2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2016, 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 CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
48 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
49 #define TIPC_FWD_MSG 1
50 #define TIPC_MAX_PORT 0xffffffff
51 #define TIPC_MIN_PORT 1
54 TIPC_LISTEN = TCP_LISTEN,
55 TIPC_ESTABLISHED = TCP_ESTABLISHED,
56 TIPC_OPEN = TCP_CLOSE,
57 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
58 TIPC_CONNECTING = TCP_SYN_SENT,
62 * struct tipc_sock - TIPC socket structure
63 * @sk: socket - interacts with 'port' and with user via the socket API
64 * @conn_type: TIPC type used when connection was established
65 * @conn_instance: TIPC instance used when connection was established
66 * @published: non-zero if port has one or more associated names
67 * @max_pkt: maximum packet size "hint" used when building messages sent by port
68 * @portid: unique port identity in TIPC socket hash table
69 * @phdr: preformatted message header used when sending messages
70 * @publications: list of publications for port
71 * @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 * @peer: 'connected' peer for dgram/rdm
79 * @node: hash table node
80 * @rcu: rcu struct for tipc_sock
90 struct list_head sock_list;
91 struct list_head publications;
102 struct sockaddr_tipc peer;
103 struct rhash_head node;
107 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
108 static void tipc_data_ready(struct sock *sk);
109 static void tipc_write_space(struct sock *sk);
110 static void tipc_sock_destruct(struct sock *sk);
111 static int tipc_release(struct socket *sock);
112 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
113 static void tipc_sk_timeout(unsigned long data);
114 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
115 struct tipc_name_seq const *seq);
116 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
117 struct tipc_name_seq const *seq);
118 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
119 static int tipc_sk_insert(struct tipc_sock *tsk);
120 static void tipc_sk_remove(struct tipc_sock *tsk);
121 static int __tipc_send_stream(struct socket *sock, struct msghdr *m,
123 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
125 static const struct proto_ops packet_ops;
126 static const struct proto_ops stream_ops;
127 static const struct proto_ops msg_ops;
128 static struct proto tipc_proto;
129 static const struct rhashtable_params tsk_rht_params;
131 static u32 tsk_own_node(struct tipc_sock *tsk)
133 return msg_prevnode(&tsk->phdr);
136 static u32 tsk_peer_node(struct tipc_sock *tsk)
138 return msg_destnode(&tsk->phdr);
141 static u32 tsk_peer_port(struct tipc_sock *tsk)
143 return msg_destport(&tsk->phdr);
146 static bool tsk_unreliable(struct tipc_sock *tsk)
148 return msg_src_droppable(&tsk->phdr) != 0;
151 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
153 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
156 static bool tsk_unreturnable(struct tipc_sock *tsk)
158 return msg_dest_droppable(&tsk->phdr) != 0;
161 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
163 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
166 static int tsk_importance(struct tipc_sock *tsk)
168 return msg_importance(&tsk->phdr);
171 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
173 if (imp > TIPC_CRITICAL_IMPORTANCE)
175 msg_set_importance(&tsk->phdr, (u32)imp);
179 static struct tipc_sock *tipc_sk(const struct sock *sk)
181 return container_of(sk, struct tipc_sock, sk);
184 static bool tsk_conn_cong(struct tipc_sock *tsk)
186 return tsk->snt_unacked > tsk->snd_win;
189 /* tsk_blocks(): translate a buffer size in bytes to number of
190 * advertisable blocks, taking into account the ratio truesize(len)/len
191 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
193 static u16 tsk_adv_blocks(int len)
195 return len / FLOWCTL_BLK_SZ / 4;
198 /* tsk_inc(): increment counter for sent or received data
199 * - If block based flow control is not supported by peer we
200 * fall back to message based ditto, incrementing the counter
202 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
204 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
205 return ((msglen / FLOWCTL_BLK_SZ) + 1);
210 * tsk_advance_rx_queue - discard first buffer in socket receive queue
212 * Caller must hold socket lock
214 static void tsk_advance_rx_queue(struct sock *sk)
216 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
219 /* tipc_sk_respond() : send response message back to sender
221 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
225 u32 onode = tipc_own_addr(sock_net(sk));
227 if (!tipc_msg_reverse(onode, &skb, err))
230 dnode = msg_destnode(buf_msg(skb));
231 selector = msg_origport(buf_msg(skb));
232 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
236 * tsk_rej_rx_queue - reject all buffers in socket receive queue
238 * Caller must hold socket lock
240 static void tsk_rej_rx_queue(struct sock *sk)
244 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
245 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
248 static bool tipc_sk_connected(struct sock *sk)
250 return sk->sk_state == TIPC_ESTABLISHED;
253 /* tipc_sk_type_connectionless - check if the socket is datagram socket
256 * Returns true if connection less, false otherwise
258 static bool tipc_sk_type_connectionless(struct sock *sk)
260 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
263 /* tsk_peer_msg - verify if message was sent by connected port's peer
265 * Handles cases where the node's network address has changed from
266 * the default of <0.0.0> to its configured setting.
268 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
270 struct sock *sk = &tsk->sk;
271 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
272 u32 peer_port = tsk_peer_port(tsk);
276 if (unlikely(!tipc_sk_connected(sk)))
279 if (unlikely(msg_origport(msg) != peer_port))
282 orig_node = msg_orignode(msg);
283 peer_node = tsk_peer_node(tsk);
285 if (likely(orig_node == peer_node))
288 if (!orig_node && (peer_node == tn->own_addr))
291 if (!peer_node && (orig_node == tn->own_addr))
297 /* tipc_set_sk_state - set the sk_state of the socket
300 * Caller must hold socket lock
302 * Returns 0 on success, errno otherwise
304 static int tipc_set_sk_state(struct sock *sk, int state)
306 int oldsk_state = sk->sk_state;
314 case TIPC_CONNECTING:
315 if (oldsk_state == TIPC_OPEN)
318 case TIPC_ESTABLISHED:
319 if (oldsk_state == TIPC_CONNECTING ||
320 oldsk_state == TIPC_OPEN)
323 case TIPC_DISCONNECTING:
324 if (oldsk_state == TIPC_CONNECTING ||
325 oldsk_state == TIPC_ESTABLISHED)
331 sk->sk_state = state;
336 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
338 struct sock *sk = sock->sk;
339 int err = sock_error(sk);
340 int typ = sock->type;
344 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
345 if (sk->sk_state == TIPC_DISCONNECTING)
347 else if (!tipc_sk_connected(sk))
352 if (signal_pending(current))
353 return sock_intr_errno(*timeout);
358 #define tipc_wait_for_cond(sock_, timeout_, condition_) \
363 while (!(condition_) && !done_) { \
364 struct sock *sk_ = sock->sk; \
365 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
367 rc_ = tipc_sk_sock_err(sock_, timeout_); \
370 prepare_to_wait(sk_sleep(sk_), &wait_, \
371 TASK_INTERRUPTIBLE); \
372 done_ = sk_wait_event(sk_, timeout_, \
373 (condition_), &wait_); \
374 remove_wait_queue(sk_sleep(sk_), &wait_); \
380 * tipc_sk_create - create a TIPC socket
381 * @net: network namespace (must be default network)
382 * @sock: pre-allocated socket structure
383 * @protocol: protocol indicator (must be 0)
384 * @kern: caused by kernel or by userspace?
386 * This routine creates additional data structures used by the TIPC socket,
387 * initializes them, and links them together.
389 * Returns 0 on success, errno otherwise
391 static int tipc_sk_create(struct net *net, struct socket *sock,
392 int protocol, int kern)
395 const struct proto_ops *ops;
397 struct tipc_sock *tsk;
398 struct tipc_msg *msg;
400 /* Validate arguments */
401 if (unlikely(protocol != 0))
402 return -EPROTONOSUPPORT;
404 switch (sock->type) {
419 /* Allocate socket's protocol area */
420 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
425 tsk->max_pkt = MAX_PKT_DEFAULT;
426 INIT_LIST_HEAD(&tsk->publications);
428 tn = net_generic(sock_net(sk), tipc_net_id);
429 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
432 /* Finish initializing socket data structures */
434 sock_init_data(sock, sk);
435 tipc_set_sk_state(sk, TIPC_OPEN);
436 if (tipc_sk_insert(tsk)) {
437 pr_warn("Socket create failed; port number exhausted\n");
440 msg_set_origport(msg, tsk->portid);
441 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
443 sk->sk_backlog_rcv = tipc_backlog_rcv;
444 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
445 sk->sk_data_ready = tipc_data_ready;
446 sk->sk_write_space = tipc_write_space;
447 sk->sk_destruct = tipc_sock_destruct;
448 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
449 atomic_set(&tsk->dupl_rcvcnt, 0);
451 /* Start out with safe limits until we receive an advertised window */
452 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
453 tsk->rcv_win = tsk->snd_win;
455 if (tipc_sk_type_connectionless(sk)) {
456 tsk_set_unreturnable(tsk, true);
457 if (sock->type == SOCK_DGRAM)
458 tsk_set_unreliable(tsk, true);
464 static void tipc_sk_callback(struct rcu_head *head)
466 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
471 /* Caller should hold socket lock for the socket. */
472 static void __tipc_shutdown(struct socket *sock, int error)
474 struct sock *sk = sock->sk;
475 struct tipc_sock *tsk = tipc_sk(sk);
476 struct net *net = sock_net(sk);
477 u32 dnode = tsk_peer_node(tsk);
480 /* Reject all unreceived messages, except on an active connection
481 * (which disconnects locally & sends a 'FIN+' to peer).
483 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
484 if (TIPC_SKB_CB(skb)->bytes_read) {
488 if (!tipc_sk_type_connectionless(sk) &&
489 sk->sk_state != TIPC_DISCONNECTING) {
490 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
491 tipc_node_remove_conn(net, dnode, tsk->portid);
493 tipc_sk_respond(sk, skb, error);
496 if (tipc_sk_type_connectionless(sk))
499 if (sk->sk_state != TIPC_DISCONNECTING) {
500 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
501 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
502 tsk_own_node(tsk), tsk_peer_port(tsk),
505 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
506 tipc_node_remove_conn(net, dnode, tsk->portid);
507 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
512 * tipc_release - destroy a TIPC socket
513 * @sock: socket to destroy
515 * This routine cleans up any messages that are still queued on the socket.
516 * For DGRAM and RDM socket types, all queued messages are rejected.
517 * For SEQPACKET and STREAM socket types, the first message is rejected
518 * and any others are discarded. (If the first message on a STREAM socket
519 * is partially-read, it is discarded and the next one is rejected instead.)
521 * NOTE: Rejected messages are not necessarily returned to the sender! They
522 * are returned or discarded according to the "destination droppable" setting
523 * specified for the message by the sender.
525 * Returns 0 on success, errno otherwise
527 static int tipc_release(struct socket *sock)
529 struct sock *sk = sock->sk;
530 struct tipc_sock *tsk;
533 * Exit if socket isn't fully initialized (occurs when a failed accept()
534 * releases a pre-allocated child socket that was never used)
542 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
543 sk->sk_shutdown = SHUTDOWN_MASK;
544 tipc_sk_withdraw(tsk, 0, NULL);
545 sk_stop_timer(sk, &sk->sk_timer);
548 /* Reject any messages that accumulated in backlog queue */
551 call_rcu(&tsk->rcu, tipc_sk_callback);
558 * tipc_bind - associate or disassocate TIPC name(s) with a socket
559 * @sock: socket structure
560 * @uaddr: socket address describing name(s) and desired operation
561 * @uaddr_len: size of socket address data structure
563 * Name and name sequence binding is indicated using a positive scope value;
564 * a negative scope value unbinds the specified name. Specifying no name
565 * (i.e. a socket address length of 0) unbinds all names from the socket.
567 * Returns 0 on success, errno otherwise
569 * NOTE: This routine doesn't need to take the socket lock since it doesn't
570 * access any non-constant socket information.
572 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
575 struct sock *sk = sock->sk;
576 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
577 struct tipc_sock *tsk = tipc_sk(sk);
581 if (unlikely(!uaddr_len)) {
582 res = tipc_sk_withdraw(tsk, 0, NULL);
586 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
590 if (addr->family != AF_TIPC) {
595 if (addr->addrtype == TIPC_ADDR_NAME)
596 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
597 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
602 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
603 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
604 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
609 res = (addr->scope > 0) ?
610 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
611 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
618 * tipc_getname - get port ID of socket or peer socket
619 * @sock: socket structure
620 * @uaddr: area for returned socket address
621 * @uaddr_len: area for returned length of socket address
622 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
624 * Returns 0 on success, errno otherwise
626 * NOTE: This routine doesn't need to take the socket lock since it only
627 * accesses socket information that is unchanging (or which changes in
628 * a completely predictable manner).
630 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
631 int *uaddr_len, int peer)
633 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
634 struct sock *sk = sock->sk;
635 struct tipc_sock *tsk = tipc_sk(sk);
636 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
638 memset(addr, 0, sizeof(*addr));
640 if ((!tipc_sk_connected(sk)) &&
641 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
643 addr->addr.id.ref = tsk_peer_port(tsk);
644 addr->addr.id.node = tsk_peer_node(tsk);
646 addr->addr.id.ref = tsk->portid;
647 addr->addr.id.node = tn->own_addr;
650 *uaddr_len = sizeof(*addr);
651 addr->addrtype = TIPC_ADDR_ID;
652 addr->family = AF_TIPC;
654 addr->addr.name.domain = 0;
660 * tipc_poll - read and possibly block on pollmask
661 * @file: file structure associated with the socket
662 * @sock: socket for which to calculate the poll bits
665 * Returns pollmask value
668 * It appears that the usual socket locking mechanisms are not useful here
669 * since the pollmask info is potentially out-of-date the moment this routine
670 * exits. TCP and other protocols seem to rely on higher level poll routines
671 * to handle any preventable race conditions, so TIPC will do the same ...
673 * IMPORTANT: The fact that a read or write operation is indicated does NOT
674 * imply that the operation will succeed, merely that it should be performed
675 * and will not block.
677 static unsigned int tipc_poll(struct file *file, struct socket *sock,
680 struct sock *sk = sock->sk;
681 struct tipc_sock *tsk = tipc_sk(sk);
684 sock_poll_wait(file, sk_sleep(sk), wait);
686 if (sk->sk_shutdown & RCV_SHUTDOWN)
687 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
688 if (sk->sk_shutdown == SHUTDOWN_MASK)
691 switch (sk->sk_state) {
692 case TIPC_ESTABLISHED:
693 if (!tsk->link_cong && !tsk_conn_cong(tsk))
697 case TIPC_CONNECTING:
698 if (!skb_queue_empty(&sk->sk_receive_queue))
699 mask |= (POLLIN | POLLRDNORM);
704 if (tipc_sk_type_connectionless(sk) &&
705 (!skb_queue_empty(&sk->sk_receive_queue)))
706 mask |= (POLLIN | POLLRDNORM);
708 case TIPC_DISCONNECTING:
709 mask = (POLLIN | POLLRDNORM | POLLHUP);
717 * tipc_sendmcast - send multicast message
718 * @sock: socket structure
719 * @seq: destination address
720 * @msg: message to send
721 * @dsz: total length of message data
722 * @timeo: timeout to wait for wakeup
724 * Called from function tipc_sendmsg(), which has done all sanity checks
725 * Returns the number of bytes sent on success, or errno
727 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
728 struct msghdr *msg, size_t dsz, long timeo)
730 struct sock *sk = sock->sk;
731 struct tipc_sock *tsk = tipc_sk(sk);
732 struct net *net = sock_net(sk);
733 struct tipc_msg *mhdr = &tsk->phdr;
734 struct sk_buff_head pktchain;
735 struct iov_iter save = msg->msg_iter;
739 if (!timeo && tsk->link_cong)
742 msg_set_type(mhdr, TIPC_MCAST_MSG);
743 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
744 msg_set_destport(mhdr, 0);
745 msg_set_destnode(mhdr, 0);
746 msg_set_nametype(mhdr, seq->type);
747 msg_set_namelower(mhdr, seq->lower);
748 msg_set_nameupper(mhdr, seq->upper);
749 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
751 skb_queue_head_init(&pktchain);
754 mtu = tipc_bcast_get_mtu(net);
755 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
756 if (unlikely(rc < 0))
760 rc = tipc_bcast_xmit(net, &pktchain);
764 if (rc == -ELINKCONG) {
766 rc = tipc_wait_for_cond(sock, &timeo, !tsk->link_cong);
770 __skb_queue_purge(&pktchain);
771 if (rc == -EMSGSIZE) {
772 msg->msg_iter = save;
781 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
782 * @arrvq: queue with arriving messages, to be cloned after destination lookup
783 * @inputq: queue with cloned messages, delivered to socket after dest lookup
785 * Multi-threaded: parallel calls with reference to same queues may occur
787 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
788 struct sk_buff_head *inputq)
790 struct tipc_msg *msg;
791 struct list_head dports;
793 u32 scope = TIPC_CLUSTER_SCOPE;
794 struct sk_buff_head tmpq;
796 struct sk_buff *skb, *_skb;
798 __skb_queue_head_init(&tmpq);
799 INIT_LIST_HEAD(&dports);
801 skb = tipc_skb_peek(arrvq, &inputq->lock);
802 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
804 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
806 if (in_own_node(net, msg_orignode(msg)))
807 scope = TIPC_NODE_SCOPE;
809 /* Create destination port list and message clones: */
810 tipc_nametbl_mc_translate(net,
811 msg_nametype(msg), msg_namelower(msg),
812 msg_nameupper(msg), scope, &dports);
813 portid = u32_pop(&dports);
814 for (; portid; portid = u32_pop(&dports)) {
815 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
817 msg_set_destport(buf_msg(_skb), portid);
818 __skb_queue_tail(&tmpq, _skb);
821 pr_warn("Failed to clone mcast rcv buffer\n");
823 /* Append to inputq if not already done by other thread */
824 spin_lock_bh(&inputq->lock);
825 if (skb_peek(arrvq) == skb) {
826 skb_queue_splice_tail_init(&tmpq, inputq);
827 kfree_skb(__skb_dequeue(arrvq));
829 spin_unlock_bh(&inputq->lock);
830 __skb_queue_purge(&tmpq);
833 tipc_sk_rcv(net, inputq);
837 * tipc_sk_proto_rcv - receive a connection mng protocol message
838 * @tsk: receiving socket
839 * @skb: pointer to message buffer.
841 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
842 struct sk_buff_head *xmitq)
844 struct sock *sk = &tsk->sk;
845 u32 onode = tsk_own_node(tsk);
846 struct tipc_msg *hdr = buf_msg(skb);
847 int mtyp = msg_type(hdr);
850 /* Ignore if connection cannot be validated: */
851 if (!tsk_peer_msg(tsk, hdr))
854 tsk->probe_unacked = false;
856 if (mtyp == CONN_PROBE) {
857 msg_set_type(hdr, CONN_PROBE_REPLY);
858 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
859 __skb_queue_tail(xmitq, skb);
861 } else if (mtyp == CONN_ACK) {
862 conn_cong = tsk_conn_cong(tsk);
863 tsk->snt_unacked -= msg_conn_ack(hdr);
864 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
865 tsk->snd_win = msg_adv_win(hdr);
867 sk->sk_write_space(sk);
868 } else if (mtyp != CONN_PROBE_REPLY) {
869 pr_warn("Received unknown CONN_PROTO msg\n");
876 * tipc_sendmsg - send message in connectionless manner
877 * @sock: socket structure
878 * @m: message to send
879 * @dsz: amount of user data to be sent
881 * Message must have an destination specified explicitly.
882 * Used for SOCK_RDM and SOCK_DGRAM messages,
883 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
884 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
886 * Returns the number of bytes sent on success, or errno otherwise
888 static int tipc_sendmsg(struct socket *sock,
889 struct msghdr *m, size_t dsz)
891 struct sock *sk = sock->sk;
895 ret = __tipc_sendmsg(sock, m, dsz);
901 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
903 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
904 struct sock *sk = sock->sk;
905 struct tipc_sock *tsk = tipc_sk(sk);
906 struct net *net = sock_net(sk);
907 struct tipc_msg *mhdr = &tsk->phdr;
909 struct sk_buff_head pktchain;
910 bool is_connectionless = tipc_sk_type_connectionless(sk);
912 struct tipc_name_seq *seq;
913 struct iov_iter save;
918 if (dsz > TIPC_MAX_USER_MSG_SIZE)
920 if (unlikely(!dest)) {
921 if (is_connectionless && tsk->peer.family == AF_TIPC)
924 return -EDESTADDRREQ;
925 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
926 dest->family != AF_TIPC) {
929 if (!is_connectionless) {
930 if (sk->sk_state == TIPC_LISTEN)
932 if (sk->sk_state != TIPC_OPEN)
936 if (dest->addrtype == TIPC_ADDR_NAME) {
937 tsk->conn_type = dest->addr.name.name.type;
938 tsk->conn_instance = dest->addr.name.name.instance;
941 seq = &dest->addr.nameseq;
942 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
944 if (dest->addrtype == TIPC_ADDR_MCAST) {
945 return tipc_sendmcast(sock, seq, m, dsz, timeo);
946 } else if (dest->addrtype == TIPC_ADDR_NAME) {
947 u32 type = dest->addr.name.name.type;
948 u32 inst = dest->addr.name.name.instance;
949 u32 domain = dest->addr.name.domain;
952 msg_set_type(mhdr, TIPC_NAMED_MSG);
953 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
954 msg_set_nametype(mhdr, type);
955 msg_set_nameinst(mhdr, inst);
956 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
957 dport = tipc_nametbl_translate(net, type, inst, &dnode);
958 msg_set_destnode(mhdr, dnode);
959 msg_set_destport(mhdr, dport);
960 if (unlikely(!dport && !dnode))
961 return -EHOSTUNREACH;
962 } else if (dest->addrtype == TIPC_ADDR_ID) {
963 dnode = dest->addr.id.node;
964 msg_set_type(mhdr, TIPC_DIRECT_MSG);
965 msg_set_lookup_scope(mhdr, 0);
966 msg_set_destnode(mhdr, dnode);
967 msg_set_destport(mhdr, dest->addr.id.ref);
968 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
971 skb_queue_head_init(&pktchain);
974 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
975 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
980 skb = skb_peek(&pktchain);
981 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
982 rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
984 if (!is_connectionless)
985 tipc_set_sk_state(sk, TIPC_CONNECTING);
988 if (rc == -ELINKCONG) {
990 rc = tipc_wait_for_cond(sock, &timeo, !tsk->link_cong);
994 __skb_queue_purge(&pktchain);
995 if (rc == -EMSGSIZE) {
1006 * tipc_send_stream - send stream-oriented data
1007 * @sock: socket structure
1009 * @dsz: total length of data to be transmitted
1011 * Used for SOCK_STREAM data.
1013 * Returns the number of bytes sent on success (or partial success),
1014 * or errno if no data sent
1016 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1018 struct sock *sk = sock->sk;
1022 ret = __tipc_send_stream(sock, m, dsz);
1028 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1030 struct sock *sk = sock->sk;
1031 struct net *net = sock_net(sk);
1032 struct tipc_sock *tsk = tipc_sk(sk);
1033 struct tipc_msg *mhdr = &tsk->phdr;
1034 struct sk_buff_head pktchain;
1035 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1036 u32 portid = tsk->portid;
1040 uint mtu, send, sent = 0;
1041 struct iov_iter save;
1042 int hlen = MIN_H_SIZE;
1044 /* Handle implied connection establishment */
1045 if (unlikely(dest)) {
1046 rc = __tipc_sendmsg(sock, m, dsz);
1047 hlen = msg_hdr_sz(mhdr);
1048 if (dsz && (dsz == rc))
1049 tsk->snt_unacked = tsk_inc(tsk, dsz + hlen);
1052 if (dsz > (uint)INT_MAX)
1055 if (unlikely(!tipc_sk_connected(sk))) {
1056 if (sk->sk_state == TIPC_DISCONNECTING)
1062 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1063 if (!timeo && tsk->link_cong)
1066 dnode = tsk_peer_node(tsk);
1067 skb_queue_head_init(&pktchain);
1072 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1073 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1074 if (unlikely(rc < 0))
1078 if (likely(!tsk_conn_cong(tsk))) {
1079 rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1081 tsk->snt_unacked += tsk_inc(tsk, send + hlen);
1087 if (rc == -EMSGSIZE) {
1088 __skb_queue_purge(&pktchain);
1089 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1094 if (rc != -ELINKCONG)
1099 rc = tipc_wait_for_cond(sock, &timeo,
1101 !tsk_conn_cong(tsk) &&
1102 tipc_sk_connected(sk)));
1105 __skb_queue_purge(&pktchain);
1106 return sent ? sent : rc;
1110 * tipc_send_packet - send a connection-oriented message
1111 * @sock: socket structure
1112 * @m: message to send
1113 * @dsz: length of data to be transmitted
1115 * Used for SOCK_SEQPACKET messages.
1117 * Returns the number of bytes sent on success, or errno otherwise
1119 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1121 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1124 return tipc_send_stream(sock, m, dsz);
1127 /* tipc_sk_finish_conn - complete the setup of a connection
1129 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1132 struct sock *sk = &tsk->sk;
1133 struct net *net = sock_net(sk);
1134 struct tipc_msg *msg = &tsk->phdr;
1136 msg_set_destnode(msg, peer_node);
1137 msg_set_destport(msg, peer_port);
1138 msg_set_type(msg, TIPC_CONN_MSG);
1139 msg_set_lookup_scope(msg, 0);
1140 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1142 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
1143 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1144 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1145 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1146 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1147 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1150 /* Fall back to message based flow control */
1151 tsk->rcv_win = FLOWCTL_MSG_WIN;
1152 tsk->snd_win = FLOWCTL_MSG_WIN;
1156 * set_orig_addr - capture sender's address for received message
1157 * @m: descriptor for message info
1158 * @msg: received message header
1160 * Note: Address is not captured if not requested by receiver.
1162 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1164 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1167 addr->family = AF_TIPC;
1168 addr->addrtype = TIPC_ADDR_ID;
1169 memset(&addr->addr, 0, sizeof(addr->addr));
1170 addr->addr.id.ref = msg_origport(msg);
1171 addr->addr.id.node = msg_orignode(msg);
1172 addr->addr.name.domain = 0; /* could leave uninitialized */
1173 addr->scope = 0; /* could leave uninitialized */
1174 m->msg_namelen = sizeof(struct sockaddr_tipc);
1179 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1180 * @m: descriptor for message info
1181 * @msg: received message header
1182 * @tsk: TIPC port associated with message
1184 * Note: Ancillary data is not captured if not requested by receiver.
1186 * Returns 0 if successful, otherwise errno
1188 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1189 struct tipc_sock *tsk)
1197 if (likely(m->msg_controllen == 0))
1200 /* Optionally capture errored message object(s) */
1201 err = msg ? msg_errcode(msg) : 0;
1202 if (unlikely(err)) {
1204 anc_data[1] = msg_data_sz(msg);
1205 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1209 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1216 /* Optionally capture message destination object */
1217 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1218 switch (dest_type) {
1219 case TIPC_NAMED_MSG:
1221 anc_data[0] = msg_nametype(msg);
1222 anc_data[1] = msg_namelower(msg);
1223 anc_data[2] = msg_namelower(msg);
1225 case TIPC_MCAST_MSG:
1227 anc_data[0] = msg_nametype(msg);
1228 anc_data[1] = msg_namelower(msg);
1229 anc_data[2] = msg_nameupper(msg);
1232 has_name = (tsk->conn_type != 0);
1233 anc_data[0] = tsk->conn_type;
1234 anc_data[1] = tsk->conn_instance;
1235 anc_data[2] = tsk->conn_instance;
1241 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1249 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1251 struct sock *sk = &tsk->sk;
1252 struct net *net = sock_net(sk);
1253 struct sk_buff *skb = NULL;
1254 struct tipc_msg *msg;
1255 u32 peer_port = tsk_peer_port(tsk);
1256 u32 dnode = tsk_peer_node(tsk);
1258 if (!tipc_sk_connected(sk))
1260 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1261 dnode, tsk_own_node(tsk), peer_port,
1262 tsk->portid, TIPC_OK);
1266 msg_set_conn_ack(msg, tsk->rcv_unacked);
1267 tsk->rcv_unacked = 0;
1269 /* Adjust to and advertize the correct window limit */
1270 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1271 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1272 msg_set_adv_win(msg, tsk->rcv_win);
1274 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1277 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1279 struct sock *sk = sock->sk;
1281 long timeo = *timeop;
1285 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1286 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1287 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1292 timeo = schedule_timeout(timeo);
1296 if (!skb_queue_empty(&sk->sk_receive_queue))
1301 err = sock_intr_errno(timeo);
1302 if (signal_pending(current))
1305 finish_wait(sk_sleep(sk), &wait);
1311 * tipc_recvmsg - receive packet-oriented message
1312 * @m: descriptor for message info
1313 * @buf_len: total size of user buffer area
1314 * @flags: receive flags
1316 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1317 * If the complete message doesn't fit in user area, truncate it.
1319 * Returns size of returned message data, errno otherwise
1321 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1324 struct sock *sk = sock->sk;
1325 struct tipc_sock *tsk = tipc_sk(sk);
1326 struct sk_buff *buf;
1327 struct tipc_msg *msg;
1328 bool is_connectionless = tipc_sk_type_connectionless(sk);
1334 /* Catch invalid receive requests */
1335 if (unlikely(!buf_len))
1340 if (!is_connectionless && unlikely(sk->sk_state == TIPC_OPEN)) {
1345 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1348 /* Look for a message in receive queue; wait if necessary */
1349 res = tipc_wait_for_rcvmsg(sock, &timeo);
1353 /* Look at first message in receive queue */
1354 buf = skb_peek(&sk->sk_receive_queue);
1356 sz = msg_data_sz(msg);
1357 hlen = msg_hdr_sz(msg);
1358 err = msg_errcode(msg);
1360 /* Discard an empty non-errored message & try again */
1361 if ((!sz) && (!err)) {
1362 tsk_advance_rx_queue(sk);
1366 /* Capture sender's address (optional) */
1367 set_orig_addr(m, msg);
1369 /* Capture ancillary data (optional) */
1370 res = tipc_sk_anc_data_recv(m, msg, tsk);
1374 /* Capture message data (if valid) & compute return value (always) */
1376 if (unlikely(buf_len < sz)) {
1378 m->msg_flags |= MSG_TRUNC;
1380 res = skb_copy_datagram_msg(buf, hlen, m, sz);
1385 if (is_connectionless || err == TIPC_CONN_SHUTDOWN ||
1392 if (unlikely(flags & MSG_PEEK))
1395 if (likely(!is_connectionless)) {
1396 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1397 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1398 tipc_sk_send_ack(tsk);
1400 tsk_advance_rx_queue(sk);
1407 * tipc_recv_stream - receive stream-oriented data
1408 * @m: descriptor for message info
1409 * @buf_len: total size of user buffer area
1410 * @flags: receive flags
1412 * Used for SOCK_STREAM messages only. If not enough data is available
1413 * will optionally wait for more; never truncates data.
1415 * Returns size of returned message data, errno otherwise
1417 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1418 size_t buf_len, int flags)
1420 struct sock *sk = sock->sk;
1421 struct tipc_sock *tsk = tipc_sk(sk);
1422 struct sk_buff *buf;
1423 struct tipc_msg *msg;
1431 /* Catch invalid receive attempts */
1432 if (unlikely(!buf_len))
1437 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1442 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1443 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1446 /* Look for a message in receive queue; wait if necessary */
1447 res = tipc_wait_for_rcvmsg(sock, &timeo);
1451 /* Look at first message in receive queue */
1452 buf = skb_peek(&sk->sk_receive_queue);
1454 sz = msg_data_sz(msg);
1455 hlen = msg_hdr_sz(msg);
1456 err = msg_errcode(msg);
1458 /* Discard an empty non-errored message & try again */
1459 if ((!sz) && (!err)) {
1460 tsk_advance_rx_queue(sk);
1464 /* Optionally capture sender's address & ancillary data of first msg */
1465 if (sz_copied == 0) {
1466 set_orig_addr(m, msg);
1467 res = tipc_sk_anc_data_recv(m, msg, tsk);
1472 /* Capture message data (if valid) & compute return value (always) */
1474 u32 offset = TIPC_SKB_CB(buf)->bytes_read;
1479 needed = (buf_len - sz_copied);
1480 sz_to_copy = min(sz, needed);
1482 res = skb_copy_datagram_msg(buf, hlen + offset, m, sz_to_copy);
1486 sz_copied += sz_to_copy;
1488 if (sz_to_copy < sz) {
1489 if (!(flags & MSG_PEEK))
1490 TIPC_SKB_CB(buf)->bytes_read =
1491 offset + sz_to_copy;
1496 goto exit; /* can't add error msg to valid data */
1498 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1504 if (unlikely(flags & MSG_PEEK))
1507 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1508 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1509 tipc_sk_send_ack(tsk);
1510 tsk_advance_rx_queue(sk);
1512 /* Loop around if more data is required */
1513 if ((sz_copied < buf_len) && /* didn't get all requested data */
1514 (!skb_queue_empty(&sk->sk_receive_queue) ||
1515 (sz_copied < target)) && /* and more is ready or required */
1516 (!err)) /* and haven't reached a FIN */
1521 return sz_copied ? sz_copied : res;
1525 * tipc_write_space - wake up thread if port congestion is released
1528 static void tipc_write_space(struct sock *sk)
1530 struct socket_wq *wq;
1533 wq = rcu_dereference(sk->sk_wq);
1534 if (skwq_has_sleeper(wq))
1535 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1536 POLLWRNORM | POLLWRBAND);
1541 * tipc_data_ready - wake up threads to indicate messages have been received
1543 * @len: the length of messages
1545 static void tipc_data_ready(struct sock *sk)
1547 struct socket_wq *wq;
1550 wq = rcu_dereference(sk->sk_wq);
1551 if (skwq_has_sleeper(wq))
1552 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1553 POLLRDNORM | POLLRDBAND);
1557 static void tipc_sock_destruct(struct sock *sk)
1559 __skb_queue_purge(&sk->sk_receive_queue);
1563 * filter_connect - Handle all incoming messages for a connection-based socket
1565 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1567 * Returns true if everything ok, false otherwise
1569 static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1571 struct sock *sk = &tsk->sk;
1572 struct net *net = sock_net(sk);
1573 struct tipc_msg *hdr = buf_msg(skb);
1575 if (unlikely(msg_mcast(hdr)))
1578 switch (sk->sk_state) {
1579 case TIPC_CONNECTING:
1580 /* Accept only ACK or NACK message */
1581 if (unlikely(!msg_connected(hdr)))
1584 if (unlikely(msg_errcode(hdr))) {
1585 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1586 sk->sk_err = ECONNREFUSED;
1590 if (unlikely(!msg_isdata(hdr))) {
1591 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1592 sk->sk_err = EINVAL;
1596 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1597 msg_set_importance(&tsk->phdr, msg_importance(hdr));
1599 /* If 'ACK+' message, add to socket receive queue */
1600 if (msg_data_sz(hdr))
1603 /* If empty 'ACK-' message, wake up sleeping connect() */
1604 if (waitqueue_active(sk_sleep(sk)))
1605 wake_up_interruptible(sk_sleep(sk));
1607 /* 'ACK-' message is neither accepted nor rejected: */
1608 msg_set_dest_droppable(hdr, 1);
1612 case TIPC_DISCONNECTING:
1615 /* Accept only SYN message */
1616 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1619 case TIPC_ESTABLISHED:
1620 /* Accept only connection-based messages sent by peer */
1621 if (unlikely(!tsk_peer_msg(tsk, hdr)))
1624 if (unlikely(msg_errcode(hdr))) {
1625 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1626 /* Let timer expire on it's own */
1627 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1629 sk->sk_state_change(sk);
1633 pr_err("Unknown sk_state %u\n", sk->sk_state);
1640 * rcvbuf_limit - get proper overload limit of socket receive queue
1644 * For connection oriented messages, irrespective of importance,
1645 * default queue limit is 2 MB.
1647 * For connectionless messages, queue limits are based on message
1648 * importance as follows:
1650 * TIPC_LOW_IMPORTANCE (2 MB)
1651 * TIPC_MEDIUM_IMPORTANCE (4 MB)
1652 * TIPC_HIGH_IMPORTANCE (8 MB)
1653 * TIPC_CRITICAL_IMPORTANCE (16 MB)
1655 * Returns overload limit according to corresponding message importance
1657 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
1659 struct tipc_sock *tsk = tipc_sk(sk);
1660 struct tipc_msg *hdr = buf_msg(skb);
1662 if (unlikely(!msg_connected(hdr)))
1663 return sk->sk_rcvbuf << msg_importance(hdr);
1665 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
1666 return sk->sk_rcvbuf;
1668 return FLOWCTL_MSG_LIM;
1672 * filter_rcv - validate incoming message
1674 * @skb: pointer to message.
1676 * Enqueues message on receive queue if acceptable; optionally handles
1677 * disconnect indication for a connected socket.
1679 * Called with socket lock already taken
1681 * Returns true if message was added to socket receive queue, otherwise false
1683 static bool filter_rcv(struct sock *sk, struct sk_buff *skb,
1684 struct sk_buff_head *xmitq)
1686 struct tipc_sock *tsk = tipc_sk(sk);
1687 struct tipc_msg *hdr = buf_msg(skb);
1688 unsigned int limit = rcvbuf_limit(sk, skb);
1690 int usr = msg_user(hdr);
1692 if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1693 tipc_sk_proto_rcv(tsk, skb, xmitq);
1697 if (unlikely(usr == SOCK_WAKEUP)) {
1700 sk->sk_write_space(sk);
1704 /* Drop if illegal message type */
1705 if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1710 /* Reject if wrong message type for current socket state */
1711 if (tipc_sk_type_connectionless(sk)) {
1712 if (msg_connected(hdr)) {
1713 err = TIPC_ERR_NO_PORT;
1716 } else if (unlikely(!filter_connect(tsk, skb))) {
1717 err = TIPC_ERR_NO_PORT;
1721 /* Reject message if there isn't room to queue it */
1722 if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1723 err = TIPC_ERR_OVERLOAD;
1727 /* Enqueue message */
1728 TIPC_SKB_CB(skb)->bytes_read = 0;
1729 __skb_queue_tail(&sk->sk_receive_queue, skb);
1730 skb_set_owner_r(skb, sk);
1732 sk->sk_data_ready(sk);
1736 if (tipc_msg_reverse(tsk_own_node(tsk), &skb, err))
1737 __skb_queue_tail(xmitq, skb);
1742 * tipc_backlog_rcv - handle incoming message from backlog queue
1746 * Caller must hold socket lock
1750 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1752 unsigned int truesize = skb->truesize;
1753 struct sk_buff_head xmitq;
1754 u32 dnode, selector;
1756 __skb_queue_head_init(&xmitq);
1758 if (likely(filter_rcv(sk, skb, &xmitq))) {
1759 atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1763 if (skb_queue_empty(&xmitq))
1766 /* Send response/rejected message */
1767 skb = __skb_dequeue(&xmitq);
1768 dnode = msg_destnode(buf_msg(skb));
1769 selector = msg_origport(buf_msg(skb));
1770 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
1775 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1776 * inputq and try adding them to socket or backlog queue
1777 * @inputq: list of incoming buffers with potentially different destinations
1778 * @sk: socket where the buffers should be enqueued
1779 * @dport: port number for the socket
1781 * Caller must hold socket lock
1783 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1784 u32 dport, struct sk_buff_head *xmitq)
1786 unsigned long time_limit = jiffies + 2;
1787 struct sk_buff *skb;
1792 while (skb_queue_len(inputq)) {
1793 if (unlikely(time_after_eq(jiffies, time_limit)))
1796 skb = tipc_skb_dequeue(inputq, dport);
1800 /* Add message directly to receive queue if possible */
1801 if (!sock_owned_by_user(sk)) {
1802 filter_rcv(sk, skb, xmitq);
1806 /* Try backlog, compensating for double-counted bytes */
1807 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1808 if (!sk->sk_backlog.len)
1809 atomic_set(dcnt, 0);
1810 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1811 if (likely(!sk_add_backlog(sk, skb, lim)))
1814 /* Overload => reject message back to sender */
1815 onode = tipc_own_addr(sock_net(sk));
1816 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
1817 __skb_queue_tail(xmitq, skb);
1823 * tipc_sk_rcv - handle a chain of incoming buffers
1824 * @inputq: buffer list containing the buffers
1825 * Consumes all buffers in list until inputq is empty
1826 * Note: may be called in multiple threads referring to the same queue
1828 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1830 struct sk_buff_head xmitq;
1831 u32 dnode, dport = 0;
1833 struct tipc_sock *tsk;
1835 struct sk_buff *skb;
1837 __skb_queue_head_init(&xmitq);
1838 while (skb_queue_len(inputq)) {
1839 dport = tipc_skb_peek_port(inputq, dport);
1840 tsk = tipc_sk_lookup(net, dport);
1844 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1845 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
1846 spin_unlock_bh(&sk->sk_lock.slock);
1848 /* Send pending response/rejected messages, if any */
1849 while ((skb = __skb_dequeue(&xmitq))) {
1850 dnode = msg_destnode(buf_msg(skb));
1851 tipc_node_xmit_skb(net, skb, dnode, dport);
1857 /* No destination socket => dequeue skb if still there */
1858 skb = tipc_skb_dequeue(inputq, dport);
1862 /* Try secondary lookup if unresolved named message */
1863 err = TIPC_ERR_NO_PORT;
1864 if (tipc_msg_lookup_dest(net, skb, &err))
1867 /* Prepare for message rejection */
1868 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1871 dnode = msg_destnode(buf_msg(skb));
1872 tipc_node_xmit_skb(net, skb, dnode, dport);
1876 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1878 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1879 struct sock *sk = sock->sk;
1883 int err = sock_error(sk);
1888 if (signal_pending(current))
1889 return sock_intr_errno(*timeo_p);
1891 add_wait_queue(sk_sleep(sk), &wait);
1892 done = sk_wait_event(sk, timeo_p,
1893 sk->sk_state != TIPC_CONNECTING, &wait);
1894 remove_wait_queue(sk_sleep(sk), &wait);
1900 * tipc_connect - establish a connection to another TIPC port
1901 * @sock: socket structure
1902 * @dest: socket address for destination port
1903 * @destlen: size of socket address data structure
1904 * @flags: file-related flags associated with socket
1906 * Returns 0 on success, errno otherwise
1908 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1909 int destlen, int flags)
1911 struct sock *sk = sock->sk;
1912 struct tipc_sock *tsk = tipc_sk(sk);
1913 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1914 struct msghdr m = {NULL,};
1915 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1921 /* DGRAM/RDM connect(), just save the destaddr */
1922 if (tipc_sk_type_connectionless(sk)) {
1923 if (dst->family == AF_UNSPEC) {
1924 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
1925 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1928 memcpy(&tsk->peer, dest, destlen);
1934 * Reject connection attempt using multicast address
1936 * Note: send_msg() validates the rest of the address fields,
1937 * so there's no need to do it here
1939 if (dst->addrtype == TIPC_ADDR_MCAST) {
1944 previous = sk->sk_state;
1946 switch (sk->sk_state) {
1948 /* Send a 'SYN-' to destination */
1950 m.msg_namelen = destlen;
1952 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1953 * indicate send_msg() is never blocked.
1956 m.msg_flags = MSG_DONTWAIT;
1958 res = __tipc_sendmsg(sock, &m, 0);
1959 if ((res < 0) && (res != -EWOULDBLOCK))
1962 /* Just entered TIPC_CONNECTING state; the only
1963 * difference is that return value in non-blocking
1964 * case is EINPROGRESS, rather than EALREADY.
1968 case TIPC_CONNECTING:
1970 if (previous == TIPC_CONNECTING)
1974 timeout = msecs_to_jiffies(timeout);
1975 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1976 res = tipc_wait_for_connect(sock, &timeout);
1978 case TIPC_ESTABLISHED:
1991 * tipc_listen - allow socket to listen for incoming connections
1992 * @sock: socket structure
1995 * Returns 0 on success, errno otherwise
1997 static int tipc_listen(struct socket *sock, int len)
1999 struct sock *sk = sock->sk;
2003 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2009 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2011 struct sock *sk = sock->sk;
2015 /* True wake-one mechanism for incoming connections: only
2016 * one process gets woken up, not the 'whole herd'.
2017 * Since we do not 'race & poll' for established sockets
2018 * anymore, the common case will execute the loop only once.
2021 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2022 TASK_INTERRUPTIBLE);
2023 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2025 timeo = schedule_timeout(timeo);
2029 if (!skb_queue_empty(&sk->sk_receive_queue))
2034 err = sock_intr_errno(timeo);
2035 if (signal_pending(current))
2038 finish_wait(sk_sleep(sk), &wait);
2043 * tipc_accept - wait for connection request
2044 * @sock: listening socket
2045 * @newsock: new socket that is to be connected
2046 * @flags: file-related flags associated with socket
2048 * Returns 0 on success, errno otherwise
2050 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2052 struct sock *new_sk, *sk = sock->sk;
2053 struct sk_buff *buf;
2054 struct tipc_sock *new_tsock;
2055 struct tipc_msg *msg;
2061 if (sk->sk_state != TIPC_LISTEN) {
2065 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2066 res = tipc_wait_for_accept(sock, timeo);
2070 buf = skb_peek(&sk->sk_receive_queue);
2072 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 0);
2075 security_sk_clone(sock->sk, new_sock->sk);
2077 new_sk = new_sock->sk;
2078 new_tsock = tipc_sk(new_sk);
2081 /* we lock on new_sk; but lockdep sees the lock on sk */
2082 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2085 * Reject any stray messages received by new socket
2086 * before the socket lock was taken (very, very unlikely)
2088 tsk_rej_rx_queue(new_sk);
2090 /* Connect new socket to it's peer */
2091 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2093 tsk_set_importance(new_tsock, msg_importance(msg));
2094 if (msg_named(msg)) {
2095 new_tsock->conn_type = msg_nametype(msg);
2096 new_tsock->conn_instance = msg_nameinst(msg);
2100 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2101 * Respond to 'SYN+' by queuing it on new socket.
2103 if (!msg_data_sz(msg)) {
2104 struct msghdr m = {NULL,};
2106 tsk_advance_rx_queue(sk);
2107 __tipc_send_stream(new_sock, &m, 0);
2109 __skb_dequeue(&sk->sk_receive_queue);
2110 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2111 skb_set_owner_r(buf, new_sk);
2113 release_sock(new_sk);
2120 * tipc_shutdown - shutdown socket connection
2121 * @sock: socket structure
2122 * @how: direction to close (must be SHUT_RDWR)
2124 * Terminates connection (if necessary), then purges socket's receive queue.
2126 * Returns 0 on success, errno otherwise
2128 static int tipc_shutdown(struct socket *sock, int how)
2130 struct sock *sk = sock->sk;
2133 if (how != SHUT_RDWR)
2138 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2139 sk->sk_shutdown = SEND_SHUTDOWN;
2141 if (sk->sk_state == TIPC_DISCONNECTING) {
2142 /* Discard any unreceived messages */
2143 __skb_queue_purge(&sk->sk_receive_queue);
2145 /* Wake up anyone sleeping in poll */
2146 sk->sk_state_change(sk);
2156 static void tipc_sk_timeout(unsigned long data)
2158 struct tipc_sock *tsk = (struct tipc_sock *)data;
2159 struct sock *sk = &tsk->sk;
2160 struct sk_buff *skb = NULL;
2161 u32 peer_port, peer_node;
2162 u32 own_node = tsk_own_node(tsk);
2165 if (!tipc_sk_connected(sk)) {
2169 peer_port = tsk_peer_port(tsk);
2170 peer_node = tsk_peer_node(tsk);
2172 if (tsk->probe_unacked) {
2173 if (!sock_owned_by_user(sk)) {
2174 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2175 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2176 tsk_peer_port(tsk));
2177 sk->sk_state_change(sk);
2179 /* Try again later */
2180 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2187 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2188 INT_H_SIZE, 0, peer_node, own_node,
2189 peer_port, tsk->portid, TIPC_OK);
2190 tsk->probe_unacked = true;
2191 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTERVAL);
2194 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2199 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2200 struct tipc_name_seq const *seq)
2202 struct sock *sk = &tsk->sk;
2203 struct net *net = sock_net(sk);
2204 struct publication *publ;
2207 if (tipc_sk_connected(sk))
2209 key = tsk->portid + tsk->pub_count + 1;
2210 if (key == tsk->portid)
2213 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2214 scope, tsk->portid, key);
2215 if (unlikely(!publ))
2218 list_add(&publ->pport_list, &tsk->publications);
2224 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2225 struct tipc_name_seq const *seq)
2227 struct net *net = sock_net(&tsk->sk);
2228 struct publication *publ;
2229 struct publication *safe;
2232 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2234 if (publ->scope != scope)
2236 if (publ->type != seq->type)
2238 if (publ->lower != seq->lower)
2240 if (publ->upper != seq->upper)
2242 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2243 publ->ref, publ->key);
2247 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2248 publ->ref, publ->key);
2251 if (list_empty(&tsk->publications))
2256 /* tipc_sk_reinit: set non-zero address in all existing sockets
2257 * when we go from standalone to network mode.
2259 void tipc_sk_reinit(struct net *net)
2261 struct tipc_net *tn = net_generic(net, tipc_net_id);
2262 const struct bucket_table *tbl;
2263 struct rhash_head *pos;
2264 struct tipc_sock *tsk;
2265 struct tipc_msg *msg;
2269 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2270 for (i = 0; i < tbl->size; i++) {
2271 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2272 spin_lock_bh(&tsk->sk.sk_lock.slock);
2274 msg_set_prevnode(msg, tn->own_addr);
2275 msg_set_orignode(msg, tn->own_addr);
2276 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2282 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2284 struct tipc_net *tn = net_generic(net, tipc_net_id);
2285 struct tipc_sock *tsk;
2288 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2290 sock_hold(&tsk->sk);
2296 static int tipc_sk_insert(struct tipc_sock *tsk)
2298 struct sock *sk = &tsk->sk;
2299 struct net *net = sock_net(sk);
2300 struct tipc_net *tn = net_generic(net, tipc_net_id);
2301 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2302 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2304 while (remaining--) {
2306 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2307 portid = TIPC_MIN_PORT;
2308 tsk->portid = portid;
2309 sock_hold(&tsk->sk);
2310 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2319 static void tipc_sk_remove(struct tipc_sock *tsk)
2321 struct sock *sk = &tsk->sk;
2322 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2324 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2325 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2330 static const struct rhashtable_params tsk_rht_params = {
2332 .head_offset = offsetof(struct tipc_sock, node),
2333 .key_offset = offsetof(struct tipc_sock, portid),
2334 .key_len = sizeof(u32), /* portid */
2335 .max_size = 1048576,
2337 .automatic_shrinking = true,
2340 int tipc_sk_rht_init(struct net *net)
2342 struct tipc_net *tn = net_generic(net, tipc_net_id);
2344 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2347 void tipc_sk_rht_destroy(struct net *net)
2349 struct tipc_net *tn = net_generic(net, tipc_net_id);
2351 /* Wait for socket readers to complete */
2354 rhashtable_destroy(&tn->sk_rht);
2358 * tipc_setsockopt - set socket option
2359 * @sock: socket structure
2360 * @lvl: option level
2361 * @opt: option identifier
2362 * @ov: pointer to new option value
2363 * @ol: length of option value
2365 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2366 * (to ease compatibility).
2368 * Returns 0 on success, errno otherwise
2370 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2371 char __user *ov, unsigned int ol)
2373 struct sock *sk = sock->sk;
2374 struct tipc_sock *tsk = tipc_sk(sk);
2378 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2380 if (lvl != SOL_TIPC)
2381 return -ENOPROTOOPT;
2382 if (ol < sizeof(value))
2384 res = get_user(value, (u32 __user *)ov);
2391 case TIPC_IMPORTANCE:
2392 res = tsk_set_importance(tsk, value);
2394 case TIPC_SRC_DROPPABLE:
2395 if (sock->type != SOCK_STREAM)
2396 tsk_set_unreliable(tsk, value);
2400 case TIPC_DEST_DROPPABLE:
2401 tsk_set_unreturnable(tsk, value);
2403 case TIPC_CONN_TIMEOUT:
2404 tipc_sk(sk)->conn_timeout = value;
2405 /* no need to set "res", since already 0 at this point */
2417 * tipc_getsockopt - get socket option
2418 * @sock: socket structure
2419 * @lvl: option level
2420 * @opt: option identifier
2421 * @ov: receptacle for option value
2422 * @ol: receptacle for length of option value
2424 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2425 * (to ease compatibility).
2427 * Returns 0 on success, errno otherwise
2429 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2430 char __user *ov, int __user *ol)
2432 struct sock *sk = sock->sk;
2433 struct tipc_sock *tsk = tipc_sk(sk);
2438 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2439 return put_user(0, ol);
2440 if (lvl != SOL_TIPC)
2441 return -ENOPROTOOPT;
2442 res = get_user(len, ol);
2449 case TIPC_IMPORTANCE:
2450 value = tsk_importance(tsk);
2452 case TIPC_SRC_DROPPABLE:
2453 value = tsk_unreliable(tsk);
2455 case TIPC_DEST_DROPPABLE:
2456 value = tsk_unreturnable(tsk);
2458 case TIPC_CONN_TIMEOUT:
2459 value = tsk->conn_timeout;
2460 /* no need to set "res", since already 0 at this point */
2462 case TIPC_NODE_RECVQ_DEPTH:
2463 value = 0; /* was tipc_queue_size, now obsolete */
2465 case TIPC_SOCK_RECVQ_DEPTH:
2466 value = skb_queue_len(&sk->sk_receive_queue);
2475 return res; /* "get" failed */
2477 if (len < sizeof(value))
2480 if (copy_to_user(ov, &value, sizeof(value)))
2483 return put_user(sizeof(value), ol);
2486 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2488 struct sock *sk = sock->sk;
2489 struct tipc_sioc_ln_req lnr;
2490 void __user *argp = (void __user *)arg;
2493 case SIOCGETLINKNAME:
2494 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2496 if (!tipc_node_get_linkname(sock_net(sk),
2497 lnr.bearer_id & 0xffff, lnr.peer,
2498 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2499 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2503 return -EADDRNOTAVAIL;
2505 return -ENOIOCTLCMD;
2509 /* Protocol switches for the various types of TIPC sockets */
2511 static const struct proto_ops msg_ops = {
2512 .owner = THIS_MODULE,
2514 .release = tipc_release,
2516 .connect = tipc_connect,
2517 .socketpair = sock_no_socketpair,
2518 .accept = sock_no_accept,
2519 .getname = tipc_getname,
2521 .ioctl = tipc_ioctl,
2522 .listen = sock_no_listen,
2523 .shutdown = tipc_shutdown,
2524 .setsockopt = tipc_setsockopt,
2525 .getsockopt = tipc_getsockopt,
2526 .sendmsg = tipc_sendmsg,
2527 .recvmsg = tipc_recvmsg,
2528 .mmap = sock_no_mmap,
2529 .sendpage = sock_no_sendpage
2532 static const struct proto_ops packet_ops = {
2533 .owner = THIS_MODULE,
2535 .release = tipc_release,
2537 .connect = tipc_connect,
2538 .socketpair = sock_no_socketpair,
2539 .accept = tipc_accept,
2540 .getname = tipc_getname,
2542 .ioctl = tipc_ioctl,
2543 .listen = tipc_listen,
2544 .shutdown = tipc_shutdown,
2545 .setsockopt = tipc_setsockopt,
2546 .getsockopt = tipc_getsockopt,
2547 .sendmsg = tipc_send_packet,
2548 .recvmsg = tipc_recvmsg,
2549 .mmap = sock_no_mmap,
2550 .sendpage = sock_no_sendpage
2553 static const struct proto_ops stream_ops = {
2554 .owner = THIS_MODULE,
2556 .release = tipc_release,
2558 .connect = tipc_connect,
2559 .socketpair = sock_no_socketpair,
2560 .accept = tipc_accept,
2561 .getname = tipc_getname,
2563 .ioctl = tipc_ioctl,
2564 .listen = tipc_listen,
2565 .shutdown = tipc_shutdown,
2566 .setsockopt = tipc_setsockopt,
2567 .getsockopt = tipc_getsockopt,
2568 .sendmsg = tipc_send_stream,
2569 .recvmsg = tipc_recv_stream,
2570 .mmap = sock_no_mmap,
2571 .sendpage = sock_no_sendpage
2574 static const struct net_proto_family tipc_family_ops = {
2575 .owner = THIS_MODULE,
2577 .create = tipc_sk_create
2580 static struct proto tipc_proto = {
2582 .owner = THIS_MODULE,
2583 .obj_size = sizeof(struct tipc_sock),
2584 .sysctl_rmem = sysctl_tipc_rmem
2588 * tipc_socket_init - initialize TIPC socket interface
2590 * Returns 0 on success, errno otherwise
2592 int tipc_socket_init(void)
2596 res = proto_register(&tipc_proto, 1);
2598 pr_err("Failed to register TIPC protocol type\n");
2602 res = sock_register(&tipc_family_ops);
2604 pr_err("Failed to register TIPC socket type\n");
2605 proto_unregister(&tipc_proto);
2613 * tipc_socket_stop - stop TIPC socket interface
2615 void tipc_socket_stop(void)
2617 sock_unregister(tipc_family_ops.family);
2618 proto_unregister(&tipc_proto);
2621 /* Caller should hold socket lock for the passed tipc socket. */
2622 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2626 struct nlattr *nest;
2628 peer_node = tsk_peer_node(tsk);
2629 peer_port = tsk_peer_port(tsk);
2631 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2633 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2635 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2638 if (tsk->conn_type != 0) {
2639 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2641 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2643 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2646 nla_nest_end(skb, nest);
2651 nla_nest_cancel(skb, nest);
2656 /* Caller should hold socket lock for the passed tipc socket. */
2657 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2658 struct tipc_sock *tsk)
2662 struct nlattr *attrs;
2663 struct net *net = sock_net(skb->sk);
2664 struct tipc_net *tn = net_generic(net, tipc_net_id);
2665 struct sock *sk = &tsk->sk;
2667 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2668 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2672 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2674 goto genlmsg_cancel;
2675 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2676 goto attr_msg_cancel;
2677 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2678 goto attr_msg_cancel;
2680 if (tipc_sk_connected(sk)) {
2681 err = __tipc_nl_add_sk_con(skb, tsk);
2683 goto attr_msg_cancel;
2684 } else if (!list_empty(&tsk->publications)) {
2685 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2686 goto attr_msg_cancel;
2688 nla_nest_end(skb, attrs);
2689 genlmsg_end(skb, hdr);
2694 nla_nest_cancel(skb, attrs);
2696 genlmsg_cancel(skb, hdr);
2701 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2704 struct tipc_sock *tsk;
2705 const struct bucket_table *tbl;
2706 struct rhash_head *pos;
2707 struct net *net = sock_net(skb->sk);
2708 struct tipc_net *tn = net_generic(net, tipc_net_id);
2709 u32 tbl_id = cb->args[0];
2710 u32 prev_portid = cb->args[1];
2713 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2714 for (; tbl_id < tbl->size; tbl_id++) {
2715 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2716 spin_lock_bh(&tsk->sk.sk_lock.slock);
2717 if (prev_portid && prev_portid != tsk->portid) {
2718 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2722 err = __tipc_nl_add_sk(skb, cb, tsk);
2724 prev_portid = tsk->portid;
2725 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2729 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2734 cb->args[0] = tbl_id;
2735 cb->args[1] = prev_portid;
2740 /* Caller should hold socket lock for the passed tipc socket. */
2741 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2742 struct netlink_callback *cb,
2743 struct publication *publ)
2746 struct nlattr *attrs;
2748 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2749 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2753 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2755 goto genlmsg_cancel;
2757 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2758 goto attr_msg_cancel;
2759 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2760 goto attr_msg_cancel;
2761 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2762 goto attr_msg_cancel;
2763 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2764 goto attr_msg_cancel;
2766 nla_nest_end(skb, attrs);
2767 genlmsg_end(skb, hdr);
2772 nla_nest_cancel(skb, attrs);
2774 genlmsg_cancel(skb, hdr);
2779 /* Caller should hold socket lock for the passed tipc socket. */
2780 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2781 struct netlink_callback *cb,
2782 struct tipc_sock *tsk, u32 *last_publ)
2785 struct publication *p;
2788 list_for_each_entry(p, &tsk->publications, pport_list) {
2789 if (p->key == *last_publ)
2792 if (p->key != *last_publ) {
2793 /* We never set seq or call nl_dump_check_consistent()
2794 * this means that setting prev_seq here will cause the
2795 * consistence check to fail in the netlink callback
2796 * handler. Resulting in the last NLMSG_DONE message
2797 * having the NLM_F_DUMP_INTR flag set.
2804 p = list_first_entry(&tsk->publications, struct publication,
2808 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2809 err = __tipc_nl_add_sk_publ(skb, cb, p);
2811 *last_publ = p->key;
2820 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2823 u32 tsk_portid = cb->args[0];
2824 u32 last_publ = cb->args[1];
2825 u32 done = cb->args[2];
2826 struct net *net = sock_net(skb->sk);
2827 struct tipc_sock *tsk;
2830 struct nlattr **attrs;
2831 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2833 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2837 if (!attrs[TIPC_NLA_SOCK])
2840 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2841 attrs[TIPC_NLA_SOCK],
2842 tipc_nl_sock_policy);
2846 if (!sock[TIPC_NLA_SOCK_REF])
2849 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2855 tsk = tipc_sk_lookup(net, tsk_portid);
2859 lock_sock(&tsk->sk);
2860 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2863 release_sock(&tsk->sk);
2866 cb->args[0] = tsk_portid;
2867 cb->args[1] = last_publ;