2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2015, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/rhashtable.h>
38 #include <linux/jhash.h>
40 #include "name_table.h"
43 #include "name_distr.h"
46 #define SS_LISTENING -1 /* socket is listening */
47 #define SS_READY -2 /* socket is connectionless */
49 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
50 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
51 #define TIPC_FWD_MSG 1
52 #define TIPC_CONN_OK 0
53 #define TIPC_CONN_PROBING 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
58 * struct tipc_sock - TIPC socket structure
59 * @sk: socket - interacts with 'port' and with user via the socket API
60 * @connected: non-zero if port is currently connected to a peer port
61 * @conn_type: TIPC type used when connection was established
62 * @conn_instance: TIPC instance used when connection was established
63 * @published: non-zero if port has one or more associated names
64 * @max_pkt: maximum packet size "hint" used when building messages sent by port
65 * @portid: unique port identity in TIPC socket hash table
66 * @phdr: preformatted message header used when sending messages
67 * @port_list: adjacent ports in TIPC's global list of ports
68 * @publications: list of publications for port
69 * @pub_count: total # of publications port has made during its lifetime
72 * @conn_timeout: the time we can wait for an unresponded setup request
73 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
74 * @link_cong: non-zero if owner must sleep because of link congestion
75 * @sent_unacked: # messages sent by socket, and not yet acked by peer
76 * @rcv_unacked: # messages read by user, but not yet acked back to peer
77 * @node: hash table node
78 * @rcu: rcu struct for tipc_sock
89 struct list_head sock_list;
90 struct list_head publications;
93 unsigned long probing_intv;
99 struct rhash_head node;
103 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
104 static void tipc_data_ready(struct sock *sk);
105 static void tipc_write_space(struct sock *sk);
106 static int tipc_release(struct socket *sock);
107 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
108 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
109 static void tipc_sk_timeout(unsigned long data);
110 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
111 struct tipc_name_seq const *seq);
112 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
113 struct tipc_name_seq const *seq);
114 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
115 static int tipc_sk_insert(struct tipc_sock *tsk);
116 static void tipc_sk_remove(struct tipc_sock *tsk);
118 static const struct proto_ops packet_ops;
119 static const struct proto_ops stream_ops;
120 static const struct proto_ops msg_ops;
122 static struct proto tipc_proto;
123 static struct proto tipc_proto_kern;
125 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
126 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
127 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
128 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
129 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
130 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
134 * Revised TIPC socket locking policy:
136 * Most socket operations take the standard socket lock when they start
137 * and hold it until they finish (or until they need to sleep). Acquiring
138 * this lock grants the owner exclusive access to the fields of the socket
139 * data structures, with the exception of the backlog queue. A few socket
140 * operations can be done without taking the socket lock because they only
141 * read socket information that never changes during the life of the socket.
143 * Socket operations may acquire the lock for the associated TIPC port if they
144 * need to perform an operation on the port. If any routine needs to acquire
145 * both the socket lock and the port lock it must take the socket lock first
146 * to avoid the risk of deadlock.
148 * The dispatcher handling incoming messages cannot grab the socket lock in
149 * the standard fashion, since invoked it runs at the BH level and cannot block.
150 * Instead, it checks to see if the socket lock is currently owned by someone,
151 * and either handles the message itself or adds it to the socket's backlog
152 * queue; in the latter case the queued message is processed once the process
153 * owning the socket lock releases it.
155 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
156 * the problem of a blocked socket operation preventing any other operations
157 * from occurring. However, applications must be careful if they have
158 * multiple threads trying to send (or receive) on the same socket, as these
159 * operations might interfere with each other. For example, doing a connect
160 * and a receive at the same time might allow the receive to consume the
161 * ACK message meant for the connect. While additional work could be done
162 * to try and overcome this, it doesn't seem to be worthwhile at the present.
164 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
165 * that another operation that must be performed in a non-blocking manner is
166 * not delayed for very long because the lock has already been taken.
168 * NOTE: This code assumes that certain fields of a port/socket pair are
169 * constant over its lifetime; such fields can be examined without taking
170 * the socket lock and/or port lock, and do not need to be re-read even
171 * after resuming processing after waiting. These fields include:
173 * - pointer to socket sk structure (aka tipc_sock structure)
174 * - pointer to port structure
178 static u32 tsk_own_node(struct tipc_sock *tsk)
180 return msg_prevnode(&tsk->phdr);
183 static u32 tsk_peer_node(struct tipc_sock *tsk)
185 return msg_destnode(&tsk->phdr);
188 static u32 tsk_peer_port(struct tipc_sock *tsk)
190 return msg_destport(&tsk->phdr);
193 static bool tsk_unreliable(struct tipc_sock *tsk)
195 return msg_src_droppable(&tsk->phdr) != 0;
198 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
200 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
203 static bool tsk_unreturnable(struct tipc_sock *tsk)
205 return msg_dest_droppable(&tsk->phdr) != 0;
208 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
210 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
213 static int tsk_importance(struct tipc_sock *tsk)
215 return msg_importance(&tsk->phdr);
218 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
220 if (imp > TIPC_CRITICAL_IMPORTANCE)
222 msg_set_importance(&tsk->phdr, (u32)imp);
226 static struct tipc_sock *tipc_sk(const struct sock *sk)
228 return container_of(sk, struct tipc_sock, sk);
231 static int tsk_conn_cong(struct tipc_sock *tsk)
233 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
237 * tsk_advance_rx_queue - discard first buffer in socket receive queue
239 * Caller must hold socket lock
241 static void tsk_advance_rx_queue(struct sock *sk)
243 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
247 * tsk_rej_rx_queue - reject all buffers in socket receive queue
249 * Caller must hold socket lock
251 static void tsk_rej_rx_queue(struct sock *sk)
255 u32 own_node = tsk_own_node(tipc_sk(sk));
257 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
258 if (tipc_msg_reverse(own_node, skb, &dnode, TIPC_ERR_NO_PORT))
259 tipc_link_xmit_skb(sock_net(sk), skb, dnode, 0);
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 tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
271 u32 peer_port = tsk_peer_port(tsk);
275 if (unlikely(!tsk->connected))
278 if (unlikely(msg_origport(msg) != peer_port))
281 orig_node = msg_orignode(msg);
282 peer_node = tsk_peer_node(tsk);
284 if (likely(orig_node == peer_node))
287 if (!orig_node && (peer_node == tn->own_addr))
290 if (!peer_node && (orig_node == tn->own_addr))
297 * tipc_sk_create - create a TIPC socket
298 * @net: network namespace (must be default network)
299 * @sock: pre-allocated socket structure
300 * @protocol: protocol indicator (must be 0)
301 * @kern: caused by kernel or by userspace?
303 * This routine creates additional data structures used by the TIPC socket,
304 * initializes them, and links them together.
306 * Returns 0 on success, errno otherwise
308 static int tipc_sk_create(struct net *net, struct socket *sock,
309 int protocol, int kern)
312 const struct proto_ops *ops;
315 struct tipc_sock *tsk;
316 struct tipc_msg *msg;
318 /* Validate arguments */
319 if (unlikely(protocol != 0))
320 return -EPROTONOSUPPORT;
322 switch (sock->type) {
325 state = SS_UNCONNECTED;
329 state = SS_UNCONNECTED;
340 /* Allocate socket's protocol area */
342 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
344 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
350 tsk->max_pkt = MAX_PKT_DEFAULT;
351 INIT_LIST_HEAD(&tsk->publications);
353 tn = net_generic(sock_net(sk), tipc_net_id);
354 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
357 /* Finish initializing socket data structures */
360 sock_init_data(sock, sk);
361 if (tipc_sk_insert(tsk)) {
362 pr_warn("Socket create failed; port numbrer exhausted\n");
365 msg_set_origport(msg, tsk->portid);
366 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
367 sk->sk_backlog_rcv = tipc_backlog_rcv;
368 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
369 sk->sk_data_ready = tipc_data_ready;
370 sk->sk_write_space = tipc_write_space;
371 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
372 tsk->sent_unacked = 0;
373 atomic_set(&tsk->dupl_rcvcnt, 0);
375 if (sock->state == SS_READY) {
376 tsk_set_unreturnable(tsk, true);
377 if (sock->type == SOCK_DGRAM)
378 tsk_set_unreliable(tsk, true);
384 * tipc_sock_create_local - create TIPC socket from inside TIPC module
385 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
387 * We cannot use sock_creat_kern here because it bumps module user count.
388 * Since socket owner and creator is the same module we must make sure
389 * that module count remains zero for module local sockets, otherwise
390 * we cannot do rmmod.
392 * Returns 0 on success, errno otherwise
394 int tipc_sock_create_local(struct net *net, int type, struct socket **res)
398 rc = sock_create_lite(AF_TIPC, type, 0, res);
400 pr_err("Failed to create kernel socket\n");
403 tipc_sk_create(net, *res, 0, 1);
409 * tipc_sock_release_local - release socket created by tipc_sock_create_local
410 * @sock: the socket to be released.
412 * Module reference count is not incremented when such sockets are created,
413 * so we must keep it from being decremented when they are released.
415 void tipc_sock_release_local(struct socket *sock)
423 * tipc_sock_accept_local - accept a connection on a socket created
424 * with tipc_sock_create_local. Use this function to avoid that
425 * module reference count is inadvertently incremented.
427 * @sock: the accepting socket
428 * @newsock: reference to the new socket to be created
429 * @flags: socket flags
432 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
435 struct sock *sk = sock->sk;
438 ret = sock_create_lite(sk->sk_family, sk->sk_type,
439 sk->sk_protocol, newsock);
443 ret = tipc_accept(sock, *newsock, flags);
445 sock_release(*newsock);
448 (*newsock)->ops = sock->ops;
452 static void tipc_sk_callback(struct rcu_head *head)
454 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
460 * tipc_release - destroy a TIPC socket
461 * @sock: socket to destroy
463 * This routine cleans up any messages that are still queued on the socket.
464 * For DGRAM and RDM socket types, all queued messages are rejected.
465 * For SEQPACKET and STREAM socket types, the first message is rejected
466 * and any others are discarded. (If the first message on a STREAM socket
467 * is partially-read, it is discarded and the next one is rejected instead.)
469 * NOTE: Rejected messages are not necessarily returned to the sender! They
470 * are returned or discarded according to the "destination droppable" setting
471 * specified for the message by the sender.
473 * Returns 0 on success, errno otherwise
475 static int tipc_release(struct socket *sock)
477 struct sock *sk = sock->sk;
479 struct tipc_sock *tsk;
481 u32 dnode, probing_state;
484 * Exit if socket isn't fully initialized (occurs when a failed accept()
485 * releases a pre-allocated child socket that was never used)
495 * Reject all unreceived messages, except on an active connection
496 * (which disconnects locally & sends a 'FIN+' to peer)
498 dnode = tsk_peer_node(tsk);
499 while (sock->state != SS_DISCONNECTING) {
500 skb = __skb_dequeue(&sk->sk_receive_queue);
503 if (TIPC_SKB_CB(skb)->handle != NULL)
506 if ((sock->state == SS_CONNECTING) ||
507 (sock->state == SS_CONNECTED)) {
508 sock->state = SS_DISCONNECTING;
510 tipc_node_remove_conn(net, dnode, tsk->portid);
512 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
514 tipc_link_xmit_skb(net, skb, dnode, 0);
518 tipc_sk_withdraw(tsk, 0, NULL);
519 probing_state = tsk->probing_state;
520 if (del_timer_sync(&sk->sk_timer) &&
521 probing_state != TIPC_CONN_PROBING)
524 if (tsk->connected) {
525 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
526 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
527 tsk_own_node(tsk), tsk_peer_port(tsk),
528 tsk->portid, TIPC_ERR_NO_PORT);
530 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
531 tipc_node_remove_conn(net, dnode, tsk->portid);
534 /* Discard any remaining (connection-based) messages in receive queue */
535 __skb_queue_purge(&sk->sk_receive_queue);
537 /* Reject any messages that accumulated in backlog queue */
538 sock->state = SS_DISCONNECTING;
541 call_rcu(&tsk->rcu, tipc_sk_callback);
548 * tipc_bind - associate or disassocate TIPC name(s) with a socket
549 * @sock: socket structure
550 * @uaddr: socket address describing name(s) and desired operation
551 * @uaddr_len: size of socket address data structure
553 * Name and name sequence binding is indicated using a positive scope value;
554 * a negative scope value unbinds the specified name. Specifying no name
555 * (i.e. a socket address length of 0) unbinds all names from the socket.
557 * Returns 0 on success, errno otherwise
559 * NOTE: This routine doesn't need to take the socket lock since it doesn't
560 * access any non-constant socket information.
562 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
565 struct sock *sk = sock->sk;
566 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
567 struct tipc_sock *tsk = tipc_sk(sk);
571 if (unlikely(!uaddr_len)) {
572 res = tipc_sk_withdraw(tsk, 0, NULL);
576 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
580 if (addr->family != AF_TIPC) {
585 if (addr->addrtype == TIPC_ADDR_NAME)
586 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
587 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
592 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
593 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
594 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
599 res = (addr->scope > 0) ?
600 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
601 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
608 * tipc_getname - get port ID of socket or peer socket
609 * @sock: socket structure
610 * @uaddr: area for returned socket address
611 * @uaddr_len: area for returned length of socket address
612 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
614 * Returns 0 on success, errno otherwise
616 * NOTE: This routine doesn't need to take the socket lock since it only
617 * accesses socket information that is unchanging (or which changes in
618 * a completely predictable manner).
620 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
621 int *uaddr_len, int peer)
623 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
624 struct tipc_sock *tsk = tipc_sk(sock->sk);
625 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
627 memset(addr, 0, sizeof(*addr));
629 if ((sock->state != SS_CONNECTED) &&
630 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
632 addr->addr.id.ref = tsk_peer_port(tsk);
633 addr->addr.id.node = tsk_peer_node(tsk);
635 addr->addr.id.ref = tsk->portid;
636 addr->addr.id.node = tn->own_addr;
639 *uaddr_len = sizeof(*addr);
640 addr->addrtype = TIPC_ADDR_ID;
641 addr->family = AF_TIPC;
643 addr->addr.name.domain = 0;
649 * tipc_poll - read and possibly block on pollmask
650 * @file: file structure associated with the socket
651 * @sock: socket for which to calculate the poll bits
654 * Returns pollmask value
657 * It appears that the usual socket locking mechanisms are not useful here
658 * since the pollmask info is potentially out-of-date the moment this routine
659 * exits. TCP and other protocols seem to rely on higher level poll routines
660 * to handle any preventable race conditions, so TIPC will do the same ...
662 * TIPC sets the returned events as follows:
664 * socket state flags set
665 * ------------ ---------
666 * unconnected no read flags
667 * POLLOUT if port is not congested
669 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
672 * connected POLLIN/POLLRDNORM if data in rx queue
673 * POLLOUT if port is not congested
675 * disconnecting POLLIN/POLLRDNORM/POLLHUP
678 * listening POLLIN if SYN in rx queue
681 * ready POLLIN/POLLRDNORM if data in rx queue
682 * [connectionless] POLLOUT (since port cannot be congested)
684 * IMPORTANT: The fact that a read or write operation is indicated does NOT
685 * imply that the operation will succeed, merely that it should be performed
686 * and will not block.
688 static unsigned int tipc_poll(struct file *file, struct socket *sock,
691 struct sock *sk = sock->sk;
692 struct tipc_sock *tsk = tipc_sk(sk);
695 sock_poll_wait(file, sk_sleep(sk), wait);
697 switch ((int)sock->state) {
704 if (!tsk->link_cong && !tsk_conn_cong(tsk))
709 if (!skb_queue_empty(&sk->sk_receive_queue))
710 mask |= (POLLIN | POLLRDNORM);
712 case SS_DISCONNECTING:
713 mask = (POLLIN | POLLRDNORM | POLLHUP);
721 * tipc_sendmcast - send multicast message
722 * @sock: socket structure
723 * @seq: destination address
724 * @msg: message to send
725 * @dsz: total length of message data
726 * @timeo: timeout to wait for wakeup
728 * Called from function tipc_sendmsg(), which has done all sanity checks
729 * Returns the number of bytes sent on success, or errno
731 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
732 struct msghdr *msg, size_t dsz, long timeo)
734 struct sock *sk = sock->sk;
735 struct tipc_sock *tsk = tipc_sk(sk);
736 struct net *net = sock_net(sk);
737 struct tipc_msg *mhdr = &tsk->phdr;
738 struct sk_buff_head *pktchain = &sk->sk_write_queue;
739 struct iov_iter save = msg->msg_iter;
743 msg_set_type(mhdr, TIPC_MCAST_MSG);
744 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
745 msg_set_destport(mhdr, 0);
746 msg_set_destnode(mhdr, 0);
747 msg_set_nametype(mhdr, seq->type);
748 msg_set_namelower(mhdr, seq->lower);
749 msg_set_nameupper(mhdr, seq->upper);
750 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
753 mtu = tipc_bclink_get_mtu();
754 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, pktchain);
755 if (unlikely(rc < 0))
759 rc = tipc_bclink_xmit(net, pktchain);
760 if (likely(rc >= 0)) {
764 if (rc == -EMSGSIZE) {
765 msg->msg_iter = save;
768 if (rc != -ELINKCONG)
770 tipc_sk(sk)->link_cong = 1;
771 rc = tipc_wait_for_sndmsg(sock, &timeo);
773 __skb_queue_purge(pktchain);
779 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
780 * @arrvq: queue with arriving messages, to be cloned after destination lookup
781 * @inputq: queue with cloned messages, delivered to socket after dest lookup
783 * Multi-threaded: parallel calls with reference to same queues may occur
785 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
786 struct sk_buff_head *inputq)
788 struct tipc_msg *msg;
789 struct tipc_plist dports;
791 u32 scope = TIPC_CLUSTER_SCOPE;
792 struct sk_buff_head tmpq;
794 struct sk_buff *skb, *_skb;
796 __skb_queue_head_init(&tmpq);
797 tipc_plist_init(&dports);
799 skb = tipc_skb_peek(arrvq, &inputq->lock);
800 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
802 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
804 if (in_own_node(net, msg_orignode(msg)))
805 scope = TIPC_NODE_SCOPE;
807 /* Create destination port list and message clones: */
808 tipc_nametbl_mc_translate(net,
809 msg_nametype(msg), msg_namelower(msg),
810 msg_nameupper(msg), scope, &dports);
811 portid = tipc_plist_pop(&dports);
812 for (; portid; portid = tipc_plist_pop(&dports)) {
813 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
815 msg_set_destport(buf_msg(_skb), portid);
816 __skb_queue_tail(&tmpq, _skb);
819 pr_warn("Failed to clone mcast rcv buffer\n");
821 /* Append to inputq if not already done by other thread */
822 spin_lock_bh(&inputq->lock);
823 if (skb_peek(arrvq) == skb) {
824 skb_queue_splice_tail_init(&tmpq, inputq);
825 kfree_skb(__skb_dequeue(arrvq));
827 spin_unlock_bh(&inputq->lock);
828 __skb_queue_purge(&tmpq);
831 tipc_sk_rcv(net, inputq);
835 * tipc_sk_proto_rcv - receive a connection mng protocol message
836 * @tsk: receiving socket
837 * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
839 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff **skb)
841 struct tipc_msg *msg = buf_msg(*skb);
844 u32 own_node = tsk_own_node(tsk);
845 /* Ignore if connection cannot be validated: */
846 if (!tsk_peer_msg(tsk, msg))
849 tsk->probing_state = TIPC_CONN_OK;
851 if (msg_type(msg) == CONN_ACK) {
852 conn_cong = tsk_conn_cong(tsk);
853 tsk->sent_unacked -= msg_msgcnt(msg);
855 tsk->sk.sk_write_space(&tsk->sk);
856 } else if (msg_type(msg) == CONN_PROBE) {
857 if (tipc_msg_reverse(own_node, *skb, &dnode, TIPC_OK)) {
858 msg_set_type(msg, CONN_PROBE_REPLY);
862 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
868 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
870 struct sock *sk = sock->sk;
871 struct tipc_sock *tsk = tipc_sk(sk);
876 int err = sock_error(sk);
879 if (sock->state == SS_DISCONNECTING)
883 if (signal_pending(current))
884 return sock_intr_errno(*timeo_p);
886 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
887 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
888 finish_wait(sk_sleep(sk), &wait);
894 * tipc_sendmsg - send message in connectionless manner
895 * @iocb: if NULL, indicates that socket lock is already held
896 * @sock: socket structure
897 * @m: message to send
898 * @dsz: amount of user data to be sent
900 * Message must have an destination specified explicitly.
901 * Used for SOCK_RDM and SOCK_DGRAM messages,
902 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
903 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
905 * Returns the number of bytes sent on success, or errno otherwise
907 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
908 struct msghdr *m, size_t dsz)
910 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
911 struct sock *sk = sock->sk;
912 struct tipc_sock *tsk = tipc_sk(sk);
913 struct net *net = sock_net(sk);
914 struct tipc_msg *mhdr = &tsk->phdr;
916 struct sk_buff_head *pktchain = &sk->sk_write_queue;
918 struct tipc_name_seq *seq = &dest->addr.nameseq;
919 struct iov_iter save;
925 return -EDESTADDRREQ;
927 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
928 (dest->family != AF_TIPC)))
931 if (dsz > TIPC_MAX_USER_MSG_SIZE)
937 if (unlikely(sock->state != SS_READY)) {
938 if (sock->state == SS_LISTENING) {
942 if (sock->state != SS_UNCONNECTED) {
946 if (tsk->published) {
950 if (dest->addrtype == TIPC_ADDR_NAME) {
951 tsk->conn_type = dest->addr.name.name.type;
952 tsk->conn_instance = dest->addr.name.name.instance;
956 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
958 if (dest->addrtype == TIPC_ADDR_MCAST) {
959 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
961 } else if (dest->addrtype == TIPC_ADDR_NAME) {
962 u32 type = dest->addr.name.name.type;
963 u32 inst = dest->addr.name.name.instance;
964 u32 domain = dest->addr.name.domain;
967 msg_set_type(mhdr, TIPC_NAMED_MSG);
968 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
969 msg_set_nametype(mhdr, type);
970 msg_set_nameinst(mhdr, inst);
971 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
972 dport = tipc_nametbl_translate(net, type, inst, &dnode);
973 msg_set_destnode(mhdr, dnode);
974 msg_set_destport(mhdr, dport);
975 if (unlikely(!dport && !dnode)) {
979 } else if (dest->addrtype == TIPC_ADDR_ID) {
980 dnode = dest->addr.id.node;
981 msg_set_type(mhdr, TIPC_DIRECT_MSG);
982 msg_set_lookup_scope(mhdr, 0);
983 msg_set_destnode(mhdr, dnode);
984 msg_set_destport(mhdr, dest->addr.id.ref);
985 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
990 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
991 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, pktchain);
996 skb = skb_peek(pktchain);
997 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
998 rc = tipc_link_xmit(net, pktchain, dnode, tsk->portid);
999 if (likely(rc >= 0)) {
1000 if (sock->state != SS_READY)
1001 sock->state = SS_CONNECTING;
1005 if (rc == -EMSGSIZE) {
1009 if (rc != -ELINKCONG)
1012 rc = tipc_wait_for_sndmsg(sock, &timeo);
1014 __skb_queue_purge(pktchain);
1023 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
1025 struct sock *sk = sock->sk;
1026 struct tipc_sock *tsk = tipc_sk(sk);
1031 int err = sock_error(sk);
1034 if (sock->state == SS_DISCONNECTING)
1036 else if (sock->state != SS_CONNECTED)
1040 if (signal_pending(current))
1041 return sock_intr_errno(*timeo_p);
1043 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1044 done = sk_wait_event(sk, timeo_p,
1046 !tsk_conn_cong(tsk)) ||
1048 finish_wait(sk_sleep(sk), &wait);
1054 * tipc_send_stream - send stream-oriented data
1056 * @sock: socket structure
1058 * @dsz: total length of data to be transmitted
1060 * Used for SOCK_STREAM data.
1062 * Returns the number of bytes sent on success (or partial success),
1063 * or errno if no data sent
1065 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1066 struct msghdr *m, size_t dsz)
1068 struct sock *sk = sock->sk;
1069 struct net *net = sock_net(sk);
1070 struct tipc_sock *tsk = tipc_sk(sk);
1071 struct tipc_msg *mhdr = &tsk->phdr;
1072 struct sk_buff_head *pktchain = &sk->sk_write_queue;
1073 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1074 u32 portid = tsk->portid;
1078 uint mtu, send, sent = 0;
1079 struct iov_iter save;
1081 /* Handle implied connection establishment */
1082 if (unlikely(dest)) {
1083 rc = tipc_sendmsg(iocb, sock, m, dsz);
1084 if (dsz && (dsz == rc))
1085 tsk->sent_unacked = 1;
1088 if (dsz > (uint)INT_MAX)
1094 if (unlikely(sock->state != SS_CONNECTED)) {
1095 if (sock->state == SS_DISCONNECTING)
1102 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1103 dnode = tsk_peer_node(tsk);
1108 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1109 rc = tipc_msg_build(mhdr, m, sent, send, mtu, pktchain);
1110 if (unlikely(rc < 0))
1113 if (likely(!tsk_conn_cong(tsk))) {
1114 rc = tipc_link_xmit(net, pktchain, dnode, portid);
1116 tsk->sent_unacked++;
1122 if (rc == -EMSGSIZE) {
1123 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1128 if (rc != -ELINKCONG)
1132 rc = tipc_wait_for_sndpkt(sock, &timeo);
1134 __skb_queue_purge(pktchain);
1139 return sent ? sent : rc;
1143 * tipc_send_packet - send a connection-oriented message
1144 * @iocb: if NULL, indicates that socket lock is already held
1145 * @sock: socket structure
1146 * @m: message to send
1147 * @dsz: length of data to be transmitted
1149 * Used for SOCK_SEQPACKET messages.
1151 * Returns the number of bytes sent on success, or errno otherwise
1153 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1154 struct msghdr *m, size_t dsz)
1156 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1159 return tipc_send_stream(iocb, sock, m, dsz);
1162 /* tipc_sk_finish_conn - complete the setup of a connection
1164 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1167 struct sock *sk = &tsk->sk;
1168 struct net *net = sock_net(sk);
1169 struct tipc_msg *msg = &tsk->phdr;
1171 msg_set_destnode(msg, peer_node);
1172 msg_set_destport(msg, peer_port);
1173 msg_set_type(msg, TIPC_CONN_MSG);
1174 msg_set_lookup_scope(msg, 0);
1175 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1177 tsk->probing_intv = CONN_PROBING_INTERVAL;
1178 tsk->probing_state = TIPC_CONN_OK;
1180 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1181 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1182 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1186 * set_orig_addr - capture sender's address for received message
1187 * @m: descriptor for message info
1188 * @msg: received message header
1190 * Note: Address is not captured if not requested by receiver.
1192 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1194 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1197 addr->family = AF_TIPC;
1198 addr->addrtype = TIPC_ADDR_ID;
1199 memset(&addr->addr, 0, sizeof(addr->addr));
1200 addr->addr.id.ref = msg_origport(msg);
1201 addr->addr.id.node = msg_orignode(msg);
1202 addr->addr.name.domain = 0; /* could leave uninitialized */
1203 addr->scope = 0; /* could leave uninitialized */
1204 m->msg_namelen = sizeof(struct sockaddr_tipc);
1209 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1210 * @m: descriptor for message info
1211 * @msg: received message header
1212 * @tsk: TIPC port associated with message
1214 * Note: Ancillary data is not captured if not requested by receiver.
1216 * Returns 0 if successful, otherwise errno
1218 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1219 struct tipc_sock *tsk)
1227 if (likely(m->msg_controllen == 0))
1230 /* Optionally capture errored message object(s) */
1231 err = msg ? msg_errcode(msg) : 0;
1232 if (unlikely(err)) {
1234 anc_data[1] = msg_data_sz(msg);
1235 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1239 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1246 /* Optionally capture message destination object */
1247 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1248 switch (dest_type) {
1249 case TIPC_NAMED_MSG:
1251 anc_data[0] = msg_nametype(msg);
1252 anc_data[1] = msg_namelower(msg);
1253 anc_data[2] = msg_namelower(msg);
1255 case TIPC_MCAST_MSG:
1257 anc_data[0] = msg_nametype(msg);
1258 anc_data[1] = msg_namelower(msg);
1259 anc_data[2] = msg_nameupper(msg);
1262 has_name = (tsk->conn_type != 0);
1263 anc_data[0] = tsk->conn_type;
1264 anc_data[1] = tsk->conn_instance;
1265 anc_data[2] = tsk->conn_instance;
1271 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1279 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1281 struct net *net = sock_net(&tsk->sk);
1282 struct sk_buff *skb = NULL;
1283 struct tipc_msg *msg;
1284 u32 peer_port = tsk_peer_port(tsk);
1285 u32 dnode = tsk_peer_node(tsk);
1287 if (!tsk->connected)
1289 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1290 dnode, tsk_own_node(tsk), peer_port,
1291 tsk->portid, TIPC_OK);
1295 msg_set_msgcnt(msg, ack);
1296 tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1299 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1301 struct sock *sk = sock->sk;
1303 long timeo = *timeop;
1307 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1308 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1309 if (sock->state == SS_DISCONNECTING) {
1314 timeo = schedule_timeout(timeo);
1318 if (!skb_queue_empty(&sk->sk_receive_queue))
1320 err = sock_intr_errno(timeo);
1321 if (signal_pending(current))
1327 finish_wait(sk_sleep(sk), &wait);
1333 * tipc_recvmsg - receive packet-oriented message
1335 * @m: descriptor for message info
1336 * @buf_len: total size of user buffer area
1337 * @flags: receive flags
1339 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1340 * If the complete message doesn't fit in user area, truncate it.
1342 * Returns size of returned message data, errno otherwise
1344 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1345 struct msghdr *m, size_t buf_len, int flags)
1347 struct sock *sk = sock->sk;
1348 struct tipc_sock *tsk = tipc_sk(sk);
1349 struct sk_buff *buf;
1350 struct tipc_msg *msg;
1356 /* Catch invalid receive requests */
1357 if (unlikely(!buf_len))
1362 if (unlikely(sock->state == SS_UNCONNECTED)) {
1367 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1370 /* Look for a message in receive queue; wait if necessary */
1371 res = tipc_wait_for_rcvmsg(sock, &timeo);
1375 /* Look at first message in receive queue */
1376 buf = skb_peek(&sk->sk_receive_queue);
1378 sz = msg_data_sz(msg);
1379 err = msg_errcode(msg);
1381 /* Discard an empty non-errored message & try again */
1382 if ((!sz) && (!err)) {
1383 tsk_advance_rx_queue(sk);
1387 /* Capture sender's address (optional) */
1388 set_orig_addr(m, msg);
1390 /* Capture ancillary data (optional) */
1391 res = tipc_sk_anc_data_recv(m, msg, tsk);
1395 /* Capture message data (if valid) & compute return value (always) */
1397 if (unlikely(buf_len < sz)) {
1399 m->msg_flags |= MSG_TRUNC;
1401 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1406 if ((sock->state == SS_READY) ||
1407 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1413 /* Consume received message (optional) */
1414 if (likely(!(flags & MSG_PEEK))) {
1415 if ((sock->state != SS_READY) &&
1416 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1417 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1418 tsk->rcv_unacked = 0;
1420 tsk_advance_rx_queue(sk);
1428 * tipc_recv_stream - receive stream-oriented data
1430 * @m: descriptor for message info
1431 * @buf_len: total size of user buffer area
1432 * @flags: receive flags
1434 * Used for SOCK_STREAM messages only. If not enough data is available
1435 * will optionally wait for more; never truncates data.
1437 * Returns size of returned message data, errno otherwise
1439 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1440 struct msghdr *m, size_t buf_len, int flags)
1442 struct sock *sk = sock->sk;
1443 struct tipc_sock *tsk = tipc_sk(sk);
1444 struct sk_buff *buf;
1445 struct tipc_msg *msg;
1448 int sz_to_copy, target, needed;
1453 /* Catch invalid receive attempts */
1454 if (unlikely(!buf_len))
1459 if (unlikely(sock->state == SS_UNCONNECTED)) {
1464 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1465 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1468 /* Look for a message in receive queue; wait if necessary */
1469 res = tipc_wait_for_rcvmsg(sock, &timeo);
1473 /* Look at first message in receive queue */
1474 buf = skb_peek(&sk->sk_receive_queue);
1476 sz = msg_data_sz(msg);
1477 err = msg_errcode(msg);
1479 /* Discard an empty non-errored message & try again */
1480 if ((!sz) && (!err)) {
1481 tsk_advance_rx_queue(sk);
1485 /* Optionally capture sender's address & ancillary data of first msg */
1486 if (sz_copied == 0) {
1487 set_orig_addr(m, msg);
1488 res = tipc_sk_anc_data_recv(m, msg, tsk);
1493 /* Capture message data (if valid) & compute return value (always) */
1495 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1498 needed = (buf_len - sz_copied);
1499 sz_to_copy = (sz <= needed) ? sz : needed;
1501 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1506 sz_copied += sz_to_copy;
1508 if (sz_to_copy < sz) {
1509 if (!(flags & MSG_PEEK))
1510 TIPC_SKB_CB(buf)->handle =
1511 (void *)(unsigned long)(offset + sz_to_copy);
1516 goto exit; /* can't add error msg to valid data */
1518 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1524 /* Consume received message (optional) */
1525 if (likely(!(flags & MSG_PEEK))) {
1526 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1527 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1528 tsk->rcv_unacked = 0;
1530 tsk_advance_rx_queue(sk);
1533 /* Loop around if more data is required */
1534 if ((sz_copied < buf_len) && /* didn't get all requested data */
1535 (!skb_queue_empty(&sk->sk_receive_queue) ||
1536 (sz_copied < target)) && /* and more is ready or required */
1537 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1538 (!err)) /* and haven't reached a FIN */
1543 return sz_copied ? sz_copied : res;
1547 * tipc_write_space - wake up thread if port congestion is released
1550 static void tipc_write_space(struct sock *sk)
1552 struct socket_wq *wq;
1555 wq = rcu_dereference(sk->sk_wq);
1556 if (wq_has_sleeper(wq))
1557 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1558 POLLWRNORM | POLLWRBAND);
1563 * tipc_data_ready - wake up threads to indicate messages have been received
1565 * @len: the length of messages
1567 static void tipc_data_ready(struct sock *sk)
1569 struct socket_wq *wq;
1572 wq = rcu_dereference(sk->sk_wq);
1573 if (wq_has_sleeper(wq))
1574 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1575 POLLRDNORM | POLLRDBAND);
1580 * filter_connect - Handle all incoming messages for a connection-based socket
1582 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1584 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1586 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **skb)
1588 struct sock *sk = &tsk->sk;
1589 struct net *net = sock_net(sk);
1590 struct socket *sock = sk->sk_socket;
1591 struct tipc_msg *msg = buf_msg(*skb);
1592 int retval = -TIPC_ERR_NO_PORT;
1597 switch ((int)sock->state) {
1599 /* Accept only connection-based messages sent by peer */
1600 if (tsk_peer_msg(tsk, msg)) {
1601 if (unlikely(msg_errcode(msg))) {
1602 sock->state = SS_DISCONNECTING;
1604 /* let timer expire on it's own */
1605 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1612 /* Accept only ACK or NACK message */
1614 if (unlikely(!msg_connected(msg)))
1617 if (unlikely(msg_errcode(msg))) {
1618 sock->state = SS_DISCONNECTING;
1619 sk->sk_err = ECONNREFUSED;
1624 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1625 sock->state = SS_DISCONNECTING;
1626 sk->sk_err = EINVAL;
1631 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1632 msg_set_importance(&tsk->phdr, msg_importance(msg));
1633 sock->state = SS_CONNECTED;
1635 /* If an incoming message is an 'ACK-', it should be
1636 * discarded here because it doesn't contain useful
1637 * data. In addition, we should try to wake up
1638 * connect() routine if sleeping.
1640 if (msg_data_sz(msg) == 0) {
1643 if (waitqueue_active(sk_sleep(sk)))
1644 wake_up_interruptible(sk_sleep(sk));
1649 case SS_UNCONNECTED:
1650 /* Accept only SYN message */
1651 if (!msg_connected(msg) && !(msg_errcode(msg)))
1654 case SS_DISCONNECTING:
1657 pr_err("Unknown socket state %u\n", sock->state);
1663 * rcvbuf_limit - get proper overload limit of socket receive queue
1667 * For all connection oriented messages, irrespective of importance,
1668 * the default overload value (i.e. 67MB) is set as limit.
1670 * For all connectionless messages, by default new queue limits are
1673 * TIPC_LOW_IMPORTANCE (4 MB)
1674 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1675 * TIPC_HIGH_IMPORTANCE (16 MB)
1676 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1678 * Returns overload limit according to corresponding message importance
1680 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1682 struct tipc_msg *msg = buf_msg(buf);
1684 if (msg_connected(msg))
1685 return sysctl_tipc_rmem[2];
1687 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1688 msg_importance(msg);
1692 * filter_rcv - validate incoming message
1694 * @skb: pointer to message. Set to NULL if buffer is consumed.
1696 * Enqueues message on receive queue if acceptable; optionally handles
1697 * disconnect indication for a connected socket.
1699 * Called with socket lock already taken
1701 * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1703 static int filter_rcv(struct sock *sk, struct sk_buff **skb)
1705 struct socket *sock = sk->sk_socket;
1706 struct tipc_sock *tsk = tipc_sk(sk);
1707 struct tipc_msg *msg = buf_msg(*skb);
1708 unsigned int limit = rcvbuf_limit(sk, *skb);
1711 if (unlikely(msg_user(msg) == CONN_MANAGER)) {
1712 tipc_sk_proto_rcv(tsk, skb);
1716 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1719 sk->sk_write_space(sk);
1724 /* Reject message if it is wrong sort of message for socket */
1725 if (msg_type(msg) > TIPC_DIRECT_MSG)
1726 return -TIPC_ERR_NO_PORT;
1728 if (sock->state == SS_READY) {
1729 if (msg_connected(msg))
1730 return -TIPC_ERR_NO_PORT;
1732 rc = filter_connect(tsk, skb);
1733 if (rc != TIPC_OK || !*skb)
1737 /* Reject message if there isn't room to queue it */
1738 if (sk_rmem_alloc_get(sk) + (*skb)->truesize >= limit)
1739 return -TIPC_ERR_OVERLOAD;
1741 /* Enqueue message */
1742 TIPC_SKB_CB(*skb)->handle = NULL;
1743 __skb_queue_tail(&sk->sk_receive_queue, *skb);
1744 skb_set_owner_r(*skb, sk);
1746 sk->sk_data_ready(sk);
1752 * tipc_backlog_rcv - handle incoming message from backlog queue
1756 * Caller must hold socket lock
1760 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1765 struct tipc_sock *tsk = tipc_sk(sk);
1766 struct net *net = sock_net(sk);
1767 uint truesize = skb->truesize;
1769 err = filter_rcv(sk, &skb);
1771 dcnt = &tsk->dupl_rcvcnt;
1772 if (atomic_read(dcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1773 atomic_add(truesize, dcnt);
1776 if (!err || tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode, -err))
1777 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
1782 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1783 * inputq and try adding them to socket or backlog queue
1784 * @inputq: list of incoming buffers with potentially different destinations
1785 * @sk: socket where the buffers should be enqueued
1786 * @dport: port number for the socket
1787 * @_skb: returned buffer to be forwarded or rejected, if applicable
1789 * Caller must hold socket lock
1791 * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1792 * or -TIPC_ERR_NO_PORT
1794 static int tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1795 u32 dport, struct sk_buff **_skb)
1800 struct sk_buff *skb;
1801 unsigned long time_limit = jiffies + 2;
1803 while (skb_queue_len(inputq)) {
1804 if (unlikely(time_after_eq(jiffies, time_limit)))
1806 skb = tipc_skb_dequeue(inputq, dport);
1809 if (!sock_owned_by_user(sk)) {
1810 err = filter_rcv(sk, &skb);
1816 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1817 if (sk->sk_backlog.len)
1818 atomic_set(dcnt, 0);
1819 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1820 if (likely(!sk_add_backlog(sk, skb, lim)))
1823 return -TIPC_ERR_OVERLOAD;
1829 * tipc_sk_rcv - handle a chain of incoming buffers
1830 * @inputq: buffer list containing the buffers
1831 * Consumes all buffers in list until inputq is empty
1832 * Note: may be called in multiple threads referring to the same queue
1833 * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1834 * Only node local calls check the return value, sending single-buffer queues
1836 int tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1838 u32 dnode, dport = 0;
1839 int err = -TIPC_ERR_NO_PORT;
1840 struct sk_buff *skb;
1841 struct tipc_sock *tsk;
1842 struct tipc_net *tn;
1845 while (skb_queue_len(inputq)) {
1847 dport = tipc_skb_peek_port(inputq, dport);
1848 tsk = tipc_sk_lookup(net, dport);
1851 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1852 err = tipc_sk_enqueue(inputq, sk, dport, &skb);
1853 spin_unlock_bh(&sk->sk_lock.slock);
1858 skb = tipc_skb_dequeue(inputq, dport);
1862 if (tipc_msg_lookup_dest(net, skb, &dnode, &err))
1865 dnode = msg_destnode(buf_msg(skb));
1868 tn = net_generic(net, tipc_net_id);
1869 if (!tipc_msg_reverse(tn->own_addr, skb, &dnode, -err))
1872 tipc_link_xmit_skb(net, skb, dnode, dport);
1874 return err ? -EHOSTUNREACH : 0;
1877 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1879 struct sock *sk = sock->sk;
1884 int err = sock_error(sk);
1889 if (signal_pending(current))
1890 return sock_intr_errno(*timeo_p);
1892 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1893 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1894 finish_wait(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 sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1913 struct msghdr m = {NULL,};
1914 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1915 socket_state previous;
1920 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1921 if (sock->state == SS_READY) {
1927 * Reject connection attempt using multicast address
1929 * Note: send_msg() validates the rest of the address fields,
1930 * so there's no need to do it here
1932 if (dst->addrtype == TIPC_ADDR_MCAST) {
1937 previous = sock->state;
1938 switch (sock->state) {
1939 case SS_UNCONNECTED:
1940 /* Send a 'SYN-' to destination */
1942 m.msg_namelen = destlen;
1944 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1945 * indicate send_msg() is never blocked.
1948 m.msg_flags = MSG_DONTWAIT;
1950 res = tipc_sendmsg(NULL, sock, &m, 0);
1951 if ((res < 0) && (res != -EWOULDBLOCK))
1954 /* Just entered SS_CONNECTING state; the only
1955 * difference is that return value in non-blocking
1956 * case is EINPROGRESS, rather than EALREADY.
1960 if (previous == SS_CONNECTING)
1964 timeout = msecs_to_jiffies(timeout);
1965 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1966 res = tipc_wait_for_connect(sock, &timeout);
1981 * tipc_listen - allow socket to listen for incoming connections
1982 * @sock: socket structure
1985 * Returns 0 on success, errno otherwise
1987 static int tipc_listen(struct socket *sock, int len)
1989 struct sock *sk = sock->sk;
1994 if (sock->state != SS_UNCONNECTED)
1997 sock->state = SS_LISTENING;
2005 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2007 struct sock *sk = sock->sk;
2011 /* True wake-one mechanism for incoming connections: only
2012 * one process gets woken up, not the 'whole herd'.
2013 * Since we do not 'race & poll' for established sockets
2014 * anymore, the common case will execute the loop only once.
2017 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2018 TASK_INTERRUPTIBLE);
2019 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2021 timeo = schedule_timeout(timeo);
2025 if (!skb_queue_empty(&sk->sk_receive_queue))
2028 if (sock->state != SS_LISTENING)
2030 err = sock_intr_errno(timeo);
2031 if (signal_pending(current))
2037 finish_wait(sk_sleep(sk), &wait);
2042 * tipc_accept - wait for connection request
2043 * @sock: listening socket
2044 * @newsock: new socket that is to be connected
2045 * @flags: file-related flags associated with socket
2047 * Returns 0 on success, errno otherwise
2049 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2051 struct sock *new_sk, *sk = sock->sk;
2052 struct sk_buff *buf;
2053 struct tipc_sock *new_tsock;
2054 struct tipc_msg *msg;
2060 if (sock->state != SS_LISTENING) {
2064 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2065 res = tipc_wait_for_accept(sock, timeo);
2069 buf = skb_peek(&sk->sk_receive_queue);
2071 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2075 new_sk = new_sock->sk;
2076 new_tsock = tipc_sk(new_sk);
2079 /* we lock on new_sk; but lockdep sees the lock on sk */
2080 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2083 * Reject any stray messages received by new socket
2084 * before the socket lock was taken (very, very unlikely)
2086 tsk_rej_rx_queue(new_sk);
2088 /* Connect new socket to it's peer */
2089 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2090 new_sock->state = SS_CONNECTED;
2092 tsk_set_importance(new_tsock, msg_importance(msg));
2093 if (msg_named(msg)) {
2094 new_tsock->conn_type = msg_nametype(msg);
2095 new_tsock->conn_instance = msg_nameinst(msg);
2099 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2100 * Respond to 'SYN+' by queuing it on new socket.
2102 if (!msg_data_sz(msg)) {
2103 struct msghdr m = {NULL,};
2105 tsk_advance_rx_queue(sk);
2106 tipc_send_packet(NULL, new_sock, &m, 0);
2108 __skb_dequeue(&sk->sk_receive_queue);
2109 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2110 skb_set_owner_r(buf, new_sk);
2112 release_sock(new_sk);
2119 * tipc_shutdown - shutdown socket connection
2120 * @sock: socket structure
2121 * @how: direction to close (must be SHUT_RDWR)
2123 * Terminates connection (if necessary), then purges socket's receive queue.
2125 * Returns 0 on success, errno otherwise
2127 static int tipc_shutdown(struct socket *sock, int how)
2129 struct sock *sk = sock->sk;
2130 struct net *net = sock_net(sk);
2131 struct tipc_sock *tsk = tipc_sk(sk);
2132 struct sk_buff *skb;
2136 if (how != SHUT_RDWR)
2141 switch (sock->state) {
2146 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2147 skb = __skb_dequeue(&sk->sk_receive_queue);
2149 if (TIPC_SKB_CB(skb)->handle != NULL) {
2153 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
2154 TIPC_CONN_SHUTDOWN))
2155 tipc_link_xmit_skb(net, skb, dnode,
2157 tipc_node_remove_conn(net, dnode, tsk->portid);
2159 dnode = tsk_peer_node(tsk);
2161 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2162 TIPC_CONN_MSG, SHORT_H_SIZE,
2163 0, dnode, tsk_own_node(tsk),
2165 tsk->portid, TIPC_CONN_SHUTDOWN);
2166 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2169 sock->state = SS_DISCONNECTING;
2170 tipc_node_remove_conn(net, dnode, tsk->portid);
2173 case SS_DISCONNECTING:
2175 /* Discard any unreceived messages */
2176 __skb_queue_purge(&sk->sk_receive_queue);
2178 /* Wake up anyone sleeping in poll */
2179 sk->sk_state_change(sk);
2191 static void tipc_sk_timeout(unsigned long data)
2193 struct tipc_sock *tsk = (struct tipc_sock *)data;
2194 struct sock *sk = &tsk->sk;
2195 struct sk_buff *skb = NULL;
2196 u32 peer_port, peer_node;
2197 u32 own_node = tsk_own_node(tsk);
2200 if (!tsk->connected) {
2204 peer_port = tsk_peer_port(tsk);
2205 peer_node = tsk_peer_node(tsk);
2207 if (tsk->probing_state == TIPC_CONN_PROBING) {
2208 /* Previous probe not answered -> self abort */
2209 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2210 TIPC_CONN_MSG, SHORT_H_SIZE, 0,
2211 own_node, peer_node, tsk->portid,
2212 peer_port, TIPC_ERR_NO_PORT);
2214 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2215 INT_H_SIZE, 0, peer_node, own_node,
2216 peer_port, tsk->portid, TIPC_OK);
2217 tsk->probing_state = TIPC_CONN_PROBING;
2218 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2222 tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2227 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2228 struct tipc_name_seq const *seq)
2230 struct net *net = sock_net(&tsk->sk);
2231 struct publication *publ;
2236 key = tsk->portid + tsk->pub_count + 1;
2237 if (key == tsk->portid)
2240 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2241 scope, tsk->portid, key);
2242 if (unlikely(!publ))
2245 list_add(&publ->pport_list, &tsk->publications);
2251 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2252 struct tipc_name_seq const *seq)
2254 struct net *net = sock_net(&tsk->sk);
2255 struct publication *publ;
2256 struct publication *safe;
2259 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2261 if (publ->scope != scope)
2263 if (publ->type != seq->type)
2265 if (publ->lower != seq->lower)
2267 if (publ->upper != seq->upper)
2269 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2270 publ->ref, publ->key);
2274 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2275 publ->ref, publ->key);
2278 if (list_empty(&tsk->publications))
2283 /* tipc_sk_reinit: set non-zero address in all existing sockets
2284 * when we go from standalone to network mode.
2286 void tipc_sk_reinit(struct net *net)
2288 struct tipc_net *tn = net_generic(net, tipc_net_id);
2289 const struct bucket_table *tbl;
2290 struct rhash_head *pos;
2291 struct tipc_sock *tsk;
2292 struct tipc_msg *msg;
2296 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2297 for (i = 0; i < tbl->size; i++) {
2298 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2299 spin_lock_bh(&tsk->sk.sk_lock.slock);
2301 msg_set_prevnode(msg, tn->own_addr);
2302 msg_set_orignode(msg, tn->own_addr);
2303 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2309 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2311 struct tipc_net *tn = net_generic(net, tipc_net_id);
2312 struct tipc_sock *tsk;
2315 tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2317 sock_hold(&tsk->sk);
2323 static int tipc_sk_insert(struct tipc_sock *tsk)
2325 struct sock *sk = &tsk->sk;
2326 struct net *net = sock_net(sk);
2327 struct tipc_net *tn = net_generic(net, tipc_net_id);
2328 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2329 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2331 while (remaining--) {
2333 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2334 portid = TIPC_MIN_PORT;
2335 tsk->portid = portid;
2336 sock_hold(&tsk->sk);
2337 if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2345 static void tipc_sk_remove(struct tipc_sock *tsk)
2347 struct sock *sk = &tsk->sk;
2348 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2350 if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2351 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2356 int tipc_sk_rht_init(struct net *net)
2358 struct tipc_net *tn = net_generic(net, tipc_net_id);
2359 struct rhashtable_params rht_params = {
2361 .head_offset = offsetof(struct tipc_sock, node),
2362 .key_offset = offsetof(struct tipc_sock, portid),
2363 .key_len = sizeof(u32), /* portid */
2365 .max_shift = 20, /* 1M */
2366 .min_shift = 8, /* 256 */
2367 .grow_decision = rht_grow_above_75,
2368 .shrink_decision = rht_shrink_below_30,
2371 return rhashtable_init(&tn->sk_rht, &rht_params);
2374 void tipc_sk_rht_destroy(struct net *net)
2376 struct tipc_net *tn = net_generic(net, tipc_net_id);
2378 /* Wait for socket readers to complete */
2381 rhashtable_destroy(&tn->sk_rht);
2385 * tipc_setsockopt - set socket option
2386 * @sock: socket structure
2387 * @lvl: option level
2388 * @opt: option identifier
2389 * @ov: pointer to new option value
2390 * @ol: length of option value
2392 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2393 * (to ease compatibility).
2395 * Returns 0 on success, errno otherwise
2397 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2398 char __user *ov, unsigned int ol)
2400 struct sock *sk = sock->sk;
2401 struct tipc_sock *tsk = tipc_sk(sk);
2405 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2407 if (lvl != SOL_TIPC)
2408 return -ENOPROTOOPT;
2409 if (ol < sizeof(value))
2411 res = get_user(value, (u32 __user *)ov);
2418 case TIPC_IMPORTANCE:
2419 res = tsk_set_importance(tsk, value);
2421 case TIPC_SRC_DROPPABLE:
2422 if (sock->type != SOCK_STREAM)
2423 tsk_set_unreliable(tsk, value);
2427 case TIPC_DEST_DROPPABLE:
2428 tsk_set_unreturnable(tsk, value);
2430 case TIPC_CONN_TIMEOUT:
2431 tipc_sk(sk)->conn_timeout = value;
2432 /* no need to set "res", since already 0 at this point */
2444 * tipc_getsockopt - get socket option
2445 * @sock: socket structure
2446 * @lvl: option level
2447 * @opt: option identifier
2448 * @ov: receptacle for option value
2449 * @ol: receptacle for length of option value
2451 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2452 * (to ease compatibility).
2454 * Returns 0 on success, errno otherwise
2456 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2457 char __user *ov, int __user *ol)
2459 struct sock *sk = sock->sk;
2460 struct tipc_sock *tsk = tipc_sk(sk);
2465 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2466 return put_user(0, ol);
2467 if (lvl != SOL_TIPC)
2468 return -ENOPROTOOPT;
2469 res = get_user(len, ol);
2476 case TIPC_IMPORTANCE:
2477 value = tsk_importance(tsk);
2479 case TIPC_SRC_DROPPABLE:
2480 value = tsk_unreliable(tsk);
2482 case TIPC_DEST_DROPPABLE:
2483 value = tsk_unreturnable(tsk);
2485 case TIPC_CONN_TIMEOUT:
2486 value = tsk->conn_timeout;
2487 /* no need to set "res", since already 0 at this point */
2489 case TIPC_NODE_RECVQ_DEPTH:
2490 value = 0; /* was tipc_queue_size, now obsolete */
2492 case TIPC_SOCK_RECVQ_DEPTH:
2493 value = skb_queue_len(&sk->sk_receive_queue);
2502 return res; /* "get" failed */
2504 if (len < sizeof(value))
2507 if (copy_to_user(ov, &value, sizeof(value)))
2510 return put_user(sizeof(value), ol);
2513 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2515 struct sock *sk = sock->sk;
2516 struct tipc_sioc_ln_req lnr;
2517 void __user *argp = (void __user *)arg;
2520 case SIOCGETLINKNAME:
2521 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2523 if (!tipc_node_get_linkname(sock_net(sk),
2524 lnr.bearer_id & 0xffff, lnr.peer,
2525 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2526 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2530 return -EADDRNOTAVAIL;
2532 return -ENOIOCTLCMD;
2536 /* Protocol switches for the various types of TIPC sockets */
2538 static const struct proto_ops msg_ops = {
2539 .owner = THIS_MODULE,
2541 .release = tipc_release,
2543 .connect = tipc_connect,
2544 .socketpair = sock_no_socketpair,
2545 .accept = sock_no_accept,
2546 .getname = tipc_getname,
2548 .ioctl = tipc_ioctl,
2549 .listen = sock_no_listen,
2550 .shutdown = tipc_shutdown,
2551 .setsockopt = tipc_setsockopt,
2552 .getsockopt = tipc_getsockopt,
2553 .sendmsg = tipc_sendmsg,
2554 .recvmsg = tipc_recvmsg,
2555 .mmap = sock_no_mmap,
2556 .sendpage = sock_no_sendpage
2559 static const struct proto_ops packet_ops = {
2560 .owner = THIS_MODULE,
2562 .release = tipc_release,
2564 .connect = tipc_connect,
2565 .socketpair = sock_no_socketpair,
2566 .accept = tipc_accept,
2567 .getname = tipc_getname,
2569 .ioctl = tipc_ioctl,
2570 .listen = tipc_listen,
2571 .shutdown = tipc_shutdown,
2572 .setsockopt = tipc_setsockopt,
2573 .getsockopt = tipc_getsockopt,
2574 .sendmsg = tipc_send_packet,
2575 .recvmsg = tipc_recvmsg,
2576 .mmap = sock_no_mmap,
2577 .sendpage = sock_no_sendpage
2580 static const struct proto_ops stream_ops = {
2581 .owner = THIS_MODULE,
2583 .release = tipc_release,
2585 .connect = tipc_connect,
2586 .socketpair = sock_no_socketpair,
2587 .accept = tipc_accept,
2588 .getname = tipc_getname,
2590 .ioctl = tipc_ioctl,
2591 .listen = tipc_listen,
2592 .shutdown = tipc_shutdown,
2593 .setsockopt = tipc_setsockopt,
2594 .getsockopt = tipc_getsockopt,
2595 .sendmsg = tipc_send_stream,
2596 .recvmsg = tipc_recv_stream,
2597 .mmap = sock_no_mmap,
2598 .sendpage = sock_no_sendpage
2601 static const struct net_proto_family tipc_family_ops = {
2602 .owner = THIS_MODULE,
2604 .create = tipc_sk_create
2607 static struct proto tipc_proto = {
2609 .owner = THIS_MODULE,
2610 .obj_size = sizeof(struct tipc_sock),
2611 .sysctl_rmem = sysctl_tipc_rmem
2614 static struct proto tipc_proto_kern = {
2616 .obj_size = sizeof(struct tipc_sock),
2617 .sysctl_rmem = sysctl_tipc_rmem
2621 * tipc_socket_init - initialize TIPC socket interface
2623 * Returns 0 on success, errno otherwise
2625 int tipc_socket_init(void)
2629 res = proto_register(&tipc_proto, 1);
2631 pr_err("Failed to register TIPC protocol type\n");
2635 res = sock_register(&tipc_family_ops);
2637 pr_err("Failed to register TIPC socket type\n");
2638 proto_unregister(&tipc_proto);
2646 * tipc_socket_stop - stop TIPC socket interface
2648 void tipc_socket_stop(void)
2650 sock_unregister(tipc_family_ops.family);
2651 proto_unregister(&tipc_proto);
2654 /* Caller should hold socket lock for the passed tipc socket. */
2655 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2659 struct nlattr *nest;
2661 peer_node = tsk_peer_node(tsk);
2662 peer_port = tsk_peer_port(tsk);
2664 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2666 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2668 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2671 if (tsk->conn_type != 0) {
2672 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2674 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2676 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2679 nla_nest_end(skb, nest);
2684 nla_nest_cancel(skb, nest);
2689 /* Caller should hold socket lock for the passed tipc socket. */
2690 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2691 struct tipc_sock *tsk)
2695 struct nlattr *attrs;
2696 struct net *net = sock_net(skb->sk);
2697 struct tipc_net *tn = net_generic(net, tipc_net_id);
2699 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2700 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2704 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2706 goto genlmsg_cancel;
2707 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2708 goto attr_msg_cancel;
2709 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2710 goto attr_msg_cancel;
2712 if (tsk->connected) {
2713 err = __tipc_nl_add_sk_con(skb, tsk);
2715 goto attr_msg_cancel;
2716 } else if (!list_empty(&tsk->publications)) {
2717 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2718 goto attr_msg_cancel;
2720 nla_nest_end(skb, attrs);
2721 genlmsg_end(skb, hdr);
2726 nla_nest_cancel(skb, attrs);
2728 genlmsg_cancel(skb, hdr);
2733 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2736 struct tipc_sock *tsk;
2737 const struct bucket_table *tbl;
2738 struct rhash_head *pos;
2739 struct net *net = sock_net(skb->sk);
2740 struct tipc_net *tn = net_generic(net, tipc_net_id);
2741 u32 tbl_id = cb->args[0];
2742 u32 prev_portid = cb->args[1];
2745 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2746 for (; tbl_id < tbl->size; tbl_id++) {
2747 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2748 spin_lock_bh(&tsk->sk.sk_lock.slock);
2749 if (prev_portid && prev_portid != tsk->portid) {
2750 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2754 err = __tipc_nl_add_sk(skb, cb, tsk);
2756 prev_portid = tsk->portid;
2757 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2761 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2766 cb->args[0] = tbl_id;
2767 cb->args[1] = prev_portid;
2772 /* Caller should hold socket lock for the passed tipc socket. */
2773 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2774 struct netlink_callback *cb,
2775 struct publication *publ)
2778 struct nlattr *attrs;
2780 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2781 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2785 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2787 goto genlmsg_cancel;
2789 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2790 goto attr_msg_cancel;
2791 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2792 goto attr_msg_cancel;
2793 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2794 goto attr_msg_cancel;
2795 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2796 goto attr_msg_cancel;
2798 nla_nest_end(skb, attrs);
2799 genlmsg_end(skb, hdr);
2804 nla_nest_cancel(skb, attrs);
2806 genlmsg_cancel(skb, hdr);
2811 /* Caller should hold socket lock for the passed tipc socket. */
2812 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2813 struct netlink_callback *cb,
2814 struct tipc_sock *tsk, u32 *last_publ)
2817 struct publication *p;
2820 list_for_each_entry(p, &tsk->publications, pport_list) {
2821 if (p->key == *last_publ)
2824 if (p->key != *last_publ) {
2825 /* We never set seq or call nl_dump_check_consistent()
2826 * this means that setting prev_seq here will cause the
2827 * consistence check to fail in the netlink callback
2828 * handler. Resulting in the last NLMSG_DONE message
2829 * having the NLM_F_DUMP_INTR flag set.
2836 p = list_first_entry(&tsk->publications, struct publication,
2840 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2841 err = __tipc_nl_add_sk_publ(skb, cb, p);
2843 *last_publ = p->key;
2852 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2855 u32 tsk_portid = cb->args[0];
2856 u32 last_publ = cb->args[1];
2857 u32 done = cb->args[2];
2858 struct net *net = sock_net(skb->sk);
2859 struct tipc_sock *tsk;
2862 struct nlattr **attrs;
2863 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2865 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2869 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2870 attrs[TIPC_NLA_SOCK],
2871 tipc_nl_sock_policy);
2875 if (!sock[TIPC_NLA_SOCK_REF])
2878 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2884 tsk = tipc_sk_lookup(net, tsk_portid);
2888 lock_sock(&tsk->sk);
2889 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2892 release_sock(&tsk->sk);
2895 cb->args[0] = tsk_portid;
2896 cb->args[1] = last_publ;