid;
2760 cb->args[0] = prev_portid;
2765 /* Caller should hold socket lock for the passed tipc socket. */
2766 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2767 struct netlink_callback *cb,
2768 struct publication *publ)
2771 struct nlattr *attrs;
2773 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2774 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2778 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2780 goto genlmsg_cancel;
2782 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2783 goto attr_msg_cancel;
2784 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2785 goto attr_msg_cancel;
2786 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2787 goto attr_msg_cancel;
2788 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2789 goto attr_msg_cancel;
2791 nla_nest_end(skb, attrs);
2792 genlmsg_end(skb, hdr);
2797 nla_nest_cancel(skb, attrs);
2799 genlmsg_cancel(skb, hdr);
2804 /* Caller should hold socket lock for the passed tipc socket. */
2805 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2806 struct netlink_callback *cb,
2807 struct tipc_sock *tsk, u32 *last_publ)
2810 struct publication *p;
2813 list_for_each_entry(p, &tsk->publications, pport_list) {
2814 if (p->key == *last_publ)
2817 if (p->key != *last_publ) {
2818 /* We never set seq or call nl_dump_check_consistent()
2819 * this means that setting prev_seq here will cause the
2820 * consistence check to fail in the netlink callback
2821 * handler. Resulting in the last NLMSG_DONE message
2822 * having the NLM_F_DUMP_INTR flag set.
2829 p = list_first_entry(&tsk->publications, struct publication,
2833 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2834 err = __tipc_nl_add_sk_publ(skb, cb, p);
2836 *last_publ = p->key;
2845 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2848 u32 tsk_portid = cb->args[0];
2849 u32 last_publ = cb->args[1];
2850 u32 done = cb->args[2];
2851 struct net *net = sock_net(skb->sk);
2852 struct tipc_sock *tsk;
2855 struct nlattr **attrs;
2856 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2858 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2862 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2863 attrs[TIPC_NLA_SOCK],
2864 tipc_nl_sock_policy);
2868 if (!sock[TIPC_NLA_SOCK_REF])
2871 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2877 tsk = tipc_sk_lookup(net, tsk_portid);
2881 lock_sock(&tsk->sk);
2882 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2885 release_sock(&tsk->sk);
2888 cb->args[0] = tsk_portid;
2889 cb->args[1] = last_publ;