2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2014, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/rhashtable.h>
38 #include <linux/jhash.h>
40 #include "name_table.h"
46 #define SS_LISTENING -1 /* socket is listening */
47 #define SS_READY -2 /* socket is connectionless */
49 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
50 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
51 #define TIPC_FWD_MSG 1
52 #define TIPC_CONN_OK 0
53 #define TIPC_CONN_PROBING 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
58 * struct tipc_sock - TIPC socket structure
59 * @sk: socket - interacts with 'port' and with user via the socket API
60 * @connected: non-zero if port is currently connected to a peer port
61 * @conn_type: TIPC type used when connection was established
62 * @conn_instance: TIPC instance used when connection was established
63 * @published: non-zero if port has one or more associated names
64 * @max_pkt: maximum packet size "hint" used when building messages sent by port
65 * @portid: unique port identity in TIPC socket hash table
66 * @phdr: preformatted message header used when sending messages
67 * @port_list: adjacent ports in TIPC's global list of ports
68 * @publications: list of publications for port
69 * @pub_count: total # of publications port has made during its lifetime
72 * @port: port - interacts with 'sk' and with the rest of the TIPC stack
73 * @peer_name: the peer of the connection, if any
74 * @conn_timeout: the time we can wait for an unresponded setup request
75 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
76 * @link_cong: non-zero if owner must sleep because of link congestion
77 * @sent_unacked: # messages sent by socket, and not yet acked by peer
78 * @rcv_unacked: # messages read by user, but not yet acked back to peer
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
91 struct list_head sock_list;
92 struct list_head publications;
95 unsigned long probing_intv;
101 struct rhash_head node;
105 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
106 static void tipc_data_ready(struct sock *sk);
107 static void tipc_write_space(struct sock *sk);
108 static int tipc_release(struct socket *sock);
109 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
110 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
111 static void tipc_sk_timeout(unsigned long data);
112 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
113 struct tipc_name_seq const *seq);
114 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
115 struct tipc_name_seq const *seq);
116 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
117 static int tipc_sk_insert(struct tipc_sock *tsk);
118 static void tipc_sk_remove(struct tipc_sock *tsk);
120 static const struct proto_ops packet_ops;
121 static const struct proto_ops stream_ops;
122 static const struct proto_ops msg_ops;
124 static struct proto tipc_proto;
125 static struct proto tipc_proto_kern;
127 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
128 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
129 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
130 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
131 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
132 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
136 * Revised TIPC socket locking policy:
138 * Most socket operations take the standard socket lock when they start
139 * and hold it until they finish (or until they need to sleep). Acquiring
140 * this lock grants the owner exclusive access to the fields of the socket
141 * data structures, with the exception of the backlog queue. A few socket
142 * operations can be done without taking the socket lock because they only
143 * read socket information that never changes during the life of the socket.
145 * Socket operations may acquire the lock for the associated TIPC port if they
146 * need to perform an operation on the port. If any routine needs to acquire
147 * both the socket lock and the port lock it must take the socket lock first
148 * to avoid the risk of deadlock.
150 * The dispatcher handling incoming messages cannot grab the socket lock in
151 * the standard fashion, since invoked it runs at the BH level and cannot block.
152 * Instead, it checks to see if the socket lock is currently owned by someone,
153 * and either handles the message itself or adds it to the socket's backlog
154 * queue; in the latter case the queued message is processed once the process
155 * owning the socket lock releases it.
157 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
158 * the problem of a blocked socket operation preventing any other operations
159 * from occurring. However, applications must be careful if they have
160 * multiple threads trying to send (or receive) on the same socket, as these
161 * operations might interfere with each other. For example, doing a connect
162 * and a receive at the same time might allow the receive to consume the
163 * ACK message meant for the connect. While additional work could be done
164 * to try and overcome this, it doesn't seem to be worthwhile at the present.
166 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
167 * that another operation that must be performed in a non-blocking manner is
168 * not delayed for very long because the lock has already been taken.
170 * NOTE: This code assumes that certain fields of a port/socket pair are
171 * constant over its lifetime; such fields can be examined without taking
172 * the socket lock and/or port lock, and do not need to be re-read even
173 * after resuming processing after waiting. These fields include:
175 * - pointer to socket sk structure (aka tipc_sock structure)
176 * - pointer to port structure
180 static u32 tsk_peer_node(struct tipc_sock *tsk)
182 return msg_destnode(&tsk->phdr);
185 static u32 tsk_peer_port(struct tipc_sock *tsk)
187 return msg_destport(&tsk->phdr);
190 static bool tsk_unreliable(struct tipc_sock *tsk)
192 return msg_src_droppable(&tsk->phdr) != 0;
195 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
197 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
200 static bool tsk_unreturnable(struct tipc_sock *tsk)
202 return msg_dest_droppable(&tsk->phdr) != 0;
205 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
207 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
210 static int tsk_importance(struct tipc_sock *tsk)
212 return msg_importance(&tsk->phdr);
215 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
217 if (imp > TIPC_CRITICAL_IMPORTANCE)
219 msg_set_importance(&tsk->phdr, (u32)imp);
223 static struct tipc_sock *tipc_sk(const struct sock *sk)
225 return container_of(sk, struct tipc_sock, sk);
228 static int tsk_conn_cong(struct tipc_sock *tsk)
230 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
234 * tsk_advance_rx_queue - discard first buffer in socket receive queue
236 * Caller must hold socket lock
238 static void tsk_advance_rx_queue(struct sock *sk)
240 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
244 * tsk_rej_rx_queue - reject all buffers in socket receive queue
246 * Caller must hold socket lock
248 static void tsk_rej_rx_queue(struct sock *sk)
252 struct net *net = sock_net(sk);
254 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
255 if (tipc_msg_reverse(net, skb, &dnode, TIPC_ERR_NO_PORT))
256 tipc_link_xmit_skb(net, skb, dnode, 0);
260 /* tsk_peer_msg - verify if message was sent by connected port's peer
262 * Handles cases where the node's network address has changed from
263 * the default of <0.0.0> to its configured setting.
265 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
267 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
268 u32 peer_port = tsk_peer_port(tsk);
272 if (unlikely(!tsk->connected))
275 if (unlikely(msg_origport(msg) != peer_port))
278 orig_node = msg_orignode(msg);
279 peer_node = tsk_peer_node(tsk);
281 if (likely(orig_node == peer_node))
284 if (!orig_node && (peer_node == tn->own_addr))
287 if (!peer_node && (orig_node == tn->own_addr))
294 * tipc_sk_create - create a TIPC socket
295 * @net: network namespace (must be default network)
296 * @sock: pre-allocated socket structure
297 * @protocol: protocol indicator (must be 0)
298 * @kern: caused by kernel or by userspace?
300 * This routine creates additional data structures used by the TIPC socket,
301 * initializes them, and links them together.
303 * Returns 0 on success, errno otherwise
305 static int tipc_sk_create(struct net *net, struct socket *sock,
306 int protocol, int kern)
308 const struct proto_ops *ops;
311 struct tipc_sock *tsk;
312 struct tipc_msg *msg;
314 /* Validate arguments */
315 if (unlikely(protocol != 0))
316 return -EPROTONOSUPPORT;
318 switch (sock->type) {
321 state = SS_UNCONNECTED;
325 state = SS_UNCONNECTED;
336 /* Allocate socket's protocol area */
338 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
340 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
346 tsk->max_pkt = MAX_PKT_DEFAULT;
347 INIT_LIST_HEAD(&tsk->publications);
349 tipc_msg_init(net, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
352 /* Finish initializing socket data structures */
355 sock_init_data(sock, sk);
356 if (tipc_sk_insert(tsk)) {
357 pr_warn("Socket create failed; port numbrer exhausted\n");
360 msg_set_origport(msg, tsk->portid);
361 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
362 sk->sk_backlog_rcv = tipc_backlog_rcv;
363 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
364 sk->sk_data_ready = tipc_data_ready;
365 sk->sk_write_space = tipc_write_space;
366 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
367 tsk->sent_unacked = 0;
368 atomic_set(&tsk->dupl_rcvcnt, 0);
370 if (sock->state == SS_READY) {
371 tsk_set_unreturnable(tsk, true);
372 if (sock->type == SOCK_DGRAM)
373 tsk_set_unreliable(tsk, true);
379 * tipc_sock_create_local - create TIPC socket from inside TIPC module
380 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
382 * We cannot use sock_creat_kern here because it bumps module user count.
383 * Since socket owner and creator is the same module we must make sure
384 * that module count remains zero for module local sockets, otherwise
385 * we cannot do rmmod.
387 * Returns 0 on success, errno otherwise
389 int tipc_sock_create_local(struct net *net, int type, struct socket **res)
393 rc = sock_create_lite(AF_TIPC, type, 0, res);
395 pr_err("Failed to create kernel socket\n");
398 tipc_sk_create(net, *res, 0, 1);
404 * tipc_sock_release_local - release socket created by tipc_sock_create_local
405 * @sock: the socket to be released.
407 * Module reference count is not incremented when such sockets are created,
408 * so we must keep it from being decremented when they are released.
410 void tipc_sock_release_local(struct socket *sock)
418 * tipc_sock_accept_local - accept a connection on a socket created
419 * with tipc_sock_create_local. Use this function to avoid that
420 * module reference count is inadvertently incremented.
422 * @sock: the accepting socket
423 * @newsock: reference to the new socket to be created
424 * @flags: socket flags
427 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
430 struct sock *sk = sock->sk;
433 ret = sock_create_lite(sk->sk_family, sk->sk_type,
434 sk->sk_protocol, newsock);
438 ret = tipc_accept(sock, *newsock, flags);
440 sock_release(*newsock);
443 (*newsock)->ops = sock->ops;
447 static void tipc_sk_callback(struct rcu_head *head)
449 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
455 * tipc_release - destroy a TIPC socket
456 * @sock: socket to destroy
458 * This routine cleans up any messages that are still queued on the socket.
459 * For DGRAM and RDM socket types, all queued messages are rejected.
460 * For SEQPACKET and STREAM socket types, the first message is rejected
461 * and any others are discarded. (If the first message on a STREAM socket
462 * is partially-read, it is discarded and the next one is rejected instead.)
464 * NOTE: Rejected messages are not necessarily returned to the sender! They
465 * are returned or discarded according to the "destination droppable" setting
466 * specified for the message by the sender.
468 * Returns 0 on success, errno otherwise
470 static int tipc_release(struct socket *sock)
472 struct sock *sk = sock->sk;
475 struct tipc_sock *tsk;
477 u32 dnode, probing_state;
480 * Exit if socket isn't fully initialized (occurs when a failed accept()
481 * releases a pre-allocated child socket that was never used)
487 tn = net_generic(net, tipc_net_id);
493 * Reject all unreceived messages, except on an active connection
494 * (which disconnects locally & sends a 'FIN+' to peer)
496 dnode = tsk_peer_node(tsk);
497 while (sock->state != SS_DISCONNECTING) {
498 skb = __skb_dequeue(&sk->sk_receive_queue);
501 if (TIPC_SKB_CB(skb)->handle != NULL)
504 if ((sock->state == SS_CONNECTING) ||
505 (sock->state == SS_CONNECTED)) {
506 sock->state = SS_DISCONNECTING;
508 tipc_node_remove_conn(net, dnode, tsk->portid);
510 if (tipc_msg_reverse(net, skb, &dnode,
512 tipc_link_xmit_skb(net, skb, dnode, 0);
516 tipc_sk_withdraw(tsk, 0, NULL);
517 probing_state = tsk->probing_state;
518 if (del_timer_sync(&sk->sk_timer) &&
519 probing_state != TIPC_CONN_PROBING)
522 if (tsk->connected) {
523 skb = tipc_msg_create(net, TIPC_CRITICAL_IMPORTANCE,
524 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
525 tn->own_addr, tsk_peer_port(tsk),
526 tsk->portid, TIPC_ERR_NO_PORT);
528 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
529 tipc_node_remove_conn(net, dnode, tsk->portid);
532 /* Discard any remaining (connection-based) messages in receive queue */
533 __skb_queue_purge(&sk->sk_receive_queue);
535 /* Reject any messages that accumulated in backlog queue */
536 sock->state = SS_DISCONNECTING;
539 call_rcu(&tsk->rcu, tipc_sk_callback);
546 * tipc_bind - associate or disassocate TIPC name(s) with a socket
547 * @sock: socket structure
548 * @uaddr: socket address describing name(s) and desired operation
549 * @uaddr_len: size of socket address data structure
551 * Name and name sequence binding is indicated using a positive scope value;
552 * a negative scope value unbinds the specified name. Specifying no name
553 * (i.e. a socket address length of 0) unbinds all names from the socket.
555 * Returns 0 on success, errno otherwise
557 * NOTE: This routine doesn't need to take the socket lock since it doesn't
558 * access any non-constant socket information.
560 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
563 struct sock *sk = sock->sk;
564 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
565 struct tipc_sock *tsk = tipc_sk(sk);
569 if (unlikely(!uaddr_len)) {
570 res = tipc_sk_withdraw(tsk, 0, NULL);
574 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
578 if (addr->family != AF_TIPC) {
583 if (addr->addrtype == TIPC_ADDR_NAME)
584 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
585 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
590 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
591 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
592 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
597 res = (addr->scope > 0) ?
598 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
599 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
606 * tipc_getname - get port ID of socket or peer socket
607 * @sock: socket structure
608 * @uaddr: area for returned socket address
609 * @uaddr_len: area for returned length of socket address
610 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
612 * Returns 0 on success, errno otherwise
614 * NOTE: This routine doesn't need to take the socket lock since it only
615 * accesses socket information that is unchanging (or which changes in
616 * a completely predictable manner).
618 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
619 int *uaddr_len, int peer)
621 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
622 struct tipc_sock *tsk = tipc_sk(sock->sk);
623 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
625 memset(addr, 0, sizeof(*addr));
627 if ((sock->state != SS_CONNECTED) &&
628 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
630 addr->addr.id.ref = tsk_peer_port(tsk);
631 addr->addr.id.node = tsk_peer_node(tsk);
633 addr->addr.id.ref = tsk->portid;
634 addr->addr.id.node = tn->own_addr;
637 *uaddr_len = sizeof(*addr);
638 addr->addrtype = TIPC_ADDR_ID;
639 addr->family = AF_TIPC;
641 addr->addr.name.domain = 0;
647 * tipc_poll - read and possibly block on pollmask
648 * @file: file structure associated with the socket
649 * @sock: socket for which to calculate the poll bits
652 * Returns pollmask value
655 * It appears that the usual socket locking mechanisms are not useful here
656 * since the pollmask info is potentially out-of-date the moment this routine
657 * exits. TCP and other protocols seem to rely on higher level poll routines
658 * to handle any preventable race conditions, so TIPC will do the same ...
660 * TIPC sets the returned events as follows:
662 * socket state flags set
663 * ------------ ---------
664 * unconnected no read flags
665 * POLLOUT if port is not congested
667 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
670 * connected POLLIN/POLLRDNORM if data in rx queue
671 * POLLOUT if port is not congested
673 * disconnecting POLLIN/POLLRDNORM/POLLHUP
676 * listening POLLIN if SYN in rx queue
679 * ready POLLIN/POLLRDNORM if data in rx queue
680 * [connectionless] POLLOUT (since port cannot be congested)
682 * IMPORTANT: The fact that a read or write operation is indicated does NOT
683 * imply that the operation will succeed, merely that it should be performed
684 * and will not block.
686 static unsigned int tipc_poll(struct file *file, struct socket *sock,
689 struct sock *sk = sock->sk;
690 struct tipc_sock *tsk = tipc_sk(sk);
693 sock_poll_wait(file, sk_sleep(sk), wait);
695 switch ((int)sock->state) {
702 if (!tsk->link_cong && !tsk_conn_cong(tsk))
707 if (!skb_queue_empty(&sk->sk_receive_queue))
708 mask |= (POLLIN | POLLRDNORM);
710 case SS_DISCONNECTING:
711 mask = (POLLIN | POLLRDNORM | POLLHUP);
719 * tipc_sendmcast - send multicast message
720 * @sock: socket structure
721 * @seq: destination address
722 * @msg: message to send
723 * @dsz: total length of message data
724 * @timeo: timeout to wait for wakeup
726 * Called from function tipc_sendmsg(), which has done all sanity checks
727 * Returns the number of bytes sent on success, or errno
729 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
730 struct msghdr *msg, size_t dsz, long timeo)
732 struct sock *sk = sock->sk;
733 struct net *net = sock_net(sk);
734 struct tipc_msg *mhdr = &tipc_sk(sk)->phdr;
735 struct sk_buff_head head;
739 msg_set_type(mhdr, TIPC_MCAST_MSG);
740 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
741 msg_set_destport(mhdr, 0);
742 msg_set_destnode(mhdr, 0);
743 msg_set_nametype(mhdr, seq->type);
744 msg_set_namelower(mhdr, seq->lower);
745 msg_set_nameupper(mhdr, seq->upper);
746 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
749 mtu = tipc_bclink_get_mtu();
750 __skb_queue_head_init(&head);
751 rc = tipc_msg_build(net, mhdr, msg, 0, dsz, mtu, &head);
752 if (unlikely(rc < 0))
756 rc = tipc_bclink_xmit(net, &head);
757 if (likely(rc >= 0)) {
763 if (rc != -ELINKCONG)
765 tipc_sk(sk)->link_cong = 1;
766 rc = tipc_wait_for_sndmsg(sock, &timeo);
768 __skb_queue_purge(&head);
773 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
775 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff *buf)
777 struct tipc_msg *msg = buf_msg(buf);
778 struct tipc_port_list dports = {0, NULL, };
779 struct tipc_port_list *item;
781 uint i, last, dst = 0;
782 u32 scope = TIPC_CLUSTER_SCOPE;
784 if (in_own_node(net, msg_orignode(msg)))
785 scope = TIPC_NODE_SCOPE;
787 /* Create destination port list: */
788 tipc_nametbl_mc_translate(net, msg_nametype(msg), msg_namelower(msg),
789 msg_nameupper(msg), scope, &dports);
796 for (item = &dports; item; item = item->next) {
797 for (i = 0; i < PLSIZE && ++dst <= last; i++) {
798 b = (dst != last) ? skb_clone(buf, GFP_ATOMIC) : buf;
800 pr_warn("Failed do clone mcast rcv buffer\n");
803 msg_set_destport(msg, item->ports[i]);
807 tipc_port_list_free(&dports);
811 * tipc_sk_proto_rcv - receive a connection mng protocol message
812 * @tsk: receiving socket
813 * @dnode: node to send response message to, if any
814 * @buf: buffer containing protocol message
815 * Returns 0 (TIPC_OK) if message was consumed, 1 (TIPC_FWD_MSG) if
816 * (CONN_PROBE_REPLY) message should be forwarded.
818 static int tipc_sk_proto_rcv(struct tipc_sock *tsk, u32 *dnode,
821 struct tipc_msg *msg = buf_msg(buf);
824 /* Ignore if connection cannot be validated: */
825 if (!tsk_peer_msg(tsk, msg))
828 tsk->probing_state = TIPC_CONN_OK;
830 if (msg_type(msg) == CONN_ACK) {
831 conn_cong = tsk_conn_cong(tsk);
832 tsk->sent_unacked -= msg_msgcnt(msg);
834 tsk->sk.sk_write_space(&tsk->sk);
835 } else if (msg_type(msg) == CONN_PROBE) {
836 if (!tipc_msg_reverse(sock_net(&tsk->sk), buf, dnode, TIPC_OK))
838 msg_set_type(msg, CONN_PROBE_REPLY);
841 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
847 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
849 struct sock *sk = sock->sk;
850 struct tipc_sock *tsk = tipc_sk(sk);
855 int err = sock_error(sk);
858 if (sock->state == SS_DISCONNECTING)
862 if (signal_pending(current))
863 return sock_intr_errno(*timeo_p);
865 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
866 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
867 finish_wait(sk_sleep(sk), &wait);
873 * tipc_sendmsg - send message in connectionless manner
874 * @iocb: if NULL, indicates that socket lock is already held
875 * @sock: socket structure
876 * @m: message to send
877 * @dsz: amount of user data to be sent
879 * Message must have an destination specified explicitly.
880 * Used for SOCK_RDM and SOCK_DGRAM messages,
881 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
882 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
884 * Returns the number of bytes sent on success, or errno otherwise
886 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
887 struct msghdr *m, size_t dsz)
889 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
890 struct sock *sk = sock->sk;
891 struct tipc_sock *tsk = tipc_sk(sk);
892 struct net *net = sock_net(sk);
893 struct tipc_msg *mhdr = &tsk->phdr;
895 struct sk_buff_head head;
897 struct tipc_name_seq *seq = &dest->addr.nameseq;
903 return -EDESTADDRREQ;
905 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
906 (dest->family != AF_TIPC)))
909 if (dsz > TIPC_MAX_USER_MSG_SIZE)
915 if (unlikely(sock->state != SS_READY)) {
916 if (sock->state == SS_LISTENING) {
920 if (sock->state != SS_UNCONNECTED) {
924 if (tsk->published) {
928 if (dest->addrtype == TIPC_ADDR_NAME) {
929 tsk->conn_type = dest->addr.name.name.type;
930 tsk->conn_instance = dest->addr.name.name.instance;
934 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
936 if (dest->addrtype == TIPC_ADDR_MCAST) {
937 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
939 } else if (dest->addrtype == TIPC_ADDR_NAME) {
940 u32 type = dest->addr.name.name.type;
941 u32 inst = dest->addr.name.name.instance;
942 u32 domain = dest->addr.name.domain;
945 msg_set_type(mhdr, TIPC_NAMED_MSG);
946 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
947 msg_set_nametype(mhdr, type);
948 msg_set_nameinst(mhdr, inst);
949 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
950 dport = tipc_nametbl_translate(net, type, inst, &dnode);
951 msg_set_destnode(mhdr, dnode);
952 msg_set_destport(mhdr, dport);
953 if (unlikely(!dport && !dnode)) {
957 } else if (dest->addrtype == TIPC_ADDR_ID) {
958 dnode = dest->addr.id.node;
959 msg_set_type(mhdr, TIPC_DIRECT_MSG);
960 msg_set_lookup_scope(mhdr, 0);
961 msg_set_destnode(mhdr, dnode);
962 msg_set_destport(mhdr, dest->addr.id.ref);
963 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
967 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
968 __skb_queue_head_init(&head);
969 rc = tipc_msg_build(net, mhdr, m, 0, dsz, mtu, &head);
974 skb = skb_peek(&head);
975 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
976 rc = tipc_link_xmit(net, &head, dnode, tsk->portid);
977 if (likely(rc >= 0)) {
978 if (sock->state != SS_READY)
979 sock->state = SS_CONNECTING;
985 if (rc != -ELINKCONG)
988 rc = tipc_wait_for_sndmsg(sock, &timeo);
990 __skb_queue_purge(&head);
999 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
1001 struct sock *sk = sock->sk;
1002 struct tipc_sock *tsk = tipc_sk(sk);
1007 int err = sock_error(sk);
1010 if (sock->state == SS_DISCONNECTING)
1012 else if (sock->state != SS_CONNECTED)
1016 if (signal_pending(current))
1017 return sock_intr_errno(*timeo_p);
1019 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1020 done = sk_wait_event(sk, timeo_p,
1022 !tsk_conn_cong(tsk)) ||
1024 finish_wait(sk_sleep(sk), &wait);
1030 * tipc_send_stream - send stream-oriented data
1032 * @sock: socket structure
1034 * @dsz: total length of data to be transmitted
1036 * Used for SOCK_STREAM data.
1038 * Returns the number of bytes sent on success (or partial success),
1039 * or errno if no data sent
1041 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1042 struct msghdr *m, size_t dsz)
1044 struct sock *sk = sock->sk;
1045 struct net *net = sock_net(sk);
1046 struct tipc_sock *tsk = tipc_sk(sk);
1047 struct tipc_msg *mhdr = &tsk->phdr;
1048 struct sk_buff_head head;
1049 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1050 u32 portid = tsk->portid;
1054 uint mtu, send, sent = 0;
1056 /* Handle implied connection establishment */
1057 if (unlikely(dest)) {
1058 rc = tipc_sendmsg(iocb, sock, m, dsz);
1059 if (dsz && (dsz == rc))
1060 tsk->sent_unacked = 1;
1063 if (dsz > (uint)INT_MAX)
1069 if (unlikely(sock->state != SS_CONNECTED)) {
1070 if (sock->state == SS_DISCONNECTING)
1077 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1078 dnode = tsk_peer_node(tsk);
1082 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1083 __skb_queue_head_init(&head);
1084 rc = tipc_msg_build(net, mhdr, m, sent, send, mtu, &head);
1085 if (unlikely(rc < 0))
1088 if (likely(!tsk_conn_cong(tsk))) {
1089 rc = tipc_link_xmit(net, &head, dnode, portid);
1091 tsk->sent_unacked++;
1097 if (rc == -EMSGSIZE) {
1098 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1102 if (rc != -ELINKCONG)
1106 rc = tipc_wait_for_sndpkt(sock, &timeo);
1108 __skb_queue_purge(&head);
1113 return sent ? sent : rc;
1117 * tipc_send_packet - send a connection-oriented message
1118 * @iocb: if NULL, indicates that socket lock is already held
1119 * @sock: socket structure
1120 * @m: message to send
1121 * @dsz: length of data to be transmitted
1123 * Used for SOCK_SEQPACKET messages.
1125 * Returns the number of bytes sent on success, or errno otherwise
1127 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1128 struct msghdr *m, size_t dsz)
1130 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1133 return tipc_send_stream(iocb, sock, m, dsz);
1136 /* tipc_sk_finish_conn - complete the setup of a connection
1138 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1141 struct sock *sk = &tsk->sk;
1142 struct net *net = sock_net(sk);
1143 struct tipc_msg *msg = &tsk->phdr;
1145 msg_set_destnode(msg, peer_node);
1146 msg_set_destport(msg, peer_port);
1147 msg_set_type(msg, TIPC_CONN_MSG);
1148 msg_set_lookup_scope(msg, 0);
1149 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1151 tsk->probing_intv = CONN_PROBING_INTERVAL;
1152 tsk->probing_state = TIPC_CONN_OK;
1154 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1155 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1156 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1160 * set_orig_addr - capture sender's address for received message
1161 * @m: descriptor for message info
1162 * @msg: received message header
1164 * Note: Address is not captured if not requested by receiver.
1166 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1168 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1171 addr->family = AF_TIPC;
1172 addr->addrtype = TIPC_ADDR_ID;
1173 memset(&addr->addr, 0, sizeof(addr->addr));
1174 addr->addr.id.ref = msg_origport(msg);
1175 addr->addr.id.node = msg_orignode(msg);
1176 addr->addr.name.domain = 0; /* could leave uninitialized */
1177 addr->scope = 0; /* could leave uninitialized */
1178 m->msg_namelen = sizeof(struct sockaddr_tipc);
1183 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1184 * @m: descriptor for message info
1185 * @msg: received message header
1186 * @tsk: TIPC port associated with message
1188 * Note: Ancillary data is not captured if not requested by receiver.
1190 * Returns 0 if successful, otherwise errno
1192 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1193 struct tipc_sock *tsk)
1201 if (likely(m->msg_controllen == 0))
1204 /* Optionally capture errored message object(s) */
1205 err = msg ? msg_errcode(msg) : 0;
1206 if (unlikely(err)) {
1208 anc_data[1] = msg_data_sz(msg);
1209 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1213 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1220 /* Optionally capture message destination object */
1221 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1222 switch (dest_type) {
1223 case TIPC_NAMED_MSG:
1225 anc_data[0] = msg_nametype(msg);
1226 anc_data[1] = msg_namelower(msg);
1227 anc_data[2] = msg_namelower(msg);
1229 case TIPC_MCAST_MSG:
1231 anc_data[0] = msg_nametype(msg);
1232 anc_data[1] = msg_namelower(msg);
1233 anc_data[2] = msg_nameupper(msg);
1236 has_name = (tsk->conn_type != 0);
1237 anc_data[0] = tsk->conn_type;
1238 anc_data[1] = tsk->conn_instance;
1239 anc_data[2] = tsk->conn_instance;
1245 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1253 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1255 struct net *net = sock_net(&tsk->sk);
1256 struct tipc_net *tn = net_generic(net, tipc_net_id);
1257 struct sk_buff *skb = NULL;
1258 struct tipc_msg *msg;
1259 u32 peer_port = tsk_peer_port(tsk);
1260 u32 dnode = tsk_peer_node(tsk);
1262 if (!tsk->connected)
1264 skb = tipc_msg_create(net, CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1265 dnode, tn->own_addr, peer_port, tsk->portid,
1270 msg_set_msgcnt(msg, ack);
1271 tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1274 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1276 struct sock *sk = sock->sk;
1278 long timeo = *timeop;
1282 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1283 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1284 if (sock->state == SS_DISCONNECTING) {
1289 timeo = schedule_timeout(timeo);
1293 if (!skb_queue_empty(&sk->sk_receive_queue))
1295 err = sock_intr_errno(timeo);
1296 if (signal_pending(current))
1302 finish_wait(sk_sleep(sk), &wait);
1308 * tipc_recvmsg - receive packet-oriented message
1310 * @m: descriptor for message info
1311 * @buf_len: total size of user buffer area
1312 * @flags: receive flags
1314 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1315 * If the complete message doesn't fit in user area, truncate it.
1317 * Returns size of returned message data, errno otherwise
1319 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1320 struct msghdr *m, size_t buf_len, int flags)
1322 struct sock *sk = sock->sk;
1323 struct tipc_sock *tsk = tipc_sk(sk);
1324 struct sk_buff *buf;
1325 struct tipc_msg *msg;
1331 /* Catch invalid receive requests */
1332 if (unlikely(!buf_len))
1337 if (unlikely(sock->state == SS_UNCONNECTED)) {
1342 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1345 /* Look for a message in receive queue; wait if necessary */
1346 res = tipc_wait_for_rcvmsg(sock, &timeo);
1350 /* Look at first message in receive queue */
1351 buf = skb_peek(&sk->sk_receive_queue);
1353 sz = msg_data_sz(msg);
1354 err = msg_errcode(msg);
1356 /* Discard an empty non-errored message & try again */
1357 if ((!sz) && (!err)) {
1358 tsk_advance_rx_queue(sk);
1362 /* Capture sender's address (optional) */
1363 set_orig_addr(m, msg);
1365 /* Capture ancillary data (optional) */
1366 res = tipc_sk_anc_data_recv(m, msg, tsk);
1370 /* Capture message data (if valid) & compute return value (always) */
1372 if (unlikely(buf_len < sz)) {
1374 m->msg_flags |= MSG_TRUNC;
1376 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1381 if ((sock->state == SS_READY) ||
1382 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1388 /* Consume received message (optional) */
1389 if (likely(!(flags & MSG_PEEK))) {
1390 if ((sock->state != SS_READY) &&
1391 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1392 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1393 tsk->rcv_unacked = 0;
1395 tsk_advance_rx_queue(sk);
1403 * tipc_recv_stream - receive stream-oriented data
1405 * @m: descriptor for message info
1406 * @buf_len: total size of user buffer area
1407 * @flags: receive flags
1409 * Used for SOCK_STREAM messages only. If not enough data is available
1410 * will optionally wait for more; never truncates data.
1412 * Returns size of returned message data, errno otherwise
1414 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1415 struct msghdr *m, size_t buf_len, int flags)
1417 struct sock *sk = sock->sk;
1418 struct tipc_sock *tsk = tipc_sk(sk);
1419 struct sk_buff *buf;
1420 struct tipc_msg *msg;
1423 int sz_to_copy, target, needed;
1428 /* Catch invalid receive attempts */
1429 if (unlikely(!buf_len))
1434 if (unlikely(sock->state == SS_UNCONNECTED)) {
1439 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1440 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1443 /* Look for a message in receive queue; wait if necessary */
1444 res = tipc_wait_for_rcvmsg(sock, &timeo);
1448 /* Look at first message in receive queue */
1449 buf = skb_peek(&sk->sk_receive_queue);
1451 sz = msg_data_sz(msg);
1452 err = msg_errcode(msg);
1454 /* Discard an empty non-errored message & try again */
1455 if ((!sz) && (!err)) {
1456 tsk_advance_rx_queue(sk);
1460 /* Optionally capture sender's address & ancillary data of first msg */
1461 if (sz_copied == 0) {
1462 set_orig_addr(m, msg);
1463 res = tipc_sk_anc_data_recv(m, msg, tsk);
1468 /* Capture message data (if valid) & compute return value (always) */
1470 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1473 needed = (buf_len - sz_copied);
1474 sz_to_copy = (sz <= needed) ? sz : needed;
1476 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1481 sz_copied += sz_to_copy;
1483 if (sz_to_copy < sz) {
1484 if (!(flags & MSG_PEEK))
1485 TIPC_SKB_CB(buf)->handle =
1486 (void *)(unsigned long)(offset + sz_to_copy);
1491 goto exit; /* can't add error msg to valid data */
1493 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1499 /* Consume received message (optional) */
1500 if (likely(!(flags & MSG_PEEK))) {
1501 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1502 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1503 tsk->rcv_unacked = 0;
1505 tsk_advance_rx_queue(sk);
1508 /* Loop around if more data is required */
1509 if ((sz_copied < buf_len) && /* didn't get all requested data */
1510 (!skb_queue_empty(&sk->sk_receive_queue) ||
1511 (sz_copied < target)) && /* and more is ready or required */
1512 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1513 (!err)) /* and haven't reached a FIN */
1518 return sz_copied ? sz_copied : res;
1522 * tipc_write_space - wake up thread if port congestion is released
1525 static void tipc_write_space(struct sock *sk)
1527 struct socket_wq *wq;
1530 wq = rcu_dereference(sk->sk_wq);
1531 if (wq_has_sleeper(wq))
1532 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1533 POLLWRNORM | POLLWRBAND);
1538 * tipc_data_ready - wake up threads to indicate messages have been received
1540 * @len: the length of messages
1542 static void tipc_data_ready(struct sock *sk)
1544 struct socket_wq *wq;
1547 wq = rcu_dereference(sk->sk_wq);
1548 if (wq_has_sleeper(wq))
1549 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1550 POLLRDNORM | POLLRDBAND);
1555 * filter_connect - Handle all incoming messages for a connection-based socket
1559 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1561 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **buf)
1563 struct sock *sk = &tsk->sk;
1564 struct net *net = sock_net(sk);
1565 struct socket *sock = sk->sk_socket;
1566 struct tipc_msg *msg = buf_msg(*buf);
1567 int retval = -TIPC_ERR_NO_PORT;
1572 switch ((int)sock->state) {
1574 /* Accept only connection-based messages sent by peer */
1575 if (tsk_peer_msg(tsk, msg)) {
1576 if (unlikely(msg_errcode(msg))) {
1577 sock->state = SS_DISCONNECTING;
1579 /* let timer expire on it's own */
1580 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1587 /* Accept only ACK or NACK message */
1589 if (unlikely(!msg_connected(msg)))
1592 if (unlikely(msg_errcode(msg))) {
1593 sock->state = SS_DISCONNECTING;
1594 sk->sk_err = ECONNREFUSED;
1599 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1600 sock->state = SS_DISCONNECTING;
1601 sk->sk_err = EINVAL;
1606 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1607 msg_set_importance(&tsk->phdr, msg_importance(msg));
1608 sock->state = SS_CONNECTED;
1610 /* If an incoming message is an 'ACK-', it should be
1611 * discarded here because it doesn't contain useful
1612 * data. In addition, we should try to wake up
1613 * connect() routine if sleeping.
1615 if (msg_data_sz(msg) == 0) {
1618 if (waitqueue_active(sk_sleep(sk)))
1619 wake_up_interruptible(sk_sleep(sk));
1624 case SS_UNCONNECTED:
1625 /* Accept only SYN message */
1626 if (!msg_connected(msg) && !(msg_errcode(msg)))
1629 case SS_DISCONNECTING:
1632 pr_err("Unknown socket state %u\n", sock->state);
1638 * rcvbuf_limit - get proper overload limit of socket receive queue
1642 * For all connection oriented messages, irrespective of importance,
1643 * the default overload value (i.e. 67MB) is set as limit.
1645 * For all connectionless messages, by default new queue limits are
1648 * TIPC_LOW_IMPORTANCE (4 MB)
1649 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1650 * TIPC_HIGH_IMPORTANCE (16 MB)
1651 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1653 * Returns overload limit according to corresponding message importance
1655 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1657 struct tipc_msg *msg = buf_msg(buf);
1659 if (msg_connected(msg))
1660 return sysctl_tipc_rmem[2];
1662 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1663 msg_importance(msg);
1667 * filter_rcv - validate incoming message
1671 * Enqueues message on receive queue if acceptable; optionally handles
1672 * disconnect indication for a connected socket.
1674 * Called with socket lock already taken; port lock may also be taken.
1676 * Returns 0 (TIPC_OK) if message was consumed, -TIPC error code if message
1677 * to be rejected, 1 (TIPC_FWD_MSG) if (CONN_MANAGER) message to be forwarded
1679 static int filter_rcv(struct sock *sk, struct sk_buff *buf)
1681 struct socket *sock = sk->sk_socket;
1682 struct tipc_sock *tsk = tipc_sk(sk);
1683 struct tipc_msg *msg = buf_msg(buf);
1684 unsigned int limit = rcvbuf_limit(sk, buf);
1688 if (unlikely(msg_user(msg) == CONN_MANAGER))
1689 return tipc_sk_proto_rcv(tsk, &onode, buf);
1691 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1694 sk->sk_write_space(sk);
1698 /* Reject message if it is wrong sort of message for socket */
1699 if (msg_type(msg) > TIPC_DIRECT_MSG)
1700 return -TIPC_ERR_NO_PORT;
1702 if (sock->state == SS_READY) {
1703 if (msg_connected(msg))
1704 return -TIPC_ERR_NO_PORT;
1706 rc = filter_connect(tsk, &buf);
1707 if (rc != TIPC_OK || buf == NULL)
1711 /* Reject message if there isn't room to queue it */
1712 if (sk_rmem_alloc_get(sk) + buf->truesize >= limit)
1713 return -TIPC_ERR_OVERLOAD;
1715 /* Enqueue message */
1716 TIPC_SKB_CB(buf)->handle = NULL;
1717 __skb_queue_tail(&sk->sk_receive_queue, buf);
1718 skb_set_owner_r(buf, sk);
1720 sk->sk_data_ready(sk);
1725 * tipc_backlog_rcv - handle incoming message from backlog queue
1729 * Caller must hold socket lock, but not port lock.
1733 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1737 struct tipc_sock *tsk = tipc_sk(sk);
1738 struct net *net = sock_net(sk);
1739 uint truesize = skb->truesize;
1741 rc = filter_rcv(sk, skb);
1744 if (atomic_read(&tsk->dupl_rcvcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1745 atomic_add(truesize, &tsk->dupl_rcvcnt);
1749 if ((rc < 0) && !tipc_msg_reverse(net, skb, &onode, -rc))
1752 tipc_link_xmit_skb(net, skb, onode, 0);
1758 * tipc_sk_rcv - handle incoming message
1759 * @skb: buffer containing arriving message
1761 * Returns 0 if success, or errno: -EHOSTUNREACH
1763 int tipc_sk_rcv(struct net *net, struct sk_buff *skb)
1765 struct tipc_sock *tsk;
1767 u32 dport = msg_destport(buf_msg(skb));
1772 /* Validate destination and message */
1773 tsk = tipc_sk_lookup(net, dport);
1774 if (unlikely(!tsk)) {
1775 rc = tipc_msg_eval(net, skb, &dnode);
1781 spin_lock_bh(&sk->sk_lock.slock);
1783 if (!sock_owned_by_user(sk)) {
1784 rc = filter_rcv(sk, skb);
1786 if (sk->sk_backlog.len == 0)
1787 atomic_set(&tsk->dupl_rcvcnt, 0);
1788 limit = rcvbuf_limit(sk, skb) + atomic_read(&tsk->dupl_rcvcnt);
1789 if (sk_add_backlog(sk, skb, limit))
1790 rc = -TIPC_ERR_OVERLOAD;
1792 spin_unlock_bh(&sk->sk_lock.slock);
1797 if ((rc < 0) && !tipc_msg_reverse(net, skb, &dnode, -rc))
1798 return -EHOSTUNREACH;
1800 tipc_link_xmit_skb(net, skb, dnode, 0);
1801 return (rc < 0) ? -EHOSTUNREACH : 0;
1804 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1806 struct sock *sk = sock->sk;
1811 int err = sock_error(sk);
1816 if (signal_pending(current))
1817 return sock_intr_errno(*timeo_p);
1819 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1820 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1821 finish_wait(sk_sleep(sk), &wait);
1827 * tipc_connect - establish a connection to another TIPC port
1828 * @sock: socket structure
1829 * @dest: socket address for destination port
1830 * @destlen: size of socket address data structure
1831 * @flags: file-related flags associated with socket
1833 * Returns 0 on success, errno otherwise
1835 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1836 int destlen, int flags)
1838 struct sock *sk = sock->sk;
1839 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1840 struct msghdr m = {NULL,};
1841 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1842 socket_state previous;
1847 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1848 if (sock->state == SS_READY) {
1854 * Reject connection attempt using multicast address
1856 * Note: send_msg() validates the rest of the address fields,
1857 * so there's no need to do it here
1859 if (dst->addrtype == TIPC_ADDR_MCAST) {
1864 previous = sock->state;
1865 switch (sock->state) {
1866 case SS_UNCONNECTED:
1867 /* Send a 'SYN-' to destination */
1869 m.msg_namelen = destlen;
1871 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1872 * indicate send_msg() is never blocked.
1875 m.msg_flags = MSG_DONTWAIT;
1877 res = tipc_sendmsg(NULL, sock, &m, 0);
1878 if ((res < 0) && (res != -EWOULDBLOCK))
1881 /* Just entered SS_CONNECTING state; the only
1882 * difference is that return value in non-blocking
1883 * case is EINPROGRESS, rather than EALREADY.
1887 if (previous == SS_CONNECTING)
1891 timeout = msecs_to_jiffies(timeout);
1892 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1893 res = tipc_wait_for_connect(sock, &timeout);
1908 * tipc_listen - allow socket to listen for incoming connections
1909 * @sock: socket structure
1912 * Returns 0 on success, errno otherwise
1914 static int tipc_listen(struct socket *sock, int len)
1916 struct sock *sk = sock->sk;
1921 if (sock->state != SS_UNCONNECTED)
1924 sock->state = SS_LISTENING;
1932 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1934 struct sock *sk = sock->sk;
1938 /* True wake-one mechanism for incoming connections: only
1939 * one process gets woken up, not the 'whole herd'.
1940 * Since we do not 'race & poll' for established sockets
1941 * anymore, the common case will execute the loop only once.
1944 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1945 TASK_INTERRUPTIBLE);
1946 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1948 timeo = schedule_timeout(timeo);
1952 if (!skb_queue_empty(&sk->sk_receive_queue))
1955 if (sock->state != SS_LISTENING)
1957 err = sock_intr_errno(timeo);
1958 if (signal_pending(current))
1964 finish_wait(sk_sleep(sk), &wait);
1969 * tipc_accept - wait for connection request
1970 * @sock: listening socket
1971 * @newsock: new socket that is to be connected
1972 * @flags: file-related flags associated with socket
1974 * Returns 0 on success, errno otherwise
1976 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1978 struct sock *new_sk, *sk = sock->sk;
1979 struct sk_buff *buf;
1980 struct tipc_sock *new_tsock;
1981 struct tipc_msg *msg;
1987 if (sock->state != SS_LISTENING) {
1991 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
1992 res = tipc_wait_for_accept(sock, timeo);
1996 buf = skb_peek(&sk->sk_receive_queue);
1998 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2002 new_sk = new_sock->sk;
2003 new_tsock = tipc_sk(new_sk);
2006 /* we lock on new_sk; but lockdep sees the lock on sk */
2007 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2010 * Reject any stray messages received by new socket
2011 * before the socket lock was taken (very, very unlikely)
2013 tsk_rej_rx_queue(new_sk);
2015 /* Connect new socket to it's peer */
2016 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2017 new_sock->state = SS_CONNECTED;
2019 tsk_set_importance(new_tsock, msg_importance(msg));
2020 if (msg_named(msg)) {
2021 new_tsock->conn_type = msg_nametype(msg);
2022 new_tsock->conn_instance = msg_nameinst(msg);
2026 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2027 * Respond to 'SYN+' by queuing it on new socket.
2029 if (!msg_data_sz(msg)) {
2030 struct msghdr m = {NULL,};
2032 tsk_advance_rx_queue(sk);
2033 tipc_send_packet(NULL, new_sock, &m, 0);
2035 __skb_dequeue(&sk->sk_receive_queue);
2036 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2037 skb_set_owner_r(buf, new_sk);
2039 release_sock(new_sk);
2046 * tipc_shutdown - shutdown socket connection
2047 * @sock: socket structure
2048 * @how: direction to close (must be SHUT_RDWR)
2050 * Terminates connection (if necessary), then purges socket's receive queue.
2052 * Returns 0 on success, errno otherwise
2054 static int tipc_shutdown(struct socket *sock, int how)
2056 struct sock *sk = sock->sk;
2057 struct net *net = sock_net(sk);
2058 struct tipc_net *tn = net_generic(net, tipc_net_id);
2059 struct tipc_sock *tsk = tipc_sk(sk);
2060 struct sk_buff *skb;
2064 if (how != SHUT_RDWR)
2069 switch (sock->state) {
2074 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2075 skb = __skb_dequeue(&sk->sk_receive_queue);
2077 if (TIPC_SKB_CB(skb)->handle != NULL) {
2081 if (tipc_msg_reverse(net, skb, &dnode,
2082 TIPC_CONN_SHUTDOWN))
2083 tipc_link_xmit_skb(net, skb, dnode,
2085 tipc_node_remove_conn(net, dnode, tsk->portid);
2087 dnode = tsk_peer_node(tsk);
2088 skb = tipc_msg_create(net, TIPC_CRITICAL_IMPORTANCE,
2089 TIPC_CONN_MSG, SHORT_H_SIZE,
2090 0, dnode, tn->own_addr,
2092 tsk->portid, TIPC_CONN_SHUTDOWN);
2093 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2096 sock->state = SS_DISCONNECTING;
2097 tipc_node_remove_conn(net, dnode, tsk->portid);
2100 case SS_DISCONNECTING:
2102 /* Discard any unreceived messages */
2103 __skb_queue_purge(&sk->sk_receive_queue);
2105 /* Wake up anyone sleeping in poll */
2106 sk->sk_state_change(sk);
2118 static void tipc_sk_timeout(unsigned long data)
2120 struct tipc_sock *tsk = (struct tipc_sock *)data;
2121 struct sock *sk = &tsk->sk;
2122 struct net *net = sock_net(sk);
2123 struct tipc_net *tn = net_generic(net, tipc_net_id);
2124 struct sk_buff *skb = NULL;
2125 u32 peer_port, peer_node;
2128 if (!tsk->connected) {
2132 peer_port = tsk_peer_port(tsk);
2133 peer_node = tsk_peer_node(tsk);
2135 if (tsk->probing_state == TIPC_CONN_PROBING) {
2136 /* Previous probe not answered -> self abort */
2137 skb = tipc_msg_create(net, TIPC_CRITICAL_IMPORTANCE,
2138 TIPC_CONN_MSG, SHORT_H_SIZE, 0,
2139 tn->own_addr, peer_node, tsk->portid,
2140 peer_port, TIPC_ERR_NO_PORT);
2142 skb = tipc_msg_create(net, CONN_MANAGER, CONN_PROBE, INT_H_SIZE,
2143 0, peer_node, tn->own_addr,
2144 peer_port, tsk->portid, TIPC_OK);
2145 tsk->probing_state = TIPC_CONN_PROBING;
2146 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2150 tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2155 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2156 struct tipc_name_seq const *seq)
2158 struct net *net = sock_net(&tsk->sk);
2159 struct publication *publ;
2164 key = tsk->portid + tsk->pub_count + 1;
2165 if (key == tsk->portid)
2168 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2169 scope, tsk->portid, key);
2170 if (unlikely(!publ))
2173 list_add(&publ->pport_list, &tsk->publications);
2179 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2180 struct tipc_name_seq const *seq)
2182 struct net *net = sock_net(&tsk->sk);
2183 struct publication *publ;
2184 struct publication *safe;
2187 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2189 if (publ->scope != scope)
2191 if (publ->type != seq->type)
2193 if (publ->lower != seq->lower)
2195 if (publ->upper != seq->upper)
2197 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2198 publ->ref, publ->key);
2202 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2203 publ->ref, publ->key);
2206 if (list_empty(&tsk->publications))
2211 static int tipc_sk_show(struct tipc_sock *tsk, char *buf,
2212 int len, int full_id)
2214 struct net *net = sock_net(&tsk->sk);
2215 struct tipc_net *tn = net_generic(net, tipc_net_id);
2216 struct publication *publ;
2220 ret = tipc_snprintf(buf, len, "<%u.%u.%u:%u>:",
2221 tipc_zone(tn->own_addr),
2222 tipc_cluster(tn->own_addr),
2223 tipc_node(tn->own_addr), tsk->portid);
2225 ret = tipc_snprintf(buf, len, "%-10u:", tsk->portid);
2227 if (tsk->connected) {
2228 u32 dport = tsk_peer_port(tsk);
2229 u32 destnode = tsk_peer_node(tsk);
2231 ret += tipc_snprintf(buf + ret, len - ret,
2232 " connected to <%u.%u.%u:%u>",
2233 tipc_zone(destnode),
2234 tipc_cluster(destnode),
2235 tipc_node(destnode), dport);
2236 if (tsk->conn_type != 0)
2237 ret += tipc_snprintf(buf + ret, len - ret,
2238 " via {%u,%u}", tsk->conn_type,
2239 tsk->conn_instance);
2240 } else if (tsk->published) {
2241 ret += tipc_snprintf(buf + ret, len - ret, " bound to");
2242 list_for_each_entry(publ, &tsk->publications, pport_list) {
2243 if (publ->lower == publ->upper)
2244 ret += tipc_snprintf(buf + ret, len - ret,
2245 " {%u,%u}", publ->type,
2248 ret += tipc_snprintf(buf + ret, len - ret,
2249 " {%u,%u,%u}", publ->type,
2250 publ->lower, publ->upper);
2253 ret += tipc_snprintf(buf + ret, len - ret, "\n");
2257 struct sk_buff *tipc_sk_socks_show(struct net *net)
2259 struct tipc_net *tn = net_generic(net, tipc_net_id);
2260 const struct bucket_table *tbl;
2261 struct rhash_head *pos;
2262 struct sk_buff *buf;
2263 struct tlv_desc *rep_tlv;
2266 struct tipc_sock *tsk;
2270 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2273 rep_tlv = (struct tlv_desc *)buf->data;
2274 pb = TLV_DATA(rep_tlv);
2275 pb_len = ULTRA_STRING_MAX_LEN;
2278 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2279 for (i = 0; i < tbl->size; i++) {
2280 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2281 spin_lock_bh(&tsk->sk.sk_lock.slock);
2282 str_len += tipc_sk_show(tsk, pb + str_len,
2283 pb_len - str_len, 0);
2284 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2289 str_len += 1; /* for "\0" */
2290 skb_put(buf, TLV_SPACE(str_len));
2291 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2296 /* tipc_sk_reinit: set non-zero address in all existing sockets
2297 * when we go from standalone to network mode.
2299 void tipc_sk_reinit(struct net *net)
2301 struct tipc_net *tn = net_generic(net, tipc_net_id);
2302 const struct bucket_table *tbl;
2303 struct rhash_head *pos;
2304 struct tipc_sock *tsk;
2305 struct tipc_msg *msg;
2309 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2310 for (i = 0; i < tbl->size; i++) {
2311 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2312 spin_lock_bh(&tsk->sk.sk_lock.slock);
2314 msg_set_prevnode(msg, tn->own_addr);
2315 msg_set_orignode(msg, tn->own_addr);
2316 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2322 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2324 struct tipc_net *tn = net_generic(net, tipc_net_id);
2325 struct tipc_sock *tsk;
2328 tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2330 sock_hold(&tsk->sk);
2336 static int tipc_sk_insert(struct tipc_sock *tsk)
2338 struct sock *sk = &tsk->sk;
2339 struct net *net = sock_net(sk);
2340 struct tipc_net *tn = net_generic(net, tipc_net_id);
2341 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2342 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2344 while (remaining--) {
2346 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2347 portid = TIPC_MIN_PORT;
2348 tsk->portid = portid;
2349 sock_hold(&tsk->sk);
2350 if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2358 static void tipc_sk_remove(struct tipc_sock *tsk)
2360 struct sock *sk = &tsk->sk;
2361 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2363 if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2364 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2369 int tipc_sk_rht_init(struct net *net)
2371 struct tipc_net *tn = net_generic(net, tipc_net_id);
2372 struct rhashtable_params rht_params = {
2374 .head_offset = offsetof(struct tipc_sock, node),
2375 .key_offset = offsetof(struct tipc_sock, portid),
2376 .key_len = sizeof(u32), /* portid */
2378 .max_shift = 20, /* 1M */
2379 .min_shift = 8, /* 256 */
2380 .grow_decision = rht_grow_above_75,
2381 .shrink_decision = rht_shrink_below_30,
2384 return rhashtable_init(&tn->sk_rht, &rht_params);
2387 void tipc_sk_rht_destroy(struct net *net)
2389 struct tipc_net *tn = net_generic(net, tipc_net_id);
2391 /* Wait for socket readers to complete */
2394 rhashtable_destroy(&tn->sk_rht);
2398 * tipc_setsockopt - set socket option
2399 * @sock: socket structure
2400 * @lvl: option level
2401 * @opt: option identifier
2402 * @ov: pointer to new option value
2403 * @ol: length of option value
2405 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2406 * (to ease compatibility).
2408 * Returns 0 on success, errno otherwise
2410 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2411 char __user *ov, unsigned int ol)
2413 struct sock *sk = sock->sk;
2414 struct tipc_sock *tsk = tipc_sk(sk);
2418 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2420 if (lvl != SOL_TIPC)
2421 return -ENOPROTOOPT;
2422 if (ol < sizeof(value))
2424 res = get_user(value, (u32 __user *)ov);
2431 case TIPC_IMPORTANCE:
2432 res = tsk_set_importance(tsk, value);
2434 case TIPC_SRC_DROPPABLE:
2435 if (sock->type != SOCK_STREAM)
2436 tsk_set_unreliable(tsk, value);
2440 case TIPC_DEST_DROPPABLE:
2441 tsk_set_unreturnable(tsk, value);
2443 case TIPC_CONN_TIMEOUT:
2444 tipc_sk(sk)->conn_timeout = value;
2445 /* no need to set "res", since already 0 at this point */
2457 * tipc_getsockopt - get socket option
2458 * @sock: socket structure
2459 * @lvl: option level
2460 * @opt: option identifier
2461 * @ov: receptacle for option value
2462 * @ol: receptacle for length of option value
2464 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2465 * (to ease compatibility).
2467 * Returns 0 on success, errno otherwise
2469 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2470 char __user *ov, int __user *ol)
2472 struct sock *sk = sock->sk;
2473 struct tipc_sock *tsk = tipc_sk(sk);
2478 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2479 return put_user(0, ol);
2480 if (lvl != SOL_TIPC)
2481 return -ENOPROTOOPT;
2482 res = get_user(len, ol);
2489 case TIPC_IMPORTANCE:
2490 value = tsk_importance(tsk);
2492 case TIPC_SRC_DROPPABLE:
2493 value = tsk_unreliable(tsk);
2495 case TIPC_DEST_DROPPABLE:
2496 value = tsk_unreturnable(tsk);
2498 case TIPC_CONN_TIMEOUT:
2499 value = tsk->conn_timeout;
2500 /* no need to set "res", since already 0 at this point */
2502 case TIPC_NODE_RECVQ_DEPTH:
2503 value = 0; /* was tipc_queue_size, now obsolete */
2505 case TIPC_SOCK_RECVQ_DEPTH:
2506 value = skb_queue_len(&sk->sk_receive_queue);
2515 return res; /* "get" failed */
2517 if (len < sizeof(value))
2520 if (copy_to_user(ov, &value, sizeof(value)))
2523 return put_user(sizeof(value), ol);
2526 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2528 struct sock *sk = sock->sk;
2529 struct tipc_sioc_ln_req lnr;
2530 void __user *argp = (void __user *)arg;
2533 case SIOCGETLINKNAME:
2534 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2536 if (!tipc_node_get_linkname(sock_net(sk),
2537 lnr.bearer_id & 0xffff, lnr.peer,
2538 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2539 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2543 return -EADDRNOTAVAIL;
2545 return -ENOIOCTLCMD;
2549 /* Protocol switches for the various types of TIPC sockets */
2551 static const struct proto_ops msg_ops = {
2552 .owner = THIS_MODULE,
2554 .release = tipc_release,
2556 .connect = tipc_connect,
2557 .socketpair = sock_no_socketpair,
2558 .accept = sock_no_accept,
2559 .getname = tipc_getname,
2561 .ioctl = tipc_ioctl,
2562 .listen = sock_no_listen,
2563 .shutdown = tipc_shutdown,
2564 .setsockopt = tipc_setsockopt,
2565 .getsockopt = tipc_getsockopt,
2566 .sendmsg = tipc_sendmsg,
2567 .recvmsg = tipc_recvmsg,
2568 .mmap = sock_no_mmap,
2569 .sendpage = sock_no_sendpage
2572 static const struct proto_ops packet_ops = {
2573 .owner = THIS_MODULE,
2575 .release = tipc_release,
2577 .connect = tipc_connect,
2578 .socketpair = sock_no_socketpair,
2579 .accept = tipc_accept,
2580 .getname = tipc_getname,
2582 .ioctl = tipc_ioctl,
2583 .listen = tipc_listen,
2584 .shutdown = tipc_shutdown,
2585 .setsockopt = tipc_setsockopt,
2586 .getsockopt = tipc_getsockopt,
2587 .sendmsg = tipc_send_packet,
2588 .recvmsg = tipc_recvmsg,
2589 .mmap = sock_no_mmap,
2590 .sendpage = sock_no_sendpage
2593 static const struct proto_ops stream_ops = {
2594 .owner = THIS_MODULE,
2596 .release = tipc_release,
2598 .connect = tipc_connect,
2599 .socketpair = sock_no_socketpair,
2600 .accept = tipc_accept,
2601 .getname = tipc_getname,
2603 .ioctl = tipc_ioctl,
2604 .listen = tipc_listen,
2605 .shutdown = tipc_shutdown,
2606 .setsockopt = tipc_setsockopt,
2607 .getsockopt = tipc_getsockopt,
2608 .sendmsg = tipc_send_stream,
2609 .recvmsg = tipc_recv_stream,
2610 .mmap = sock_no_mmap,
2611 .sendpage = sock_no_sendpage
2614 static const struct net_proto_family tipc_family_ops = {
2615 .owner = THIS_MODULE,
2617 .create = tipc_sk_create
2620 static struct proto tipc_proto = {
2622 .owner = THIS_MODULE,
2623 .obj_size = sizeof(struct tipc_sock),
2624 .sysctl_rmem = sysctl_tipc_rmem
2627 static struct proto tipc_proto_kern = {
2629 .obj_size = sizeof(struct tipc_sock),
2630 .sysctl_rmem = sysctl_tipc_rmem
2634 * tipc_socket_init - initialize TIPC socket interface
2636 * Returns 0 on success, errno otherwise
2638 int tipc_socket_init(void)
2642 res = proto_register(&tipc_proto, 1);
2644 pr_err("Failed to register TIPC protocol type\n");
2648 res = sock_register(&tipc_family_ops);
2650 pr_err("Failed to register TIPC socket type\n");
2651 proto_unregister(&tipc_proto);
2659 * tipc_socket_stop - stop TIPC socket interface
2661 void tipc_socket_stop(void)
2663 sock_unregister(tipc_family_ops.family);
2664 proto_unregister(&tipc_proto);
2667 /* Caller should hold socket lock for the passed tipc socket. */
2668 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2672 struct nlattr *nest;
2674 peer_node = tsk_peer_node(tsk);
2675 peer_port = tsk_peer_port(tsk);
2677 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2679 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2681 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2684 if (tsk->conn_type != 0) {
2685 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2687 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2689 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2692 nla_nest_end(skb, nest);
2697 nla_nest_cancel(skb, nest);
2702 /* Caller should hold socket lock for the passed tipc socket. */
2703 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2704 struct tipc_sock *tsk)
2708 struct nlattr *attrs;
2709 struct net *net = sock_net(skb->sk);
2710 struct tipc_net *tn = net_generic(net, tipc_net_id);
2712 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2713 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2717 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2719 goto genlmsg_cancel;
2720 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2721 goto attr_msg_cancel;
2722 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2723 goto attr_msg_cancel;
2725 if (tsk->connected) {
2726 err = __tipc_nl_add_sk_con(skb, tsk);
2728 goto attr_msg_cancel;
2729 } else if (!list_empty(&tsk->publications)) {
2730 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2731 goto attr_msg_cancel;
2733 nla_nest_end(skb, attrs);
2734 genlmsg_end(skb, hdr);
2739 nla_nest_cancel(skb, attrs);
2741 genlmsg_cancel(skb, hdr);
2746 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2749 struct tipc_sock *tsk;
2750 const struct bucket_table *tbl;
2751 struct rhash_head *pos;
2752 u32 prev_portid = cb->args[0];
2753 u32 portid = prev_portid;
2754 struct net *net = sock_net(skb->sk);
2755 struct tipc_net *tn = net_generic(net, tipc_net_id);
2759 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2760 for (i = 0; i < tbl->size; i++) {
2761 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2762 spin_lock_bh(&tsk->sk.sk_lock.slock);
2763 portid = tsk->portid;
2764 err = __tipc_nl_add_sk(skb, cb, tsk);
2765 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2769 prev_portid = portid;
2774 cb->args[0] = prev_portid;
2779 /* Caller should hold socket lock for the passed tipc socket. */
2780 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2781 struct netlink_callback *cb,
2782 struct publication *publ)
2785 struct nlattr *attrs;
2787 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2788 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2792 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2794 goto genlmsg_cancel;
2796 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2797 goto attr_msg_cancel;
2798 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2799 goto attr_msg_cancel;
2800 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2801 goto attr_msg_cancel;
2802 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2803 goto attr_msg_cancel;
2805 nla_nest_end(skb, attrs);
2806 genlmsg_end(skb, hdr);
2811 nla_nest_cancel(skb, attrs);
2813 genlmsg_cancel(skb, hdr);
2818 /* Caller should hold socket lock for the passed tipc socket. */
2819 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2820 struct netlink_callback *cb,
2821 struct tipc_sock *tsk, u32 *last_publ)
2824 struct publication *p;
2827 list_for_each_entry(p, &tsk->publications, pport_list) {
2828 if (p->key == *last_publ)
2831 if (p->key != *last_publ) {
2832 /* We never set seq or call nl_dump_check_consistent()
2833 * this means that setting prev_seq here will cause the
2834 * consistence check to fail in the netlink callback
2835 * handler. Resulting in the last NLMSG_DONE message
2836 * having the NLM_F_DUMP_INTR flag set.
2843 p = list_first_entry(&tsk->publications, struct publication,
2847 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2848 err = __tipc_nl_add_sk_publ(skb, cb, p);
2850 *last_publ = p->key;
2859 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2862 u32 tsk_portid = cb->args[0];
2863 u32 last_publ = cb->args[1];
2864 u32 done = cb->args[2];
2865 struct net *net = sock_net(skb->sk);
2866 struct tipc_sock *tsk;
2869 struct nlattr **attrs;
2870 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2872 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2876 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2877 attrs[TIPC_NLA_SOCK],
2878 tipc_nl_sock_policy);
2882 if (!sock[TIPC_NLA_SOCK_REF])
2885 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2891 tsk = tipc_sk_lookup(net, tsk_portid);
2895 lock_sock(&tsk->sk);
2896 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2899 release_sock(&tsk->sk);
2902 cb->args[0] = tsk_portid;
2903 cb->args[1] = last_publ;