2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2015, 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>
39 #include "name_table.h"
42 #include "name_distr.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 * @conn_timeout: the time we can wait for an unresponded setup request
73 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
74 * @link_cong: non-zero if owner must sleep because of link congestion
75 * @sent_unacked: # messages sent by socket, and not yet acked by peer
76 * @rcv_unacked: # messages read by user, but not yet acked back to peer
77 * @remote: 'connected' peer for dgram/rdm
78 * @node: hash table node
79 * @rcu: rcu struct for tipc_sock
90 struct list_head sock_list;
91 struct list_head publications;
94 unsigned long probing_intv;
100 struct sockaddr_tipc remote;
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);
119 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
121 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
123 static const struct proto_ops packet_ops;
124 static const struct proto_ops stream_ops;
125 static const struct proto_ops msg_ops;
126 static struct proto tipc_proto;
128 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
129 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
130 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
131 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
132 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
133 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
136 static const struct rhashtable_params tsk_rht_params;
139 * Revised TIPC socket locking policy:
141 * Most socket operations take the standard socket lock when they start
142 * and hold it until they finish (or until they need to sleep). Acquiring
143 * this lock grants the owner exclusive access to the fields of the socket
144 * data structures, with the exception of the backlog queue. A few socket
145 * operations can be done without taking the socket lock because they only
146 * read socket information that never changes during the life of the socket.
148 * Socket operations may acquire the lock for the associated TIPC port if they
149 * need to perform an operation on the port. If any routine needs to acquire
150 * both the socket lock and the port lock it must take the socket lock first
151 * to avoid the risk of deadlock.
153 * The dispatcher handling incoming messages cannot grab the socket lock in
154 * the standard fashion, since invoked it runs at the BH level and cannot block.
155 * Instead, it checks to see if the socket lock is currently owned by someone,
156 * and either handles the message itself or adds it to the socket's backlog
157 * queue; in the latter case the queued message is processed once the process
158 * owning the socket lock releases it.
160 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
161 * the problem of a blocked socket operation preventing any other operations
162 * from occurring. However, applications must be careful if they have
163 * multiple threads trying to send (or receive) on the same socket, as these
164 * operations might interfere with each other. For example, doing a connect
165 * and a receive at the same time might allow the receive to consume the
166 * ACK message meant for the connect. While additional work could be done
167 * to try and overcome this, it doesn't seem to be worthwhile at the present.
169 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
170 * that another operation that must be performed in a non-blocking manner is
171 * not delayed for very long because the lock has already been taken.
173 * NOTE: This code assumes that certain fields of a port/socket pair are
174 * constant over its lifetime; such fields can be examined without taking
175 * the socket lock and/or port lock, and do not need to be re-read even
176 * after resuming processing after waiting. These fields include:
178 * - pointer to socket sk structure (aka tipc_sock structure)
179 * - pointer to port structure
183 static u32 tsk_own_node(struct tipc_sock *tsk)
185 return msg_prevnode(&tsk->phdr);
188 static u32 tsk_peer_node(struct tipc_sock *tsk)
190 return msg_destnode(&tsk->phdr);
193 static u32 tsk_peer_port(struct tipc_sock *tsk)
195 return msg_destport(&tsk->phdr);
198 static bool tsk_unreliable(struct tipc_sock *tsk)
200 return msg_src_droppable(&tsk->phdr) != 0;
203 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
205 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
208 static bool tsk_unreturnable(struct tipc_sock *tsk)
210 return msg_dest_droppable(&tsk->phdr) != 0;
213 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
215 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
218 static int tsk_importance(struct tipc_sock *tsk)
220 return msg_importance(&tsk->phdr);
223 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
225 if (imp > TIPC_CRITICAL_IMPORTANCE)
227 msg_set_importance(&tsk->phdr, (u32)imp);
231 static struct tipc_sock *tipc_sk(const struct sock *sk)
233 return container_of(sk, struct tipc_sock, sk);
236 static int tsk_conn_cong(struct tipc_sock *tsk)
238 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
242 * tsk_advance_rx_queue - discard first buffer in socket receive queue
244 * Caller must hold socket lock
246 static void tsk_advance_rx_queue(struct sock *sk)
248 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
252 * tsk_rej_rx_queue - reject all buffers in socket receive queue
254 * Caller must hold socket lock
256 static void tsk_rej_rx_queue(struct sock *sk)
260 u32 own_node = tsk_own_node(tipc_sk(sk));
262 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
263 if (tipc_msg_reverse(own_node, skb, &dnode, TIPC_ERR_NO_PORT))
264 tipc_node_xmit_skb(sock_net(sk), skb, dnode, 0);
268 /* tsk_peer_msg - verify if message was sent by connected port's peer
270 * Handles cases where the node's network address has changed from
271 * the default of <0.0.0> to its configured setting.
273 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
275 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
276 u32 peer_port = tsk_peer_port(tsk);
280 if (unlikely(!tsk->connected))
283 if (unlikely(msg_origport(msg) != peer_port))
286 orig_node = msg_orignode(msg);
287 peer_node = tsk_peer_node(tsk);
289 if (likely(orig_node == peer_node))
292 if (!orig_node && (peer_node == tn->own_addr))
295 if (!peer_node && (orig_node == tn->own_addr))
302 * tipc_sk_create - create a TIPC socket
303 * @net: network namespace (must be default network)
304 * @sock: pre-allocated socket structure
305 * @protocol: protocol indicator (must be 0)
306 * @kern: caused by kernel or by userspace?
308 * This routine creates additional data structures used by the TIPC socket,
309 * initializes them, and links them together.
311 * Returns 0 on success, errno otherwise
313 static int tipc_sk_create(struct net *net, struct socket *sock,
314 int protocol, int kern)
317 const struct proto_ops *ops;
320 struct tipc_sock *tsk;
321 struct tipc_msg *msg;
323 /* Validate arguments */
324 if (unlikely(protocol != 0))
325 return -EPROTONOSUPPORT;
327 switch (sock->type) {
330 state = SS_UNCONNECTED;
334 state = SS_UNCONNECTED;
345 /* Allocate socket's protocol area */
346 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
351 tsk->max_pkt = MAX_PKT_DEFAULT;
352 INIT_LIST_HEAD(&tsk->publications);
354 tn = net_generic(sock_net(sk), tipc_net_id);
355 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
358 /* Finish initializing socket data structures */
361 sock_init_data(sock, sk);
362 if (tipc_sk_insert(tsk)) {
363 pr_warn("Socket create failed; port numbrer exhausted\n");
366 msg_set_origport(msg, tsk->portid);
367 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
368 sk->sk_backlog_rcv = tipc_backlog_rcv;
369 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
370 sk->sk_data_ready = tipc_data_ready;
371 sk->sk_write_space = tipc_write_space;
372 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
373 tsk->sent_unacked = 0;
374 atomic_set(&tsk->dupl_rcvcnt, 0);
376 if (sock->state == SS_READY) {
377 tsk_set_unreturnable(tsk, true);
378 if (sock->type == SOCK_DGRAM)
379 tsk_set_unreliable(tsk, true);
384 static void tipc_sk_callback(struct rcu_head *head)
386 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
392 * tipc_release - destroy a TIPC socket
393 * @sock: socket to destroy
395 * This routine cleans up any messages that are still queued on the socket.
396 * For DGRAM and RDM socket types, all queued messages are rejected.
397 * For SEQPACKET and STREAM socket types, the first message is rejected
398 * and any others are discarded. (If the first message on a STREAM socket
399 * is partially-read, it is discarded and the next one is rejected instead.)
401 * NOTE: Rejected messages are not necessarily returned to the sender! They
402 * are returned or discarded according to the "destination droppable" setting
403 * specified for the message by the sender.
405 * Returns 0 on success, errno otherwise
407 static int tipc_release(struct socket *sock)
409 struct sock *sk = sock->sk;
411 struct tipc_sock *tsk;
416 * Exit if socket isn't fully initialized (occurs when a failed accept()
417 * releases a pre-allocated child socket that was never used)
427 * Reject all unreceived messages, except on an active connection
428 * (which disconnects locally & sends a 'FIN+' to peer)
430 dnode = tsk_peer_node(tsk);
431 while (sock->state != SS_DISCONNECTING) {
432 skb = __skb_dequeue(&sk->sk_receive_queue);
435 if (TIPC_SKB_CB(skb)->handle != NULL)
438 if ((sock->state == SS_CONNECTING) ||
439 (sock->state == SS_CONNECTED)) {
440 sock->state = SS_DISCONNECTING;
442 tipc_node_remove_conn(net, dnode, tsk->portid);
444 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
446 tipc_node_xmit_skb(net, skb, dnode, 0);
450 tipc_sk_withdraw(tsk, 0, NULL);
451 sk_stop_timer(sk, &sk->sk_timer);
453 if (tsk->connected) {
454 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
455 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
456 tsk_own_node(tsk), tsk_peer_port(tsk),
457 tsk->portid, TIPC_ERR_NO_PORT);
459 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
460 tipc_node_remove_conn(net, dnode, tsk->portid);
463 /* Discard any remaining (connection-based) messages in receive queue */
464 __skb_queue_purge(&sk->sk_receive_queue);
466 /* Reject any messages that accumulated in backlog queue */
467 sock->state = SS_DISCONNECTING;
470 call_rcu(&tsk->rcu, tipc_sk_callback);
477 * tipc_bind - associate or disassocate TIPC name(s) with a socket
478 * @sock: socket structure
479 * @uaddr: socket address describing name(s) and desired operation
480 * @uaddr_len: size of socket address data structure
482 * Name and name sequence binding is indicated using a positive scope value;
483 * a negative scope value unbinds the specified name. Specifying no name
484 * (i.e. a socket address length of 0) unbinds all names from the socket.
486 * Returns 0 on success, errno otherwise
488 * NOTE: This routine doesn't need to take the socket lock since it doesn't
489 * access any non-constant socket information.
491 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
494 struct sock *sk = sock->sk;
495 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
496 struct tipc_sock *tsk = tipc_sk(sk);
500 if (unlikely(!uaddr_len)) {
501 res = tipc_sk_withdraw(tsk, 0, NULL);
505 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
509 if (addr->family != AF_TIPC) {
514 if (addr->addrtype == TIPC_ADDR_NAME)
515 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
516 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
521 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
522 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
523 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
528 res = (addr->scope > 0) ?
529 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
530 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
537 * tipc_getname - get port ID of socket or peer socket
538 * @sock: socket structure
539 * @uaddr: area for returned socket address
540 * @uaddr_len: area for returned length of socket address
541 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
543 * Returns 0 on success, errno otherwise
545 * NOTE: This routine doesn't need to take the socket lock since it only
546 * accesses socket information that is unchanging (or which changes in
547 * a completely predictable manner).
549 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
550 int *uaddr_len, int peer)
552 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
553 struct tipc_sock *tsk = tipc_sk(sock->sk);
554 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
556 memset(addr, 0, sizeof(*addr));
558 if ((sock->state != SS_CONNECTED) &&
559 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
561 addr->addr.id.ref = tsk_peer_port(tsk);
562 addr->addr.id.node = tsk_peer_node(tsk);
564 addr->addr.id.ref = tsk->portid;
565 addr->addr.id.node = tn->own_addr;
568 *uaddr_len = sizeof(*addr);
569 addr->addrtype = TIPC_ADDR_ID;
570 addr->family = AF_TIPC;
572 addr->addr.name.domain = 0;
578 * tipc_poll - read and possibly block on pollmask
579 * @file: file structure associated with the socket
580 * @sock: socket for which to calculate the poll bits
583 * Returns pollmask value
586 * It appears that the usual socket locking mechanisms are not useful here
587 * since the pollmask info is potentially out-of-date the moment this routine
588 * exits. TCP and other protocols seem to rely on higher level poll routines
589 * to handle any preventable race conditions, so TIPC will do the same ...
591 * TIPC sets the returned events as follows:
593 * socket state flags set
594 * ------------ ---------
595 * unconnected no read flags
596 * POLLOUT if port is not congested
598 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
601 * connected POLLIN/POLLRDNORM if data in rx queue
602 * POLLOUT if port is not congested
604 * disconnecting POLLIN/POLLRDNORM/POLLHUP
607 * listening POLLIN if SYN in rx queue
610 * ready POLLIN/POLLRDNORM if data in rx queue
611 * [connectionless] POLLOUT (since port cannot be congested)
613 * IMPORTANT: The fact that a read or write operation is indicated does NOT
614 * imply that the operation will succeed, merely that it should be performed
615 * and will not block.
617 static unsigned int tipc_poll(struct file *file, struct socket *sock,
620 struct sock *sk = sock->sk;
621 struct tipc_sock *tsk = tipc_sk(sk);
624 sock_poll_wait(file, sk_sleep(sk), wait);
626 switch ((int)sock->state) {
633 if (!tsk->link_cong && !tsk_conn_cong(tsk))
638 if (!skb_queue_empty(&sk->sk_receive_queue))
639 mask |= (POLLIN | POLLRDNORM);
641 case SS_DISCONNECTING:
642 mask = (POLLIN | POLLRDNORM | POLLHUP);
650 * tipc_sendmcast - send multicast message
651 * @sock: socket structure
652 * @seq: destination address
653 * @msg: message to send
654 * @dsz: total length of message data
655 * @timeo: timeout to wait for wakeup
657 * Called from function tipc_sendmsg(), which has done all sanity checks
658 * Returns the number of bytes sent on success, or errno
660 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
661 struct msghdr *msg, size_t dsz, long timeo)
663 struct sock *sk = sock->sk;
664 struct tipc_sock *tsk = tipc_sk(sk);
665 struct net *net = sock_net(sk);
666 struct tipc_msg *mhdr = &tsk->phdr;
667 struct sk_buff_head *pktchain = &sk->sk_write_queue;
668 struct iov_iter save = msg->msg_iter;
672 msg_set_type(mhdr, TIPC_MCAST_MSG);
673 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
674 msg_set_destport(mhdr, 0);
675 msg_set_destnode(mhdr, 0);
676 msg_set_nametype(mhdr, seq->type);
677 msg_set_namelower(mhdr, seq->lower);
678 msg_set_nameupper(mhdr, seq->upper);
679 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
682 mtu = tipc_bclink_get_mtu();
683 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, pktchain);
684 if (unlikely(rc < 0))
688 rc = tipc_bclink_xmit(net, pktchain);
692 if (rc == -ELINKCONG) {
694 rc = tipc_wait_for_sndmsg(sock, &timeo);
698 __skb_queue_purge(pktchain);
699 if (rc == -EMSGSIZE) {
700 msg->msg_iter = save;
709 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
710 * @arrvq: queue with arriving messages, to be cloned after destination lookup
711 * @inputq: queue with cloned messages, delivered to socket after dest lookup
713 * Multi-threaded: parallel calls with reference to same queues may occur
715 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
716 struct sk_buff_head *inputq)
718 struct tipc_msg *msg;
719 struct tipc_plist dports;
721 u32 scope = TIPC_CLUSTER_SCOPE;
722 struct sk_buff_head tmpq;
724 struct sk_buff *skb, *_skb;
726 __skb_queue_head_init(&tmpq);
727 tipc_plist_init(&dports);
729 skb = tipc_skb_peek(arrvq, &inputq->lock);
730 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
732 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
734 if (in_own_node(net, msg_orignode(msg)))
735 scope = TIPC_NODE_SCOPE;
737 /* Create destination port list and message clones: */
738 tipc_nametbl_mc_translate(net,
739 msg_nametype(msg), msg_namelower(msg),
740 msg_nameupper(msg), scope, &dports);
741 portid = tipc_plist_pop(&dports);
742 for (; portid; portid = tipc_plist_pop(&dports)) {
743 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
745 msg_set_destport(buf_msg(_skb), portid);
746 __skb_queue_tail(&tmpq, _skb);
749 pr_warn("Failed to clone mcast rcv buffer\n");
751 /* Append to inputq if not already done by other thread */
752 spin_lock_bh(&inputq->lock);
753 if (skb_peek(arrvq) == skb) {
754 skb_queue_splice_tail_init(&tmpq, inputq);
755 kfree_skb(__skb_dequeue(arrvq));
757 spin_unlock_bh(&inputq->lock);
758 __skb_queue_purge(&tmpq);
761 tipc_sk_rcv(net, inputq);
765 * tipc_sk_proto_rcv - receive a connection mng protocol message
766 * @tsk: receiving socket
767 * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
769 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff **skb)
771 struct tipc_msg *msg = buf_msg(*skb);
774 u32 own_node = tsk_own_node(tsk);
775 /* Ignore if connection cannot be validated: */
776 if (!tsk_peer_msg(tsk, msg))
779 tsk->probing_state = TIPC_CONN_OK;
781 if (msg_type(msg) == CONN_ACK) {
782 conn_cong = tsk_conn_cong(tsk);
783 tsk->sent_unacked -= msg_msgcnt(msg);
785 tsk->sk.sk_write_space(&tsk->sk);
786 } else if (msg_type(msg) == CONN_PROBE) {
787 if (tipc_msg_reverse(own_node, *skb, &dnode, TIPC_OK)) {
788 msg_set_type(msg, CONN_PROBE_REPLY);
792 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
798 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
800 struct sock *sk = sock->sk;
801 struct tipc_sock *tsk = tipc_sk(sk);
806 int err = sock_error(sk);
809 if (sock->state == SS_DISCONNECTING)
813 if (signal_pending(current))
814 return sock_intr_errno(*timeo_p);
816 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
817 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
818 finish_wait(sk_sleep(sk), &wait);
824 * tipc_sendmsg - send message in connectionless manner
825 * @sock: socket structure
826 * @m: message to send
827 * @dsz: amount of user data to be sent
829 * Message must have an destination specified explicitly.
830 * Used for SOCK_RDM and SOCK_DGRAM messages,
831 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
832 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
834 * Returns the number of bytes sent on success, or errno otherwise
836 static int tipc_sendmsg(struct socket *sock,
837 struct msghdr *m, size_t dsz)
839 struct sock *sk = sock->sk;
843 ret = __tipc_sendmsg(sock, m, dsz);
849 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
851 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
852 struct sock *sk = sock->sk;
853 struct tipc_sock *tsk = tipc_sk(sk);
854 struct net *net = sock_net(sk);
855 struct tipc_msg *mhdr = &tsk->phdr;
857 struct sk_buff_head *pktchain = &sk->sk_write_queue;
859 struct tipc_name_seq *seq;
860 struct iov_iter save;
865 if (dsz > TIPC_MAX_USER_MSG_SIZE)
867 if (unlikely(!dest)) {
868 if (tsk->connected && sock->state == SS_READY)
871 return -EDESTADDRREQ;
872 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
873 dest->family != AF_TIPC) {
876 if (unlikely(sock->state != SS_READY)) {
877 if (sock->state == SS_LISTENING)
879 if (sock->state != SS_UNCONNECTED)
883 if (dest->addrtype == TIPC_ADDR_NAME) {
884 tsk->conn_type = dest->addr.name.name.type;
885 tsk->conn_instance = dest->addr.name.name.instance;
888 seq = &dest->addr.nameseq;
889 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
891 if (dest->addrtype == TIPC_ADDR_MCAST) {
892 return tipc_sendmcast(sock, seq, m, dsz, timeo);
893 } else if (dest->addrtype == TIPC_ADDR_NAME) {
894 u32 type = dest->addr.name.name.type;
895 u32 inst = dest->addr.name.name.instance;
896 u32 domain = dest->addr.name.domain;
899 msg_set_type(mhdr, TIPC_NAMED_MSG);
900 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
901 msg_set_nametype(mhdr, type);
902 msg_set_nameinst(mhdr, inst);
903 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
904 dport = tipc_nametbl_translate(net, type, inst, &dnode);
905 msg_set_destnode(mhdr, dnode);
906 msg_set_destport(mhdr, dport);
907 if (unlikely(!dport && !dnode))
908 return -EHOSTUNREACH;
909 } else if (dest->addrtype == TIPC_ADDR_ID) {
910 dnode = dest->addr.id.node;
911 msg_set_type(mhdr, TIPC_DIRECT_MSG);
912 msg_set_lookup_scope(mhdr, 0);
913 msg_set_destnode(mhdr, dnode);
914 msg_set_destport(mhdr, dest->addr.id.ref);
915 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
920 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
921 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, pktchain);
926 skb = skb_peek(pktchain);
927 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
928 rc = tipc_node_xmit(net, pktchain, dnode, tsk->portid);
930 if (sock->state != SS_READY)
931 sock->state = SS_CONNECTING;
934 if (rc == -ELINKCONG) {
936 rc = tipc_wait_for_sndmsg(sock, &timeo);
940 __skb_queue_purge(pktchain);
941 if (rc == -EMSGSIZE) {
951 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
953 struct sock *sk = sock->sk;
954 struct tipc_sock *tsk = tipc_sk(sk);
959 int err = sock_error(sk);
962 if (sock->state == SS_DISCONNECTING)
964 else if (sock->state != SS_CONNECTED)
968 if (signal_pending(current))
969 return sock_intr_errno(*timeo_p);
971 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
972 done = sk_wait_event(sk, timeo_p,
974 !tsk_conn_cong(tsk)) ||
976 finish_wait(sk_sleep(sk), &wait);
982 * tipc_send_stream - send stream-oriented data
983 * @sock: socket structure
985 * @dsz: total length of data to be transmitted
987 * Used for SOCK_STREAM data.
989 * Returns the number of bytes sent on success (or partial success),
990 * or errno if no data sent
992 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
994 struct sock *sk = sock->sk;
998 ret = __tipc_send_stream(sock, m, dsz);
1004 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1006 struct sock *sk = sock->sk;
1007 struct net *net = sock_net(sk);
1008 struct tipc_sock *tsk = tipc_sk(sk);
1009 struct tipc_msg *mhdr = &tsk->phdr;
1010 struct sk_buff_head *pktchain = &sk->sk_write_queue;
1011 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1012 u32 portid = tsk->portid;
1016 uint mtu, send, sent = 0;
1017 struct iov_iter save;
1019 /* Handle implied connection establishment */
1020 if (unlikely(dest)) {
1021 rc = __tipc_sendmsg(sock, m, dsz);
1022 if (dsz && (dsz == rc))
1023 tsk->sent_unacked = 1;
1026 if (dsz > (uint)INT_MAX)
1029 if (unlikely(sock->state != SS_CONNECTED)) {
1030 if (sock->state == SS_DISCONNECTING)
1036 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1037 dnode = tsk_peer_node(tsk);
1042 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1043 rc = tipc_msg_build(mhdr, m, sent, send, mtu, pktchain);
1044 if (unlikely(rc < 0))
1047 if (likely(!tsk_conn_cong(tsk))) {
1048 rc = tipc_node_xmit(net, pktchain, dnode, portid);
1050 tsk->sent_unacked++;
1056 if (rc == -EMSGSIZE) {
1057 __skb_queue_purge(pktchain);
1058 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1063 if (rc != -ELINKCONG)
1068 rc = tipc_wait_for_sndpkt(sock, &timeo);
1071 __skb_queue_purge(pktchain);
1072 return sent ? sent : rc;
1076 * tipc_send_packet - send a connection-oriented message
1077 * @sock: socket structure
1078 * @m: message to send
1079 * @dsz: length of data to be transmitted
1081 * Used for SOCK_SEQPACKET messages.
1083 * Returns the number of bytes sent on success, or errno otherwise
1085 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1087 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1090 return tipc_send_stream(sock, m, dsz);
1093 /* tipc_sk_finish_conn - complete the setup of a connection
1095 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1098 struct sock *sk = &tsk->sk;
1099 struct net *net = sock_net(sk);
1100 struct tipc_msg *msg = &tsk->phdr;
1102 msg_set_destnode(msg, peer_node);
1103 msg_set_destport(msg, peer_port);
1104 msg_set_type(msg, TIPC_CONN_MSG);
1105 msg_set_lookup_scope(msg, 0);
1106 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1108 tsk->probing_intv = CONN_PROBING_INTERVAL;
1109 tsk->probing_state = TIPC_CONN_OK;
1111 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1112 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1113 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1117 * set_orig_addr - capture sender's address for received message
1118 * @m: descriptor for message info
1119 * @msg: received message header
1121 * Note: Address is not captured if not requested by receiver.
1123 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1125 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1128 addr->family = AF_TIPC;
1129 addr->addrtype = TIPC_ADDR_ID;
1130 memset(&addr->addr, 0, sizeof(addr->addr));
1131 addr->addr.id.ref = msg_origport(msg);
1132 addr->addr.id.node = msg_orignode(msg);
1133 addr->addr.name.domain = 0; /* could leave uninitialized */
1134 addr->scope = 0; /* could leave uninitialized */
1135 m->msg_namelen = sizeof(struct sockaddr_tipc);
1140 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1141 * @m: descriptor for message info
1142 * @msg: received message header
1143 * @tsk: TIPC port associated with message
1145 * Note: Ancillary data is not captured if not requested by receiver.
1147 * Returns 0 if successful, otherwise errno
1149 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1150 struct tipc_sock *tsk)
1158 if (likely(m->msg_controllen == 0))
1161 /* Optionally capture errored message object(s) */
1162 err = msg ? msg_errcode(msg) : 0;
1163 if (unlikely(err)) {
1165 anc_data[1] = msg_data_sz(msg);
1166 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1170 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1177 /* Optionally capture message destination object */
1178 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1179 switch (dest_type) {
1180 case TIPC_NAMED_MSG:
1182 anc_data[0] = msg_nametype(msg);
1183 anc_data[1] = msg_namelower(msg);
1184 anc_data[2] = msg_namelower(msg);
1186 case TIPC_MCAST_MSG:
1188 anc_data[0] = msg_nametype(msg);
1189 anc_data[1] = msg_namelower(msg);
1190 anc_data[2] = msg_nameupper(msg);
1193 has_name = (tsk->conn_type != 0);
1194 anc_data[0] = tsk->conn_type;
1195 anc_data[1] = tsk->conn_instance;
1196 anc_data[2] = tsk->conn_instance;
1202 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1210 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1212 struct net *net = sock_net(&tsk->sk);
1213 struct sk_buff *skb = NULL;
1214 struct tipc_msg *msg;
1215 u32 peer_port = tsk_peer_port(tsk);
1216 u32 dnode = tsk_peer_node(tsk);
1218 if (!tsk->connected)
1220 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1221 dnode, tsk_own_node(tsk), peer_port,
1222 tsk->portid, TIPC_OK);
1226 msg_set_msgcnt(msg, ack);
1227 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1230 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1232 struct sock *sk = sock->sk;
1234 long timeo = *timeop;
1238 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1239 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1240 if (sock->state == SS_DISCONNECTING) {
1245 timeo = schedule_timeout(timeo);
1249 if (!skb_queue_empty(&sk->sk_receive_queue))
1254 err = sock_intr_errno(timeo);
1255 if (signal_pending(current))
1258 finish_wait(sk_sleep(sk), &wait);
1264 * tipc_recvmsg - receive packet-oriented message
1265 * @m: descriptor for message info
1266 * @buf_len: total size of user buffer area
1267 * @flags: receive flags
1269 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1270 * If the complete message doesn't fit in user area, truncate it.
1272 * Returns size of returned message data, errno otherwise
1274 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1277 struct sock *sk = sock->sk;
1278 struct tipc_sock *tsk = tipc_sk(sk);
1279 struct sk_buff *buf;
1280 struct tipc_msg *msg;
1286 /* Catch invalid receive requests */
1287 if (unlikely(!buf_len))
1292 if (unlikely(sock->state == SS_UNCONNECTED)) {
1297 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1300 /* Look for a message in receive queue; wait if necessary */
1301 res = tipc_wait_for_rcvmsg(sock, &timeo);
1305 /* Look at first message in receive queue */
1306 buf = skb_peek(&sk->sk_receive_queue);
1308 sz = msg_data_sz(msg);
1309 err = msg_errcode(msg);
1311 /* Discard an empty non-errored message & try again */
1312 if ((!sz) && (!err)) {
1313 tsk_advance_rx_queue(sk);
1317 /* Capture sender's address (optional) */
1318 set_orig_addr(m, msg);
1320 /* Capture ancillary data (optional) */
1321 res = tipc_sk_anc_data_recv(m, msg, tsk);
1325 /* Capture message data (if valid) & compute return value (always) */
1327 if (unlikely(buf_len < sz)) {
1329 m->msg_flags |= MSG_TRUNC;
1331 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1336 if ((sock->state == SS_READY) ||
1337 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1343 /* Consume received message (optional) */
1344 if (likely(!(flags & MSG_PEEK))) {
1345 if ((sock->state != SS_READY) &&
1346 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1347 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1348 tsk->rcv_unacked = 0;
1350 tsk_advance_rx_queue(sk);
1358 * tipc_recv_stream - receive stream-oriented data
1359 * @m: descriptor for message info
1360 * @buf_len: total size of user buffer area
1361 * @flags: receive flags
1363 * Used for SOCK_STREAM messages only. If not enough data is available
1364 * will optionally wait for more; never truncates data.
1366 * Returns size of returned message data, errno otherwise
1368 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1369 size_t buf_len, int flags)
1371 struct sock *sk = sock->sk;
1372 struct tipc_sock *tsk = tipc_sk(sk);
1373 struct sk_buff *buf;
1374 struct tipc_msg *msg;
1377 int sz_to_copy, target, needed;
1382 /* Catch invalid receive attempts */
1383 if (unlikely(!buf_len))
1388 if (unlikely(sock->state == SS_UNCONNECTED)) {
1393 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1394 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1397 /* Look for a message in receive queue; wait if necessary */
1398 res = tipc_wait_for_rcvmsg(sock, &timeo);
1402 /* Look at first message in receive queue */
1403 buf = skb_peek(&sk->sk_receive_queue);
1405 sz = msg_data_sz(msg);
1406 err = msg_errcode(msg);
1408 /* Discard an empty non-errored message & try again */
1409 if ((!sz) && (!err)) {
1410 tsk_advance_rx_queue(sk);
1414 /* Optionally capture sender's address & ancillary data of first msg */
1415 if (sz_copied == 0) {
1416 set_orig_addr(m, msg);
1417 res = tipc_sk_anc_data_recv(m, msg, tsk);
1422 /* Capture message data (if valid) & compute return value (always) */
1424 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1427 needed = (buf_len - sz_copied);
1428 sz_to_copy = (sz <= needed) ? sz : needed;
1430 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1435 sz_copied += sz_to_copy;
1437 if (sz_to_copy < sz) {
1438 if (!(flags & MSG_PEEK))
1439 TIPC_SKB_CB(buf)->handle =
1440 (void *)(unsigned long)(offset + sz_to_copy);
1445 goto exit; /* can't add error msg to valid data */
1447 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1453 /* Consume received message (optional) */
1454 if (likely(!(flags & MSG_PEEK))) {
1455 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1456 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1457 tsk->rcv_unacked = 0;
1459 tsk_advance_rx_queue(sk);
1462 /* Loop around if more data is required */
1463 if ((sz_copied < buf_len) && /* didn't get all requested data */
1464 (!skb_queue_empty(&sk->sk_receive_queue) ||
1465 (sz_copied < target)) && /* and more is ready or required */
1466 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1467 (!err)) /* and haven't reached a FIN */
1472 return sz_copied ? sz_copied : res;
1476 * tipc_write_space - wake up thread if port congestion is released
1479 static void tipc_write_space(struct sock *sk)
1481 struct socket_wq *wq;
1484 wq = rcu_dereference(sk->sk_wq);
1485 if (wq_has_sleeper(wq))
1486 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1487 POLLWRNORM | POLLWRBAND);
1492 * tipc_data_ready - wake up threads to indicate messages have been received
1494 * @len: the length of messages
1496 static void tipc_data_ready(struct sock *sk)
1498 struct socket_wq *wq;
1501 wq = rcu_dereference(sk->sk_wq);
1502 if (wq_has_sleeper(wq))
1503 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1504 POLLRDNORM | POLLRDBAND);
1509 * filter_connect - Handle all incoming messages for a connection-based socket
1511 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1513 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1515 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **skb)
1517 struct sock *sk = &tsk->sk;
1518 struct net *net = sock_net(sk);
1519 struct socket *sock = sk->sk_socket;
1520 struct tipc_msg *msg = buf_msg(*skb);
1521 int retval = -TIPC_ERR_NO_PORT;
1526 switch ((int)sock->state) {
1528 /* Accept only connection-based messages sent by peer */
1529 if (tsk_peer_msg(tsk, msg)) {
1530 if (unlikely(msg_errcode(msg))) {
1531 sock->state = SS_DISCONNECTING;
1533 /* let timer expire on it's own */
1534 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1541 /* Accept only ACK or NACK message */
1543 if (unlikely(!msg_connected(msg)))
1546 if (unlikely(msg_errcode(msg))) {
1547 sock->state = SS_DISCONNECTING;
1548 sk->sk_err = ECONNREFUSED;
1553 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1554 sock->state = SS_DISCONNECTING;
1555 sk->sk_err = EINVAL;
1560 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1561 msg_set_importance(&tsk->phdr, msg_importance(msg));
1562 sock->state = SS_CONNECTED;
1564 /* If an incoming message is an 'ACK-', it should be
1565 * discarded here because it doesn't contain useful
1566 * data. In addition, we should try to wake up
1567 * connect() routine if sleeping.
1569 if (msg_data_sz(msg) == 0) {
1572 if (waitqueue_active(sk_sleep(sk)))
1573 wake_up_interruptible(sk_sleep(sk));
1578 case SS_UNCONNECTED:
1579 /* Accept only SYN message */
1580 if (!msg_connected(msg) && !(msg_errcode(msg)))
1583 case SS_DISCONNECTING:
1586 pr_err("Unknown socket state %u\n", sock->state);
1592 * rcvbuf_limit - get proper overload limit of socket receive queue
1596 * For all connection oriented messages, irrespective of importance,
1597 * the default overload value (i.e. 67MB) is set as limit.
1599 * For all connectionless messages, by default new queue limits are
1602 * TIPC_LOW_IMPORTANCE (4 MB)
1603 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1604 * TIPC_HIGH_IMPORTANCE (16 MB)
1605 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1607 * Returns overload limit according to corresponding message importance
1609 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1611 struct tipc_msg *msg = buf_msg(buf);
1613 if (msg_connected(msg))
1614 return sysctl_tipc_rmem[2];
1616 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1617 msg_importance(msg);
1621 * filter_rcv - validate incoming message
1623 * @skb: pointer to message. Set to NULL if buffer is consumed.
1625 * Enqueues message on receive queue if acceptable; optionally handles
1626 * disconnect indication for a connected socket.
1628 * Called with socket lock already taken
1630 * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1632 static int filter_rcv(struct sock *sk, struct sk_buff **skb)
1634 struct socket *sock = sk->sk_socket;
1635 struct tipc_sock *tsk = tipc_sk(sk);
1636 struct tipc_msg *msg = buf_msg(*skb);
1637 unsigned int limit = rcvbuf_limit(sk, *skb);
1640 if (unlikely(msg_user(msg) == CONN_MANAGER)) {
1641 tipc_sk_proto_rcv(tsk, skb);
1645 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1648 sk->sk_write_space(sk);
1653 /* Reject message if it is wrong sort of message for socket */
1654 if (msg_type(msg) > TIPC_DIRECT_MSG)
1655 return -TIPC_ERR_NO_PORT;
1657 if (sock->state == SS_READY) {
1658 if (msg_connected(msg))
1659 return -TIPC_ERR_NO_PORT;
1661 rc = filter_connect(tsk, skb);
1662 if (rc != TIPC_OK || !*skb)
1666 /* Reject message if there isn't room to queue it */
1667 if (sk_rmem_alloc_get(sk) + (*skb)->truesize >= limit)
1668 return -TIPC_ERR_OVERLOAD;
1670 /* Enqueue message */
1671 TIPC_SKB_CB(*skb)->handle = NULL;
1672 __skb_queue_tail(&sk->sk_receive_queue, *skb);
1673 skb_set_owner_r(*skb, sk);
1675 sk->sk_data_ready(sk);
1681 * tipc_backlog_rcv - handle incoming message from backlog queue
1685 * Caller must hold socket lock
1689 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1694 struct tipc_sock *tsk = tipc_sk(sk);
1695 struct net *net = sock_net(sk);
1696 uint truesize = skb->truesize;
1698 err = filter_rcv(sk, &skb);
1700 dcnt = &tsk->dupl_rcvcnt;
1701 if (atomic_read(dcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1702 atomic_add(truesize, dcnt);
1705 if (!err || tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode, -err))
1706 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
1711 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1712 * inputq and try adding them to socket or backlog queue
1713 * @inputq: list of incoming buffers with potentially different destinations
1714 * @sk: socket where the buffers should be enqueued
1715 * @dport: port number for the socket
1716 * @_skb: returned buffer to be forwarded or rejected, if applicable
1718 * Caller must hold socket lock
1720 * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1721 * or -TIPC_ERR_NO_PORT
1723 static int tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1724 u32 dport, struct sk_buff **_skb)
1729 struct sk_buff *skb;
1730 unsigned long time_limit = jiffies + 2;
1732 while (skb_queue_len(inputq)) {
1733 if (unlikely(time_after_eq(jiffies, time_limit)))
1735 skb = tipc_skb_dequeue(inputq, dport);
1738 if (!sock_owned_by_user(sk)) {
1739 err = filter_rcv(sk, &skb);
1745 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1746 if (sk->sk_backlog.len)
1747 atomic_set(dcnt, 0);
1748 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1749 if (likely(!sk_add_backlog(sk, skb, lim)))
1752 return -TIPC_ERR_OVERLOAD;
1758 * tipc_sk_rcv - handle a chain of incoming buffers
1759 * @inputq: buffer list containing the buffers
1760 * Consumes all buffers in list until inputq is empty
1761 * Note: may be called in multiple threads referring to the same queue
1762 * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1763 * Only node local calls check the return value, sending single-buffer queues
1765 int tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1767 u32 dnode, dport = 0;
1769 struct sk_buff *skb;
1770 struct tipc_sock *tsk;
1771 struct tipc_net *tn;
1774 while (skb_queue_len(inputq)) {
1775 err = -TIPC_ERR_NO_PORT;
1777 dport = tipc_skb_peek_port(inputq, dport);
1778 tsk = tipc_sk_lookup(net, dport);
1781 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1782 err = tipc_sk_enqueue(inputq, sk, dport, &skb);
1783 spin_unlock_bh(&sk->sk_lock.slock);
1788 skb = tipc_skb_dequeue(inputq, dport);
1792 if (tipc_msg_lookup_dest(net, skb, &dnode, &err))
1795 dnode = msg_destnode(buf_msg(skb));
1798 tn = net_generic(net, tipc_net_id);
1799 if (!tipc_msg_reverse(tn->own_addr, skb, &dnode, -err))
1802 tipc_node_xmit_skb(net, skb, dnode, dport);
1804 return err ? -EHOSTUNREACH : 0;
1807 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1809 struct sock *sk = sock->sk;
1814 int err = sock_error(sk);
1819 if (signal_pending(current))
1820 return sock_intr_errno(*timeo_p);
1822 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1823 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1824 finish_wait(sk_sleep(sk), &wait);
1830 * tipc_connect - establish a connection to another TIPC port
1831 * @sock: socket structure
1832 * @dest: socket address for destination port
1833 * @destlen: size of socket address data structure
1834 * @flags: file-related flags associated with socket
1836 * Returns 0 on success, errno otherwise
1838 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1839 int destlen, int flags)
1841 struct sock *sk = sock->sk;
1842 struct tipc_sock *tsk = tipc_sk(sk);
1843 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1844 struct msghdr m = {NULL,};
1845 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1846 socket_state previous;
1851 /* DGRAM/RDM connect(), just save the destaddr */
1852 if (sock->state == SS_READY) {
1853 if (dst->family == AF_UNSPEC) {
1854 memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1856 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1859 memcpy(&tsk->remote, dest, destlen);
1866 * Reject connection attempt using multicast address
1868 * Note: send_msg() validates the rest of the address fields,
1869 * so there's no need to do it here
1871 if (dst->addrtype == TIPC_ADDR_MCAST) {
1876 previous = sock->state;
1877 switch (sock->state) {
1878 case SS_UNCONNECTED:
1879 /* Send a 'SYN-' to destination */
1881 m.msg_namelen = destlen;
1883 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1884 * indicate send_msg() is never blocked.
1887 m.msg_flags = MSG_DONTWAIT;
1889 res = __tipc_sendmsg(sock, &m, 0);
1890 if ((res < 0) && (res != -EWOULDBLOCK))
1893 /* Just entered SS_CONNECTING state; the only
1894 * difference is that return value in non-blocking
1895 * case is EINPROGRESS, rather than EALREADY.
1899 if (previous == SS_CONNECTING)
1903 timeout = msecs_to_jiffies(timeout);
1904 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1905 res = tipc_wait_for_connect(sock, &timeout);
1920 * tipc_listen - allow socket to listen for incoming connections
1921 * @sock: socket structure
1924 * Returns 0 on success, errno otherwise
1926 static int tipc_listen(struct socket *sock, int len)
1928 struct sock *sk = sock->sk;
1933 if (sock->state != SS_UNCONNECTED)
1936 sock->state = SS_LISTENING;
1944 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1946 struct sock *sk = sock->sk;
1950 /* True wake-one mechanism for incoming connections: only
1951 * one process gets woken up, not the 'whole herd'.
1952 * Since we do not 'race & poll' for established sockets
1953 * anymore, the common case will execute the loop only once.
1956 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1957 TASK_INTERRUPTIBLE);
1958 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1960 timeo = schedule_timeout(timeo);
1964 if (!skb_queue_empty(&sk->sk_receive_queue))
1967 if (sock->state != SS_LISTENING)
1972 err = sock_intr_errno(timeo);
1973 if (signal_pending(current))
1976 finish_wait(sk_sleep(sk), &wait);
1981 * tipc_accept - wait for connection request
1982 * @sock: listening socket
1983 * @newsock: new socket that is to be connected
1984 * @flags: file-related flags associated with socket
1986 * Returns 0 on success, errno otherwise
1988 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
1990 struct sock *new_sk, *sk = sock->sk;
1991 struct sk_buff *buf;
1992 struct tipc_sock *new_tsock;
1993 struct tipc_msg *msg;
1999 if (sock->state != SS_LISTENING) {
2003 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2004 res = tipc_wait_for_accept(sock, timeo);
2008 buf = skb_peek(&sk->sk_receive_queue);
2010 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2013 security_sk_clone(sock->sk, new_sock->sk);
2015 new_sk = new_sock->sk;
2016 new_tsock = tipc_sk(new_sk);
2019 /* we lock on new_sk; but lockdep sees the lock on sk */
2020 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2023 * Reject any stray messages received by new socket
2024 * before the socket lock was taken (very, very unlikely)
2026 tsk_rej_rx_queue(new_sk);
2028 /* Connect new socket to it's peer */
2029 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2030 new_sock->state = SS_CONNECTED;
2032 tsk_set_importance(new_tsock, msg_importance(msg));
2033 if (msg_named(msg)) {
2034 new_tsock->conn_type = msg_nametype(msg);
2035 new_tsock->conn_instance = msg_nameinst(msg);
2039 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2040 * Respond to 'SYN+' by queuing it on new socket.
2042 if (!msg_data_sz(msg)) {
2043 struct msghdr m = {NULL,};
2045 tsk_advance_rx_queue(sk);
2046 __tipc_send_stream(new_sock, &m, 0);
2048 __skb_dequeue(&sk->sk_receive_queue);
2049 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2050 skb_set_owner_r(buf, new_sk);
2052 release_sock(new_sk);
2059 * tipc_shutdown - shutdown socket connection
2060 * @sock: socket structure
2061 * @how: direction to close (must be SHUT_RDWR)
2063 * Terminates connection (if necessary), then purges socket's receive queue.
2065 * Returns 0 on success, errno otherwise
2067 static int tipc_shutdown(struct socket *sock, int how)
2069 struct sock *sk = sock->sk;
2070 struct net *net = sock_net(sk);
2071 struct tipc_sock *tsk = tipc_sk(sk);
2072 struct sk_buff *skb;
2076 if (how != SHUT_RDWR)
2081 switch (sock->state) {
2086 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2087 skb = __skb_dequeue(&sk->sk_receive_queue);
2089 if (TIPC_SKB_CB(skb)->handle != NULL) {
2093 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
2094 TIPC_CONN_SHUTDOWN))
2095 tipc_node_xmit_skb(net, skb, dnode,
2098 dnode = tsk_peer_node(tsk);
2100 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2101 TIPC_CONN_MSG, SHORT_H_SIZE,
2102 0, dnode, tsk_own_node(tsk),
2104 tsk->portid, TIPC_CONN_SHUTDOWN);
2105 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
2108 sock->state = SS_DISCONNECTING;
2109 tipc_node_remove_conn(net, dnode, tsk->portid);
2112 case SS_DISCONNECTING:
2114 /* Discard any unreceived messages */
2115 __skb_queue_purge(&sk->sk_receive_queue);
2117 /* Wake up anyone sleeping in poll */
2118 sk->sk_state_change(sk);
2130 static void tipc_sk_timeout(unsigned long data)
2132 struct tipc_sock *tsk = (struct tipc_sock *)data;
2133 struct sock *sk = &tsk->sk;
2134 struct sk_buff *skb = NULL;
2135 u32 peer_port, peer_node;
2136 u32 own_node = tsk_own_node(tsk);
2139 if (!tsk->connected) {
2143 peer_port = tsk_peer_port(tsk);
2144 peer_node = tsk_peer_node(tsk);
2146 if (tsk->probing_state == TIPC_CONN_PROBING) {
2147 if (!sock_owned_by_user(sk)) {
2148 sk->sk_socket->state = SS_DISCONNECTING;
2150 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2151 tsk_peer_port(tsk));
2152 sk->sk_state_change(sk);
2154 /* Try again later */
2155 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2159 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2160 INT_H_SIZE, 0, peer_node, own_node,
2161 peer_port, tsk->portid, TIPC_OK);
2162 tsk->probing_state = TIPC_CONN_PROBING;
2163 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2167 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2172 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2173 struct tipc_name_seq const *seq)
2175 struct net *net = sock_net(&tsk->sk);
2176 struct publication *publ;
2181 key = tsk->portid + tsk->pub_count + 1;
2182 if (key == tsk->portid)
2185 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2186 scope, tsk->portid, key);
2187 if (unlikely(!publ))
2190 list_add(&publ->pport_list, &tsk->publications);
2196 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2197 struct tipc_name_seq const *seq)
2199 struct net *net = sock_net(&tsk->sk);
2200 struct publication *publ;
2201 struct publication *safe;
2204 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2206 if (publ->scope != scope)
2208 if (publ->type != seq->type)
2210 if (publ->lower != seq->lower)
2212 if (publ->upper != seq->upper)
2214 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2215 publ->ref, publ->key);
2219 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2220 publ->ref, publ->key);
2223 if (list_empty(&tsk->publications))
2228 /* tipc_sk_reinit: set non-zero address in all existing sockets
2229 * when we go from standalone to network mode.
2231 void tipc_sk_reinit(struct net *net)
2233 struct tipc_net *tn = net_generic(net, tipc_net_id);
2234 const struct bucket_table *tbl;
2235 struct rhash_head *pos;
2236 struct tipc_sock *tsk;
2237 struct tipc_msg *msg;
2241 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2242 for (i = 0; i < tbl->size; i++) {
2243 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2244 spin_lock_bh(&tsk->sk.sk_lock.slock);
2246 msg_set_prevnode(msg, tn->own_addr);
2247 msg_set_orignode(msg, tn->own_addr);
2248 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2254 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2256 struct tipc_net *tn = net_generic(net, tipc_net_id);
2257 struct tipc_sock *tsk;
2260 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2262 sock_hold(&tsk->sk);
2268 static int tipc_sk_insert(struct tipc_sock *tsk)
2270 struct sock *sk = &tsk->sk;
2271 struct net *net = sock_net(sk);
2272 struct tipc_net *tn = net_generic(net, tipc_net_id);
2273 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2274 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2276 while (remaining--) {
2278 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2279 portid = TIPC_MIN_PORT;
2280 tsk->portid = portid;
2281 sock_hold(&tsk->sk);
2282 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2291 static void tipc_sk_remove(struct tipc_sock *tsk)
2293 struct sock *sk = &tsk->sk;
2294 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2296 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2297 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2302 static const struct rhashtable_params tsk_rht_params = {
2304 .head_offset = offsetof(struct tipc_sock, node),
2305 .key_offset = offsetof(struct tipc_sock, portid),
2306 .key_len = sizeof(u32), /* portid */
2307 .max_size = 1048576,
2309 .automatic_shrinking = true,
2312 int tipc_sk_rht_init(struct net *net)
2314 struct tipc_net *tn = net_generic(net, tipc_net_id);
2316 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2319 void tipc_sk_rht_destroy(struct net *net)
2321 struct tipc_net *tn = net_generic(net, tipc_net_id);
2323 /* Wait for socket readers to complete */
2326 rhashtable_destroy(&tn->sk_rht);
2330 * tipc_setsockopt - set socket option
2331 * @sock: socket structure
2332 * @lvl: option level
2333 * @opt: option identifier
2334 * @ov: pointer to new option value
2335 * @ol: length of option value
2337 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2338 * (to ease compatibility).
2340 * Returns 0 on success, errno otherwise
2342 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2343 char __user *ov, unsigned int ol)
2345 struct sock *sk = sock->sk;
2346 struct tipc_sock *tsk = tipc_sk(sk);
2350 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2352 if (lvl != SOL_TIPC)
2353 return -ENOPROTOOPT;
2354 if (ol < sizeof(value))
2356 res = get_user(value, (u32 __user *)ov);
2363 case TIPC_IMPORTANCE:
2364 res = tsk_set_importance(tsk, value);
2366 case TIPC_SRC_DROPPABLE:
2367 if (sock->type != SOCK_STREAM)
2368 tsk_set_unreliable(tsk, value);
2372 case TIPC_DEST_DROPPABLE:
2373 tsk_set_unreturnable(tsk, value);
2375 case TIPC_CONN_TIMEOUT:
2376 tipc_sk(sk)->conn_timeout = value;
2377 /* no need to set "res", since already 0 at this point */
2389 * tipc_getsockopt - get socket option
2390 * @sock: socket structure
2391 * @lvl: option level
2392 * @opt: option identifier
2393 * @ov: receptacle for option value
2394 * @ol: receptacle for length of option value
2396 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2397 * (to ease compatibility).
2399 * Returns 0 on success, errno otherwise
2401 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2402 char __user *ov, int __user *ol)
2404 struct sock *sk = sock->sk;
2405 struct tipc_sock *tsk = tipc_sk(sk);
2410 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2411 return put_user(0, ol);
2412 if (lvl != SOL_TIPC)
2413 return -ENOPROTOOPT;
2414 res = get_user(len, ol);
2421 case TIPC_IMPORTANCE:
2422 value = tsk_importance(tsk);
2424 case TIPC_SRC_DROPPABLE:
2425 value = tsk_unreliable(tsk);
2427 case TIPC_DEST_DROPPABLE:
2428 value = tsk_unreturnable(tsk);
2430 case TIPC_CONN_TIMEOUT:
2431 value = tsk->conn_timeout;
2432 /* no need to set "res", since already 0 at this point */
2434 case TIPC_NODE_RECVQ_DEPTH:
2435 value = 0; /* was tipc_queue_size, now obsolete */
2437 case TIPC_SOCK_RECVQ_DEPTH:
2438 value = skb_queue_len(&sk->sk_receive_queue);
2447 return res; /* "get" failed */
2449 if (len < sizeof(value))
2452 if (copy_to_user(ov, &value, sizeof(value)))
2455 return put_user(sizeof(value), ol);
2458 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2460 struct sock *sk = sock->sk;
2461 struct tipc_sioc_ln_req lnr;
2462 void __user *argp = (void __user *)arg;
2465 case SIOCGETLINKNAME:
2466 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2468 if (!tipc_node_get_linkname(sock_net(sk),
2469 lnr.bearer_id & 0xffff, lnr.peer,
2470 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2471 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2475 return -EADDRNOTAVAIL;
2477 return -ENOIOCTLCMD;
2481 /* Protocol switches for the various types of TIPC sockets */
2483 static const struct proto_ops msg_ops = {
2484 .owner = THIS_MODULE,
2486 .release = tipc_release,
2488 .connect = tipc_connect,
2489 .socketpair = sock_no_socketpair,
2490 .accept = sock_no_accept,
2491 .getname = tipc_getname,
2493 .ioctl = tipc_ioctl,
2494 .listen = sock_no_listen,
2495 .shutdown = tipc_shutdown,
2496 .setsockopt = tipc_setsockopt,
2497 .getsockopt = tipc_getsockopt,
2498 .sendmsg = tipc_sendmsg,
2499 .recvmsg = tipc_recvmsg,
2500 .mmap = sock_no_mmap,
2501 .sendpage = sock_no_sendpage
2504 static const struct proto_ops packet_ops = {
2505 .owner = THIS_MODULE,
2507 .release = tipc_release,
2509 .connect = tipc_connect,
2510 .socketpair = sock_no_socketpair,
2511 .accept = tipc_accept,
2512 .getname = tipc_getname,
2514 .ioctl = tipc_ioctl,
2515 .listen = tipc_listen,
2516 .shutdown = tipc_shutdown,
2517 .setsockopt = tipc_setsockopt,
2518 .getsockopt = tipc_getsockopt,
2519 .sendmsg = tipc_send_packet,
2520 .recvmsg = tipc_recvmsg,
2521 .mmap = sock_no_mmap,
2522 .sendpage = sock_no_sendpage
2525 static const struct proto_ops stream_ops = {
2526 .owner = THIS_MODULE,
2528 .release = tipc_release,
2530 .connect = tipc_connect,
2531 .socketpair = sock_no_socketpair,
2532 .accept = tipc_accept,
2533 .getname = tipc_getname,
2535 .ioctl = tipc_ioctl,
2536 .listen = tipc_listen,
2537 .shutdown = tipc_shutdown,
2538 .setsockopt = tipc_setsockopt,
2539 .getsockopt = tipc_getsockopt,
2540 .sendmsg = tipc_send_stream,
2541 .recvmsg = tipc_recv_stream,
2542 .mmap = sock_no_mmap,
2543 .sendpage = sock_no_sendpage
2546 static const struct net_proto_family tipc_family_ops = {
2547 .owner = THIS_MODULE,
2549 .create = tipc_sk_create
2552 static struct proto tipc_proto = {
2554 .owner = THIS_MODULE,
2555 .obj_size = sizeof(struct tipc_sock),
2556 .sysctl_rmem = sysctl_tipc_rmem
2560 * tipc_socket_init - initialize TIPC socket interface
2562 * Returns 0 on success, errno otherwise
2564 int tipc_socket_init(void)
2568 res = proto_register(&tipc_proto, 1);
2570 pr_err("Failed to register TIPC protocol type\n");
2574 res = sock_register(&tipc_family_ops);
2576 pr_err("Failed to register TIPC socket type\n");
2577 proto_unregister(&tipc_proto);
2585 * tipc_socket_stop - stop TIPC socket interface
2587 void tipc_socket_stop(void)
2589 sock_unregister(tipc_family_ops.family);
2590 proto_unregister(&tipc_proto);
2593 /* Caller should hold socket lock for the passed tipc socket. */
2594 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2598 struct nlattr *nest;
2600 peer_node = tsk_peer_node(tsk);
2601 peer_port = tsk_peer_port(tsk);
2603 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2605 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2607 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2610 if (tsk->conn_type != 0) {
2611 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2613 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2615 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2618 nla_nest_end(skb, nest);
2623 nla_nest_cancel(skb, nest);
2628 /* Caller should hold socket lock for the passed tipc socket. */
2629 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2630 struct tipc_sock *tsk)
2634 struct nlattr *attrs;
2635 struct net *net = sock_net(skb->sk);
2636 struct tipc_net *tn = net_generic(net, tipc_net_id);
2638 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2639 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2643 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2645 goto genlmsg_cancel;
2646 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2647 goto attr_msg_cancel;
2648 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2649 goto attr_msg_cancel;
2651 if (tsk->connected) {
2652 err = __tipc_nl_add_sk_con(skb, tsk);
2654 goto attr_msg_cancel;
2655 } else if (!list_empty(&tsk->publications)) {
2656 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2657 goto attr_msg_cancel;
2659 nla_nest_end(skb, attrs);
2660 genlmsg_end(skb, hdr);
2665 nla_nest_cancel(skb, attrs);
2667 genlmsg_cancel(skb, hdr);
2672 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2675 struct tipc_sock *tsk;
2676 const struct bucket_table *tbl;
2677 struct rhash_head *pos;
2678 struct net *net = sock_net(skb->sk);
2679 struct tipc_net *tn = net_generic(net, tipc_net_id);
2680 u32 tbl_id = cb->args[0];
2681 u32 prev_portid = cb->args[1];
2684 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2685 for (; tbl_id < tbl->size; tbl_id++) {
2686 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2687 spin_lock_bh(&tsk->sk.sk_lock.slock);
2688 if (prev_portid && prev_portid != tsk->portid) {
2689 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2693 err = __tipc_nl_add_sk(skb, cb, tsk);
2695 prev_portid = tsk->portid;
2696 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2700 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2705 cb->args[0] = tbl_id;
2706 cb->args[1] = prev_portid;
2711 /* Caller should hold socket lock for the passed tipc socket. */
2712 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2713 struct netlink_callback *cb,
2714 struct publication *publ)
2717 struct nlattr *attrs;
2719 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2720 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2724 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2726 goto genlmsg_cancel;
2728 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2729 goto attr_msg_cancel;
2730 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2731 goto attr_msg_cancel;
2732 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2733 goto attr_msg_cancel;
2734 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2735 goto attr_msg_cancel;
2737 nla_nest_end(skb, attrs);
2738 genlmsg_end(skb, hdr);
2743 nla_nest_cancel(skb, attrs);
2745 genlmsg_cancel(skb, hdr);
2750 /* Caller should hold socket lock for the passed tipc socket. */
2751 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2752 struct netlink_callback *cb,
2753 struct tipc_sock *tsk, u32 *last_publ)
2756 struct publication *p;
2759 list_for_each_entry(p, &tsk->publications, pport_list) {
2760 if (p->key == *last_publ)
2763 if (p->key != *last_publ) {
2764 /* We never set seq or call nl_dump_check_consistent()
2765 * this means that setting prev_seq here will cause the
2766 * consistence check to fail in the netlink callback
2767 * handler. Resulting in the last NLMSG_DONE message
2768 * having the NLM_F_DUMP_INTR flag set.
2775 p = list_first_entry(&tsk->publications, struct publication,
2779 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2780 err = __tipc_nl_add_sk_publ(skb, cb, p);
2782 *last_publ = p->key;
2791 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2794 u32 tsk_portid = cb->args[0];
2795 u32 last_publ = cb->args[1];
2796 u32 done = cb->args[2];
2797 struct net *net = sock_net(skb->sk);
2798 struct tipc_sock *tsk;
2801 struct nlattr **attrs;
2802 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2804 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2808 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2809 attrs[TIPC_NLA_SOCK],
2810 tipc_nl_sock_policy);
2814 if (!sock[TIPC_NLA_SOCK_REF])
2817 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2823 tsk = tipc_sk_lookup(net, tsk_portid);
2827 lock_sock(&tsk->sk);
2828 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2831 release_sock(&tsk->sk);
2834 cb->args[0] = tsk_portid;
2835 cb->args[1] = last_publ;