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"
47 #define SS_LISTENING -1 /* socket is listening */
48 #define SS_READY -2 /* socket is connectionless */
50 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
52 #define TIPC_FWD_MSG 1
53 #define TIPC_CONN_OK 0
54 #define TIPC_CONN_PROBING 1
55 #define TIPC_MAX_PORT 0xffffffff
56 #define TIPC_MIN_PORT 1
59 * struct tipc_sock - TIPC socket structure
60 * @sk: socket - interacts with 'port' and with user via the socket API
61 * @connected: non-zero if port is currently connected to a peer port
62 * @conn_type: TIPC type used when connection was established
63 * @conn_instance: TIPC instance used when connection was established
64 * @published: non-zero if port has one or more associated names
65 * @max_pkt: maximum packet size "hint" used when building messages sent by port
66 * @portid: unique port identity in TIPC socket hash table
67 * @phdr: preformatted message header used when sending messages
68 * @port_list: adjacent ports in TIPC's global list of ports
69 * @publications: list of publications for port
70 * @pub_count: total # of publications port has made during its lifetime
73 * @conn_timeout: the time we can wait for an unresponded setup request
74 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
75 * @link_cong: non-zero if owner must sleep because of link congestion
76 * @sent_unacked: # messages sent by socket, and not yet acked by peer
77 * @rcv_unacked: # messages read by user, but not yet acked back to peer
78 * @remote: 'connected' peer for dgram/rdm
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
91 struct list_head sock_list;
92 struct list_head publications;
95 unsigned long probing_intv;
104 struct sockaddr_tipc remote;
105 struct rhash_head node;
109 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
110 static void tipc_data_ready(struct sock *sk);
111 static void tipc_write_space(struct sock *sk);
112 static void tipc_sock_destruct(struct sock *sk);
113 static int tipc_release(struct socket *sock);
114 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
115 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
116 static void tipc_sk_timeout(unsigned long data);
117 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
118 struct tipc_name_seq const *seq);
119 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
120 struct tipc_name_seq const *seq);
121 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
122 static int tipc_sk_insert(struct tipc_sock *tsk);
123 static void tipc_sk_remove(struct tipc_sock *tsk);
124 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
126 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
128 static const struct proto_ops packet_ops;
129 static const struct proto_ops stream_ops;
130 static const struct proto_ops msg_ops;
131 static struct proto tipc_proto;
133 static const struct rhashtable_params tsk_rht_params;
136 * Revised TIPC socket locking policy:
138 * Most socket operations take the standard socket lock when they start
139 * and hold it until they finish (or until they need to sleep). Acquiring
140 * this lock grants the owner exclusive access to the fields of the socket
141 * data structures, with the exception of the backlog queue. A few socket
142 * operations can be done without taking the socket lock because they only
143 * read socket information that never changes during the life of the socket.
145 * Socket operations may acquire the lock for the associated TIPC port if they
146 * need to perform an operation on the port. If any routine needs to acquire
147 * both the socket lock and the port lock it must take the socket lock first
148 * to avoid the risk of deadlock.
150 * The dispatcher handling incoming messages cannot grab the socket lock in
151 * the standard fashion, since invoked it runs at the BH level and cannot block.
152 * Instead, it checks to see if the socket lock is currently owned by someone,
153 * and either handles the message itself or adds it to the socket's backlog
154 * queue; in the latter case the queued message is processed once the process
155 * owning the socket lock releases it.
157 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
158 * the problem of a blocked socket operation preventing any other operations
159 * from occurring. However, applications must be careful if they have
160 * multiple threads trying to send (or receive) on the same socket, as these
161 * operations might interfere with each other. For example, doing a connect
162 * and a receive at the same time might allow the receive to consume the
163 * ACK message meant for the connect. While additional work could be done
164 * to try and overcome this, it doesn't seem to be worthwhile at the present.
166 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
167 * that another operation that must be performed in a non-blocking manner is
168 * not delayed for very long because the lock has already been taken.
170 * NOTE: This code assumes that certain fields of a port/socket pair are
171 * constant over its lifetime; such fields can be examined without taking
172 * the socket lock and/or port lock, and do not need to be re-read even
173 * after resuming processing after waiting. These fields include:
175 * - pointer to socket sk structure (aka tipc_sock structure)
176 * - pointer to port structure
180 static u32 tsk_own_node(struct tipc_sock *tsk)
182 return msg_prevnode(&tsk->phdr);
185 static u32 tsk_peer_node(struct tipc_sock *tsk)
187 return msg_destnode(&tsk->phdr);
190 static u32 tsk_peer_port(struct tipc_sock *tsk)
192 return msg_destport(&tsk->phdr);
195 static bool tsk_unreliable(struct tipc_sock *tsk)
197 return msg_src_droppable(&tsk->phdr) != 0;
200 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
202 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
205 static bool tsk_unreturnable(struct tipc_sock *tsk)
207 return msg_dest_droppable(&tsk->phdr) != 0;
210 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
212 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
215 static int tsk_importance(struct tipc_sock *tsk)
217 return msg_importance(&tsk->phdr);
220 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
222 if (imp > TIPC_CRITICAL_IMPORTANCE)
224 msg_set_importance(&tsk->phdr, (u32)imp);
228 static struct tipc_sock *tipc_sk(const struct sock *sk)
230 return container_of(sk, struct tipc_sock, sk);
233 static bool tsk_conn_cong(struct tipc_sock *tsk)
235 return tsk->snt_unacked >= tsk->snd_win;
238 /* tsk_blocks(): translate a buffer size in bytes to number of
239 * advertisable blocks, taking into account the ratio truesize(len)/len
240 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
242 static u16 tsk_adv_blocks(int len)
244 return len / FLOWCTL_BLK_SZ / 4;
247 /* tsk_inc(): increment counter for sent or received data
248 * - If block based flow control is not supported by peer we
249 * fall back to message based ditto, incrementing the counter
251 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
253 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
254 return ((msglen / FLOWCTL_BLK_SZ) + 1);
259 * tsk_advance_rx_queue - discard first buffer in socket receive queue
261 * Caller must hold socket lock
263 static void tsk_advance_rx_queue(struct sock *sk)
265 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
268 /* tipc_sk_respond() : send response message back to sender
270 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
274 u32 onode = tipc_own_addr(sock_net(sk));
276 if (!tipc_msg_reverse(onode, &skb, err))
279 dnode = msg_destnode(buf_msg(skb));
280 selector = msg_origport(buf_msg(skb));
281 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
285 * tsk_rej_rx_queue - reject all buffers in socket receive queue
287 * Caller must hold socket lock
289 static void tsk_rej_rx_queue(struct sock *sk)
293 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
294 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
297 /* tsk_peer_msg - verify if message was sent by connected port's peer
299 * Handles cases where the node's network address has changed from
300 * the default of <0.0.0> to its configured setting.
302 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
304 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
305 u32 peer_port = tsk_peer_port(tsk);
309 if (unlikely(!tsk->connected))
312 if (unlikely(msg_origport(msg) != peer_port))
315 orig_node = msg_orignode(msg);
316 peer_node = tsk_peer_node(tsk);
318 if (likely(orig_node == peer_node))
321 if (!orig_node && (peer_node == tn->own_addr))
324 if (!peer_node && (orig_node == tn->own_addr))
331 * tipc_sk_create - create a TIPC socket
332 * @net: network namespace (must be default network)
333 * @sock: pre-allocated socket structure
334 * @protocol: protocol indicator (must be 0)
335 * @kern: caused by kernel or by userspace?
337 * This routine creates additional data structures used by the TIPC socket,
338 * initializes them, and links them together.
340 * Returns 0 on success, errno otherwise
342 static int tipc_sk_create(struct net *net, struct socket *sock,
343 int protocol, int kern)
346 const struct proto_ops *ops;
349 struct tipc_sock *tsk;
350 struct tipc_msg *msg;
352 /* Validate arguments */
353 if (unlikely(protocol != 0))
354 return -EPROTONOSUPPORT;
356 switch (sock->type) {
359 state = SS_UNCONNECTED;
363 state = SS_UNCONNECTED;
374 /* Allocate socket's protocol area */
375 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
380 tsk->max_pkt = MAX_PKT_DEFAULT;
381 INIT_LIST_HEAD(&tsk->publications);
383 tn = net_generic(sock_net(sk), tipc_net_id);
384 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
387 /* Finish initializing socket data structures */
390 sock_init_data(sock, sk);
391 if (tipc_sk_insert(tsk)) {
392 pr_warn("Socket create failed; port number exhausted\n");
395 msg_set_origport(msg, tsk->portid);
396 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
397 sk->sk_backlog_rcv = tipc_backlog_rcv;
398 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
399 sk->sk_data_ready = tipc_data_ready;
400 sk->sk_write_space = tipc_write_space;
401 sk->sk_destruct = tipc_sock_destruct;
402 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
403 atomic_set(&tsk->dupl_rcvcnt, 0);
405 /* Start out with safe limits until we receive an advertised window */
406 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
407 tsk->rcv_win = tsk->snd_win;
409 if (sock->state == SS_READY) {
410 tsk_set_unreturnable(tsk, true);
411 if (sock->type == SOCK_DGRAM)
412 tsk_set_unreliable(tsk, true);
417 static void tipc_sk_callback(struct rcu_head *head)
419 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
425 * tipc_release - destroy a TIPC socket
426 * @sock: socket to destroy
428 * This routine cleans up any messages that are still queued on the socket.
429 * For DGRAM and RDM socket types, all queued messages are rejected.
430 * For SEQPACKET and STREAM socket types, the first message is rejected
431 * and any others are discarded. (If the first message on a STREAM socket
432 * is partially-read, it is discarded and the next one is rejected instead.)
434 * NOTE: Rejected messages are not necessarily returned to the sender! They
435 * are returned or discarded according to the "destination droppable" setting
436 * specified for the message by the sender.
438 * Returns 0 on success, errno otherwise
440 static int tipc_release(struct socket *sock)
442 struct sock *sk = sock->sk;
444 struct tipc_sock *tsk;
449 * Exit if socket isn't fully initialized (occurs when a failed accept()
450 * releases a pre-allocated child socket that was never used)
460 * Reject all unreceived messages, except on an active connection
461 * (which disconnects locally & sends a 'FIN+' to peer)
463 dnode = tsk_peer_node(tsk);
464 while (sock->state != SS_DISCONNECTING) {
465 skb = __skb_dequeue(&sk->sk_receive_queue);
468 if (TIPC_SKB_CB(skb)->bytes_read)
471 if ((sock->state == SS_CONNECTING) ||
472 (sock->state == SS_CONNECTED)) {
473 sock->state = SS_DISCONNECTING;
475 tipc_node_remove_conn(net, dnode, tsk->portid);
477 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
481 tipc_sk_withdraw(tsk, 0, NULL);
482 sk_stop_timer(sk, &sk->sk_timer);
484 if (tsk->connected) {
485 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
486 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
487 tsk_own_node(tsk), tsk_peer_port(tsk),
488 tsk->portid, TIPC_ERR_NO_PORT);
490 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
491 tipc_node_remove_conn(net, dnode, tsk->portid);
494 /* Reject any messages that accumulated in backlog queue */
495 sock->state = SS_DISCONNECTING;
498 call_rcu(&tsk->rcu, tipc_sk_callback);
505 * tipc_bind - associate or disassocate TIPC name(s) with a socket
506 * @sock: socket structure
507 * @uaddr: socket address describing name(s) and desired operation
508 * @uaddr_len: size of socket address data structure
510 * Name and name sequence binding is indicated using a positive scope value;
511 * a negative scope value unbinds the specified name. Specifying no name
512 * (i.e. a socket address length of 0) unbinds all names from the socket.
514 * Returns 0 on success, errno otherwise
516 * NOTE: This routine doesn't need to take the socket lock since it doesn't
517 * access any non-constant socket information.
519 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
522 struct sock *sk = sock->sk;
523 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
524 struct tipc_sock *tsk = tipc_sk(sk);
528 if (unlikely(!uaddr_len)) {
529 res = tipc_sk_withdraw(tsk, 0, NULL);
533 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
537 if (addr->family != AF_TIPC) {
542 if (addr->addrtype == TIPC_ADDR_NAME)
543 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
544 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
549 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
550 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
551 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
556 res = (addr->scope > 0) ?
557 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
558 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
565 * tipc_getname - get port ID of socket or peer socket
566 * @sock: socket structure
567 * @uaddr: area for returned socket address
568 * @uaddr_len: area for returned length of socket address
569 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
571 * Returns 0 on success, errno otherwise
573 * NOTE: This routine doesn't need to take the socket lock since it only
574 * accesses socket information that is unchanging (or which changes in
575 * a completely predictable manner).
577 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
578 int *uaddr_len, int peer)
580 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
581 struct tipc_sock *tsk = tipc_sk(sock->sk);
582 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
584 memset(addr, 0, sizeof(*addr));
586 if ((sock->state != SS_CONNECTED) &&
587 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
589 addr->addr.id.ref = tsk_peer_port(tsk);
590 addr->addr.id.node = tsk_peer_node(tsk);
592 addr->addr.id.ref = tsk->portid;
593 addr->addr.id.node = tn->own_addr;
596 *uaddr_len = sizeof(*addr);
597 addr->addrtype = TIPC_ADDR_ID;
598 addr->family = AF_TIPC;
600 addr->addr.name.domain = 0;
606 * tipc_poll - read and possibly block on pollmask
607 * @file: file structure associated with the socket
608 * @sock: socket for which to calculate the poll bits
611 * Returns pollmask value
614 * It appears that the usual socket locking mechanisms are not useful here
615 * since the pollmask info is potentially out-of-date the moment this routine
616 * exits. TCP and other protocols seem to rely on higher level poll routines
617 * to handle any preventable race conditions, so TIPC will do the same ...
619 * TIPC sets the returned events as follows:
621 * socket state flags set
622 * ------------ ---------
623 * unconnected no read flags
624 * POLLOUT if port is not congested
626 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
629 * connected POLLIN/POLLRDNORM if data in rx queue
630 * POLLOUT if port is not congested
632 * disconnecting POLLIN/POLLRDNORM/POLLHUP
635 * listening POLLIN if SYN in rx queue
638 * ready POLLIN/POLLRDNORM if data in rx queue
639 * [connectionless] POLLOUT (since port cannot be congested)
641 * IMPORTANT: The fact that a read or write operation is indicated does NOT
642 * imply that the operation will succeed, merely that it should be performed
643 * and will not block.
645 static unsigned int tipc_poll(struct file *file, struct socket *sock,
648 struct sock *sk = sock->sk;
649 struct tipc_sock *tsk = tipc_sk(sk);
652 sock_poll_wait(file, sk_sleep(sk), wait);
654 switch ((int)sock->state) {
661 if (!tsk->link_cong && !tsk_conn_cong(tsk))
666 if (!skb_queue_empty(&sk->sk_receive_queue))
667 mask |= (POLLIN | POLLRDNORM);
669 case SS_DISCONNECTING:
670 mask = (POLLIN | POLLRDNORM | POLLHUP);
678 * tipc_sendmcast - send multicast message
679 * @sock: socket structure
680 * @seq: destination address
681 * @msg: message to send
682 * @dsz: total length of message data
683 * @timeo: timeout to wait for wakeup
685 * Called from function tipc_sendmsg(), which has done all sanity checks
686 * Returns the number of bytes sent on success, or errno
688 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
689 struct msghdr *msg, size_t dsz, long timeo)
691 struct sock *sk = sock->sk;
692 struct tipc_sock *tsk = tipc_sk(sk);
693 struct net *net = sock_net(sk);
694 struct tipc_msg *mhdr = &tsk->phdr;
695 struct sk_buff_head pktchain;
696 struct iov_iter save = msg->msg_iter;
700 if (!timeo && tsk->link_cong)
703 msg_set_type(mhdr, TIPC_MCAST_MSG);
704 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
705 msg_set_destport(mhdr, 0);
706 msg_set_destnode(mhdr, 0);
707 msg_set_nametype(mhdr, seq->type);
708 msg_set_namelower(mhdr, seq->lower);
709 msg_set_nameupper(mhdr, seq->upper);
710 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
712 skb_queue_head_init(&pktchain);
715 mtu = tipc_bcast_get_mtu(net);
716 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
717 if (unlikely(rc < 0))
721 rc = tipc_bcast_xmit(net, &pktchain);
725 if (rc == -ELINKCONG) {
727 rc = tipc_wait_for_sndmsg(sock, &timeo);
731 __skb_queue_purge(&pktchain);
732 if (rc == -EMSGSIZE) {
733 msg->msg_iter = save;
742 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
743 * @arrvq: queue with arriving messages, to be cloned after destination lookup
744 * @inputq: queue with cloned messages, delivered to socket after dest lookup
746 * Multi-threaded: parallel calls with reference to same queues may occur
748 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
749 struct sk_buff_head *inputq)
751 struct tipc_msg *msg;
752 struct tipc_plist dports;
754 u32 scope = TIPC_CLUSTER_SCOPE;
755 struct sk_buff_head tmpq;
757 struct sk_buff *skb, *_skb;
759 __skb_queue_head_init(&tmpq);
760 tipc_plist_init(&dports);
762 skb = tipc_skb_peek(arrvq, &inputq->lock);
763 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
765 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
767 if (in_own_node(net, msg_orignode(msg)))
768 scope = TIPC_NODE_SCOPE;
770 /* Create destination port list and message clones: */
771 tipc_nametbl_mc_translate(net,
772 msg_nametype(msg), msg_namelower(msg),
773 msg_nameupper(msg), scope, &dports);
774 portid = tipc_plist_pop(&dports);
775 for (; portid; portid = tipc_plist_pop(&dports)) {
776 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
778 msg_set_destport(buf_msg(_skb), portid);
779 __skb_queue_tail(&tmpq, _skb);
782 pr_warn("Failed to clone mcast rcv buffer\n");
784 /* Append to inputq if not already done by other thread */
785 spin_lock_bh(&inputq->lock);
786 if (skb_peek(arrvq) == skb) {
787 skb_queue_splice_tail_init(&tmpq, inputq);
788 kfree_skb(__skb_dequeue(arrvq));
790 spin_unlock_bh(&inputq->lock);
791 __skb_queue_purge(&tmpq);
794 tipc_sk_rcv(net, inputq);
798 * tipc_sk_proto_rcv - receive a connection mng protocol message
799 * @tsk: receiving socket
800 * @skb: pointer to message buffer.
802 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
803 struct sk_buff_head *xmitq)
805 struct sock *sk = &tsk->sk;
806 u32 onode = tsk_own_node(tsk);
807 struct tipc_msg *hdr = buf_msg(skb);
808 int mtyp = msg_type(hdr);
811 /* Ignore if connection cannot be validated: */
812 if (!tsk_peer_msg(tsk, hdr))
815 tsk->probing_state = TIPC_CONN_OK;
817 if (mtyp == CONN_PROBE) {
818 msg_set_type(hdr, CONN_PROBE_REPLY);
819 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
820 __skb_queue_tail(xmitq, skb);
822 } else if (mtyp == CONN_ACK) {
823 conn_cong = tsk_conn_cong(tsk);
824 tsk->snt_unacked -= msg_conn_ack(hdr);
825 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
826 tsk->snd_win = msg_adv_win(hdr);
828 sk->sk_write_space(sk);
829 } else if (mtyp != CONN_PROBE_REPLY) {
830 pr_warn("Received unknown CONN_PROTO msg\n");
836 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
838 struct sock *sk = sock->sk;
839 struct tipc_sock *tsk = tipc_sk(sk);
844 int err = sock_error(sk);
847 if (sock->state == SS_DISCONNECTING)
851 if (signal_pending(current))
852 return sock_intr_errno(*timeo_p);
854 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
855 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
856 finish_wait(sk_sleep(sk), &wait);
862 * tipc_sendmsg - send message in connectionless manner
863 * @sock: socket structure
864 * @m: message to send
865 * @dsz: amount of user data to be sent
867 * Message must have an destination specified explicitly.
868 * Used for SOCK_RDM and SOCK_DGRAM messages,
869 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
870 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
872 * Returns the number of bytes sent on success, or errno otherwise
874 static int tipc_sendmsg(struct socket *sock,
875 struct msghdr *m, size_t dsz)
877 struct sock *sk = sock->sk;
881 ret = __tipc_sendmsg(sock, m, dsz);
887 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
889 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
890 struct sock *sk = sock->sk;
891 struct tipc_sock *tsk = tipc_sk(sk);
892 struct net *net = sock_net(sk);
893 struct tipc_msg *mhdr = &tsk->phdr;
895 struct sk_buff_head pktchain;
897 struct tipc_name_seq *seq;
898 struct iov_iter save;
903 if (dsz > TIPC_MAX_USER_MSG_SIZE)
905 if (unlikely(!dest)) {
906 if (tsk->connected && sock->state == SS_READY)
909 return -EDESTADDRREQ;
910 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
911 dest->family != AF_TIPC) {
914 if (unlikely(sock->state != SS_READY)) {
915 if (sock->state == SS_LISTENING)
917 if (sock->state != SS_UNCONNECTED)
921 if (dest->addrtype == TIPC_ADDR_NAME) {
922 tsk->conn_type = dest->addr.name.name.type;
923 tsk->conn_instance = dest->addr.name.name.instance;
926 seq = &dest->addr.nameseq;
927 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
929 if (dest->addrtype == TIPC_ADDR_MCAST) {
930 return tipc_sendmcast(sock, seq, m, dsz, timeo);
931 } else if (dest->addrtype == TIPC_ADDR_NAME) {
932 u32 type = dest->addr.name.name.type;
933 u32 inst = dest->addr.name.name.instance;
934 u32 domain = dest->addr.name.domain;
937 msg_set_type(mhdr, TIPC_NAMED_MSG);
938 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
939 msg_set_nametype(mhdr, type);
940 msg_set_nameinst(mhdr, inst);
941 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
942 dport = tipc_nametbl_translate(net, type, inst, &dnode);
943 msg_set_destnode(mhdr, dnode);
944 msg_set_destport(mhdr, dport);
945 if (unlikely(!dport && !dnode))
946 return -EHOSTUNREACH;
947 } else if (dest->addrtype == TIPC_ADDR_ID) {
948 dnode = dest->addr.id.node;
949 msg_set_type(mhdr, TIPC_DIRECT_MSG);
950 msg_set_lookup_scope(mhdr, 0);
951 msg_set_destnode(mhdr, dnode);
952 msg_set_destport(mhdr, dest->addr.id.ref);
953 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
956 skb_queue_head_init(&pktchain);
959 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
960 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
965 skb = skb_peek(&pktchain);
966 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
967 rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
969 if (sock->state != SS_READY)
970 sock->state = SS_CONNECTING;
973 if (rc == -ELINKCONG) {
975 rc = tipc_wait_for_sndmsg(sock, &timeo);
979 __skb_queue_purge(&pktchain);
980 if (rc == -EMSGSIZE) {
990 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
992 struct sock *sk = sock->sk;
993 struct tipc_sock *tsk = tipc_sk(sk);
998 int err = sock_error(sk);
1001 if (sock->state == SS_DISCONNECTING)
1003 else if (sock->state != SS_CONNECTED)
1007 if (signal_pending(current))
1008 return sock_intr_errno(*timeo_p);
1010 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1011 done = sk_wait_event(sk, timeo_p,
1013 !tsk_conn_cong(tsk)) ||
1015 finish_wait(sk_sleep(sk), &wait);
1021 * tipc_send_stream - send stream-oriented data
1022 * @sock: socket structure
1024 * @dsz: total length of data to be transmitted
1026 * Used for SOCK_STREAM data.
1028 * Returns the number of bytes sent on success (or partial success),
1029 * or errno if no data sent
1031 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1033 struct sock *sk = sock->sk;
1037 ret = __tipc_send_stream(sock, m, dsz);
1043 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1045 struct sock *sk = sock->sk;
1046 struct net *net = sock_net(sk);
1047 struct tipc_sock *tsk = tipc_sk(sk);
1048 struct tipc_msg *mhdr = &tsk->phdr;
1049 struct sk_buff_head pktchain;
1050 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1051 u32 portid = tsk->portid;
1055 uint mtu, send, sent = 0;
1056 struct iov_iter save;
1057 int hlen = MIN_H_SIZE;
1059 /* Handle implied connection establishment */
1060 if (unlikely(dest)) {
1061 rc = __tipc_sendmsg(sock, m, dsz);
1062 hlen = msg_hdr_sz(mhdr);
1063 if (dsz && (dsz == rc))
1064 tsk->snt_unacked = tsk_inc(tsk, dsz + hlen);
1067 if (dsz > (uint)INT_MAX)
1070 if (unlikely(sock->state != SS_CONNECTED)) {
1071 if (sock->state == SS_DISCONNECTING)
1077 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1078 if (!timeo && tsk->link_cong)
1081 dnode = tsk_peer_node(tsk);
1082 skb_queue_head_init(&pktchain);
1087 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1088 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1089 if (unlikely(rc < 0))
1093 if (likely(!tsk_conn_cong(tsk))) {
1094 rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1096 tsk->snt_unacked += tsk_inc(tsk, send + hlen);
1102 if (rc == -EMSGSIZE) {
1103 __skb_queue_purge(&pktchain);
1104 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1109 if (rc != -ELINKCONG)
1114 rc = tipc_wait_for_sndpkt(sock, &timeo);
1117 __skb_queue_purge(&pktchain);
1118 return sent ? sent : rc;
1122 * tipc_send_packet - send a connection-oriented message
1123 * @sock: socket structure
1124 * @m: message to send
1125 * @dsz: length of data to be transmitted
1127 * Used for SOCK_SEQPACKET messages.
1129 * Returns the number of bytes sent on success, or errno otherwise
1131 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1133 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1136 return tipc_send_stream(sock, m, dsz);
1139 /* tipc_sk_finish_conn - complete the setup of a connection
1141 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1144 struct sock *sk = &tsk->sk;
1145 struct net *net = sock_net(sk);
1146 struct tipc_msg *msg = &tsk->phdr;
1148 msg_set_destnode(msg, peer_node);
1149 msg_set_destport(msg, peer_port);
1150 msg_set_type(msg, TIPC_CONN_MSG);
1151 msg_set_lookup_scope(msg, 0);
1152 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1154 tsk->probing_intv = CONN_PROBING_INTERVAL;
1155 tsk->probing_state = TIPC_CONN_OK;
1157 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1158 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1159 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1160 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1161 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1164 /* Fall back to message based flow control */
1165 tsk->rcv_win = FLOWCTL_MSG_WIN;
1166 tsk->snd_win = FLOWCTL_MSG_WIN;
1170 * set_orig_addr - capture sender's address for received message
1171 * @m: descriptor for message info
1172 * @msg: received message header
1174 * Note: Address is not captured if not requested by receiver.
1176 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1178 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1181 addr->family = AF_TIPC;
1182 addr->addrtype = TIPC_ADDR_ID;
1183 memset(&addr->addr, 0, sizeof(addr->addr));
1184 addr->addr.id.ref = msg_origport(msg);
1185 addr->addr.id.node = msg_orignode(msg);
1186 addr->addr.name.domain = 0; /* could leave uninitialized */
1187 addr->scope = 0; /* could leave uninitialized */
1188 m->msg_namelen = sizeof(struct sockaddr_tipc);
1193 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1194 * @m: descriptor for message info
1195 * @msg: received message header
1196 * @tsk: TIPC port associated with message
1198 * Note: Ancillary data is not captured if not requested by receiver.
1200 * Returns 0 if successful, otherwise errno
1202 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1203 struct tipc_sock *tsk)
1211 if (likely(m->msg_controllen == 0))
1214 /* Optionally capture errored message object(s) */
1215 err = msg ? msg_errcode(msg) : 0;
1216 if (unlikely(err)) {
1218 anc_data[1] = msg_data_sz(msg);
1219 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1223 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1230 /* Optionally capture message destination object */
1231 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1232 switch (dest_type) {
1233 case TIPC_NAMED_MSG:
1235 anc_data[0] = msg_nametype(msg);
1236 anc_data[1] = msg_namelower(msg);
1237 anc_data[2] = msg_namelower(msg);
1239 case TIPC_MCAST_MSG:
1241 anc_data[0] = msg_nametype(msg);
1242 anc_data[1] = msg_namelower(msg);
1243 anc_data[2] = msg_nameupper(msg);
1246 has_name = (tsk->conn_type != 0);
1247 anc_data[0] = tsk->conn_type;
1248 anc_data[1] = tsk->conn_instance;
1249 anc_data[2] = tsk->conn_instance;
1255 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1263 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1265 struct net *net = sock_net(&tsk->sk);
1266 struct sk_buff *skb = NULL;
1267 struct tipc_msg *msg;
1268 u32 peer_port = tsk_peer_port(tsk);
1269 u32 dnode = tsk_peer_node(tsk);
1271 if (!tsk->connected)
1273 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1274 dnode, tsk_own_node(tsk), peer_port,
1275 tsk->portid, TIPC_OK);
1279 msg_set_conn_ack(msg, tsk->rcv_unacked);
1280 tsk->rcv_unacked = 0;
1282 /* Adjust to and advertize the correct window limit */
1283 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1284 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1285 msg_set_adv_win(msg, tsk->rcv_win);
1287 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1290 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1292 struct sock *sk = sock->sk;
1294 long timeo = *timeop;
1298 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1299 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1300 if (sock->state == SS_DISCONNECTING) {
1305 timeo = schedule_timeout(timeo);
1309 if (!skb_queue_empty(&sk->sk_receive_queue))
1314 err = sock_intr_errno(timeo);
1315 if (signal_pending(current))
1318 finish_wait(sk_sleep(sk), &wait);
1324 * tipc_recvmsg - receive packet-oriented message
1325 * @m: descriptor for message info
1326 * @buf_len: total size of user buffer area
1327 * @flags: receive flags
1329 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1330 * If the complete message doesn't fit in user area, truncate it.
1332 * Returns size of returned message data, errno otherwise
1334 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1337 struct sock *sk = sock->sk;
1338 struct tipc_sock *tsk = tipc_sk(sk);
1339 struct sk_buff *buf;
1340 struct tipc_msg *msg;
1346 /* Catch invalid receive requests */
1347 if (unlikely(!buf_len))
1352 if (unlikely(sock->state == SS_UNCONNECTED)) {
1357 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1360 /* Look for a message in receive queue; wait if necessary */
1361 res = tipc_wait_for_rcvmsg(sock, &timeo);
1365 /* Look at first message in receive queue */
1366 buf = skb_peek(&sk->sk_receive_queue);
1368 sz = msg_data_sz(msg);
1369 hlen = msg_hdr_sz(msg);
1370 err = msg_errcode(msg);
1372 /* Discard an empty non-errored message & try again */
1373 if ((!sz) && (!err)) {
1374 tsk_advance_rx_queue(sk);
1378 /* Capture sender's address (optional) */
1379 set_orig_addr(m, msg);
1381 /* Capture ancillary data (optional) */
1382 res = tipc_sk_anc_data_recv(m, msg, tsk);
1386 /* Capture message data (if valid) & compute return value (always) */
1388 if (unlikely(buf_len < sz)) {
1390 m->msg_flags |= MSG_TRUNC;
1392 res = skb_copy_datagram_msg(buf, hlen, m, sz);
1397 if ((sock->state == SS_READY) ||
1398 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1404 if (unlikely(flags & MSG_PEEK))
1407 if (likely(sock->state != SS_READY)) {
1408 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1409 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1410 tipc_sk_send_ack(tsk);
1412 tsk_advance_rx_queue(sk);
1419 * tipc_recv_stream - receive stream-oriented data
1420 * @m: descriptor for message info
1421 * @buf_len: total size of user buffer area
1422 * @flags: receive flags
1424 * Used for SOCK_STREAM messages only. If not enough data is available
1425 * will optionally wait for more; never truncates data.
1427 * Returns size of returned message data, errno otherwise
1429 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1430 size_t buf_len, int flags)
1432 struct sock *sk = sock->sk;
1433 struct tipc_sock *tsk = tipc_sk(sk);
1434 struct sk_buff *buf;
1435 struct tipc_msg *msg;
1443 /* Catch invalid receive attempts */
1444 if (unlikely(!buf_len))
1449 if (unlikely(sock->state == SS_UNCONNECTED)) {
1454 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1455 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1458 /* Look for a message in receive queue; wait if necessary */
1459 res = tipc_wait_for_rcvmsg(sock, &timeo);
1463 /* Look at first message in receive queue */
1464 buf = skb_peek(&sk->sk_receive_queue);
1466 sz = msg_data_sz(msg);
1467 hlen = msg_hdr_sz(msg);
1468 err = msg_errcode(msg);
1470 /* Discard an empty non-errored message & try again */
1471 if ((!sz) && (!err)) {
1472 tsk_advance_rx_queue(sk);
1476 /* Optionally capture sender's address & ancillary data of first msg */
1477 if (sz_copied == 0) {
1478 set_orig_addr(m, msg);
1479 res = tipc_sk_anc_data_recv(m, msg, tsk);
1484 /* Capture message data (if valid) & compute return value (always) */
1486 u32 offset = TIPC_SKB_CB(buf)->bytes_read;
1491 needed = (buf_len - sz_copied);
1492 sz_to_copy = min(sz, needed);
1494 res = skb_copy_datagram_msg(buf, hlen + offset, m, sz_to_copy);
1498 sz_copied += sz_to_copy;
1500 if (sz_to_copy < sz) {
1501 if (!(flags & MSG_PEEK))
1502 TIPC_SKB_CB(buf)->bytes_read =
1503 offset + sz_to_copy;
1508 goto exit; /* can't add error msg to valid data */
1510 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1516 if (unlikely(flags & MSG_PEEK))
1519 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1520 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1521 tipc_sk_send_ack(tsk);
1522 tsk_advance_rx_queue(sk);
1524 /* Loop around if more data is required */
1525 if ((sz_copied < buf_len) && /* didn't get all requested data */
1526 (!skb_queue_empty(&sk->sk_receive_queue) ||
1527 (sz_copied < target)) && /* and more is ready or required */
1528 (!err)) /* and haven't reached a FIN */
1533 return sz_copied ? sz_copied : res;
1537 * tipc_write_space - wake up thread if port congestion is released
1540 static void tipc_write_space(struct sock *sk)
1542 struct socket_wq *wq;
1545 wq = rcu_dereference(sk->sk_wq);
1546 if (skwq_has_sleeper(wq))
1547 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1548 POLLWRNORM | POLLWRBAND);
1553 * tipc_data_ready - wake up threads to indicate messages have been received
1555 * @len: the length of messages
1557 static void tipc_data_ready(struct sock *sk)
1559 struct socket_wq *wq;
1562 wq = rcu_dereference(sk->sk_wq);
1563 if (skwq_has_sleeper(wq))
1564 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1565 POLLRDNORM | POLLRDBAND);
1569 static void tipc_sock_destruct(struct sock *sk)
1571 __skb_queue_purge(&sk->sk_receive_queue);
1575 * filter_connect - Handle all incoming messages for a connection-based socket
1577 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1579 * Returns true if everything ok, false otherwise
1581 static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1583 struct sock *sk = &tsk->sk;
1584 struct net *net = sock_net(sk);
1585 struct socket *sock = sk->sk_socket;
1586 struct tipc_msg *hdr = buf_msg(skb);
1588 if (unlikely(msg_mcast(hdr)))
1591 switch ((int)sock->state) {
1594 /* Accept only connection-based messages sent by peer */
1595 if (unlikely(!tsk_peer_msg(tsk, hdr)))
1598 if (unlikely(msg_errcode(hdr))) {
1599 sock->state = SS_DISCONNECTING;
1601 /* Let timer expire on it's own */
1602 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1604 sk->sk_state_change(sk);
1610 /* Accept only ACK or NACK message */
1611 if (unlikely(!msg_connected(hdr)))
1614 if (unlikely(msg_errcode(hdr))) {
1615 sock->state = SS_DISCONNECTING;
1616 sk->sk_err = ECONNREFUSED;
1620 if (unlikely(!msg_isdata(hdr))) {
1621 sock->state = SS_DISCONNECTING;
1622 sk->sk_err = EINVAL;
1626 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1627 msg_set_importance(&tsk->phdr, msg_importance(hdr));
1628 sock->state = SS_CONNECTED;
1630 /* If 'ACK+' message, add to socket receive queue */
1631 if (msg_data_sz(hdr))
1634 /* If empty 'ACK-' message, wake up sleeping connect() */
1635 if (waitqueue_active(sk_sleep(sk)))
1636 wake_up_interruptible(sk_sleep(sk));
1638 /* 'ACK-' message is neither accepted nor rejected: */
1639 msg_set_dest_droppable(hdr, 1);
1643 case SS_UNCONNECTED:
1645 /* Accept only SYN message */
1646 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1649 case SS_DISCONNECTING:
1652 pr_err("Unknown socket state %u\n", sock->state);
1658 * rcvbuf_limit - get proper overload limit of socket receive queue
1662 * For connection oriented messages, irrespective of importance,
1663 * default queue limit is 2 MB.
1665 * For connectionless messages, queue limits are based on message
1666 * importance as follows:
1668 * TIPC_LOW_IMPORTANCE (2 MB)
1669 * TIPC_MEDIUM_IMPORTANCE (4 MB)
1670 * TIPC_HIGH_IMPORTANCE (8 MB)
1671 * TIPC_CRITICAL_IMPORTANCE (16 MB)
1673 * Returns overload limit according to corresponding message importance
1675 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
1677 struct tipc_sock *tsk = tipc_sk(sk);
1678 struct tipc_msg *hdr = buf_msg(skb);
1680 if (unlikely(!msg_connected(hdr)))
1681 return sk->sk_rcvbuf << msg_importance(hdr);
1683 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
1684 return sk->sk_rcvbuf;
1686 return FLOWCTL_MSG_LIM;
1690 * filter_rcv - validate incoming message
1692 * @skb: pointer to message.
1694 * Enqueues message on receive queue if acceptable; optionally handles
1695 * disconnect indication for a connected socket.
1697 * Called with socket lock already taken
1699 * Returns true if message was added to socket receive queue, otherwise false
1701 static bool filter_rcv(struct sock *sk, struct sk_buff *skb,
1702 struct sk_buff_head *xmitq)
1704 struct socket *sock = sk->sk_socket;
1705 struct tipc_sock *tsk = tipc_sk(sk);
1706 struct tipc_msg *hdr = buf_msg(skb);
1707 unsigned int limit = rcvbuf_limit(sk, skb);
1709 int usr = msg_user(hdr);
1711 if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1712 tipc_sk_proto_rcv(tsk, skb, xmitq);
1716 if (unlikely(usr == SOCK_WAKEUP)) {
1719 sk->sk_write_space(sk);
1723 /* Drop if illegal message type */
1724 if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1729 /* Reject if wrong message type for current socket state */
1730 if (unlikely(sock->state == SS_READY)) {
1731 if (msg_connected(hdr)) {
1732 err = TIPC_ERR_NO_PORT;
1735 } else if (unlikely(!filter_connect(tsk, skb))) {
1736 err = TIPC_ERR_NO_PORT;
1740 /* Reject message if there isn't room to queue it */
1741 if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1742 err = TIPC_ERR_OVERLOAD;
1746 /* Enqueue message */
1747 TIPC_SKB_CB(skb)->bytes_read = 0;
1748 __skb_queue_tail(&sk->sk_receive_queue, skb);
1749 skb_set_owner_r(skb, sk);
1751 sk->sk_data_ready(sk);
1755 if (tipc_msg_reverse(tsk_own_node(tsk), &skb, err))
1756 __skb_queue_tail(xmitq, skb);
1761 * tipc_backlog_rcv - handle incoming message from backlog queue
1765 * Caller must hold socket lock
1769 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1771 unsigned int truesize = skb->truesize;
1772 struct sk_buff_head xmitq;
1773 u32 dnode, selector;
1775 __skb_queue_head_init(&xmitq);
1777 if (likely(filter_rcv(sk, skb, &xmitq))) {
1778 atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1782 if (skb_queue_empty(&xmitq))
1785 /* Send response/rejected message */
1786 skb = __skb_dequeue(&xmitq);
1787 dnode = msg_destnode(buf_msg(skb));
1788 selector = msg_origport(buf_msg(skb));
1789 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
1794 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1795 * inputq and try adding them to socket or backlog queue
1796 * @inputq: list of incoming buffers with potentially different destinations
1797 * @sk: socket where the buffers should be enqueued
1798 * @dport: port number for the socket
1800 * Caller must hold socket lock
1802 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1803 u32 dport, struct sk_buff_head *xmitq)
1805 unsigned long time_limit = jiffies + 2;
1806 struct sk_buff *skb;
1811 while (skb_queue_len(inputq)) {
1812 if (unlikely(time_after_eq(jiffies, time_limit)))
1815 skb = tipc_skb_dequeue(inputq, dport);
1819 /* Add message directly to receive queue if possible */
1820 if (!sock_owned_by_user(sk)) {
1821 filter_rcv(sk, skb, xmitq);
1825 /* Try backlog, compensating for double-counted bytes */
1826 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1827 if (!sk->sk_backlog.len)
1828 atomic_set(dcnt, 0);
1829 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1830 if (likely(!sk_add_backlog(sk, skb, lim)))
1833 /* Overload => reject message back to sender */
1834 onode = tipc_own_addr(sock_net(sk));
1835 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
1836 __skb_queue_tail(xmitq, skb);
1842 * tipc_sk_rcv - handle a chain of incoming buffers
1843 * @inputq: buffer list containing the buffers
1844 * Consumes all buffers in list until inputq is empty
1845 * Note: may be called in multiple threads referring to the same queue
1847 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1849 struct sk_buff_head xmitq;
1850 u32 dnode, dport = 0;
1852 struct tipc_sock *tsk;
1854 struct sk_buff *skb;
1856 __skb_queue_head_init(&xmitq);
1857 while (skb_queue_len(inputq)) {
1858 dport = tipc_skb_peek_port(inputq, dport);
1859 tsk = tipc_sk_lookup(net, dport);
1863 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1864 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
1865 spin_unlock_bh(&sk->sk_lock.slock);
1867 /* Send pending response/rejected messages, if any */
1868 while ((skb = __skb_dequeue(&xmitq))) {
1869 dnode = msg_destnode(buf_msg(skb));
1870 tipc_node_xmit_skb(net, skb, dnode, dport);
1876 /* No destination socket => dequeue skb if still there */
1877 skb = tipc_skb_dequeue(inputq, dport);
1881 /* Try secondary lookup if unresolved named message */
1882 err = TIPC_ERR_NO_PORT;
1883 if (tipc_msg_lookup_dest(net, skb, &err))
1886 /* Prepare for message rejection */
1887 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1890 dnode = msg_destnode(buf_msg(skb));
1891 tipc_node_xmit_skb(net, skb, dnode, dport);
1895 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1897 struct sock *sk = sock->sk;
1902 int err = sock_error(sk);
1907 if (signal_pending(current))
1908 return sock_intr_errno(*timeo_p);
1910 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1911 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1912 finish_wait(sk_sleep(sk), &wait);
1918 * tipc_connect - establish a connection to another TIPC port
1919 * @sock: socket structure
1920 * @dest: socket address for destination port
1921 * @destlen: size of socket address data structure
1922 * @flags: file-related flags associated with socket
1924 * Returns 0 on success, errno otherwise
1926 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1927 int destlen, int flags)
1929 struct sock *sk = sock->sk;
1930 struct tipc_sock *tsk = tipc_sk(sk);
1931 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1932 struct msghdr m = {NULL,};
1933 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1934 socket_state previous;
1939 /* DGRAM/RDM connect(), just save the destaddr */
1940 if (sock->state == SS_READY) {
1941 if (dst->family == AF_UNSPEC) {
1942 memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1944 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1947 memcpy(&tsk->remote, dest, destlen);
1954 * Reject connection attempt using multicast address
1956 * Note: send_msg() validates the rest of the address fields,
1957 * so there's no need to do it here
1959 if (dst->addrtype == TIPC_ADDR_MCAST) {
1964 previous = sock->state;
1965 switch (sock->state) {
1966 case SS_UNCONNECTED:
1967 /* Send a 'SYN-' to destination */
1969 m.msg_namelen = destlen;
1971 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1972 * indicate send_msg() is never blocked.
1975 m.msg_flags = MSG_DONTWAIT;
1977 res = __tipc_sendmsg(sock, &m, 0);
1978 if ((res < 0) && (res != -EWOULDBLOCK))
1981 /* Just entered SS_CONNECTING state; the only
1982 * difference is that return value in non-blocking
1983 * case is EINPROGRESS, rather than EALREADY.
1987 if (previous == SS_CONNECTING)
1991 timeout = msecs_to_jiffies(timeout);
1992 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1993 res = tipc_wait_for_connect(sock, &timeout);
2008 * tipc_listen - allow socket to listen for incoming connections
2009 * @sock: socket structure
2012 * Returns 0 on success, errno otherwise
2014 static int tipc_listen(struct socket *sock, int len)
2016 struct sock *sk = sock->sk;
2021 if (sock->state != SS_UNCONNECTED)
2024 sock->state = SS_LISTENING;
2032 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2034 struct sock *sk = sock->sk;
2038 /* True wake-one mechanism for incoming connections: only
2039 * one process gets woken up, not the 'whole herd'.
2040 * Since we do not 'race & poll' for established sockets
2041 * anymore, the common case will execute the loop only once.
2044 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2045 TASK_INTERRUPTIBLE);
2046 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2048 timeo = schedule_timeout(timeo);
2052 if (!skb_queue_empty(&sk->sk_receive_queue))
2055 if (sock->state != SS_LISTENING)
2060 err = sock_intr_errno(timeo);
2061 if (signal_pending(current))
2064 finish_wait(sk_sleep(sk), &wait);
2069 * tipc_accept - wait for connection request
2070 * @sock: listening socket
2071 * @newsock: new socket that is to be connected
2072 * @flags: file-related flags associated with socket
2074 * Returns 0 on success, errno otherwise
2076 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2078 struct sock *new_sk, *sk = sock->sk;
2079 struct sk_buff *buf;
2080 struct tipc_sock *new_tsock;
2081 struct tipc_msg *msg;
2087 if (sock->state != SS_LISTENING) {
2091 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2092 res = tipc_wait_for_accept(sock, timeo);
2096 buf = skb_peek(&sk->sk_receive_queue);
2098 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 0);
2101 security_sk_clone(sock->sk, new_sock->sk);
2103 new_sk = new_sock->sk;
2104 new_tsock = tipc_sk(new_sk);
2107 /* we lock on new_sk; but lockdep sees the lock on sk */
2108 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2111 * Reject any stray messages received by new socket
2112 * before the socket lock was taken (very, very unlikely)
2114 tsk_rej_rx_queue(new_sk);
2116 /* Connect new socket to it's peer */
2117 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2118 new_sock->state = SS_CONNECTED;
2120 tsk_set_importance(new_tsock, msg_importance(msg));
2121 if (msg_named(msg)) {
2122 new_tsock->conn_type = msg_nametype(msg);
2123 new_tsock->conn_instance = msg_nameinst(msg);
2127 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2128 * Respond to 'SYN+' by queuing it on new socket.
2130 if (!msg_data_sz(msg)) {
2131 struct msghdr m = {NULL,};
2133 tsk_advance_rx_queue(sk);
2134 __tipc_send_stream(new_sock, &m, 0);
2136 __skb_dequeue(&sk->sk_receive_queue);
2137 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2138 skb_set_owner_r(buf, new_sk);
2140 release_sock(new_sk);
2147 * tipc_shutdown - shutdown socket connection
2148 * @sock: socket structure
2149 * @how: direction to close (must be SHUT_RDWR)
2151 * Terminates connection (if necessary), then purges socket's receive queue.
2153 * Returns 0 on success, errno otherwise
2155 static int tipc_shutdown(struct socket *sock, int how)
2157 struct sock *sk = sock->sk;
2158 struct net *net = sock_net(sk);
2159 struct tipc_sock *tsk = tipc_sk(sk);
2160 struct sk_buff *skb;
2161 u32 dnode = tsk_peer_node(tsk);
2162 u32 dport = tsk_peer_port(tsk);
2163 u32 onode = tipc_own_addr(net);
2164 u32 oport = tsk->portid;
2167 if (how != SHUT_RDWR)
2172 switch (sock->state) {
2177 dnode = tsk_peer_node(tsk);
2179 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2180 skb = __skb_dequeue(&sk->sk_receive_queue);
2182 if (TIPC_SKB_CB(skb)->bytes_read) {
2186 tipc_sk_respond(sk, skb, TIPC_CONN_SHUTDOWN);
2188 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2189 TIPC_CONN_MSG, SHORT_H_SIZE,
2190 0, dnode, onode, dport, oport,
2191 TIPC_CONN_SHUTDOWN);
2193 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
2196 sock->state = SS_DISCONNECTING;
2197 tipc_node_remove_conn(net, dnode, tsk->portid);
2200 case SS_DISCONNECTING:
2202 /* Discard any unreceived messages */
2203 __skb_queue_purge(&sk->sk_receive_queue);
2205 /* Wake up anyone sleeping in poll */
2206 sk->sk_state_change(sk);
2218 static void tipc_sk_timeout(unsigned long data)
2220 struct tipc_sock *tsk = (struct tipc_sock *)data;
2221 struct sock *sk = &tsk->sk;
2222 struct sk_buff *skb = NULL;
2223 u32 peer_port, peer_node;
2224 u32 own_node = tsk_own_node(tsk);
2227 if (!tsk->connected) {
2231 peer_port = tsk_peer_port(tsk);
2232 peer_node = tsk_peer_node(tsk);
2234 if (tsk->probing_state == TIPC_CONN_PROBING) {
2235 if (!sock_owned_by_user(sk)) {
2236 sk->sk_socket->state = SS_DISCONNECTING;
2238 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2239 tsk_peer_port(tsk));
2240 sk->sk_state_change(sk);
2242 /* Try again later */
2243 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2247 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2248 INT_H_SIZE, 0, peer_node, own_node,
2249 peer_port, tsk->portid, TIPC_OK);
2250 tsk->probing_state = TIPC_CONN_PROBING;
2251 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2255 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2260 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2261 struct tipc_name_seq const *seq)
2263 struct net *net = sock_net(&tsk->sk);
2264 struct publication *publ;
2269 key = tsk->portid + tsk->pub_count + 1;
2270 if (key == tsk->portid)
2273 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2274 scope, tsk->portid, key);
2275 if (unlikely(!publ))
2278 list_add(&publ->pport_list, &tsk->publications);
2284 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2285 struct tipc_name_seq const *seq)
2287 struct net *net = sock_net(&tsk->sk);
2288 struct publication *publ;
2289 struct publication *safe;
2292 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2294 if (publ->scope != scope)
2296 if (publ->type != seq->type)
2298 if (publ->lower != seq->lower)
2300 if (publ->upper != seq->upper)
2302 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2303 publ->ref, publ->key);
2307 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2308 publ->ref, publ->key);
2311 if (list_empty(&tsk->publications))
2316 /* tipc_sk_reinit: set non-zero address in all existing sockets
2317 * when we go from standalone to network mode.
2319 void tipc_sk_reinit(struct net *net)
2321 struct tipc_net *tn = net_generic(net, tipc_net_id);
2322 const struct bucket_table *tbl;
2323 struct rhash_head *pos;
2324 struct tipc_sock *tsk;
2325 struct tipc_msg *msg;
2329 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2330 for (i = 0; i < tbl->size; i++) {
2331 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2332 spin_lock_bh(&tsk->sk.sk_lock.slock);
2334 msg_set_prevnode(msg, tn->own_addr);
2335 msg_set_orignode(msg, tn->own_addr);
2336 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2342 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2344 struct tipc_net *tn = net_generic(net, tipc_net_id);
2345 struct tipc_sock *tsk;
2348 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2350 sock_hold(&tsk->sk);
2356 static int tipc_sk_insert(struct tipc_sock *tsk)
2358 struct sock *sk = &tsk->sk;
2359 struct net *net = sock_net(sk);
2360 struct tipc_net *tn = net_generic(net, tipc_net_id);
2361 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2362 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2364 while (remaining--) {
2366 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2367 portid = TIPC_MIN_PORT;
2368 tsk->portid = portid;
2369 sock_hold(&tsk->sk);
2370 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2379 static void tipc_sk_remove(struct tipc_sock *tsk)
2381 struct sock *sk = &tsk->sk;
2382 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2384 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2385 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2390 static const struct rhashtable_params tsk_rht_params = {
2392 .head_offset = offsetof(struct tipc_sock, node),
2393 .key_offset = offsetof(struct tipc_sock, portid),
2394 .key_len = sizeof(u32), /* portid */
2395 .max_size = 1048576,
2397 .automatic_shrinking = true,
2400 int tipc_sk_rht_init(struct net *net)
2402 struct tipc_net *tn = net_generic(net, tipc_net_id);
2404 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2407 void tipc_sk_rht_destroy(struct net *net)
2409 struct tipc_net *tn = net_generic(net, tipc_net_id);
2411 /* Wait for socket readers to complete */
2414 rhashtable_destroy(&tn->sk_rht);
2418 * tipc_setsockopt - set socket option
2419 * @sock: socket structure
2420 * @lvl: option level
2421 * @opt: option identifier
2422 * @ov: pointer to new option value
2423 * @ol: length of option value
2425 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2426 * (to ease compatibility).
2428 * Returns 0 on success, errno otherwise
2430 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2431 char __user *ov, unsigned int ol)
2433 struct sock *sk = sock->sk;
2434 struct tipc_sock *tsk = tipc_sk(sk);
2438 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2440 if (lvl != SOL_TIPC)
2441 return -ENOPROTOOPT;
2442 if (ol < sizeof(value))
2444 res = get_user(value, (u32 __user *)ov);
2451 case TIPC_IMPORTANCE:
2452 res = tsk_set_importance(tsk, value);
2454 case TIPC_SRC_DROPPABLE:
2455 if (sock->type != SOCK_STREAM)
2456 tsk_set_unreliable(tsk, value);
2460 case TIPC_DEST_DROPPABLE:
2461 tsk_set_unreturnable(tsk, value);
2463 case TIPC_CONN_TIMEOUT:
2464 tipc_sk(sk)->conn_timeout = value;
2465 /* no need to set "res", since already 0 at this point */
2477 * tipc_getsockopt - get socket option
2478 * @sock: socket structure
2479 * @lvl: option level
2480 * @opt: option identifier
2481 * @ov: receptacle for option value
2482 * @ol: receptacle for length of option value
2484 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2485 * (to ease compatibility).
2487 * Returns 0 on success, errno otherwise
2489 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2490 char __user *ov, int __user *ol)
2492 struct sock *sk = sock->sk;
2493 struct tipc_sock *tsk = tipc_sk(sk);
2498 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2499 return put_user(0, ol);
2500 if (lvl != SOL_TIPC)
2501 return -ENOPROTOOPT;
2502 res = get_user(len, ol);
2509 case TIPC_IMPORTANCE:
2510 value = tsk_importance(tsk);
2512 case TIPC_SRC_DROPPABLE:
2513 value = tsk_unreliable(tsk);
2515 case TIPC_DEST_DROPPABLE:
2516 value = tsk_unreturnable(tsk);
2518 case TIPC_CONN_TIMEOUT:
2519 value = tsk->conn_timeout;
2520 /* no need to set "res", since already 0 at this point */
2522 case TIPC_NODE_RECVQ_DEPTH:
2523 value = 0; /* was tipc_queue_size, now obsolete */
2525 case TIPC_SOCK_RECVQ_DEPTH:
2526 value = skb_queue_len(&sk->sk_receive_queue);
2535 return res; /* "get" failed */
2537 if (len < sizeof(value))
2540 if (copy_to_user(ov, &value, sizeof(value)))
2543 return put_user(sizeof(value), ol);
2546 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2548 struct sock *sk = sock->sk;
2549 struct tipc_sioc_ln_req lnr;
2550 void __user *argp = (void __user *)arg;
2553 case SIOCGETLINKNAME:
2554 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2556 if (!tipc_node_get_linkname(sock_net(sk),
2557 lnr.bearer_id & 0xffff, lnr.peer,
2558 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2559 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2563 return -EADDRNOTAVAIL;
2565 return -ENOIOCTLCMD;
2569 /* Protocol switches for the various types of TIPC sockets */
2571 static const struct proto_ops msg_ops = {
2572 .owner = THIS_MODULE,
2574 .release = tipc_release,
2576 .connect = tipc_connect,
2577 .socketpair = sock_no_socketpair,
2578 .accept = sock_no_accept,
2579 .getname = tipc_getname,
2581 .ioctl = tipc_ioctl,
2582 .listen = sock_no_listen,
2583 .shutdown = tipc_shutdown,
2584 .setsockopt = tipc_setsockopt,
2585 .getsockopt = tipc_getsockopt,
2586 .sendmsg = tipc_sendmsg,
2587 .recvmsg = tipc_recvmsg,
2588 .mmap = sock_no_mmap,
2589 .sendpage = sock_no_sendpage
2592 static const struct proto_ops packet_ops = {
2593 .owner = THIS_MODULE,
2595 .release = tipc_release,
2597 .connect = tipc_connect,
2598 .socketpair = sock_no_socketpair,
2599 .accept = tipc_accept,
2600 .getname = tipc_getname,
2602 .ioctl = tipc_ioctl,
2603 .listen = tipc_listen,
2604 .shutdown = tipc_shutdown,
2605 .setsockopt = tipc_setsockopt,
2606 .getsockopt = tipc_getsockopt,
2607 .sendmsg = tipc_send_packet,
2608 .recvmsg = tipc_recvmsg,
2609 .mmap = sock_no_mmap,
2610 .sendpage = sock_no_sendpage
2613 static const struct proto_ops stream_ops = {
2614 .owner = THIS_MODULE,
2616 .release = tipc_release,
2618 .connect = tipc_connect,
2619 .socketpair = sock_no_socketpair,
2620 .accept = tipc_accept,
2621 .getname = tipc_getname,
2623 .ioctl = tipc_ioctl,
2624 .listen = tipc_listen,
2625 .shutdown = tipc_shutdown,
2626 .setsockopt = tipc_setsockopt,
2627 .getsockopt = tipc_getsockopt,
2628 .sendmsg = tipc_send_stream,
2629 .recvmsg = tipc_recv_stream,
2630 .mmap = sock_no_mmap,
2631 .sendpage = sock_no_sendpage
2634 static const struct net_proto_family tipc_family_ops = {
2635 .owner = THIS_MODULE,
2637 .create = tipc_sk_create
2640 static struct proto tipc_proto = {
2642 .owner = THIS_MODULE,
2643 .obj_size = sizeof(struct tipc_sock),
2644 .sysctl_rmem = sysctl_tipc_rmem
2648 * tipc_socket_init - initialize TIPC socket interface
2650 * Returns 0 on success, errno otherwise
2652 int tipc_socket_init(void)
2656 res = proto_register(&tipc_proto, 1);
2658 pr_err("Failed to register TIPC protocol type\n");
2662 res = sock_register(&tipc_family_ops);
2664 pr_err("Failed to register TIPC socket type\n");
2665 proto_unregister(&tipc_proto);
2673 * tipc_socket_stop - stop TIPC socket interface
2675 void tipc_socket_stop(void)
2677 sock_unregister(tipc_family_ops.family);
2678 proto_unregister(&tipc_proto);
2681 /* Caller should hold socket lock for the passed tipc socket. */
2682 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2686 struct nlattr *nest;
2688 peer_node = tsk_peer_node(tsk);
2689 peer_port = tsk_peer_port(tsk);
2691 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2693 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2695 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2698 if (tsk->conn_type != 0) {
2699 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2701 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2703 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2706 nla_nest_end(skb, nest);
2711 nla_nest_cancel(skb, nest);
2716 /* Caller should hold socket lock for the passed tipc socket. */
2717 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2718 struct tipc_sock *tsk)
2722 struct nlattr *attrs;
2723 struct net *net = sock_net(skb->sk);
2724 struct tipc_net *tn = net_generic(net, tipc_net_id);
2726 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2727 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2731 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2733 goto genlmsg_cancel;
2734 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2735 goto attr_msg_cancel;
2736 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2737 goto attr_msg_cancel;
2739 if (tsk->connected) {
2740 err = __tipc_nl_add_sk_con(skb, tsk);
2742 goto attr_msg_cancel;
2743 } else if (!list_empty(&tsk->publications)) {
2744 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
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 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2763 struct tipc_sock *tsk;
2764 const struct bucket_table *tbl;
2765 struct rhash_head *pos;
2766 struct net *net = sock_net(skb->sk);
2767 struct tipc_net *tn = net_generic(net, tipc_net_id);
2768 u32 tbl_id = cb->args[0];
2769 u32 prev_portid = cb->args[1];
2772 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2773 for (; tbl_id < tbl->size; tbl_id++) {
2774 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2775 spin_lock_bh(&tsk->sk.sk_lock.slock);
2776 if (prev_portid && prev_portid != tsk->portid) {
2777 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2781 err = __tipc_nl_add_sk(skb, cb, tsk);
2783 prev_portid = tsk->portid;
2784 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2788 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2793 cb->args[0] = tbl_id;
2794 cb->args[1] = prev_portid;
2799 /* Caller should hold socket lock for the passed tipc socket. */
2800 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2801 struct netlink_callback *cb,
2802 struct publication *publ)
2805 struct nlattr *attrs;
2807 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2808 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2812 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2814 goto genlmsg_cancel;
2816 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2817 goto attr_msg_cancel;
2818 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2819 goto attr_msg_cancel;
2820 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2821 goto attr_msg_cancel;
2822 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2823 goto attr_msg_cancel;
2825 nla_nest_end(skb, attrs);
2826 genlmsg_end(skb, hdr);
2831 nla_nest_cancel(skb, attrs);
2833 genlmsg_cancel(skb, hdr);
2838 /* Caller should hold socket lock for the passed tipc socket. */
2839 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2840 struct netlink_callback *cb,
2841 struct tipc_sock *tsk, u32 *last_publ)
2844 struct publication *p;
2847 list_for_each_entry(p, &tsk->publications, pport_list) {
2848 if (p->key == *last_publ)
2851 if (p->key != *last_publ) {
2852 /* We never set seq or call nl_dump_check_consistent()
2853 * this means that setting prev_seq here will cause the
2854 * consistence check to fail in the netlink callback
2855 * handler. Resulting in the last NLMSG_DONE message
2856 * having the NLM_F_DUMP_INTR flag set.
2863 p = list_first_entry(&tsk->publications, struct publication,
2867 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2868 err = __tipc_nl_add_sk_publ(skb, cb, p);
2870 *last_publ = p->key;
2879 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2882 u32 tsk_portid = cb->args[0];
2883 u32 last_publ = cb->args[1];
2884 u32 done = cb->args[2];
2885 struct net *net = sock_net(skb->sk);
2886 struct tipc_sock *tsk;
2889 struct nlattr **attrs;
2890 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2892 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2896 if (!attrs[TIPC_NLA_SOCK])
2899 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2900 attrs[TIPC_NLA_SOCK],
2901 tipc_nl_sock_policy);
2905 if (!sock[TIPC_NLA_SOCK_REF])
2908 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2914 tsk = tipc_sk_lookup(net, tsk_portid);
2918 lock_sock(&tsk->sk);
2919 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2922 release_sock(&tsk->sk);
2925 cb->args[0] = tsk_portid;
2926 cb->args[1] = last_publ;