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 void tipc_sock_destruct(struct sock *sk);
109 static int tipc_release(struct socket *sock);
110 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
111 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
112 static void tipc_sk_timeout(unsigned long data);
113 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
114 struct tipc_name_seq const *seq);
115 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
116 struct tipc_name_seq const *seq);
117 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
118 static int tipc_sk_insert(struct tipc_sock *tsk);
119 static void tipc_sk_remove(struct tipc_sock *tsk);
120 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
122 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
124 static const struct proto_ops packet_ops;
125 static const struct proto_ops stream_ops;
126 static const struct proto_ops msg_ops;
127 static struct proto tipc_proto;
129 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
130 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
131 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
132 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
133 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
134 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
137 static const struct rhashtable_params tsk_rht_params;
140 * Revised TIPC socket locking policy:
142 * Most socket operations take the standard socket lock when they start
143 * and hold it until they finish (or until they need to sleep). Acquiring
144 * this lock grants the owner exclusive access to the fields of the socket
145 * data structures, with the exception of the backlog queue. A few socket
146 * operations can be done without taking the socket lock because they only
147 * read socket information that never changes during the life of the socket.
149 * Socket operations may acquire the lock for the associated TIPC port if they
150 * need to perform an operation on the port. If any routine needs to acquire
151 * both the socket lock and the port lock it must take the socket lock first
152 * to avoid the risk of deadlock.
154 * The dispatcher handling incoming messages cannot grab the socket lock in
155 * the standard fashion, since invoked it runs at the BH level and cannot block.
156 * Instead, it checks to see if the socket lock is currently owned by someone,
157 * and either handles the message itself or adds it to the socket's backlog
158 * queue; in the latter case the queued message is processed once the process
159 * owning the socket lock releases it.
161 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
162 * the problem of a blocked socket operation preventing any other operations
163 * from occurring. However, applications must be careful if they have
164 * multiple threads trying to send (or receive) on the same socket, as these
165 * operations might interfere with each other. For example, doing a connect
166 * and a receive at the same time might allow the receive to consume the
167 * ACK message meant for the connect. While additional work could be done
168 * to try and overcome this, it doesn't seem to be worthwhile at the present.
170 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
171 * that another operation that must be performed in a non-blocking manner is
172 * not delayed for very long because the lock has already been taken.
174 * NOTE: This code assumes that certain fields of a port/socket pair are
175 * constant over its lifetime; such fields can be examined without taking
176 * the socket lock and/or port lock, and do not need to be re-read even
177 * after resuming processing after waiting. These fields include:
179 * - pointer to socket sk structure (aka tipc_sock structure)
180 * - pointer to port structure
184 static u32 tsk_own_node(struct tipc_sock *tsk)
186 return msg_prevnode(&tsk->phdr);
189 static u32 tsk_peer_node(struct tipc_sock *tsk)
191 return msg_destnode(&tsk->phdr);
194 static u32 tsk_peer_port(struct tipc_sock *tsk)
196 return msg_destport(&tsk->phdr);
199 static bool tsk_unreliable(struct tipc_sock *tsk)
201 return msg_src_droppable(&tsk->phdr) != 0;
204 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
206 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
209 static bool tsk_unreturnable(struct tipc_sock *tsk)
211 return msg_dest_droppable(&tsk->phdr) != 0;
214 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
216 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
219 static int tsk_importance(struct tipc_sock *tsk)
221 return msg_importance(&tsk->phdr);
224 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
226 if (imp > TIPC_CRITICAL_IMPORTANCE)
228 msg_set_importance(&tsk->phdr, (u32)imp);
232 static struct tipc_sock *tipc_sk(const struct sock *sk)
234 return container_of(sk, struct tipc_sock, sk);
237 static int tsk_conn_cong(struct tipc_sock *tsk)
239 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
243 * tsk_advance_rx_queue - discard first buffer in socket receive queue
245 * Caller must hold socket lock
247 static void tsk_advance_rx_queue(struct sock *sk)
249 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
252 /* tipc_sk_respond() : send response message back to sender
254 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
258 u32 onode = tipc_own_addr(sock_net(sk));
260 if (!tipc_msg_reverse(onode, &skb, err))
263 dnode = msg_destnode(buf_msg(skb));
264 selector = msg_origport(buf_msg(skb));
265 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
269 * tsk_rej_rx_queue - reject all buffers in socket receive queue
271 * Caller must hold socket lock
273 static void tsk_rej_rx_queue(struct sock *sk)
277 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
278 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
281 /* tsk_peer_msg - verify if message was sent by connected port's peer
283 * Handles cases where the node's network address has changed from
284 * the default of <0.0.0> to its configured setting.
286 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
288 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
289 u32 peer_port = tsk_peer_port(tsk);
293 if (unlikely(!tsk->connected))
296 if (unlikely(msg_origport(msg) != peer_port))
299 orig_node = msg_orignode(msg);
300 peer_node = tsk_peer_node(tsk);
302 if (likely(orig_node == peer_node))
305 if (!orig_node && (peer_node == tn->own_addr))
308 if (!peer_node && (orig_node == tn->own_addr))
315 * tipc_sk_create - create a TIPC socket
316 * @net: network namespace (must be default network)
317 * @sock: pre-allocated socket structure
318 * @protocol: protocol indicator (must be 0)
319 * @kern: caused by kernel or by userspace?
321 * This routine creates additional data structures used by the TIPC socket,
322 * initializes them, and links them together.
324 * Returns 0 on success, errno otherwise
326 static int tipc_sk_create(struct net *net, struct socket *sock,
327 int protocol, int kern)
330 const struct proto_ops *ops;
333 struct tipc_sock *tsk;
334 struct tipc_msg *msg;
336 /* Validate arguments */
337 if (unlikely(protocol != 0))
338 return -EPROTONOSUPPORT;
340 switch (sock->type) {
343 state = SS_UNCONNECTED;
347 state = SS_UNCONNECTED;
358 /* Allocate socket's protocol area */
359 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
364 tsk->max_pkt = MAX_PKT_DEFAULT;
365 INIT_LIST_HEAD(&tsk->publications);
367 tn = net_generic(sock_net(sk), tipc_net_id);
368 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
371 /* Finish initializing socket data structures */
374 sock_init_data(sock, sk);
375 if (tipc_sk_insert(tsk)) {
376 pr_warn("Socket create failed; port numbrer exhausted\n");
379 msg_set_origport(msg, tsk->portid);
380 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
381 sk->sk_backlog_rcv = tipc_backlog_rcv;
382 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
383 sk->sk_data_ready = tipc_data_ready;
384 sk->sk_write_space = tipc_write_space;
385 sk->sk_destruct = tipc_sock_destruct;
386 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
387 tsk->sent_unacked = 0;
388 atomic_set(&tsk->dupl_rcvcnt, 0);
390 if (sock->state == SS_READY) {
391 tsk_set_unreturnable(tsk, true);
392 if (sock->type == SOCK_DGRAM)
393 tsk_set_unreliable(tsk, true);
398 static void tipc_sk_callback(struct rcu_head *head)
400 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
406 * tipc_release - destroy a TIPC socket
407 * @sock: socket to destroy
409 * This routine cleans up any messages that are still queued on the socket.
410 * For DGRAM and RDM socket types, all queued messages are rejected.
411 * For SEQPACKET and STREAM socket types, the first message is rejected
412 * and any others are discarded. (If the first message on a STREAM socket
413 * is partially-read, it is discarded and the next one is rejected instead.)
415 * NOTE: Rejected messages are not necessarily returned to the sender! They
416 * are returned or discarded according to the "destination droppable" setting
417 * specified for the message by the sender.
419 * Returns 0 on success, errno otherwise
421 static int tipc_release(struct socket *sock)
423 struct sock *sk = sock->sk;
425 struct tipc_sock *tsk;
430 * Exit if socket isn't fully initialized (occurs when a failed accept()
431 * releases a pre-allocated child socket that was never used)
441 * Reject all unreceived messages, except on an active connection
442 * (which disconnects locally & sends a 'FIN+' to peer)
444 dnode = tsk_peer_node(tsk);
445 while (sock->state != SS_DISCONNECTING) {
446 skb = __skb_dequeue(&sk->sk_receive_queue);
449 if (TIPC_SKB_CB(skb)->handle != NULL)
452 if ((sock->state == SS_CONNECTING) ||
453 (sock->state == SS_CONNECTED)) {
454 sock->state = SS_DISCONNECTING;
456 tipc_node_remove_conn(net, dnode, tsk->portid);
458 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
462 tipc_sk_withdraw(tsk, 0, NULL);
463 sk_stop_timer(sk, &sk->sk_timer);
465 if (tsk->connected) {
466 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
467 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
468 tsk_own_node(tsk), tsk_peer_port(tsk),
469 tsk->portid, TIPC_ERR_NO_PORT);
471 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
472 tipc_node_remove_conn(net, dnode, tsk->portid);
475 /* Reject any messages that accumulated in backlog queue */
476 sock->state = SS_DISCONNECTING;
479 call_rcu(&tsk->rcu, tipc_sk_callback);
486 * tipc_bind - associate or disassocate TIPC name(s) with a socket
487 * @sock: socket structure
488 * @uaddr: socket address describing name(s) and desired operation
489 * @uaddr_len: size of socket address data structure
491 * Name and name sequence binding is indicated using a positive scope value;
492 * a negative scope value unbinds the specified name. Specifying no name
493 * (i.e. a socket address length of 0) unbinds all names from the socket.
495 * Returns 0 on success, errno otherwise
497 * NOTE: This routine doesn't need to take the socket lock since it doesn't
498 * access any non-constant socket information.
500 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
503 struct sock *sk = sock->sk;
504 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
505 struct tipc_sock *tsk = tipc_sk(sk);
509 if (unlikely(!uaddr_len)) {
510 res = tipc_sk_withdraw(tsk, 0, NULL);
514 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
518 if (addr->family != AF_TIPC) {
523 if (addr->addrtype == TIPC_ADDR_NAME)
524 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
525 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
530 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
531 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
532 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
537 res = (addr->scope > 0) ?
538 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
539 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
546 * tipc_getname - get port ID of socket or peer socket
547 * @sock: socket structure
548 * @uaddr: area for returned socket address
549 * @uaddr_len: area for returned length of socket address
550 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
552 * Returns 0 on success, errno otherwise
554 * NOTE: This routine doesn't need to take the socket lock since it only
555 * accesses socket information that is unchanging (or which changes in
556 * a completely predictable manner).
558 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
559 int *uaddr_len, int peer)
561 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
562 struct tipc_sock *tsk = tipc_sk(sock->sk);
563 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
565 memset(addr, 0, sizeof(*addr));
567 if ((sock->state != SS_CONNECTED) &&
568 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
570 addr->addr.id.ref = tsk_peer_port(tsk);
571 addr->addr.id.node = tsk_peer_node(tsk);
573 addr->addr.id.ref = tsk->portid;
574 addr->addr.id.node = tn->own_addr;
577 *uaddr_len = sizeof(*addr);
578 addr->addrtype = TIPC_ADDR_ID;
579 addr->family = AF_TIPC;
581 addr->addr.name.domain = 0;
587 * tipc_poll - read and possibly block on pollmask
588 * @file: file structure associated with the socket
589 * @sock: socket for which to calculate the poll bits
592 * Returns pollmask value
595 * It appears that the usual socket locking mechanisms are not useful here
596 * since the pollmask info is potentially out-of-date the moment this routine
597 * exits. TCP and other protocols seem to rely on higher level poll routines
598 * to handle any preventable race conditions, so TIPC will do the same ...
600 * TIPC sets the returned events as follows:
602 * socket state flags set
603 * ------------ ---------
604 * unconnected no read flags
605 * POLLOUT if port is not congested
607 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
610 * connected POLLIN/POLLRDNORM if data in rx queue
611 * POLLOUT if port is not congested
613 * disconnecting POLLIN/POLLRDNORM/POLLHUP
616 * listening POLLIN if SYN in rx queue
619 * ready POLLIN/POLLRDNORM if data in rx queue
620 * [connectionless] POLLOUT (since port cannot be congested)
622 * IMPORTANT: The fact that a read or write operation is indicated does NOT
623 * imply that the operation will succeed, merely that it should be performed
624 * and will not block.
626 static unsigned int tipc_poll(struct file *file, struct socket *sock,
629 struct sock *sk = sock->sk;
630 struct tipc_sock *tsk = tipc_sk(sk);
633 sock_poll_wait(file, sk_sleep(sk), wait);
635 switch ((int)sock->state) {
642 if (!tsk->link_cong && !tsk_conn_cong(tsk))
647 if (!skb_queue_empty(&sk->sk_receive_queue))
648 mask |= (POLLIN | POLLRDNORM);
650 case SS_DISCONNECTING:
651 mask = (POLLIN | POLLRDNORM | POLLHUP);
659 * tipc_sendmcast - send multicast message
660 * @sock: socket structure
661 * @seq: destination address
662 * @msg: message to send
663 * @dsz: total length of message data
664 * @timeo: timeout to wait for wakeup
666 * Called from function tipc_sendmsg(), which has done all sanity checks
667 * Returns the number of bytes sent on success, or errno
669 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
670 struct msghdr *msg, size_t dsz, long timeo)
672 struct sock *sk = sock->sk;
673 struct tipc_sock *tsk = tipc_sk(sk);
674 struct net *net = sock_net(sk);
675 struct tipc_msg *mhdr = &tsk->phdr;
676 struct sk_buff_head pktchain;
677 struct iov_iter save = msg->msg_iter;
681 msg_set_type(mhdr, TIPC_MCAST_MSG);
682 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
683 msg_set_destport(mhdr, 0);
684 msg_set_destnode(mhdr, 0);
685 msg_set_nametype(mhdr, seq->type);
686 msg_set_namelower(mhdr, seq->lower);
687 msg_set_nameupper(mhdr, seq->upper);
688 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
690 skb_queue_head_init(&pktchain);
693 mtu = tipc_bcast_get_mtu(net);
694 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
695 if (unlikely(rc < 0))
699 rc = tipc_bcast_xmit(net, &pktchain);
703 if (rc == -ELINKCONG) {
705 rc = tipc_wait_for_sndmsg(sock, &timeo);
709 __skb_queue_purge(&pktchain);
710 if (rc == -EMSGSIZE) {
711 msg->msg_iter = save;
720 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
721 * @arrvq: queue with arriving messages, to be cloned after destination lookup
722 * @inputq: queue with cloned messages, delivered to socket after dest lookup
724 * Multi-threaded: parallel calls with reference to same queues may occur
726 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
727 struct sk_buff_head *inputq)
729 struct tipc_msg *msg;
730 struct tipc_plist dports;
732 u32 scope = TIPC_CLUSTER_SCOPE;
733 struct sk_buff_head tmpq;
735 struct sk_buff *skb, *_skb;
737 __skb_queue_head_init(&tmpq);
738 tipc_plist_init(&dports);
740 skb = tipc_skb_peek(arrvq, &inputq->lock);
741 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
743 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
745 if (in_own_node(net, msg_orignode(msg)))
746 scope = TIPC_NODE_SCOPE;
748 /* Create destination port list and message clones: */
749 tipc_nametbl_mc_translate(net,
750 msg_nametype(msg), msg_namelower(msg),
751 msg_nameupper(msg), scope, &dports);
752 portid = tipc_plist_pop(&dports);
753 for (; portid; portid = tipc_plist_pop(&dports)) {
754 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
756 msg_set_destport(buf_msg(_skb), portid);
757 __skb_queue_tail(&tmpq, _skb);
760 pr_warn("Failed to clone mcast rcv buffer\n");
762 /* Append to inputq if not already done by other thread */
763 spin_lock_bh(&inputq->lock);
764 if (skb_peek(arrvq) == skb) {
765 skb_queue_splice_tail_init(&tmpq, inputq);
766 kfree_skb(__skb_dequeue(arrvq));
768 spin_unlock_bh(&inputq->lock);
769 __skb_queue_purge(&tmpq);
772 tipc_sk_rcv(net, inputq);
776 * tipc_sk_proto_rcv - receive a connection mng protocol message
777 * @tsk: receiving socket
778 * @skb: pointer to message buffer.
780 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb)
782 struct sock *sk = &tsk->sk;
783 struct tipc_msg *hdr = buf_msg(skb);
784 int mtyp = msg_type(hdr);
787 /* Ignore if connection cannot be validated: */
788 if (!tsk_peer_msg(tsk, hdr))
791 tsk->probing_state = TIPC_CONN_OK;
793 if (mtyp == CONN_PROBE) {
794 msg_set_type(hdr, CONN_PROBE_REPLY);
795 tipc_sk_respond(sk, skb, TIPC_OK);
797 } else if (mtyp == CONN_ACK) {
798 conn_cong = tsk_conn_cong(tsk);
799 tsk->sent_unacked -= msg_msgcnt(hdr);
801 sk->sk_write_space(sk);
802 } else if (mtyp != CONN_PROBE_REPLY) {
803 pr_warn("Received unknown CONN_PROTO msg\n");
809 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
811 struct sock *sk = sock->sk;
812 struct tipc_sock *tsk = tipc_sk(sk);
817 int err = sock_error(sk);
820 if (sock->state == SS_DISCONNECTING)
824 if (signal_pending(current))
825 return sock_intr_errno(*timeo_p);
827 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
828 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
829 finish_wait(sk_sleep(sk), &wait);
835 * tipc_sendmsg - send message in connectionless manner
836 * @sock: socket structure
837 * @m: message to send
838 * @dsz: amount of user data to be sent
840 * Message must have an destination specified explicitly.
841 * Used for SOCK_RDM and SOCK_DGRAM messages,
842 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
843 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
845 * Returns the number of bytes sent on success, or errno otherwise
847 static int tipc_sendmsg(struct socket *sock,
848 struct msghdr *m, size_t dsz)
850 struct sock *sk = sock->sk;
854 ret = __tipc_sendmsg(sock, m, dsz);
860 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
862 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
863 struct sock *sk = sock->sk;
864 struct tipc_sock *tsk = tipc_sk(sk);
865 struct net *net = sock_net(sk);
866 struct tipc_msg *mhdr = &tsk->phdr;
868 struct sk_buff_head pktchain;
870 struct tipc_name_seq *seq;
871 struct iov_iter save;
876 if (dsz > TIPC_MAX_USER_MSG_SIZE)
878 if (unlikely(!dest)) {
879 if (tsk->connected && sock->state == SS_READY)
882 return -EDESTADDRREQ;
883 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
884 dest->family != AF_TIPC) {
887 if (unlikely(sock->state != SS_READY)) {
888 if (sock->state == SS_LISTENING)
890 if (sock->state != SS_UNCONNECTED)
894 if (dest->addrtype == TIPC_ADDR_NAME) {
895 tsk->conn_type = dest->addr.name.name.type;
896 tsk->conn_instance = dest->addr.name.name.instance;
899 seq = &dest->addr.nameseq;
900 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
902 if (dest->addrtype == TIPC_ADDR_MCAST) {
903 return tipc_sendmcast(sock, seq, m, dsz, timeo);
904 } else if (dest->addrtype == TIPC_ADDR_NAME) {
905 u32 type = dest->addr.name.name.type;
906 u32 inst = dest->addr.name.name.instance;
907 u32 domain = dest->addr.name.domain;
910 msg_set_type(mhdr, TIPC_NAMED_MSG);
911 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
912 msg_set_nametype(mhdr, type);
913 msg_set_nameinst(mhdr, inst);
914 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
915 dport = tipc_nametbl_translate(net, type, inst, &dnode);
916 msg_set_destnode(mhdr, dnode);
917 msg_set_destport(mhdr, dport);
918 if (unlikely(!dport && !dnode))
919 return -EHOSTUNREACH;
920 } else if (dest->addrtype == TIPC_ADDR_ID) {
921 dnode = dest->addr.id.node;
922 msg_set_type(mhdr, TIPC_DIRECT_MSG);
923 msg_set_lookup_scope(mhdr, 0);
924 msg_set_destnode(mhdr, dnode);
925 msg_set_destport(mhdr, dest->addr.id.ref);
926 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
929 skb_queue_head_init(&pktchain);
932 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
933 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
938 skb = skb_peek(&pktchain);
939 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
940 rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
942 if (sock->state != SS_READY)
943 sock->state = SS_CONNECTING;
946 if (rc == -ELINKCONG) {
948 rc = tipc_wait_for_sndmsg(sock, &timeo);
952 __skb_queue_purge(&pktchain);
953 if (rc == -EMSGSIZE) {
963 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
965 struct sock *sk = sock->sk;
966 struct tipc_sock *tsk = tipc_sk(sk);
971 int err = sock_error(sk);
974 if (sock->state == SS_DISCONNECTING)
976 else if (sock->state != SS_CONNECTED)
980 if (signal_pending(current))
981 return sock_intr_errno(*timeo_p);
983 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
984 done = sk_wait_event(sk, timeo_p,
986 !tsk_conn_cong(tsk)) ||
988 finish_wait(sk_sleep(sk), &wait);
994 * tipc_send_stream - send stream-oriented data
995 * @sock: socket structure
997 * @dsz: total length of data to be transmitted
999 * Used for SOCK_STREAM data.
1001 * Returns the number of bytes sent on success (or partial success),
1002 * or errno if no data sent
1004 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1006 struct sock *sk = sock->sk;
1010 ret = __tipc_send_stream(sock, m, dsz);
1016 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1018 struct sock *sk = sock->sk;
1019 struct net *net = sock_net(sk);
1020 struct tipc_sock *tsk = tipc_sk(sk);
1021 struct tipc_msg *mhdr = &tsk->phdr;
1022 struct sk_buff_head pktchain;
1023 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1024 u32 portid = tsk->portid;
1028 uint mtu, send, sent = 0;
1029 struct iov_iter save;
1031 /* Handle implied connection establishment */
1032 if (unlikely(dest)) {
1033 rc = __tipc_sendmsg(sock, m, dsz);
1034 if (dsz && (dsz == rc))
1035 tsk->sent_unacked = 1;
1038 if (dsz > (uint)INT_MAX)
1041 if (unlikely(sock->state != SS_CONNECTED)) {
1042 if (sock->state == SS_DISCONNECTING)
1048 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1049 dnode = tsk_peer_node(tsk);
1050 skb_queue_head_init(&pktchain);
1055 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1056 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1057 if (unlikely(rc < 0))
1061 if (likely(!tsk_conn_cong(tsk))) {
1062 rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1064 tsk->sent_unacked++;
1070 if (rc == -EMSGSIZE) {
1071 __skb_queue_purge(&pktchain);
1072 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1077 if (rc != -ELINKCONG)
1082 rc = tipc_wait_for_sndpkt(sock, &timeo);
1085 __skb_queue_purge(&pktchain);
1086 return sent ? sent : rc;
1090 * tipc_send_packet - send a connection-oriented message
1091 * @sock: socket structure
1092 * @m: message to send
1093 * @dsz: length of data to be transmitted
1095 * Used for SOCK_SEQPACKET messages.
1097 * Returns the number of bytes sent on success, or errno otherwise
1099 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1101 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1104 return tipc_send_stream(sock, m, dsz);
1107 /* tipc_sk_finish_conn - complete the setup of a connection
1109 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1112 struct sock *sk = &tsk->sk;
1113 struct net *net = sock_net(sk);
1114 struct tipc_msg *msg = &tsk->phdr;
1116 msg_set_destnode(msg, peer_node);
1117 msg_set_destport(msg, peer_port);
1118 msg_set_type(msg, TIPC_CONN_MSG);
1119 msg_set_lookup_scope(msg, 0);
1120 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1122 tsk->probing_intv = CONN_PROBING_INTERVAL;
1123 tsk->probing_state = TIPC_CONN_OK;
1125 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1126 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1127 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1131 * set_orig_addr - capture sender's address for received message
1132 * @m: descriptor for message info
1133 * @msg: received message header
1135 * Note: Address is not captured if not requested by receiver.
1137 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1139 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1142 addr->family = AF_TIPC;
1143 addr->addrtype = TIPC_ADDR_ID;
1144 memset(&addr->addr, 0, sizeof(addr->addr));
1145 addr->addr.id.ref = msg_origport(msg);
1146 addr->addr.id.node = msg_orignode(msg);
1147 addr->addr.name.domain = 0; /* could leave uninitialized */
1148 addr->scope = 0; /* could leave uninitialized */
1149 m->msg_namelen = sizeof(struct sockaddr_tipc);
1154 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1155 * @m: descriptor for message info
1156 * @msg: received message header
1157 * @tsk: TIPC port associated with message
1159 * Note: Ancillary data is not captured if not requested by receiver.
1161 * Returns 0 if successful, otherwise errno
1163 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1164 struct tipc_sock *tsk)
1172 if (likely(m->msg_controllen == 0))
1175 /* Optionally capture errored message object(s) */
1176 err = msg ? msg_errcode(msg) : 0;
1177 if (unlikely(err)) {
1179 anc_data[1] = msg_data_sz(msg);
1180 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1184 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1191 /* Optionally capture message destination object */
1192 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1193 switch (dest_type) {
1194 case TIPC_NAMED_MSG:
1196 anc_data[0] = msg_nametype(msg);
1197 anc_data[1] = msg_namelower(msg);
1198 anc_data[2] = msg_namelower(msg);
1200 case TIPC_MCAST_MSG:
1202 anc_data[0] = msg_nametype(msg);
1203 anc_data[1] = msg_namelower(msg);
1204 anc_data[2] = msg_nameupper(msg);
1207 has_name = (tsk->conn_type != 0);
1208 anc_data[0] = tsk->conn_type;
1209 anc_data[1] = tsk->conn_instance;
1210 anc_data[2] = tsk->conn_instance;
1216 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1224 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1226 struct net *net = sock_net(&tsk->sk);
1227 struct sk_buff *skb = NULL;
1228 struct tipc_msg *msg;
1229 u32 peer_port = tsk_peer_port(tsk);
1230 u32 dnode = tsk_peer_node(tsk);
1232 if (!tsk->connected)
1234 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1235 dnode, tsk_own_node(tsk), peer_port,
1236 tsk->portid, TIPC_OK);
1240 msg_set_msgcnt(msg, ack);
1241 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1244 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1246 struct sock *sk = sock->sk;
1248 long timeo = *timeop;
1252 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1253 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1254 if (sock->state == SS_DISCONNECTING) {
1259 timeo = schedule_timeout(timeo);
1263 if (!skb_queue_empty(&sk->sk_receive_queue))
1268 err = sock_intr_errno(timeo);
1269 if (signal_pending(current))
1272 finish_wait(sk_sleep(sk), &wait);
1278 * tipc_recvmsg - receive packet-oriented message
1279 * @m: descriptor for message info
1280 * @buf_len: total size of user buffer area
1281 * @flags: receive flags
1283 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1284 * If the complete message doesn't fit in user area, truncate it.
1286 * Returns size of returned message data, errno otherwise
1288 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1291 struct sock *sk = sock->sk;
1292 struct tipc_sock *tsk = tipc_sk(sk);
1293 struct sk_buff *buf;
1294 struct tipc_msg *msg;
1300 /* Catch invalid receive requests */
1301 if (unlikely(!buf_len))
1306 if (unlikely(sock->state == SS_UNCONNECTED)) {
1311 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1314 /* Look for a message in receive queue; wait if necessary */
1315 res = tipc_wait_for_rcvmsg(sock, &timeo);
1319 /* Look at first message in receive queue */
1320 buf = skb_peek(&sk->sk_receive_queue);
1322 sz = msg_data_sz(msg);
1323 err = msg_errcode(msg);
1325 /* Discard an empty non-errored message & try again */
1326 if ((!sz) && (!err)) {
1327 tsk_advance_rx_queue(sk);
1331 /* Capture sender's address (optional) */
1332 set_orig_addr(m, msg);
1334 /* Capture ancillary data (optional) */
1335 res = tipc_sk_anc_data_recv(m, msg, tsk);
1339 /* Capture message data (if valid) & compute return value (always) */
1341 if (unlikely(buf_len < sz)) {
1343 m->msg_flags |= MSG_TRUNC;
1345 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1350 if ((sock->state == SS_READY) ||
1351 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1357 /* Consume received message (optional) */
1358 if (likely(!(flags & MSG_PEEK))) {
1359 if ((sock->state != SS_READY) &&
1360 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1361 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1362 tsk->rcv_unacked = 0;
1364 tsk_advance_rx_queue(sk);
1372 * tipc_recv_stream - receive stream-oriented data
1373 * @m: descriptor for message info
1374 * @buf_len: total size of user buffer area
1375 * @flags: receive flags
1377 * Used for SOCK_STREAM messages only. If not enough data is available
1378 * will optionally wait for more; never truncates data.
1380 * Returns size of returned message data, errno otherwise
1382 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1383 size_t buf_len, int flags)
1385 struct sock *sk = sock->sk;
1386 struct tipc_sock *tsk = tipc_sk(sk);
1387 struct sk_buff *buf;
1388 struct tipc_msg *msg;
1391 int sz_to_copy, target, needed;
1396 /* Catch invalid receive attempts */
1397 if (unlikely(!buf_len))
1402 if (unlikely(sock->state == SS_UNCONNECTED)) {
1407 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1408 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1411 /* Look for a message in receive queue; wait if necessary */
1412 res = tipc_wait_for_rcvmsg(sock, &timeo);
1416 /* Look at first message in receive queue */
1417 buf = skb_peek(&sk->sk_receive_queue);
1419 sz = msg_data_sz(msg);
1420 err = msg_errcode(msg);
1422 /* Discard an empty non-errored message & try again */
1423 if ((!sz) && (!err)) {
1424 tsk_advance_rx_queue(sk);
1428 /* Optionally capture sender's address & ancillary data of first msg */
1429 if (sz_copied == 0) {
1430 set_orig_addr(m, msg);
1431 res = tipc_sk_anc_data_recv(m, msg, tsk);
1436 /* Capture message data (if valid) & compute return value (always) */
1438 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1441 needed = (buf_len - sz_copied);
1442 sz_to_copy = (sz <= needed) ? sz : needed;
1444 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1449 sz_copied += sz_to_copy;
1451 if (sz_to_copy < sz) {
1452 if (!(flags & MSG_PEEK))
1453 TIPC_SKB_CB(buf)->handle =
1454 (void *)(unsigned long)(offset + sz_to_copy);
1459 goto exit; /* can't add error msg to valid data */
1461 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1467 /* Consume received message (optional) */
1468 if (likely(!(flags & MSG_PEEK))) {
1469 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1470 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1471 tsk->rcv_unacked = 0;
1473 tsk_advance_rx_queue(sk);
1476 /* Loop around if more data is required */
1477 if ((sz_copied < buf_len) && /* didn't get all requested data */
1478 (!skb_queue_empty(&sk->sk_receive_queue) ||
1479 (sz_copied < target)) && /* and more is ready or required */
1480 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1481 (!err)) /* and haven't reached a FIN */
1486 return sz_copied ? sz_copied : res;
1490 * tipc_write_space - wake up thread if port congestion is released
1493 static void tipc_write_space(struct sock *sk)
1495 struct socket_wq *wq;
1498 wq = rcu_dereference(sk->sk_wq);
1499 if (wq_has_sleeper(wq))
1500 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1501 POLLWRNORM | POLLWRBAND);
1506 * tipc_data_ready - wake up threads to indicate messages have been received
1508 * @len: the length of messages
1510 static void tipc_data_ready(struct sock *sk)
1512 struct socket_wq *wq;
1515 wq = rcu_dereference(sk->sk_wq);
1516 if (wq_has_sleeper(wq))
1517 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1518 POLLRDNORM | POLLRDBAND);
1522 static void tipc_sock_destruct(struct sock *sk)
1524 __skb_queue_purge(&sk->sk_receive_queue);
1528 * filter_connect - Handle all incoming messages for a connection-based socket
1530 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1532 * Returns true if everything ok, false otherwise
1534 static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1536 struct sock *sk = &tsk->sk;
1537 struct net *net = sock_net(sk);
1538 struct socket *sock = sk->sk_socket;
1539 struct tipc_msg *hdr = buf_msg(skb);
1541 if (unlikely(msg_mcast(hdr)))
1544 switch ((int)sock->state) {
1547 /* Accept only connection-based messages sent by peer */
1548 if (unlikely(!tsk_peer_msg(tsk, hdr)))
1551 if (unlikely(msg_errcode(hdr))) {
1552 sock->state = SS_DISCONNECTING;
1554 /* Let timer expire on it's own */
1555 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1562 /* Accept only ACK or NACK message */
1563 if (unlikely(!msg_connected(hdr)))
1566 if (unlikely(msg_errcode(hdr))) {
1567 sock->state = SS_DISCONNECTING;
1568 sk->sk_err = ECONNREFUSED;
1572 if (unlikely(!msg_isdata(hdr))) {
1573 sock->state = SS_DISCONNECTING;
1574 sk->sk_err = EINVAL;
1578 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1579 msg_set_importance(&tsk->phdr, msg_importance(hdr));
1580 sock->state = SS_CONNECTED;
1582 /* If 'ACK+' message, add to socket receive queue */
1583 if (msg_data_sz(hdr))
1586 /* If empty 'ACK-' message, wake up sleeping connect() */
1587 if (waitqueue_active(sk_sleep(sk)))
1588 wake_up_interruptible(sk_sleep(sk));
1590 /* 'ACK-' message is neither accepted nor rejected: */
1591 msg_set_dest_droppable(hdr, 1);
1595 case SS_UNCONNECTED:
1597 /* Accept only SYN message */
1598 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1601 case SS_DISCONNECTING:
1604 pr_err("Unknown socket state %u\n", sock->state);
1610 * rcvbuf_limit - get proper overload limit of socket receive queue
1614 * For all connection oriented messages, irrespective of importance,
1615 * the default overload value (i.e. 67MB) is set as limit.
1617 * For all connectionless messages, by default new queue limits are
1620 * TIPC_LOW_IMPORTANCE (4 MB)
1621 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1622 * TIPC_HIGH_IMPORTANCE (16 MB)
1623 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1625 * Returns overload limit according to corresponding message importance
1627 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1629 struct tipc_msg *msg = buf_msg(buf);
1631 if (msg_connected(msg))
1632 return sysctl_tipc_rmem[2];
1634 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1635 msg_importance(msg);
1639 * filter_rcv - validate incoming message
1641 * @skb: pointer to message.
1643 * Enqueues message on receive queue if acceptable; optionally handles
1644 * disconnect indication for a connected socket.
1646 * Called with socket lock already taken
1648 * Returns true if message was added to socket receive queue, otherwise false
1650 static bool filter_rcv(struct sock *sk, struct sk_buff *skb)
1652 struct socket *sock = sk->sk_socket;
1653 struct tipc_sock *tsk = tipc_sk(sk);
1654 struct tipc_msg *hdr = buf_msg(skb);
1655 unsigned int limit = rcvbuf_limit(sk, skb);
1657 int usr = msg_user(hdr);
1659 if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1660 tipc_sk_proto_rcv(tsk, skb);
1664 if (unlikely(usr == SOCK_WAKEUP)) {
1667 sk->sk_write_space(sk);
1671 /* Drop if illegal message type */
1672 if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1677 /* Reject if wrong message type for current socket state */
1678 if (unlikely(sock->state == SS_READY)) {
1679 if (msg_connected(hdr)) {
1680 err = TIPC_ERR_NO_PORT;
1683 } else if (unlikely(!filter_connect(tsk, skb))) {
1684 err = TIPC_ERR_NO_PORT;
1688 /* Reject message if there isn't room to queue it */
1689 if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1690 err = TIPC_ERR_OVERLOAD;
1694 /* Enqueue message */
1695 TIPC_SKB_CB(skb)->handle = NULL;
1696 __skb_queue_tail(&sk->sk_receive_queue, skb);
1697 skb_set_owner_r(skb, sk);
1699 sk->sk_data_ready(sk);
1703 tipc_sk_respond(sk, skb, err);
1708 * tipc_backlog_rcv - handle incoming message from backlog queue
1712 * Caller must hold socket lock
1716 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1718 unsigned int truesize = skb->truesize;
1720 if (likely(filter_rcv(sk, skb)))
1721 atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1726 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1727 * inputq and try adding them to socket or backlog queue
1728 * @inputq: list of incoming buffers with potentially different destinations
1729 * @sk: socket where the buffers should be enqueued
1730 * @dport: port number for the socket
1732 * Caller must hold socket lock
1734 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1739 struct sk_buff *skb;
1740 unsigned long time_limit = jiffies + 2;
1742 while (skb_queue_len(inputq)) {
1743 if (unlikely(time_after_eq(jiffies, time_limit)))
1746 skb = tipc_skb_dequeue(inputq, dport);
1750 /* Add message directly to receive queue if possible */
1751 if (!sock_owned_by_user(sk)) {
1752 filter_rcv(sk, skb);
1756 /* Try backlog, compensating for double-counted bytes */
1757 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1758 if (sk->sk_backlog.len)
1759 atomic_set(dcnt, 0);
1760 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1761 if (likely(!sk_add_backlog(sk, skb, lim)))
1764 /* Overload => reject message back to sender */
1765 tipc_sk_respond(sk, skb, TIPC_ERR_OVERLOAD);
1771 * tipc_sk_rcv - handle a chain of incoming buffers
1772 * @inputq: buffer list containing the buffers
1773 * Consumes all buffers in list until inputq is empty
1774 * Note: may be called in multiple threads referring to the same queue
1776 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1778 u32 dnode, dport = 0;
1780 struct tipc_sock *tsk;
1782 struct sk_buff *skb;
1784 while (skb_queue_len(inputq)) {
1785 dport = tipc_skb_peek_port(inputq, dport);
1786 tsk = tipc_sk_lookup(net, dport);
1790 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1791 tipc_sk_enqueue(inputq, sk, dport);
1792 spin_unlock_bh(&sk->sk_lock.slock);
1798 /* No destination socket => dequeue skb if still there */
1799 skb = tipc_skb_dequeue(inputq, dport);
1803 /* Try secondary lookup if unresolved named message */
1804 err = TIPC_ERR_NO_PORT;
1805 if (tipc_msg_lookup_dest(net, skb, &err))
1808 /* Prepare for message rejection */
1809 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1812 dnode = msg_destnode(buf_msg(skb));
1813 tipc_node_xmit_skb(net, skb, dnode, dport);
1817 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1819 struct sock *sk = sock->sk;
1824 int err = sock_error(sk);
1829 if (signal_pending(current))
1830 return sock_intr_errno(*timeo_p);
1832 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1833 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1834 finish_wait(sk_sleep(sk), &wait);
1840 * tipc_connect - establish a connection to another TIPC port
1841 * @sock: socket structure
1842 * @dest: socket address for destination port
1843 * @destlen: size of socket address data structure
1844 * @flags: file-related flags associated with socket
1846 * Returns 0 on success, errno otherwise
1848 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1849 int destlen, int flags)
1851 struct sock *sk = sock->sk;
1852 struct tipc_sock *tsk = tipc_sk(sk);
1853 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1854 struct msghdr m = {NULL,};
1855 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1856 socket_state previous;
1861 /* DGRAM/RDM connect(), just save the destaddr */
1862 if (sock->state == SS_READY) {
1863 if (dst->family == AF_UNSPEC) {
1864 memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1866 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1869 memcpy(&tsk->remote, dest, destlen);
1876 * Reject connection attempt using multicast address
1878 * Note: send_msg() validates the rest of the address fields,
1879 * so there's no need to do it here
1881 if (dst->addrtype == TIPC_ADDR_MCAST) {
1886 previous = sock->state;
1887 switch (sock->state) {
1888 case SS_UNCONNECTED:
1889 /* Send a 'SYN-' to destination */
1891 m.msg_namelen = destlen;
1893 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1894 * indicate send_msg() is never blocked.
1897 m.msg_flags = MSG_DONTWAIT;
1899 res = __tipc_sendmsg(sock, &m, 0);
1900 if ((res < 0) && (res != -EWOULDBLOCK))
1903 /* Just entered SS_CONNECTING state; the only
1904 * difference is that return value in non-blocking
1905 * case is EINPROGRESS, rather than EALREADY.
1909 if (previous == SS_CONNECTING)
1913 timeout = msecs_to_jiffies(timeout);
1914 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1915 res = tipc_wait_for_connect(sock, &timeout);
1930 * tipc_listen - allow socket to listen for incoming connections
1931 * @sock: socket structure
1934 * Returns 0 on success, errno otherwise
1936 static int tipc_listen(struct socket *sock, int len)
1938 struct sock *sk = sock->sk;
1943 if (sock->state != SS_UNCONNECTED)
1946 sock->state = SS_LISTENING;
1954 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1956 struct sock *sk = sock->sk;
1960 /* True wake-one mechanism for incoming connections: only
1961 * one process gets woken up, not the 'whole herd'.
1962 * Since we do not 'race & poll' for established sockets
1963 * anymore, the common case will execute the loop only once.
1966 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
1967 TASK_INTERRUPTIBLE);
1968 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1970 timeo = schedule_timeout(timeo);
1974 if (!skb_queue_empty(&sk->sk_receive_queue))
1977 if (sock->state != SS_LISTENING)
1982 err = sock_intr_errno(timeo);
1983 if (signal_pending(current))
1986 finish_wait(sk_sleep(sk), &wait);
1991 * tipc_accept - wait for connection request
1992 * @sock: listening socket
1993 * @newsock: new socket that is to be connected
1994 * @flags: file-related flags associated with socket
1996 * Returns 0 on success, errno otherwise
1998 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2000 struct sock *new_sk, *sk = sock->sk;
2001 struct sk_buff *buf;
2002 struct tipc_sock *new_tsock;
2003 struct tipc_msg *msg;
2009 if (sock->state != SS_LISTENING) {
2013 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2014 res = tipc_wait_for_accept(sock, timeo);
2018 buf = skb_peek(&sk->sk_receive_queue);
2020 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2023 security_sk_clone(sock->sk, new_sock->sk);
2025 new_sk = new_sock->sk;
2026 new_tsock = tipc_sk(new_sk);
2029 /* we lock on new_sk; but lockdep sees the lock on sk */
2030 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2033 * Reject any stray messages received by new socket
2034 * before the socket lock was taken (very, very unlikely)
2036 tsk_rej_rx_queue(new_sk);
2038 /* Connect new socket to it's peer */
2039 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2040 new_sock->state = SS_CONNECTED;
2042 tsk_set_importance(new_tsock, msg_importance(msg));
2043 if (msg_named(msg)) {
2044 new_tsock->conn_type = msg_nametype(msg);
2045 new_tsock->conn_instance = msg_nameinst(msg);
2049 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2050 * Respond to 'SYN+' by queuing it on new socket.
2052 if (!msg_data_sz(msg)) {
2053 struct msghdr m = {NULL,};
2055 tsk_advance_rx_queue(sk);
2056 __tipc_send_stream(new_sock, &m, 0);
2058 __skb_dequeue(&sk->sk_receive_queue);
2059 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2060 skb_set_owner_r(buf, new_sk);
2062 release_sock(new_sk);
2069 * tipc_shutdown - shutdown socket connection
2070 * @sock: socket structure
2071 * @how: direction to close (must be SHUT_RDWR)
2073 * Terminates connection (if necessary), then purges socket's receive queue.
2075 * Returns 0 on success, errno otherwise
2077 static int tipc_shutdown(struct socket *sock, int how)
2079 struct sock *sk = sock->sk;
2080 struct net *net = sock_net(sk);
2081 struct tipc_sock *tsk = tipc_sk(sk);
2082 struct sk_buff *skb;
2083 u32 dnode = tsk_peer_node(tsk);
2084 u32 dport = tsk_peer_port(tsk);
2085 u32 onode = tipc_own_addr(net);
2086 u32 oport = tsk->portid;
2089 if (how != SHUT_RDWR)
2094 switch (sock->state) {
2099 dnode = tsk_peer_node(tsk);
2101 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2102 skb = __skb_dequeue(&sk->sk_receive_queue);
2104 if (TIPC_SKB_CB(skb)->handle != NULL) {
2108 tipc_sk_respond(sk, skb, TIPC_CONN_SHUTDOWN);
2110 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2111 TIPC_CONN_MSG, SHORT_H_SIZE,
2112 0, dnode, onode, dport, oport,
2113 TIPC_CONN_SHUTDOWN);
2115 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
2118 sock->state = SS_DISCONNECTING;
2119 tipc_node_remove_conn(net, dnode, tsk->portid);
2122 case SS_DISCONNECTING:
2124 /* Discard any unreceived messages */
2125 __skb_queue_purge(&sk->sk_receive_queue);
2127 /* Wake up anyone sleeping in poll */
2128 sk->sk_state_change(sk);
2140 static void tipc_sk_timeout(unsigned long data)
2142 struct tipc_sock *tsk = (struct tipc_sock *)data;
2143 struct sock *sk = &tsk->sk;
2144 struct sk_buff *skb = NULL;
2145 u32 peer_port, peer_node;
2146 u32 own_node = tsk_own_node(tsk);
2149 if (!tsk->connected) {
2153 peer_port = tsk_peer_port(tsk);
2154 peer_node = tsk_peer_node(tsk);
2156 if (tsk->probing_state == TIPC_CONN_PROBING) {
2157 if (!sock_owned_by_user(sk)) {
2158 sk->sk_socket->state = SS_DISCONNECTING;
2160 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2161 tsk_peer_port(tsk));
2162 sk->sk_state_change(sk);
2164 /* Try again later */
2165 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2169 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2170 INT_H_SIZE, 0, peer_node, own_node,
2171 peer_port, tsk->portid, TIPC_OK);
2172 tsk->probing_state = TIPC_CONN_PROBING;
2173 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2177 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2182 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2183 struct tipc_name_seq const *seq)
2185 struct net *net = sock_net(&tsk->sk);
2186 struct publication *publ;
2191 key = tsk->portid + tsk->pub_count + 1;
2192 if (key == tsk->portid)
2195 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2196 scope, tsk->portid, key);
2197 if (unlikely(!publ))
2200 list_add(&publ->pport_list, &tsk->publications);
2206 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2207 struct tipc_name_seq const *seq)
2209 struct net *net = sock_net(&tsk->sk);
2210 struct publication *publ;
2211 struct publication *safe;
2214 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2216 if (publ->scope != scope)
2218 if (publ->type != seq->type)
2220 if (publ->lower != seq->lower)
2222 if (publ->upper != seq->upper)
2224 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2225 publ->ref, publ->key);
2229 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2230 publ->ref, publ->key);
2233 if (list_empty(&tsk->publications))
2238 /* tipc_sk_reinit: set non-zero address in all existing sockets
2239 * when we go from standalone to network mode.
2241 void tipc_sk_reinit(struct net *net)
2243 struct tipc_net *tn = net_generic(net, tipc_net_id);
2244 const struct bucket_table *tbl;
2245 struct rhash_head *pos;
2246 struct tipc_sock *tsk;
2247 struct tipc_msg *msg;
2251 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2252 for (i = 0; i < tbl->size; i++) {
2253 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2254 spin_lock_bh(&tsk->sk.sk_lock.slock);
2256 msg_set_prevnode(msg, tn->own_addr);
2257 msg_set_orignode(msg, tn->own_addr);
2258 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2264 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2266 struct tipc_net *tn = net_generic(net, tipc_net_id);
2267 struct tipc_sock *tsk;
2270 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2272 sock_hold(&tsk->sk);
2278 static int tipc_sk_insert(struct tipc_sock *tsk)
2280 struct sock *sk = &tsk->sk;
2281 struct net *net = sock_net(sk);
2282 struct tipc_net *tn = net_generic(net, tipc_net_id);
2283 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2284 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2286 while (remaining--) {
2288 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2289 portid = TIPC_MIN_PORT;
2290 tsk->portid = portid;
2291 sock_hold(&tsk->sk);
2292 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2301 static void tipc_sk_remove(struct tipc_sock *tsk)
2303 struct sock *sk = &tsk->sk;
2304 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2306 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2307 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2312 static const struct rhashtable_params tsk_rht_params = {
2314 .head_offset = offsetof(struct tipc_sock, node),
2315 .key_offset = offsetof(struct tipc_sock, portid),
2316 .key_len = sizeof(u32), /* portid */
2317 .max_size = 1048576,
2319 .automatic_shrinking = true,
2322 int tipc_sk_rht_init(struct net *net)
2324 struct tipc_net *tn = net_generic(net, tipc_net_id);
2326 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2329 void tipc_sk_rht_destroy(struct net *net)
2331 struct tipc_net *tn = net_generic(net, tipc_net_id);
2333 /* Wait for socket readers to complete */
2336 rhashtable_destroy(&tn->sk_rht);
2340 * tipc_setsockopt - set socket option
2341 * @sock: socket structure
2342 * @lvl: option level
2343 * @opt: option identifier
2344 * @ov: pointer to new option value
2345 * @ol: length of option value
2347 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2348 * (to ease compatibility).
2350 * Returns 0 on success, errno otherwise
2352 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2353 char __user *ov, unsigned int ol)
2355 struct sock *sk = sock->sk;
2356 struct tipc_sock *tsk = tipc_sk(sk);
2360 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2362 if (lvl != SOL_TIPC)
2363 return -ENOPROTOOPT;
2364 if (ol < sizeof(value))
2366 res = get_user(value, (u32 __user *)ov);
2373 case TIPC_IMPORTANCE:
2374 res = tsk_set_importance(tsk, value);
2376 case TIPC_SRC_DROPPABLE:
2377 if (sock->type != SOCK_STREAM)
2378 tsk_set_unreliable(tsk, value);
2382 case TIPC_DEST_DROPPABLE:
2383 tsk_set_unreturnable(tsk, value);
2385 case TIPC_CONN_TIMEOUT:
2386 tipc_sk(sk)->conn_timeout = value;
2387 /* no need to set "res", since already 0 at this point */
2399 * tipc_getsockopt - get socket option
2400 * @sock: socket structure
2401 * @lvl: option level
2402 * @opt: option identifier
2403 * @ov: receptacle for option value
2404 * @ol: receptacle for length of option value
2406 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2407 * (to ease compatibility).
2409 * Returns 0 on success, errno otherwise
2411 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2412 char __user *ov, int __user *ol)
2414 struct sock *sk = sock->sk;
2415 struct tipc_sock *tsk = tipc_sk(sk);
2420 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2421 return put_user(0, ol);
2422 if (lvl != SOL_TIPC)
2423 return -ENOPROTOOPT;
2424 res = get_user(len, ol);
2431 case TIPC_IMPORTANCE:
2432 value = tsk_importance(tsk);
2434 case TIPC_SRC_DROPPABLE:
2435 value = tsk_unreliable(tsk);
2437 case TIPC_DEST_DROPPABLE:
2438 value = tsk_unreturnable(tsk);
2440 case TIPC_CONN_TIMEOUT:
2441 value = tsk->conn_timeout;
2442 /* no need to set "res", since already 0 at this point */
2444 case TIPC_NODE_RECVQ_DEPTH:
2445 value = 0; /* was tipc_queue_size, now obsolete */
2447 case TIPC_SOCK_RECVQ_DEPTH:
2448 value = skb_queue_len(&sk->sk_receive_queue);
2457 return res; /* "get" failed */
2459 if (len < sizeof(value))
2462 if (copy_to_user(ov, &value, sizeof(value)))
2465 return put_user(sizeof(value), ol);
2468 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2470 struct sock *sk = sock->sk;
2471 struct tipc_sioc_ln_req lnr;
2472 void __user *argp = (void __user *)arg;
2475 case SIOCGETLINKNAME:
2476 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2478 if (!tipc_node_get_linkname(sock_net(sk),
2479 lnr.bearer_id & 0xffff, lnr.peer,
2480 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2481 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2485 return -EADDRNOTAVAIL;
2487 return -ENOIOCTLCMD;
2491 /* Protocol switches for the various types of TIPC sockets */
2493 static const struct proto_ops msg_ops = {
2494 .owner = THIS_MODULE,
2496 .release = tipc_release,
2498 .connect = tipc_connect,
2499 .socketpair = sock_no_socketpair,
2500 .accept = sock_no_accept,
2501 .getname = tipc_getname,
2503 .ioctl = tipc_ioctl,
2504 .listen = sock_no_listen,
2505 .shutdown = tipc_shutdown,
2506 .setsockopt = tipc_setsockopt,
2507 .getsockopt = tipc_getsockopt,
2508 .sendmsg = tipc_sendmsg,
2509 .recvmsg = tipc_recvmsg,
2510 .mmap = sock_no_mmap,
2511 .sendpage = sock_no_sendpage
2514 static const struct proto_ops packet_ops = {
2515 .owner = THIS_MODULE,
2517 .release = tipc_release,
2519 .connect = tipc_connect,
2520 .socketpair = sock_no_socketpair,
2521 .accept = tipc_accept,
2522 .getname = tipc_getname,
2524 .ioctl = tipc_ioctl,
2525 .listen = tipc_listen,
2526 .shutdown = tipc_shutdown,
2527 .setsockopt = tipc_setsockopt,
2528 .getsockopt = tipc_getsockopt,
2529 .sendmsg = tipc_send_packet,
2530 .recvmsg = tipc_recvmsg,
2531 .mmap = sock_no_mmap,
2532 .sendpage = sock_no_sendpage
2535 static const struct proto_ops stream_ops = {
2536 .owner = THIS_MODULE,
2538 .release = tipc_release,
2540 .connect = tipc_connect,
2541 .socketpair = sock_no_socketpair,
2542 .accept = tipc_accept,
2543 .getname = tipc_getname,
2545 .ioctl = tipc_ioctl,
2546 .listen = tipc_listen,
2547 .shutdown = tipc_shutdown,
2548 .setsockopt = tipc_setsockopt,
2549 .getsockopt = tipc_getsockopt,
2550 .sendmsg = tipc_send_stream,
2551 .recvmsg = tipc_recv_stream,
2552 .mmap = sock_no_mmap,
2553 .sendpage = sock_no_sendpage
2556 static const struct net_proto_family tipc_family_ops = {
2557 .owner = THIS_MODULE,
2559 .create = tipc_sk_create
2562 static struct proto tipc_proto = {
2564 .owner = THIS_MODULE,
2565 .obj_size = sizeof(struct tipc_sock),
2566 .sysctl_rmem = sysctl_tipc_rmem
2570 * tipc_socket_init - initialize TIPC socket interface
2572 * Returns 0 on success, errno otherwise
2574 int tipc_socket_init(void)
2578 res = proto_register(&tipc_proto, 1);
2580 pr_err("Failed to register TIPC protocol type\n");
2584 res = sock_register(&tipc_family_ops);
2586 pr_err("Failed to register TIPC socket type\n");
2587 proto_unregister(&tipc_proto);
2595 * tipc_socket_stop - stop TIPC socket interface
2597 void tipc_socket_stop(void)
2599 sock_unregister(tipc_family_ops.family);
2600 proto_unregister(&tipc_proto);
2603 /* Caller should hold socket lock for the passed tipc socket. */
2604 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2608 struct nlattr *nest;
2610 peer_node = tsk_peer_node(tsk);
2611 peer_port = tsk_peer_port(tsk);
2613 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2615 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2617 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2620 if (tsk->conn_type != 0) {
2621 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2623 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2625 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2628 nla_nest_end(skb, nest);
2633 nla_nest_cancel(skb, nest);
2638 /* Caller should hold socket lock for the passed tipc socket. */
2639 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2640 struct tipc_sock *tsk)
2644 struct nlattr *attrs;
2645 struct net *net = sock_net(skb->sk);
2646 struct tipc_net *tn = net_generic(net, tipc_net_id);
2648 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2649 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2653 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2655 goto genlmsg_cancel;
2656 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2657 goto attr_msg_cancel;
2658 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2659 goto attr_msg_cancel;
2661 if (tsk->connected) {
2662 err = __tipc_nl_add_sk_con(skb, tsk);
2664 goto attr_msg_cancel;
2665 } else if (!list_empty(&tsk->publications)) {
2666 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2667 goto attr_msg_cancel;
2669 nla_nest_end(skb, attrs);
2670 genlmsg_end(skb, hdr);
2675 nla_nest_cancel(skb, attrs);
2677 genlmsg_cancel(skb, hdr);
2682 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2685 struct tipc_sock *tsk;
2686 const struct bucket_table *tbl;
2687 struct rhash_head *pos;
2688 struct net *net = sock_net(skb->sk);
2689 struct tipc_net *tn = net_generic(net, tipc_net_id);
2690 u32 tbl_id = cb->args[0];
2691 u32 prev_portid = cb->args[1];
2694 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2695 for (; tbl_id < tbl->size; tbl_id++) {
2696 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2697 spin_lock_bh(&tsk->sk.sk_lock.slock);
2698 if (prev_portid && prev_portid != tsk->portid) {
2699 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2703 err = __tipc_nl_add_sk(skb, cb, tsk);
2705 prev_portid = tsk->portid;
2706 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2710 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2715 cb->args[0] = tbl_id;
2716 cb->args[1] = prev_portid;
2721 /* Caller should hold socket lock for the passed tipc socket. */
2722 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2723 struct netlink_callback *cb,
2724 struct publication *publ)
2727 struct nlattr *attrs;
2729 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2730 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2734 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2736 goto genlmsg_cancel;
2738 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2739 goto attr_msg_cancel;
2740 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2741 goto attr_msg_cancel;
2742 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2743 goto attr_msg_cancel;
2744 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2745 goto attr_msg_cancel;
2747 nla_nest_end(skb, attrs);
2748 genlmsg_end(skb, hdr);
2753 nla_nest_cancel(skb, attrs);
2755 genlmsg_cancel(skb, hdr);
2760 /* Caller should hold socket lock for the passed tipc socket. */
2761 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2762 struct netlink_callback *cb,
2763 struct tipc_sock *tsk, u32 *last_publ)
2766 struct publication *p;
2769 list_for_each_entry(p, &tsk->publications, pport_list) {
2770 if (p->key == *last_publ)
2773 if (p->key != *last_publ) {
2774 /* We never set seq or call nl_dump_check_consistent()
2775 * this means that setting prev_seq here will cause the
2776 * consistence check to fail in the netlink callback
2777 * handler. Resulting in the last NLMSG_DONE message
2778 * having the NLM_F_DUMP_INTR flag set.
2785 p = list_first_entry(&tsk->publications, struct publication,
2789 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2790 err = __tipc_nl_add_sk_publ(skb, cb, p);
2792 *last_publ = p->key;
2801 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2804 u32 tsk_portid = cb->args[0];
2805 u32 last_publ = cb->args[1];
2806 u32 done = cb->args[2];
2807 struct net *net = sock_net(skb->sk);
2808 struct tipc_sock *tsk;
2811 struct nlattr **attrs;
2812 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2814 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2818 if (!attrs[TIPC_NLA_SOCK])
2821 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2822 attrs[TIPC_NLA_SOCK],
2823 tipc_nl_sock_policy);
2827 if (!sock[TIPC_NLA_SOCK_REF])
2830 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2836 tsk = tipc_sk_lookup(net, tsk_portid);
2840 lock_sock(&tsk->sk);
2841 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2844 release_sock(&tsk->sk);
2847 cb->args[0] = tsk_portid;
2848 cb->args[1] = last_publ;