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 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;
132 static const struct rhashtable_params tsk_rht_params;
134 static u32 tsk_own_node(struct tipc_sock *tsk)
136 return msg_prevnode(&tsk->phdr);
139 static u32 tsk_peer_node(struct tipc_sock *tsk)
141 return msg_destnode(&tsk->phdr);
144 static u32 tsk_peer_port(struct tipc_sock *tsk)
146 return msg_destport(&tsk->phdr);
149 static bool tsk_unreliable(struct tipc_sock *tsk)
151 return msg_src_droppable(&tsk->phdr) != 0;
154 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
156 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
159 static bool tsk_unreturnable(struct tipc_sock *tsk)
161 return msg_dest_droppable(&tsk->phdr) != 0;
164 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
166 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
169 static int tsk_importance(struct tipc_sock *tsk)
171 return msg_importance(&tsk->phdr);
174 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
176 if (imp > TIPC_CRITICAL_IMPORTANCE)
178 msg_set_importance(&tsk->phdr, (u32)imp);
182 static struct tipc_sock *tipc_sk(const struct sock *sk)
184 return container_of(sk, struct tipc_sock, sk);
187 static bool tsk_conn_cong(struct tipc_sock *tsk)
189 return tsk->snt_unacked > tsk->snd_win;
192 /* tsk_blocks(): translate a buffer size in bytes to number of
193 * advertisable blocks, taking into account the ratio truesize(len)/len
194 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
196 static u16 tsk_adv_blocks(int len)
198 return len / FLOWCTL_BLK_SZ / 4;
201 /* tsk_inc(): increment counter for sent or received data
202 * - If block based flow control is not supported by peer we
203 * fall back to message based ditto, incrementing the counter
205 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
207 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
208 return ((msglen / FLOWCTL_BLK_SZ) + 1);
213 * tsk_advance_rx_queue - discard first buffer in socket receive queue
215 * Caller must hold socket lock
217 static void tsk_advance_rx_queue(struct sock *sk)
219 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
222 /* tipc_sk_respond() : send response message back to sender
224 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
228 u32 onode = tipc_own_addr(sock_net(sk));
230 if (!tipc_msg_reverse(onode, &skb, err))
233 dnode = msg_destnode(buf_msg(skb));
234 selector = msg_origport(buf_msg(skb));
235 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
239 * tsk_rej_rx_queue - reject all buffers in socket receive queue
241 * Caller must hold socket lock
243 static void tsk_rej_rx_queue(struct sock *sk)
247 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
248 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
251 /* tsk_peer_msg - verify if message was sent by connected port's peer
253 * Handles cases where the node's network address has changed from
254 * the default of <0.0.0> to its configured setting.
256 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
258 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
259 u32 peer_port = tsk_peer_port(tsk);
263 if (unlikely(!tsk->connected))
266 if (unlikely(msg_origport(msg) != peer_port))
269 orig_node = msg_orignode(msg);
270 peer_node = tsk_peer_node(tsk);
272 if (likely(orig_node == peer_node))
275 if (!orig_node && (peer_node == tn->own_addr))
278 if (!peer_node && (orig_node == tn->own_addr))
285 * tipc_sk_create - create a TIPC socket
286 * @net: network namespace (must be default network)
287 * @sock: pre-allocated socket structure
288 * @protocol: protocol indicator (must be 0)
289 * @kern: caused by kernel or by userspace?
291 * This routine creates additional data structures used by the TIPC socket,
292 * initializes them, and links them together.
294 * Returns 0 on success, errno otherwise
296 static int tipc_sk_create(struct net *net, struct socket *sock,
297 int protocol, int kern)
300 const struct proto_ops *ops;
303 struct tipc_sock *tsk;
304 struct tipc_msg *msg;
306 /* Validate arguments */
307 if (unlikely(protocol != 0))
308 return -EPROTONOSUPPORT;
310 switch (sock->type) {
313 state = SS_UNCONNECTED;
317 state = SS_UNCONNECTED;
328 /* Allocate socket's protocol area */
329 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
334 tsk->max_pkt = MAX_PKT_DEFAULT;
335 INIT_LIST_HEAD(&tsk->publications);
337 tn = net_generic(sock_net(sk), tipc_net_id);
339 /* Finish initializing socket data structures */
342 sock_init_data(sock, sk);
343 if (tipc_sk_insert(tsk)) {
345 pr_warn("Socket create failed; port number exhausted\n");
349 /* Ensure tsk is visible before we read own_addr. */
352 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
355 msg_set_origport(msg, tsk->portid);
356 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
357 sk->sk_backlog_rcv = tipc_backlog_rcv;
358 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
359 sk->sk_data_ready = tipc_data_ready;
360 sk->sk_write_space = tipc_write_space;
361 sk->sk_destruct = tipc_sock_destruct;
362 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
363 atomic_set(&tsk->dupl_rcvcnt, 0);
365 /* Start out with safe limits until we receive an advertised window */
366 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
367 tsk->rcv_win = tsk->snd_win;
369 if (sock->state == SS_READY) {
370 tsk_set_unreturnable(tsk, true);
371 if (sock->type == SOCK_DGRAM)
372 tsk_set_unreliable(tsk, true);
377 static void tipc_sk_callback(struct rcu_head *head)
379 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
385 * tipc_release - destroy a TIPC socket
386 * @sock: socket to destroy
388 * This routine cleans up any messages that are still queued on the socket.
389 * For DGRAM and RDM socket types, all queued messages are rejected.
390 * For SEQPACKET and STREAM socket types, the first message is rejected
391 * and any others are discarded. (If the first message on a STREAM socket
392 * is partially-read, it is discarded and the next one is rejected instead.)
394 * NOTE: Rejected messages are not necessarily returned to the sender! They
395 * are returned or discarded according to the "destination droppable" setting
396 * specified for the message by the sender.
398 * Returns 0 on success, errno otherwise
400 static int tipc_release(struct socket *sock)
402 struct sock *sk = sock->sk;
404 struct tipc_sock *tsk;
409 * Exit if socket isn't fully initialized (occurs when a failed accept()
410 * releases a pre-allocated child socket that was never used)
420 * Reject all unreceived messages, except on an active connection
421 * (which disconnects locally & sends a 'FIN+' to peer)
423 dnode = tsk_peer_node(tsk);
424 while (sock->state != SS_DISCONNECTING) {
425 skb = __skb_dequeue(&sk->sk_receive_queue);
428 if (TIPC_SKB_CB(skb)->handle != NULL)
431 if ((sock->state == SS_CONNECTING) ||
432 (sock->state == SS_CONNECTED)) {
433 sock->state = SS_DISCONNECTING;
435 tipc_node_remove_conn(net, dnode, tsk->portid);
437 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
441 tipc_sk_withdraw(tsk, 0, NULL);
442 sk_stop_timer(sk, &sk->sk_timer);
444 if (tsk->connected) {
445 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
446 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
447 tsk_own_node(tsk), tsk_peer_port(tsk),
448 tsk->portid, TIPC_ERR_NO_PORT);
450 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
451 tipc_node_remove_conn(net, dnode, tsk->portid);
454 /* Reject any messages that accumulated in backlog queue */
455 sock->state = SS_DISCONNECTING;
458 call_rcu(&tsk->rcu, tipc_sk_callback);
465 * tipc_bind - associate or disassocate TIPC name(s) with a socket
466 * @sock: socket structure
467 * @uaddr: socket address describing name(s) and desired operation
468 * @uaddr_len: size of socket address data structure
470 * Name and name sequence binding is indicated using a positive scope value;
471 * a negative scope value unbinds the specified name. Specifying no name
472 * (i.e. a socket address length of 0) unbinds all names from the socket.
474 * Returns 0 on success, errno otherwise
476 * NOTE: This routine doesn't need to take the socket lock since it doesn't
477 * access any non-constant socket information.
479 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
482 struct sock *sk = sock->sk;
483 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
484 struct tipc_sock *tsk = tipc_sk(sk);
488 if (unlikely(!uaddr_len)) {
489 res = tipc_sk_withdraw(tsk, 0, NULL);
493 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
497 if (addr->family != AF_TIPC) {
502 if (addr->addrtype == TIPC_ADDR_NAME)
503 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
504 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
509 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
510 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
511 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
516 res = (addr->scope > 0) ?
517 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
518 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
525 * tipc_getname - get port ID of socket or peer socket
526 * @sock: socket structure
527 * @uaddr: area for returned socket address
528 * @uaddr_len: area for returned length of socket address
529 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
531 * Returns 0 on success, errno otherwise
533 * NOTE: This routine doesn't need to take the socket lock since it only
534 * accesses socket information that is unchanging (or which changes in
535 * a completely predictable manner).
537 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
538 int *uaddr_len, int peer)
540 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
541 struct tipc_sock *tsk = tipc_sk(sock->sk);
542 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
544 memset(addr, 0, sizeof(*addr));
546 if ((sock->state != SS_CONNECTED) &&
547 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
549 addr->addr.id.ref = tsk_peer_port(tsk);
550 addr->addr.id.node = tsk_peer_node(tsk);
552 addr->addr.id.ref = tsk->portid;
553 addr->addr.id.node = tn->own_addr;
556 *uaddr_len = sizeof(*addr);
557 addr->addrtype = TIPC_ADDR_ID;
558 addr->family = AF_TIPC;
560 addr->addr.name.domain = 0;
566 * tipc_poll - read and possibly block on pollmask
567 * @file: file structure associated with the socket
568 * @sock: socket for which to calculate the poll bits
571 * Returns pollmask value
574 * It appears that the usual socket locking mechanisms are not useful here
575 * since the pollmask info is potentially out-of-date the moment this routine
576 * exits. TCP and other protocols seem to rely on higher level poll routines
577 * to handle any preventable race conditions, so TIPC will do the same ...
579 * TIPC sets the returned events as follows:
581 * socket state flags set
582 * ------------ ---------
583 * unconnected no read flags
584 * POLLOUT if port is not congested
586 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
589 * connected POLLIN/POLLRDNORM if data in rx queue
590 * POLLOUT if port is not congested
592 * disconnecting POLLIN/POLLRDNORM/POLLHUP
595 * listening POLLIN if SYN in rx queue
598 * ready POLLIN/POLLRDNORM if data in rx queue
599 * [connectionless] POLLOUT (since port cannot be congested)
601 * IMPORTANT: The fact that a read or write operation is indicated does NOT
602 * imply that the operation will succeed, merely that it should be performed
603 * and will not block.
605 static unsigned int tipc_poll(struct file *file, struct socket *sock,
608 struct sock *sk = sock->sk;
609 struct tipc_sock *tsk = tipc_sk(sk);
612 sock_poll_wait(file, sk_sleep(sk), wait);
614 switch ((int)sock->state) {
621 if (!tsk->link_cong && !tsk_conn_cong(tsk))
626 if (!skb_queue_empty(&sk->sk_receive_queue))
627 mask |= (POLLIN | POLLRDNORM);
629 case SS_DISCONNECTING:
630 mask = (POLLIN | POLLRDNORM | POLLHUP);
638 * tipc_sendmcast - send multicast message
639 * @sock: socket structure
640 * @seq: destination address
641 * @msg: message to send
642 * @dsz: total length of message data
643 * @timeo: timeout to wait for wakeup
645 * Called from function tipc_sendmsg(), which has done all sanity checks
646 * Returns the number of bytes sent on success, or errno
648 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
649 struct msghdr *msg, size_t dsz, long timeo)
651 struct sock *sk = sock->sk;
652 struct tipc_sock *tsk = tipc_sk(sk);
653 struct net *net = sock_net(sk);
654 struct tipc_msg *mhdr = &tsk->phdr;
655 struct sk_buff_head pktchain;
656 struct iov_iter save = msg->msg_iter;
660 msg_set_type(mhdr, TIPC_MCAST_MSG);
661 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
662 msg_set_destport(mhdr, 0);
663 msg_set_destnode(mhdr, 0);
664 msg_set_nametype(mhdr, seq->type);
665 msg_set_namelower(mhdr, seq->lower);
666 msg_set_nameupper(mhdr, seq->upper);
667 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
669 skb_queue_head_init(&pktchain);
672 mtu = tipc_bcast_get_mtu(net);
673 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, &pktchain);
674 if (unlikely(rc < 0))
678 rc = tipc_bcast_xmit(net, &pktchain);
682 if (rc == -ELINKCONG) {
684 rc = tipc_wait_for_sndmsg(sock, &timeo);
688 __skb_queue_purge(&pktchain);
689 if (rc == -EMSGSIZE) {
690 msg->msg_iter = save;
699 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
700 * @arrvq: queue with arriving messages, to be cloned after destination lookup
701 * @inputq: queue with cloned messages, delivered to socket after dest lookup
703 * Multi-threaded: parallel calls with reference to same queues may occur
705 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
706 struct sk_buff_head *inputq)
708 struct tipc_msg *msg;
709 struct tipc_plist dports;
711 u32 scope = TIPC_CLUSTER_SCOPE;
712 struct sk_buff_head tmpq;
714 struct sk_buff *skb, *_skb;
716 __skb_queue_head_init(&tmpq);
717 tipc_plist_init(&dports);
719 skb = tipc_skb_peek(arrvq, &inputq->lock);
720 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
722 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
724 if (in_own_node(net, msg_orignode(msg)))
725 scope = TIPC_NODE_SCOPE;
727 /* Create destination port list and message clones: */
728 tipc_nametbl_mc_translate(net,
729 msg_nametype(msg), msg_namelower(msg),
730 msg_nameupper(msg), scope, &dports);
731 portid = tipc_plist_pop(&dports);
732 for (; portid; portid = tipc_plist_pop(&dports)) {
733 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
735 msg_set_destport(buf_msg(_skb), portid);
736 __skb_queue_tail(&tmpq, _skb);
739 pr_warn("Failed to clone mcast rcv buffer\n");
741 /* Append to inputq if not already done by other thread */
742 spin_lock_bh(&inputq->lock);
743 if (skb_peek(arrvq) == skb) {
744 skb_queue_splice_tail_init(&tmpq, inputq);
745 /* Decrease the skb's refcnt as increasing in the
746 * function tipc_skb_peek
748 kfree_skb(__skb_dequeue(arrvq));
750 spin_unlock_bh(&inputq->lock);
751 __skb_queue_purge(&tmpq);
754 tipc_sk_rcv(net, inputq);
758 * tipc_sk_proto_rcv - receive a connection mng protocol message
759 * @tsk: receiving socket
760 * @skb: pointer to message buffer.
762 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
763 struct sk_buff_head *xmitq)
765 struct sock *sk = &tsk->sk;
766 u32 onode = tsk_own_node(tsk);
767 struct tipc_msg *hdr = buf_msg(skb);
768 int mtyp = msg_type(hdr);
771 /* Ignore if connection cannot be validated: */
772 if (!tsk_peer_msg(tsk, hdr))
775 tsk->probing_state = TIPC_CONN_OK;
777 if (mtyp == CONN_PROBE) {
778 msg_set_type(hdr, CONN_PROBE_REPLY);
779 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
780 __skb_queue_tail(xmitq, skb);
782 } else if (mtyp == CONN_ACK) {
783 conn_cong = tsk_conn_cong(tsk);
784 tsk->snt_unacked -= msg_conn_ack(hdr);
785 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
786 tsk->snd_win = msg_adv_win(hdr);
788 sk->sk_write_space(sk);
789 } else if (mtyp != CONN_PROBE_REPLY) {
790 pr_warn("Received unknown CONN_PROTO msg\n");
796 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
798 struct sock *sk = sock->sk;
799 struct tipc_sock *tsk = tipc_sk(sk);
804 int err = sock_error(sk);
807 if (sock->state == SS_DISCONNECTING)
811 if (signal_pending(current))
812 return sock_intr_errno(*timeo_p);
814 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
815 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
816 finish_wait(sk_sleep(sk), &wait);
822 * tipc_sendmsg - send message in connectionless manner
823 * @sock: socket structure
824 * @m: message to send
825 * @dsz: amount of user data to be sent
827 * Message must have an destination specified explicitly.
828 * Used for SOCK_RDM and SOCK_DGRAM messages,
829 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
830 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
832 * Returns the number of bytes sent on success, or errno otherwise
834 static int tipc_sendmsg(struct socket *sock,
835 struct msghdr *m, size_t dsz)
837 struct sock *sk = sock->sk;
841 ret = __tipc_sendmsg(sock, m, dsz);
847 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz)
849 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
850 struct sock *sk = sock->sk;
851 struct tipc_sock *tsk = tipc_sk(sk);
852 struct net *net = sock_net(sk);
853 struct tipc_msg *mhdr = &tsk->phdr;
855 struct sk_buff_head pktchain;
857 struct tipc_name_seq *seq;
858 struct iov_iter save;
863 if (dsz > TIPC_MAX_USER_MSG_SIZE)
865 if (unlikely(!dest)) {
866 if (tsk->connected && sock->state == SS_READY)
869 return -EDESTADDRREQ;
870 } else if (unlikely(m->msg_namelen < sizeof(*dest)) ||
871 dest->family != AF_TIPC) {
874 if (unlikely(sock->state != SS_READY)) {
875 if (sock->state == SS_LISTENING)
877 if (sock->state != SS_UNCONNECTED)
881 if (dest->addrtype == TIPC_ADDR_NAME) {
882 tsk->conn_type = dest->addr.name.name.type;
883 tsk->conn_instance = dest->addr.name.name.instance;
886 seq = &dest->addr.nameseq;
887 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
889 if (dest->addrtype == TIPC_ADDR_MCAST) {
890 return tipc_sendmcast(sock, seq, m, dsz, timeo);
891 } else if (dest->addrtype == TIPC_ADDR_NAME) {
892 u32 type = dest->addr.name.name.type;
893 u32 inst = dest->addr.name.name.instance;
894 u32 domain = dest->addr.name.domain;
897 msg_set_type(mhdr, TIPC_NAMED_MSG);
898 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
899 msg_set_nametype(mhdr, type);
900 msg_set_nameinst(mhdr, inst);
901 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
902 dport = tipc_nametbl_translate(net, type, inst, &dnode);
903 msg_set_destnode(mhdr, dnode);
904 msg_set_destport(mhdr, dport);
905 if (unlikely(!dport && !dnode))
906 return -EHOSTUNREACH;
907 } else if (dest->addrtype == TIPC_ADDR_ID) {
908 dnode = dest->addr.id.node;
909 msg_set_type(mhdr, TIPC_DIRECT_MSG);
910 msg_set_lookup_scope(mhdr, 0);
911 msg_set_destnode(mhdr, dnode);
912 msg_set_destport(mhdr, dest->addr.id.ref);
913 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
916 skb_queue_head_init(&pktchain);
919 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
920 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, &pktchain);
925 skb = skb_peek(&pktchain);
926 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
927 rc = tipc_node_xmit(net, &pktchain, dnode, tsk->portid);
929 if (sock->state != SS_READY)
930 sock->state = SS_CONNECTING;
933 if (rc == -ELINKCONG) {
935 rc = tipc_wait_for_sndmsg(sock, &timeo);
939 __skb_queue_purge(&pktchain);
940 if (rc == -EMSGSIZE) {
950 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
952 struct sock *sk = sock->sk;
953 struct tipc_sock *tsk = tipc_sk(sk);
958 int err = sock_error(sk);
961 if (sock->state == SS_DISCONNECTING)
963 else if (sock->state != SS_CONNECTED)
967 if (signal_pending(current))
968 return sock_intr_errno(*timeo_p);
970 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
971 done = sk_wait_event(sk, timeo_p,
973 !tsk_conn_cong(tsk)) ||
975 finish_wait(sk_sleep(sk), &wait);
981 * tipc_send_stream - send stream-oriented data
982 * @sock: socket structure
984 * @dsz: total length of data to be transmitted
986 * Used for SOCK_STREAM data.
988 * Returns the number of bytes sent on success (or partial success),
989 * or errno if no data sent
991 static int tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
993 struct sock *sk = sock->sk;
997 ret = __tipc_send_stream(sock, m, dsz);
1003 static int __tipc_send_stream(struct socket *sock, struct msghdr *m, size_t dsz)
1005 struct sock *sk = sock->sk;
1006 struct net *net = sock_net(sk);
1007 struct tipc_sock *tsk = tipc_sk(sk);
1008 struct tipc_msg *mhdr = &tsk->phdr;
1009 struct sk_buff_head pktchain;
1010 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1011 u32 portid = tsk->portid;
1015 uint mtu, send, sent = 0;
1016 struct iov_iter save;
1017 int hlen = MIN_H_SIZE;
1019 /* Handle implied connection establishment */
1020 if (unlikely(dest)) {
1021 rc = __tipc_sendmsg(sock, m, dsz);
1022 hlen = msg_hdr_sz(mhdr);
1023 if (dsz && (dsz == rc))
1024 tsk->snt_unacked = tsk_inc(tsk, dsz + hlen);
1027 if (dsz > (uint)INT_MAX)
1030 if (unlikely(sock->state != SS_CONNECTED)) {
1031 if (sock->state == SS_DISCONNECTING)
1037 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1038 dnode = tsk_peer_node(tsk);
1039 skb_queue_head_init(&pktchain);
1044 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1045 rc = tipc_msg_build(mhdr, m, sent, send, mtu, &pktchain);
1046 if (unlikely(rc < 0))
1050 if (likely(!tsk_conn_cong(tsk))) {
1051 rc = tipc_node_xmit(net, &pktchain, dnode, portid);
1053 tsk->snt_unacked += tsk_inc(tsk, send + hlen);
1059 if (rc == -EMSGSIZE) {
1060 __skb_queue_purge(&pktchain);
1061 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1066 if (rc != -ELINKCONG)
1071 rc = tipc_wait_for_sndpkt(sock, &timeo);
1074 __skb_queue_purge(&pktchain);
1075 return sent ? sent : rc;
1079 * tipc_send_packet - send a connection-oriented message
1080 * @sock: socket structure
1081 * @m: message to send
1082 * @dsz: length of data to be transmitted
1084 * Used for SOCK_SEQPACKET messages.
1086 * Returns the number of bytes sent on success, or errno otherwise
1088 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1090 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1093 return tipc_send_stream(sock, m, dsz);
1096 /* tipc_sk_finish_conn - complete the setup of a connection
1098 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1101 struct sock *sk = &tsk->sk;
1102 struct net *net = sock_net(sk);
1103 struct tipc_msg *msg = &tsk->phdr;
1105 msg_set_destnode(msg, peer_node);
1106 msg_set_destport(msg, peer_port);
1107 msg_set_type(msg, TIPC_CONN_MSG);
1108 msg_set_lookup_scope(msg, 0);
1109 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1111 tsk->probing_intv = CONN_PROBING_INTERVAL;
1112 tsk->probing_state = TIPC_CONN_OK;
1114 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1115 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1116 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1117 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1118 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1121 /* Fall back to message based flow control */
1122 tsk->rcv_win = FLOWCTL_MSG_WIN;
1123 tsk->snd_win = FLOWCTL_MSG_WIN;
1127 * set_orig_addr - capture sender's address for received message
1128 * @m: descriptor for message info
1129 * @msg: received message header
1131 * Note: Address is not captured if not requested by receiver.
1133 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1135 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1138 addr->family = AF_TIPC;
1139 addr->addrtype = TIPC_ADDR_ID;
1140 memset(&addr->addr, 0, sizeof(addr->addr));
1141 addr->addr.id.ref = msg_origport(msg);
1142 addr->addr.id.node = msg_orignode(msg);
1143 addr->addr.name.domain = 0; /* could leave uninitialized */
1144 addr->scope = 0; /* could leave uninitialized */
1145 m->msg_namelen = sizeof(struct sockaddr_tipc);
1150 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1151 * @m: descriptor for message info
1152 * @msg: received message header
1153 * @tsk: TIPC port associated with message
1155 * Note: Ancillary data is not captured if not requested by receiver.
1157 * Returns 0 if successful, otherwise errno
1159 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1160 struct tipc_sock *tsk)
1168 if (likely(m->msg_controllen == 0))
1171 /* Optionally capture errored message object(s) */
1172 err = msg ? msg_errcode(msg) : 0;
1173 if (unlikely(err)) {
1175 anc_data[1] = msg_data_sz(msg);
1176 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1180 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1187 /* Optionally capture message destination object */
1188 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1189 switch (dest_type) {
1190 case TIPC_NAMED_MSG:
1192 anc_data[0] = msg_nametype(msg);
1193 anc_data[1] = msg_namelower(msg);
1194 anc_data[2] = msg_namelower(msg);
1196 case TIPC_MCAST_MSG:
1198 anc_data[0] = msg_nametype(msg);
1199 anc_data[1] = msg_namelower(msg);
1200 anc_data[2] = msg_nameupper(msg);
1203 has_name = (tsk->conn_type != 0);
1204 anc_data[0] = tsk->conn_type;
1205 anc_data[1] = tsk->conn_instance;
1206 anc_data[2] = tsk->conn_instance;
1212 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1220 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1222 struct net *net = sock_net(&tsk->sk);
1223 struct sk_buff *skb = NULL;
1224 struct tipc_msg *msg;
1225 u32 peer_port = tsk_peer_port(tsk);
1226 u32 dnode = tsk_peer_node(tsk);
1228 if (!tsk->connected)
1230 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1231 dnode, tsk_own_node(tsk), peer_port,
1232 tsk->portid, TIPC_OK);
1236 msg_set_conn_ack(msg, tsk->rcv_unacked);
1237 tsk->rcv_unacked = 0;
1239 /* Adjust to and advertize the correct window limit */
1240 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1241 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1242 msg_set_adv_win(msg, tsk->rcv_win);
1244 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1247 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1249 struct sock *sk = sock->sk;
1251 long timeo = *timeop;
1255 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1256 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1257 if (sock->state == SS_DISCONNECTING) {
1262 timeo = schedule_timeout(timeo);
1266 if (!skb_queue_empty(&sk->sk_receive_queue))
1271 err = sock_intr_errno(timeo);
1272 if (signal_pending(current))
1275 finish_wait(sk_sleep(sk), &wait);
1281 * tipc_recvmsg - receive packet-oriented message
1282 * @m: descriptor for message info
1283 * @buf_len: total size of user buffer area
1284 * @flags: receive flags
1286 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1287 * If the complete message doesn't fit in user area, truncate it.
1289 * Returns size of returned message data, errno otherwise
1291 static int tipc_recvmsg(struct socket *sock, struct msghdr *m, size_t buf_len,
1294 struct sock *sk = sock->sk;
1295 struct tipc_sock *tsk = tipc_sk(sk);
1296 struct sk_buff *buf;
1297 struct tipc_msg *msg;
1303 /* Catch invalid receive requests */
1304 if (unlikely(!buf_len))
1309 if (unlikely(sock->state == SS_UNCONNECTED)) {
1314 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1317 /* Look for a message in receive queue; wait if necessary */
1318 res = tipc_wait_for_rcvmsg(sock, &timeo);
1322 /* Look at first message in receive queue */
1323 buf = skb_peek(&sk->sk_receive_queue);
1325 sz = msg_data_sz(msg);
1326 hlen = msg_hdr_sz(msg);
1327 err = msg_errcode(msg);
1329 /* Discard an empty non-errored message & try again */
1330 if ((!sz) && (!err)) {
1331 tsk_advance_rx_queue(sk);
1335 /* Capture sender's address (optional) */
1336 set_orig_addr(m, msg);
1338 /* Capture ancillary data (optional) */
1339 res = tipc_sk_anc_data_recv(m, msg, tsk);
1343 /* Capture message data (if valid) & compute return value (always) */
1345 if (unlikely(buf_len < sz)) {
1347 m->msg_flags |= MSG_TRUNC;
1349 res = skb_copy_datagram_msg(buf, hlen, m, sz);
1354 if ((sock->state == SS_READY) ||
1355 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1361 if (unlikely(flags & MSG_PEEK))
1364 if (likely(sock->state != SS_READY)) {
1365 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1366 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1367 tipc_sk_send_ack(tsk);
1369 tsk_advance_rx_queue(sk);
1376 * tipc_recv_stream - receive stream-oriented data
1377 * @m: descriptor for message info
1378 * @buf_len: total size of user buffer area
1379 * @flags: receive flags
1381 * Used for SOCK_STREAM messages only. If not enough data is available
1382 * will optionally wait for more; never truncates data.
1384 * Returns size of returned message data, errno otherwise
1386 static int tipc_recv_stream(struct socket *sock, struct msghdr *m,
1387 size_t buf_len, int flags)
1389 struct sock *sk = sock->sk;
1390 struct tipc_sock *tsk = tipc_sk(sk);
1391 struct sk_buff *buf;
1392 struct tipc_msg *msg;
1395 int sz_to_copy, target, needed;
1400 /* Catch invalid receive attempts */
1401 if (unlikely(!buf_len))
1406 if (unlikely(sock->state == SS_UNCONNECTED)) {
1411 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1412 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1415 /* Look for a message in receive queue; wait if necessary */
1416 res = tipc_wait_for_rcvmsg(sock, &timeo);
1420 /* Look at first message in receive queue */
1421 buf = skb_peek(&sk->sk_receive_queue);
1423 sz = msg_data_sz(msg);
1424 hlen = msg_hdr_sz(msg);
1425 err = msg_errcode(msg);
1427 /* Discard an empty non-errored message & try again */
1428 if ((!sz) && (!err)) {
1429 tsk_advance_rx_queue(sk);
1433 /* Optionally capture sender's address & ancillary data of first msg */
1434 if (sz_copied == 0) {
1435 set_orig_addr(m, msg);
1436 res = tipc_sk_anc_data_recv(m, msg, tsk);
1441 /* Capture message data (if valid) & compute return value (always) */
1443 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1446 needed = (buf_len - sz_copied);
1447 sz_to_copy = (sz <= needed) ? sz : needed;
1449 res = skb_copy_datagram_msg(buf, hlen + offset, m, sz_to_copy);
1453 sz_copied += sz_to_copy;
1455 if (sz_to_copy < sz) {
1456 if (!(flags & MSG_PEEK))
1457 TIPC_SKB_CB(buf)->handle =
1458 (void *)(unsigned long)(offset + sz_to_copy);
1463 goto exit; /* can't add error msg to valid data */
1465 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1471 if (unlikely(flags & MSG_PEEK))
1474 tsk->rcv_unacked += tsk_inc(tsk, hlen + sz);
1475 if (unlikely(tsk->rcv_unacked >= (tsk->rcv_win / 4)))
1476 tipc_sk_send_ack(tsk);
1477 tsk_advance_rx_queue(sk);
1479 /* Loop around if more data is required */
1480 if ((sz_copied < buf_len) && /* didn't get all requested data */
1481 (!skb_queue_empty(&sk->sk_receive_queue) ||
1482 (sz_copied < target)) && /* and more is ready or required */
1483 (!err)) /* and haven't reached a FIN */
1488 return sz_copied ? sz_copied : res;
1492 * tipc_write_space - wake up thread if port congestion is released
1495 static void tipc_write_space(struct sock *sk)
1497 struct socket_wq *wq;
1500 wq = rcu_dereference(sk->sk_wq);
1501 if (skwq_has_sleeper(wq))
1502 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1503 POLLWRNORM | POLLWRBAND);
1508 * tipc_data_ready - wake up threads to indicate messages have been received
1510 * @len: the length of messages
1512 static void tipc_data_ready(struct sock *sk)
1514 struct socket_wq *wq;
1517 wq = rcu_dereference(sk->sk_wq);
1518 if (skwq_has_sleeper(wq))
1519 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1520 POLLRDNORM | POLLRDBAND);
1524 static void tipc_sock_destruct(struct sock *sk)
1526 __skb_queue_purge(&sk->sk_receive_queue);
1530 * filter_connect - Handle all incoming messages for a connection-based socket
1532 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1534 * Returns true if everything ok, false otherwise
1536 static bool filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1538 struct sock *sk = &tsk->sk;
1539 struct net *net = sock_net(sk);
1540 struct socket *sock = sk->sk_socket;
1541 struct tipc_msg *hdr = buf_msg(skb);
1543 if (unlikely(msg_mcast(hdr)))
1546 switch ((int)sock->state) {
1549 /* Accept only connection-based messages sent by peer */
1550 if (unlikely(!tsk_peer_msg(tsk, hdr)))
1553 if (unlikely(msg_errcode(hdr))) {
1554 sock->state = SS_DISCONNECTING;
1556 /* Let timer expire on it's own */
1557 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1564 /* Accept only ACK or NACK message */
1565 if (unlikely(!msg_connected(hdr)))
1568 if (unlikely(msg_errcode(hdr))) {
1569 sock->state = SS_DISCONNECTING;
1570 sk->sk_err = ECONNREFUSED;
1574 if (unlikely(!msg_isdata(hdr))) {
1575 sock->state = SS_DISCONNECTING;
1576 sk->sk_err = EINVAL;
1580 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
1581 msg_set_importance(&tsk->phdr, msg_importance(hdr));
1582 sock->state = SS_CONNECTED;
1584 /* If 'ACK+' message, add to socket receive queue */
1585 if (msg_data_sz(hdr))
1588 /* If empty 'ACK-' message, wake up sleeping connect() */
1589 if (waitqueue_active(sk_sleep(sk)))
1590 wake_up_interruptible(sk_sleep(sk));
1592 /* 'ACK-' message is neither accepted nor rejected: */
1593 msg_set_dest_droppable(hdr, 1);
1597 case SS_UNCONNECTED:
1599 /* Accept only SYN message */
1600 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
1603 case SS_DISCONNECTING:
1606 pr_err("Unknown socket state %u\n", sock->state);
1612 * rcvbuf_limit - get proper overload limit of socket receive queue
1616 * For connection oriented messages, irrespective of importance,
1617 * default queue limit is 2 MB.
1619 * For connectionless messages, queue limits are based on message
1620 * importance as follows:
1622 * TIPC_LOW_IMPORTANCE (2 MB)
1623 * TIPC_MEDIUM_IMPORTANCE (4 MB)
1624 * TIPC_HIGH_IMPORTANCE (8 MB)
1625 * TIPC_CRITICAL_IMPORTANCE (16 MB)
1627 * Returns overload limit according to corresponding message importance
1629 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
1631 struct tipc_sock *tsk = tipc_sk(sk);
1632 struct tipc_msg *hdr = buf_msg(skb);
1634 if (unlikely(!msg_connected(hdr)))
1635 return sk->sk_rcvbuf << msg_importance(hdr);
1637 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
1638 return sk->sk_rcvbuf;
1640 return FLOWCTL_MSG_LIM;
1644 * filter_rcv - validate incoming message
1646 * @skb: pointer to message.
1648 * Enqueues message on receive queue if acceptable; optionally handles
1649 * disconnect indication for a connected socket.
1651 * Called with socket lock already taken
1653 * Returns true if message was added to socket receive queue, otherwise false
1655 static bool filter_rcv(struct sock *sk, struct sk_buff *skb,
1656 struct sk_buff_head *xmitq)
1658 struct socket *sock = sk->sk_socket;
1659 struct tipc_sock *tsk = tipc_sk(sk);
1660 struct tipc_msg *hdr = buf_msg(skb);
1661 unsigned int limit = rcvbuf_limit(sk, skb);
1663 int usr = msg_user(hdr);
1665 if (unlikely(msg_user(hdr) == CONN_MANAGER)) {
1666 tipc_sk_proto_rcv(tsk, skb, xmitq);
1670 if (unlikely(usr == SOCK_WAKEUP)) {
1673 sk->sk_write_space(sk);
1677 /* Drop if illegal message type */
1678 if (unlikely(msg_type(hdr) > TIPC_DIRECT_MSG)) {
1683 /* Reject if wrong message type for current socket state */
1684 if (unlikely(sock->state == SS_READY)) {
1685 if (msg_connected(hdr)) {
1686 err = TIPC_ERR_NO_PORT;
1689 } else if (unlikely(!filter_connect(tsk, skb))) {
1690 err = TIPC_ERR_NO_PORT;
1694 /* Reject message if there isn't room to queue it */
1695 if (unlikely(sk_rmem_alloc_get(sk) + skb->truesize >= limit)) {
1696 err = TIPC_ERR_OVERLOAD;
1700 /* Enqueue message */
1701 TIPC_SKB_CB(skb)->handle = NULL;
1702 __skb_queue_tail(&sk->sk_receive_queue, skb);
1703 skb_set_owner_r(skb, sk);
1705 sk->sk_data_ready(sk);
1709 if (tipc_msg_reverse(tsk_own_node(tsk), &skb, err))
1710 __skb_queue_tail(xmitq, skb);
1715 * tipc_backlog_rcv - handle incoming message from backlog queue
1719 * Caller must hold socket lock
1723 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1725 unsigned int truesize = skb->truesize;
1726 struct sk_buff_head xmitq;
1727 u32 dnode, selector;
1729 __skb_queue_head_init(&xmitq);
1731 if (likely(filter_rcv(sk, skb, &xmitq))) {
1732 atomic_add(truesize, &tipc_sk(sk)->dupl_rcvcnt);
1736 if (skb_queue_empty(&xmitq))
1739 /* Send response/rejected message */
1740 skb = __skb_dequeue(&xmitq);
1741 dnode = msg_destnode(buf_msg(skb));
1742 selector = msg_origport(buf_msg(skb));
1743 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
1748 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1749 * inputq and try adding them to socket or backlog queue
1750 * @inputq: list of incoming buffers with potentially different destinations
1751 * @sk: socket where the buffers should be enqueued
1752 * @dport: port number for the socket
1754 * Caller must hold socket lock
1756 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1757 u32 dport, struct sk_buff_head *xmitq)
1759 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
1760 struct sk_buff *skb;
1765 while (skb_queue_len(inputq)) {
1766 if (unlikely(time_after_eq(jiffies, time_limit)))
1769 skb = tipc_skb_dequeue(inputq, dport);
1773 /* Add message directly to receive queue if possible */
1774 if (!sock_owned_by_user(sk)) {
1775 filter_rcv(sk, skb, xmitq);
1779 /* Try backlog, compensating for double-counted bytes */
1780 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1781 if (!sk->sk_backlog.len)
1782 atomic_set(dcnt, 0);
1783 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1784 if (likely(!sk_add_backlog(sk, skb, lim)))
1787 /* Overload => reject message back to sender */
1788 onode = tipc_own_addr(sock_net(sk));
1789 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
1790 __skb_queue_tail(xmitq, skb);
1796 * tipc_sk_rcv - handle a chain of incoming buffers
1797 * @inputq: buffer list containing the buffers
1798 * Consumes all buffers in list until inputq is empty
1799 * Note: may be called in multiple threads referring to the same queue
1801 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1803 struct sk_buff_head xmitq;
1804 u32 dnode, dport = 0;
1806 struct tipc_sock *tsk;
1808 struct sk_buff *skb;
1810 __skb_queue_head_init(&xmitq);
1811 while (skb_queue_len(inputq)) {
1812 dport = tipc_skb_peek_port(inputq, dport);
1813 tsk = tipc_sk_lookup(net, dport);
1817 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1818 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
1819 spin_unlock_bh(&sk->sk_lock.slock);
1821 /* Send pending response/rejected messages, if any */
1822 while ((skb = __skb_dequeue(&xmitq))) {
1823 dnode = msg_destnode(buf_msg(skb));
1824 tipc_node_xmit_skb(net, skb, dnode, dport);
1830 /* No destination socket => dequeue skb if still there */
1831 skb = tipc_skb_dequeue(inputq, dport);
1835 /* Try secondary lookup if unresolved named message */
1836 err = TIPC_ERR_NO_PORT;
1837 if (tipc_msg_lookup_dest(net, skb, &err))
1840 /* Prepare for message rejection */
1841 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
1844 dnode = msg_destnode(buf_msg(skb));
1845 tipc_node_xmit_skb(net, skb, dnode, dport);
1849 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1851 struct sock *sk = sock->sk;
1856 int err = sock_error(sk);
1861 if (signal_pending(current))
1862 return sock_intr_errno(*timeo_p);
1864 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1865 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1866 finish_wait(sk_sleep(sk), &wait);
1872 * tipc_connect - establish a connection to another TIPC port
1873 * @sock: socket structure
1874 * @dest: socket address for destination port
1875 * @destlen: size of socket address data structure
1876 * @flags: file-related flags associated with socket
1878 * Returns 0 on success, errno otherwise
1880 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1881 int destlen, int flags)
1883 struct sock *sk = sock->sk;
1884 struct tipc_sock *tsk = tipc_sk(sk);
1885 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1886 struct msghdr m = {NULL,};
1887 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
1888 socket_state previous;
1893 /* DGRAM/RDM connect(), just save the destaddr */
1894 if (sock->state == SS_READY) {
1895 if (dst->family == AF_UNSPEC) {
1896 memset(&tsk->remote, 0, sizeof(struct sockaddr_tipc));
1898 } else if (destlen != sizeof(struct sockaddr_tipc)) {
1901 memcpy(&tsk->remote, dest, destlen);
1908 * Reject connection attempt using multicast address
1910 * Note: send_msg() validates the rest of the address fields,
1911 * so there's no need to do it here
1913 if (dst->addrtype == TIPC_ADDR_MCAST) {
1918 previous = sock->state;
1919 switch (sock->state) {
1920 case SS_UNCONNECTED:
1921 /* Send a 'SYN-' to destination */
1923 m.msg_namelen = destlen;
1925 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1926 * indicate send_msg() is never blocked.
1929 m.msg_flags = MSG_DONTWAIT;
1931 res = __tipc_sendmsg(sock, &m, 0);
1932 if ((res < 0) && (res != -EWOULDBLOCK))
1935 /* Just entered SS_CONNECTING state; the only
1936 * difference is that return value in non-blocking
1937 * case is EINPROGRESS, rather than EALREADY.
1941 if (previous == SS_CONNECTING)
1945 timeout = msecs_to_jiffies(timeout);
1946 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1947 res = tipc_wait_for_connect(sock, &timeout);
1962 * tipc_listen - allow socket to listen for incoming connections
1963 * @sock: socket structure
1966 * Returns 0 on success, errno otherwise
1968 static int tipc_listen(struct socket *sock, int len)
1970 struct sock *sk = sock->sk;
1975 if (sock->state != SS_UNCONNECTED)
1978 sock->state = SS_LISTENING;
1986 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1988 struct sock *sk = sock->sk;
1989 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1992 /* True wake-one mechanism for incoming connections: only
1993 * one process gets woken up, not the 'whole herd'.
1994 * Since we do not 'race & poll' for established sockets
1995 * anymore, the common case will execute the loop only once.
1998 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1999 add_wait_queue(sk_sleep(sk), &wait);
2001 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2003 remove_wait_queue(sk_sleep(sk), &wait);
2006 if (!skb_queue_empty(&sk->sk_receive_queue))
2009 if (sock->state != SS_LISTENING)
2014 err = sock_intr_errno(timeo);
2015 if (signal_pending(current))
2022 * tipc_accept - wait for connection request
2023 * @sock: listening socket
2024 * @newsock: new socket that is to be connected
2025 * @flags: file-related flags associated with socket
2027 * Returns 0 on success, errno otherwise
2029 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2031 struct sock *new_sk, *sk = sock->sk;
2032 struct sk_buff *buf;
2033 struct tipc_sock *new_tsock;
2034 struct tipc_msg *msg;
2040 if (sock->state != SS_LISTENING) {
2044 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2045 res = tipc_wait_for_accept(sock, timeo);
2049 buf = skb_peek(&sk->sk_receive_queue);
2051 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2054 security_sk_clone(sock->sk, new_sock->sk);
2056 new_sk = new_sock->sk;
2057 new_tsock = tipc_sk(new_sk);
2060 /* we lock on new_sk; but lockdep sees the lock on sk */
2061 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2064 * Reject any stray messages received by new socket
2065 * before the socket lock was taken (very, very unlikely)
2067 tsk_rej_rx_queue(new_sk);
2069 /* Connect new socket to it's peer */
2070 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2071 new_sock->state = SS_CONNECTED;
2073 tsk_set_importance(new_tsock, msg_importance(msg));
2074 if (msg_named(msg)) {
2075 new_tsock->conn_type = msg_nametype(msg);
2076 new_tsock->conn_instance = msg_nameinst(msg);
2080 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2081 * Respond to 'SYN+' by queuing it on new socket.
2083 if (!msg_data_sz(msg)) {
2084 struct msghdr m = {NULL,};
2086 tsk_advance_rx_queue(sk);
2087 __tipc_send_stream(new_sock, &m, 0);
2089 __skb_dequeue(&sk->sk_receive_queue);
2090 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2091 skb_set_owner_r(buf, new_sk);
2093 release_sock(new_sk);
2100 * tipc_shutdown - shutdown socket connection
2101 * @sock: socket structure
2102 * @how: direction to close (must be SHUT_RDWR)
2104 * Terminates connection (if necessary), then purges socket's receive queue.
2106 * Returns 0 on success, errno otherwise
2108 static int tipc_shutdown(struct socket *sock, int how)
2110 struct sock *sk = sock->sk;
2111 struct net *net = sock_net(sk);
2112 struct tipc_sock *tsk = tipc_sk(sk);
2113 struct sk_buff *skb;
2114 u32 dnode = tsk_peer_node(tsk);
2115 u32 dport = tsk_peer_port(tsk);
2116 u32 onode = tipc_own_addr(net);
2117 u32 oport = tsk->portid;
2120 if (how != SHUT_RDWR)
2125 switch (sock->state) {
2130 dnode = tsk_peer_node(tsk);
2132 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2133 skb = __skb_dequeue(&sk->sk_receive_queue);
2135 if (TIPC_SKB_CB(skb)->handle != NULL) {
2139 tipc_sk_respond(sk, skb, TIPC_CONN_SHUTDOWN);
2141 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2142 TIPC_CONN_MSG, SHORT_H_SIZE,
2143 0, dnode, onode, dport, oport,
2144 TIPC_CONN_SHUTDOWN);
2146 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
2149 sock->state = SS_DISCONNECTING;
2150 tipc_node_remove_conn(net, dnode, tsk->portid);
2153 case SS_DISCONNECTING:
2155 /* Discard any unreceived messages */
2156 __skb_queue_purge(&sk->sk_receive_queue);
2158 /* Wake up anyone sleeping in poll */
2159 sk->sk_state_change(sk);
2171 static void tipc_sk_timeout(unsigned long data)
2173 struct tipc_sock *tsk = (struct tipc_sock *)data;
2174 struct sock *sk = &tsk->sk;
2175 struct sk_buff *skb = NULL;
2176 u32 peer_port, peer_node;
2177 u32 own_node = tsk_own_node(tsk);
2180 if (!tsk->connected) {
2184 peer_port = tsk_peer_port(tsk);
2185 peer_node = tsk_peer_node(tsk);
2187 if (tsk->probing_state == TIPC_CONN_PROBING) {
2188 if (!sock_owned_by_user(sk)) {
2189 sk->sk_socket->state = SS_DISCONNECTING;
2191 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
2192 tsk_peer_port(tsk));
2193 sk->sk_state_change(sk);
2195 /* Try again later */
2196 sk_reset_timer(sk, &sk->sk_timer, (HZ / 20));
2200 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2201 INT_H_SIZE, 0, peer_node, own_node,
2202 peer_port, tsk->portid, TIPC_OK);
2203 tsk->probing_state = TIPC_CONN_PROBING;
2204 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2208 tipc_node_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2213 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2214 struct tipc_name_seq const *seq)
2216 struct net *net = sock_net(&tsk->sk);
2217 struct publication *publ;
2222 key = tsk->portid + tsk->pub_count + 1;
2223 if (key == tsk->portid)
2226 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2227 scope, tsk->portid, key);
2228 if (unlikely(!publ))
2231 list_add(&publ->pport_list, &tsk->publications);
2237 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2238 struct tipc_name_seq const *seq)
2240 struct net *net = sock_net(&tsk->sk);
2241 struct publication *publ;
2242 struct publication *safe;
2245 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2247 if (publ->scope != scope)
2249 if (publ->type != seq->type)
2251 if (publ->lower != seq->lower)
2253 if (publ->upper != seq->upper)
2255 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2256 publ->ref, publ->key);
2260 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2261 publ->ref, publ->key);
2264 if (list_empty(&tsk->publications))
2269 /* tipc_sk_reinit: set non-zero address in all existing sockets
2270 * when we go from standalone to network mode.
2272 void tipc_sk_reinit(struct net *net)
2274 struct tipc_net *tn = net_generic(net, tipc_net_id);
2275 struct rhashtable_iter iter;
2276 struct tipc_sock *tsk;
2277 struct tipc_msg *msg;
2279 rhashtable_walk_enter(&tn->sk_rht, &iter);
2282 tsk = ERR_PTR(rhashtable_walk_start(&iter));
2286 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2287 sock_hold(&tsk->sk);
2288 rhashtable_walk_stop(&iter);
2289 lock_sock(&tsk->sk);
2291 msg_set_prevnode(msg, tn->own_addr);
2292 msg_set_orignode(msg, tn->own_addr);
2293 release_sock(&tsk->sk);
2294 rhashtable_walk_start(&iter);
2298 rhashtable_walk_stop(&iter);
2299 } while (tsk == ERR_PTR(-EAGAIN));
2302 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2304 struct tipc_net *tn = net_generic(net, tipc_net_id);
2305 struct tipc_sock *tsk;
2308 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2310 sock_hold(&tsk->sk);
2316 static int tipc_sk_insert(struct tipc_sock *tsk)
2318 struct sock *sk = &tsk->sk;
2319 struct net *net = sock_net(sk);
2320 struct tipc_net *tn = net_generic(net, tipc_net_id);
2321 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2322 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2324 while (remaining--) {
2326 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2327 portid = TIPC_MIN_PORT;
2328 tsk->portid = portid;
2329 sock_hold(&tsk->sk);
2330 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2339 static void tipc_sk_remove(struct tipc_sock *tsk)
2341 struct sock *sk = &tsk->sk;
2342 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2344 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2345 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2350 static const struct rhashtable_params tsk_rht_params = {
2352 .head_offset = offsetof(struct tipc_sock, node),
2353 .key_offset = offsetof(struct tipc_sock, portid),
2354 .key_len = sizeof(u32), /* portid */
2355 .max_size = 1048576,
2357 .automatic_shrinking = true,
2360 int tipc_sk_rht_init(struct net *net)
2362 struct tipc_net *tn = net_generic(net, tipc_net_id);
2364 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2367 void tipc_sk_rht_destroy(struct net *net)
2369 struct tipc_net *tn = net_generic(net, tipc_net_id);
2371 /* Wait for socket readers to complete */
2374 rhashtable_destroy(&tn->sk_rht);
2378 * tipc_setsockopt - set socket option
2379 * @sock: socket structure
2380 * @lvl: option level
2381 * @opt: option identifier
2382 * @ov: pointer to new option value
2383 * @ol: length of option value
2385 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2386 * (to ease compatibility).
2388 * Returns 0 on success, errno otherwise
2390 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2391 char __user *ov, unsigned int ol)
2393 struct sock *sk = sock->sk;
2394 struct tipc_sock *tsk = tipc_sk(sk);
2398 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2400 if (lvl != SOL_TIPC)
2401 return -ENOPROTOOPT;
2402 if (ol < sizeof(value))
2404 res = get_user(value, (u32 __user *)ov);
2411 case TIPC_IMPORTANCE:
2412 res = tsk_set_importance(tsk, value);
2414 case TIPC_SRC_DROPPABLE:
2415 if (sock->type != SOCK_STREAM)
2416 tsk_set_unreliable(tsk, value);
2420 case TIPC_DEST_DROPPABLE:
2421 tsk_set_unreturnable(tsk, value);
2423 case TIPC_CONN_TIMEOUT:
2424 tipc_sk(sk)->conn_timeout = value;
2425 /* no need to set "res", since already 0 at this point */
2437 * tipc_getsockopt - get socket option
2438 * @sock: socket structure
2439 * @lvl: option level
2440 * @opt: option identifier
2441 * @ov: receptacle for option value
2442 * @ol: receptacle for length of option value
2444 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2445 * (to ease compatibility).
2447 * Returns 0 on success, errno otherwise
2449 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2450 char __user *ov, int __user *ol)
2452 struct sock *sk = sock->sk;
2453 struct tipc_sock *tsk = tipc_sk(sk);
2458 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2459 return put_user(0, ol);
2460 if (lvl != SOL_TIPC)
2461 return -ENOPROTOOPT;
2462 res = get_user(len, ol);
2469 case TIPC_IMPORTANCE:
2470 value = tsk_importance(tsk);
2472 case TIPC_SRC_DROPPABLE:
2473 value = tsk_unreliable(tsk);
2475 case TIPC_DEST_DROPPABLE:
2476 value = tsk_unreturnable(tsk);
2478 case TIPC_CONN_TIMEOUT:
2479 value = tsk->conn_timeout;
2480 /* no need to set "res", since already 0 at this point */
2482 case TIPC_NODE_RECVQ_DEPTH:
2483 value = 0; /* was tipc_queue_size, now obsolete */
2485 case TIPC_SOCK_RECVQ_DEPTH:
2486 value = skb_queue_len(&sk->sk_receive_queue);
2495 return res; /* "get" failed */
2497 if (len < sizeof(value))
2500 if (copy_to_user(ov, &value, sizeof(value)))
2503 return put_user(sizeof(value), ol);
2506 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2508 struct sock *sk = sock->sk;
2509 struct tipc_sioc_ln_req lnr;
2510 void __user *argp = (void __user *)arg;
2513 case SIOCGETLINKNAME:
2514 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2516 if (!tipc_node_get_linkname(sock_net(sk),
2517 lnr.bearer_id & 0xffff, lnr.peer,
2518 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2519 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2523 return -EADDRNOTAVAIL;
2525 return -ENOIOCTLCMD;
2529 /* Protocol switches for the various types of TIPC sockets */
2531 static const struct proto_ops msg_ops = {
2532 .owner = THIS_MODULE,
2534 .release = tipc_release,
2536 .connect = tipc_connect,
2537 .socketpair = sock_no_socketpair,
2538 .accept = sock_no_accept,
2539 .getname = tipc_getname,
2541 .ioctl = tipc_ioctl,
2542 .listen = sock_no_listen,
2543 .shutdown = tipc_shutdown,
2544 .setsockopt = tipc_setsockopt,
2545 .getsockopt = tipc_getsockopt,
2546 .sendmsg = tipc_sendmsg,
2547 .recvmsg = tipc_recvmsg,
2548 .mmap = sock_no_mmap,
2549 .sendpage = sock_no_sendpage
2552 static const struct proto_ops packet_ops = {
2553 .owner = THIS_MODULE,
2555 .release = tipc_release,
2557 .connect = tipc_connect,
2558 .socketpair = sock_no_socketpair,
2559 .accept = tipc_accept,
2560 .getname = tipc_getname,
2562 .ioctl = tipc_ioctl,
2563 .listen = tipc_listen,
2564 .shutdown = tipc_shutdown,
2565 .setsockopt = tipc_setsockopt,
2566 .getsockopt = tipc_getsockopt,
2567 .sendmsg = tipc_send_packet,
2568 .recvmsg = tipc_recvmsg,
2569 .mmap = sock_no_mmap,
2570 .sendpage = sock_no_sendpage
2573 static const struct proto_ops stream_ops = {
2574 .owner = THIS_MODULE,
2576 .release = tipc_release,
2578 .connect = tipc_connect,
2579 .socketpair = sock_no_socketpair,
2580 .accept = tipc_accept,
2581 .getname = tipc_getname,
2583 .ioctl = tipc_ioctl,
2584 .listen = tipc_listen,
2585 .shutdown = tipc_shutdown,
2586 .setsockopt = tipc_setsockopt,
2587 .getsockopt = tipc_getsockopt,
2588 .sendmsg = tipc_send_stream,
2589 .recvmsg = tipc_recv_stream,
2590 .mmap = sock_no_mmap,
2591 .sendpage = sock_no_sendpage
2594 static const struct net_proto_family tipc_family_ops = {
2595 .owner = THIS_MODULE,
2597 .create = tipc_sk_create
2600 static struct proto tipc_proto = {
2602 .owner = THIS_MODULE,
2603 .obj_size = sizeof(struct tipc_sock),
2604 .sysctl_rmem = sysctl_tipc_rmem
2608 * tipc_socket_init - initialize TIPC socket interface
2610 * Returns 0 on success, errno otherwise
2612 int tipc_socket_init(void)
2616 res = proto_register(&tipc_proto, 1);
2618 pr_err("Failed to register TIPC protocol type\n");
2622 res = sock_register(&tipc_family_ops);
2624 pr_err("Failed to register TIPC socket type\n");
2625 proto_unregister(&tipc_proto);
2633 * tipc_socket_stop - stop TIPC socket interface
2635 void tipc_socket_stop(void)
2637 sock_unregister(tipc_family_ops.family);
2638 proto_unregister(&tipc_proto);
2641 /* Caller should hold socket lock for the passed tipc socket. */
2642 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2646 struct nlattr *nest;
2648 peer_node = tsk_peer_node(tsk);
2649 peer_port = tsk_peer_port(tsk);
2651 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2653 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2655 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2658 if (tsk->conn_type != 0) {
2659 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2661 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2663 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2666 nla_nest_end(skb, nest);
2671 nla_nest_cancel(skb, nest);
2676 /* Caller should hold socket lock for the passed tipc socket. */
2677 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2678 struct tipc_sock *tsk)
2682 struct nlattr *attrs;
2683 struct net *net = sock_net(skb->sk);
2684 struct tipc_net *tn = net_generic(net, tipc_net_id);
2686 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2687 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2691 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2693 goto genlmsg_cancel;
2694 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2695 goto attr_msg_cancel;
2696 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2697 goto attr_msg_cancel;
2699 if (tsk->connected) {
2700 err = __tipc_nl_add_sk_con(skb, tsk);
2702 goto attr_msg_cancel;
2703 } else if (!list_empty(&tsk->publications)) {
2704 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2705 goto attr_msg_cancel;
2707 nla_nest_end(skb, attrs);
2708 genlmsg_end(skb, hdr);
2713 nla_nest_cancel(skb, attrs);
2715 genlmsg_cancel(skb, hdr);
2720 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2723 struct tipc_sock *tsk;
2724 const struct bucket_table *tbl;
2725 struct rhash_head *pos;
2726 struct net *net = sock_net(skb->sk);
2727 struct tipc_net *tn = net_generic(net, tipc_net_id);
2728 u32 tbl_id = cb->args[0];
2729 u32 prev_portid = cb->args[1];
2732 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2733 for (; tbl_id < tbl->size; tbl_id++) {
2734 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2735 spin_lock_bh(&tsk->sk.sk_lock.slock);
2736 if (prev_portid && prev_portid != tsk->portid) {
2737 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2741 err = __tipc_nl_add_sk(skb, cb, tsk);
2743 prev_portid = tsk->portid;
2744 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2748 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2753 cb->args[0] = tbl_id;
2754 cb->args[1] = prev_portid;
2759 /* Caller should hold socket lock for the passed tipc socket. */
2760 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2761 struct netlink_callback *cb,
2762 struct publication *publ)
2765 struct nlattr *attrs;
2767 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2768 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2772 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2774 goto genlmsg_cancel;
2776 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2777 goto attr_msg_cancel;
2778 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2779 goto attr_msg_cancel;
2780 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2781 goto attr_msg_cancel;
2782 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2783 goto attr_msg_cancel;
2785 nla_nest_end(skb, attrs);
2786 genlmsg_end(skb, hdr);
2791 nla_nest_cancel(skb, attrs);
2793 genlmsg_cancel(skb, hdr);
2798 /* Caller should hold socket lock for the passed tipc socket. */
2799 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2800 struct netlink_callback *cb,
2801 struct tipc_sock *tsk, u32 *last_publ)
2804 struct publication *p;
2807 list_for_each_entry(p, &tsk->publications, pport_list) {
2808 if (p->key == *last_publ)
2811 if (p->key != *last_publ) {
2812 /* We never set seq or call nl_dump_check_consistent()
2813 * this means that setting prev_seq here will cause the
2814 * consistence check to fail in the netlink callback
2815 * handler. Resulting in the last NLMSG_DONE message
2816 * having the NLM_F_DUMP_INTR flag set.
2823 p = list_first_entry(&tsk->publications, struct publication,
2827 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2828 err = __tipc_nl_add_sk_publ(skb, cb, p);
2830 *last_publ = p->key;
2839 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2842 u32 tsk_portid = cb->args[0];
2843 u32 last_publ = cb->args[1];
2844 u32 done = cb->args[2];
2845 struct net *net = sock_net(skb->sk);
2846 struct tipc_sock *tsk;
2849 struct nlattr **attrs;
2850 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2852 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2856 if (!attrs[TIPC_NLA_SOCK])
2859 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2860 attrs[TIPC_NLA_SOCK],
2861 tipc_nl_sock_policy);
2865 if (!sock[TIPC_NLA_SOCK_REF])
2868 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2874 tsk = tipc_sk_lookup(net, tsk_portid);
2878 lock_sock(&tsk->sk);
2879 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2882 release_sock(&tsk->sk);
2885 cb->args[0] = tsk_portid;
2886 cb->args[1] = last_publ;